<%BANNER%>

Post-Harvest Recovery of Forest Structure and Spectral Properties after Selective Logging in Lowland Bolivia


PAGE 1

POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA By EBEN NORTH BROADBENT A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Eben North Broadbent

PAGE 3

Dedicated to the forests of Bolivia

PAGE 4

ACKNOWLEDGMENTS I would like to thank my advisor (Daniel J. Zarin, School of Forest Resources and Conservation). His unwavering support and guidance over these last 2.5 years made this thesis possible. I would like to thank Michael W. Binford (Land Use and Environmental Change Institute, Department of Geography) for helpful technical discussions, for my formal introduction to remote sensing and geographic information systems, and for his trust and generosity in allowing me use of the field spectrometry during my field work in summer 2003. I would like to thank Francis Jack Putz, Department of Botany, for taking the time to walk with me through the forests of the La Chonta concession; and for his valuable critiques of my field methodology. Thanks go to Ramon Littell, Department of Statistics, for insightful discussions regarding analysis of the field and remote sensing measurements. I would like to thank Gregory P. Asner (Carnegie Institution at Stanford University) for inviting me to his lab and for helpful discussions about remote sensing of selective logging. I also thank Amanda Cooper of the Carnegie Institution for processing my imagery several times, and for helpful discussions regarding the AutoMCU procedure. Great thanks go to Marielos Pea-Claros (director of forestry investigation at BOLFOR, and later IBIF) who made the field portion of my thesis possible. Our discussions of research ideas (during my internship with BOLFOR in 2002 before starting graduate school) and the many since then, helped make this thesis possible. Todd iv

PAGE 5

Frederickson, Joaqun Justiniano, and everybody else at BOLFOR helped me extensively through discussions, technical and logistical assistance, and basketball games. I thank the employees of the Superintendencia Forestal for generous use of their space during my presentation in summer 2004; and for their comments on my research. My field assistant and good friend Victor Hugo-Lopez, one of the hardest workers I know, motivated me to enter the densest liana tangles and provided insightful feedback regarding the field methodology. His passion and dedication to forestry and forest conservation in Bolivia will continue to inspire me. Lucas Fortini, Roberta Veluci-Marlow, and Kelly Keefe made my time in the lab and in Gainesville full of good memories. My friends Michael Burke, Michael McCarty, and Brad Rosenheim gave me their support and friendship when I needed it over the years. I thank my mother, Taihaku; and my father and stepmother, Jeff and Jeylan, for their love, support, and continual interest in my graduate studies over the years. Thanks also go to my sister, Leafye, for her love and support. I thank my wife, Angelica, for the uncountable nights spent laughing when we could have been stressed, for making my life as wonderful as it is, and for her continual help in making this thesis possible. Te amo siempre. v

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ...........................................................................................................viii LIST OF FIGURES .............................................................................................................x ABSTRACT ......................................................................................................................xii CHAPTER 1 INTRODUCTION AND LITERATURE REVIEW....................................................1 1.1 Deforestation, Forest Degradation, and Logging in Bolivia...................................1 1.2 Ecological Impacts of Selective Logging...............................................................3 1.2.1 Ground Area Disturbances...........................................................................5 1.2.2 Residual Stand Damage................................................................................6 1.2.3 Forest Canopy Damage................................................................................7 1.3 Remote Sensing of Selective Logging....................................................................9 1.3.1 Textural and Single Band Analysis............................................................11 1.3.2 Band Indices...............................................................................................12 1.3.3 Linear Spectral Mixture Model (LSMM) Analysis....................................14 1.3.4 Influences of Topography and Seasonality on Spectral Response.............19 2 POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA..........21 2.1 Introduction...........................................................................................................21 2.2 Site Description....................................................................................................23 2.3 Methods................................................................................................................26 2.3.1 Field Spatial Analyses................................................................................27 2.3.2 Field Measurements and Analyses.............................................................28 2.3.3 Remote Sensing Measurements and Analyses...........................................30 vi

PAGE 7

2.4 Results...................................................................................................................34 2.4.1 Field Spatial Analyses................................................................................34 2.4.2 Field Measurements....................................................................................36 2.4.2.1 Post-harvest recovery of forest structure in felling gaps..................36 2.4.2.2 Skid trails..........................................................................................43 2.4.3 Remote Sensing..........................................................................................43 2.4.3.1 Post-harvest recovery of spectral characteristics of felling gaps.....45 2.4.4 Linking Field and Remotely-Sensed Measurements..................................48 2.5 Discussion.............................................................................................................50 APPENDIX A GROUND, STAND AND CANOPY DAMAGE AFTER SELECTIVE LOGGING..................................................................................................................53 B MEAN MONTHLY PRECIPITATION IN LA CHONTA........................................56 C DISTRIBUTION OF FELLING GAP AREA SIZES WITHIN THE STUDY PARCELS...................................................................................................................57 D SKID TRAIL AREAS AND HARVEST INTENSITIES FOR ALL BOLFOR LONG TERM SILVICULTURAL RESEARCH PLOTS.........................................58 E SIGNIFICANCE AND F VALUES OF 2-WAY REPEATED MEASURES ANOVAS OF REMOTE SENSING VARIABLES OF FELLING GAP PIXELS.......................................................................................................................59 LIST OF REFERENCES...................................................................................................62 BIOGRAPHICAL SKETCH.............................................................................................80 vii

PAGE 8

LIST OF TABLES Table page 2-1. Characteristics of the selectively-logged parcels used in this study...........................26 2-2. Percent parcel area disturbed by tree fall gaps and skid trails....................................34 2-3. Sample size of large, medium, and small felling gaps within the logged parcels.......................................................................................................................36 2-4. The F and P values for the main effects of parcel, gap size, and gap zone, and their interactions for mixed 3-way ANOVAs of variables measured in felling gaps...........................................................................................................37 2-5. Mean values of field measurement variables within felling gaps for < 1-, 6-, 13-, and 19-months post-harvest parcels. Unlogged control forest values are provided for comparison.........................................................................38 2-6. Mean values (standard error) of field measurement variables for all large, medium, and small felling gaps......................................................................38 2-7. Mean values (standard error) of measured variables for all trunk and canopy felling gap zones..........................................................................................39 2-8. Mean (standard error) for field factors within skid trails < 1, 6, 13 and 19 months post-harvest.............................................................................................43 2-12. Pearson bivariate correlations between field and remote sensing measurements of felling gaps in the < 1 month post-harvest parcel........................49 2-13. Pearson bivariate correlations between field and remote sensing measurements of felling gaps in the 6 months post-harvest parcel..........................50 A-1. Section 1 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha)...................................................................................................................53 A-2. Section 2 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha)...................................................................................................................54 viii

PAGE 9

A-3. Section 3 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha)...................................................................................................................55 D-1. Skid trail areas and harvest intensities for all BOLFOR long term silvicultural research plots........................................................................................58 E-1. Significance and F values of 2-way repeated measures ANOVAs of remote sensing variables of felling gap pixels.........................................................59 E-2. Mean differences and P Values for two-way repeated measures ANOVA post-hoc comparisons (Dunnetts Test) of NDVI, PV, NPV, and soil fractions in large (> 800 m 2 ) felling gaps versus unlogged control parcel pixels..............................................................................................................60 E-3. Mean differences and P Values for two-way repeated measures ANOVA post-hoc comparisons (Dunnetts Test) of NDVI, PV, NPV, and soil fractions in medium (400 to 800 m 2 ) felling gaps versus unlogged control parcel pixels..............................................................................................................61 ix

PAGE 10

LIST OF FIGURES Figure page 1-1. Ground area disturbed for varying levels of harvest intensity in planned and unplanned logging...............................................................................................6 1-2. Residual stand damage for various levels and harvest intensity in planned and unplanned logging...............................................................................................7 1-3. Significant relationship between increasing % canopy cover loss and increasing harvest intensity for planned logging........................................................9 2-1. Location of research parcels in the La Chonta forestry concession...........................24 2-2. Locations of tree fall gaps and skid trails are shown for the logged study parcels.............................................................................................................35 2-3. Michaelis-Menten nonlinear model fit over harvest intensity versus % parcel area affected by skid trails.........................................................................36 2-4. Canopy openness of all felling gaps as affected by the interaction between gap size and gap zone................................................................................39 2-5. The field factors of Liana, NPV, soil coverage, and NPV height as affected by the interaction between parcel and gap zone.........................................41 2-6. Photosynthetic vegetation coverage as affected by the interaction between parcel and gap size.....................................................................................42 2-7. Control parcel mean values for NDVI and PV, NPV, and soil fractions versus image acquisition date...................................................................................44 2-8. Difference between spectral characteristics of large felling gap and unlogged control parcel pixels for NDVI, PV, NPV, and Soil................................46 2-9. Difference between spectral characteristics of medium felling gap and unlogged control parcel pixels for NDVI, PV, NPV, and Soil................................47 2-10. Difference between spectral characteristics of small felling gap and unlogged control parcel pixels for NDVI, PV, NPV, and Soil................................48 x

PAGE 11

B-1. Mean monthly precipitation (mm) measured in La Chonta from 1993................................................................................................................56 C-1. Gaps size classes for this study were small felling gaps < 400 m 2 medium felling gaps 400 m 2 to 800 m 2 and large felling gaps > 800 m 2 ................57 xi

PAGE 12

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA By Eben North Broadbent May 2005 Chair: Daniel J. Zarin Major Department: School of Forest Resources and Conservation Our study combined extensive field measurements of the spatial and temporal dynamics of felling gaps and skid trails < 1-19 months post-harvest in a forest in lowland Bolivia with remote sensing measurements through simultaneous ASTER satellite overflights during the summer of 2003. An advanced probabilistic spectral mixture model (referred to as AutoMCU ) was used to derive per-pixel fractional cover estimates of photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and soil. These results were compared with the normalized difference in vegetation index (NDVI) and field-derived GIS maps of felling gaps and skid trails. We found that NDVI, PV, NPV, and soil fractions were useful for identifying felling gaps > 400 m 2 for up to 6 months after logging, and for identifying felling gaps < 400 m 2 for up to 3 months after logging; but they were not useful for identifying skid trails. The PV fraction was most sensitive to felling gaps. The NPV and soil fractions were both highly correlated with topographic shade and were thus less useful for xii

PAGE 13

monitoring forest disturbances, especially in areas with more pronounced relief. These results identify important spatial and temporal thresholds relevant to monitoring selective logging with remote sensing; and may be used in the development of automated programs for identifying selectively logged forests in the region. xiii

PAGE 14

CHAPTER 1 INTRODUCTION AND LITERATURE REVIEW We examined post-harvest recovery of forest structure and spectral characteristics in felling gaps and skid trails after selective logging in a forest concession located in the Department of Santa Cruz, Bolivia. Here, is a brief overview of the forest sector in Bolivia; followed by a description of the ecological impacts of selective logging and recent approaches for monitoring selective logging in the tropics using analysis of remotely sensed imagery. 1.1 Deforestation, Forest Degradation, and Logging in Bolivia There are about 40 million ha of Amazonian lowland tropical forest in Bolivia (Steininger et al., 2001a) and over 53 million hectares (nearly 48% of the national territory) of forest cover in the country as a whole (Rodriguez, 2001). Recently, Bolivian lowland forests were placed among the top global conservation priorities (Myers et al., 2000; Steininger et al., 2001b) because of their high diversity of flora (Gentry, 1995) and fauna (Armonia., 1995; Stotz et al., 1996), and their abundance of habitat types (Prado and Gibbs, 1993; Killeen et al., 1998). Over 316 species of mammals, including 16 species of primates (Ergueta and Sarmiento, 1992); 1274 species of birds (Armonia., 1995); and over 20,000 species of flowering plants, including more than 2000 species of trees and shrubs (Malleux, 2000), have been discovered so far in Bolivian forests. Historically, deforestation rates in the Bolivian lowlands were low, with only 2.4 million ha (or 5.6%) of the original forested area cleared by 1990 (CUMAT, 1992). Recently, deforestation has accelerated from ~ 80,000 ha (or 0.2% of the forested area) 1

PAGE 15

2 per year in the late-1980s, to more than 270,000 ha per year through the mid-1990s (MDSMA, 1995; Rodriguez, 2001; Pacheco, 2002). Deforestation in the Amazon portion of the Department of Santa Cruz alone (which covers approximately 61% of the entire department) has increased dramatically from 38,000 ha per year cleared annually between 1986 to 1990 (CUMAT, 1992) to more than 200,000 ha per year in the mid-1990s (Camacho et al., 2001; Kaimowitz et al., 2002; Pacheco, 2002), partly as a result of increased colonization and expansion of cattle ranching (Pacheco, 2002) and soybean production (Kaimowitz, Thiele, and Pacheco, 1999; Kaimowitz and Smith, 2001). Analysis of the spatial patterns of deforestation in the Department of Santa Cruz, Bolivia, shows that areas near roads and population centers are most likely to become deforested (Kaimowitz et al., 2002). Selective logging, defined as the extraction of timber species (Verissimo et al., 1995) having the greatest economic value (Uhl, Barreto, and Verissimo, 1997), is also an important factor in degradation in Bolivia. Previous timber extraction in Bolivia depleted forests of mahogany (Swietenia macrophylla), oak (Amburana cearensis), cedar (Cedrela sp.), morado (Macherium sp.), tarara (Centrolobium sp.), and tajibo (Tabebuia sp.) (CORDECRUZ, 1994). In 2001, there were 40 million ha of forest designated for permanent forestry production (Rodriguez, 2001); equal to approximately 77% of the total national forest cover. Of this area only 8.5 million ha (or approximately 16 % of the total forested area) have active logging rights (Forestal, 2002). Illegal logging in Bolivia comprises much of the total selectively logged area. Cordero (2003) reports that of 133 inspections of logging operations by the Bolivian Superintendencia Forestal, 39% were found to be

PAGE 16

3 illegal with a further 19% legal, but not in compliance with regulations. In addition to causing extensive forest degradation, illegal logging makes legal operations less financially competitive; and illegally logged areas are often subsequently converted to pasture or agriculture, or burn in wildfires. 1.2 Ecological Impacts of Selective Logging Forest damage resulting from selective logging operations can be divided into the general categories of: (1) ground area disturbances; (2) residual stand damages; and, (3) canopy cover reductions (Uhl and Viera, 1989; White, 1994; Johns, Barreto, and Uhl, 1996; Jackson, Fredericksen, and Malcolm, 2002). Ground-area disturbance results from the construction and use of skid trails, logging roads, and log landings (Nicholson, 1958; Fox, 1968; Gillman et al., 1985; Jackson, Fredericksen, and Malcolm, 2002; Pereira et al., 2002), which may result in soil compaction and damage hydrological functions (Reisinger, Simmons, and Pope, 1988; Jackson, Sturm, and Ward, 2001). Residual stand damage results from harvesting of trees; which can damage or kill surrounding trees and vegetation and disturb regeneration. Stand-level disturbances can be inferred from changes in composition of forest regeneration after logging (Weaks and Creekmore, 1981; King and Chapman, 1983; Uhl and Viera, 1989; Panfil and Gullison, 1998; Fredericksen and Licona, 2000a; Fredericksen and Licona, 2000b; Fredericksen and Mostacedo, 2000; Jackson, Fredericksen, and Malcolm, 2002); from residual stand damage (Nicholson, 1958; Whitman, Brokaw, and Hagan, 1997); or simply from the overall reduction in basal area (Webb, 1997). Canopy-cover reduction results from felling of trees, which then causes further damage as they fall and remove other neighbor trees canopies through direct impact or liana intercanopy connections. Canopy disturbances, such as increases in canopy openness (Horne and Gwalter, 1982; Crome, Moore, and

PAGE 17

4 Richards, 1992; Pereira et al., 2002) due to selective logging have been insufficiently studied (Pereira et al., 2002). Though often difficult to distinguish disturbance category boundaries, they still serve as a simple foundation for comparisons between varying harvest intensities and silvicultural treatments. Reduced impact logging (RIL) techniques can significantly minimize forest damage as compared with conventional logging (CL). The main components of RIL logging (Putz and Pinard, 1993; Pinard and Putz, 1996; Bertault and Sist, 1997; Uhl, Barreto, and Verissimo, 1997; Pinard, Putz, and Tay, 2000; Sist, 2000; Pereira et al., 2002) are Inventory and mapping to reduce waste during logging Planning of roads, log decks, and skid trails Vine cutting prior to harvest Planning of extraction, and Directional felling and bucking of trunks. Many of these RIL components were mandated in 1996 when the Bolivian government implemented Forestry Law #1700, which instituted new legal and regulatory frameworks for control and monitoring of forestry operations (Griffith, 1999; Alvira, Putz, and Fredericksen, 2004). The forest damage resulting from CL and RIL techniques varies widely, and depends on factors such as basal area removed, minimum cutting diameters, and forest type (Gullison and Hardner, 1993; Pinard and Putz, 1996; Panfil and Gullison, 1998). Conventional logging has been shown to damage 10-40% of the living forest biomass (Uhl et al., 1991); may disrupt ecological processes (Uhl and Viera, 1989), including the regeneration of commercially valuable species (Fredericksen and Licona, 2000a; Fredericksen and Mostacedo, 2000); may alter species composition (Johns, 1992;

PAGE 18

5 Fredericksen et al., 1999; Lewis, 2001; Sekercioglu, 2002; Fredericksen and Fredericksen, 2002); and may affect forest biogeochemical processes (Asner, Keller, and Silvas, 2004). 1.2.1 Ground Area Disturbances Ground area disturbances (GAD) commonly result from the construction and use of log landings, logging roads, skid trails, or impacts from tree felling. Effects of GAD include increased levels of soil compaction (Whitman, Brokaw, and Hagan, 1997; Jackson, Fredericksen, and Malcolm, 2002), and altered site hydrology (Asdak et al., 1998; Fletcher and Muda, 1999; Tague and Band, 2001). Relationships between harvest intensity and GAD are difficult to establish due to the large variety of harvesting practices (Gullison and Hardner, 1993). Figure 1-1 illustrates the relationship for data obtained from 25 published articles, of which 14 were classified as having planned and 11 unplanned logging operations. Ground area disturbances were less common after reduced-impact logging (RIL) compared to conventional logging (CL) in Paragominas, Brazil (Johns, Barreto, and Uhl, 1996; Pereira et al., 2002). Panfil & Gullison (1998) found a strong relationship between increasing harvest intensities and increasing GAD within the Chimanes Forest, Bolivia, at even the relatively low harvest intensities of 1 to 6 trees/ha. Asner et al. (2004) found that between 4.8 and 11.2 % of the ground area was disturbed after RIL and CL, respectively, logging in an Amazonian forest. Harvest intensities for the RIL and CL logging in their study were 3 and 6.4 trees/ha, respectively. Skid trails comprised 2.9 to 8.8% of the parcel following harvest. Jackson et al. (2002) found that selective logging at 4.35 trees/ha in a tropical forest in Bolivia damaged approximately 50% of the total area studied.

PAGE 19

6 Figure 1-1. Ground area disturbed for varying levels of harvest intensity in planned and unplanned logging. Data from sources cited in Appendix A. Of the disturbed area half was in the form of skid trails, roads, and log landings, and half was in the form of felling gaps (Jackson, Fredericksen, and Malcolm, 2002). Pereira et al. (2002) studied differences in disturbed area and canopy openness for RIL and CL techniques between 100 ha plots that were harvested at similar intensities (~3 individuals/ha). The total ground area disturbed was twice as great for CL (8.9 to 11.2% vs. 4.6 to 4.8% for CL and RIL, respectively) as it was for RIL (Pereira et al., 2002). 1.2.2 Residual Stand Damage Aspects of residual stand damage, such as tree mortality (Johns, Barreto, and Uhl, 1996; Bertault and Sist, 1997; Webb, 1997; Panfil and Gullison, 1998; Sist et al., 1998; Sist and Nguyen-The, 2002), alterations in subsequent species composition (Panfil and Gullison, 1998), and regeneration (Fredericksen and Mostacedo, 2000) have been widely studied in the tropics. A significant linear relationship (P < 0.05, R 2 = 0.96) was found between increasing percentages of residual stand damage and increasing harvest intensity

PAGE 20

7 for unplanned logging operations for the literature reviewed in Appendix A, although the sample size (n = 4) was limited. Figure 1-2. Residual stand damage for various levels and harvest intensity in planned and unplanned logging. Data from sources cited in Appendix A. 1.2.3 Forest Canopy Damage Reductions in forest canopy cover are strongly related to silvicultural treatments, such as pre-harvest liana cutting (Putz, 1992; Vidal et al., 1997; Pinard, Putz, and Licona, 1999), as well as to increased harvesting intensity (Gullison and Hardner, 1993; Bertault and Sist, 1997). Silvicultural interventions can increase or decrease canopy coverage. For example, tree girdling or poisoning, and the cutting of unmarketable species, can lead to larger canopy reductions. Other techniques, such as vine cutting to minimize inter-canopy connections, can reduce the canopy damage per tree harvested for some forest types (Appanah and Putz, 1984; Vidal et al., 1997). Panfil & Gullison (1998) found a correlation between increasing harvest intensities and increased canopy damage, which reached an asymptote at greater harvesting

PAGE 21

8 intensities due to the re-use of previously constructed skid trails and logging roads. Pereira et al (2002), found canopy openness of 16.5% and 21.9% for two CL blocks as compared with 4.9% and 10.9% for two blocks harvested with a RIL approach that included extensive pre-planning of roads, log decks and skid trails, as well as planned directional felling and vine cutting prior to harvest (Uhl, Barreto, and Verissimo, 1997). Other studies (Hendrison, 1990; Johns, Barreto, and Uhl, 1996), have found similar results. Temporal patterns of recovery following logging operations have been studied extensively at the residual stand level (Dickinson, Whigham, and Hermann, 2000; Fredericksen and Pariona, 2002), including alterations in timber regeneration (Fredericksen and Mostacedo, 2000) and residual stand mortality rates (Sist and Nguyen-The, 2002). Canopy level damage, however, and its subsequent recovery (i.e. canopy closure), have been little studied (Pereira et al., 2002). Cannon et al. (1994) assessed canopy damage and closure for blocks in West Kalimantan, Indonesia. Three sites were selected that had undergone similar harvest intensities and had been harvested 0.5, 1 and 8 years prior to their study. The overall canopy openness for the sites decreased from 63% to 49% to 21% with increasing time since harvesting. An analysis of planned and unplanned logging in Appendix A showed a significant linear relationship (P < 0.05, R 2 = 0.72, n = 7) between increasing harvest intensity and increasing levels of canopy loss (Figure 1-3). No relationship was found for unplanned logging, though it was obvious that much greater levels of canopy loss occurred at low harvest intensities. The utility of the relationship between increasing harvest intensity and increasing canopy loss for remotely monitoring logging operations is discussed in the

PAGE 22

9 following section. Forest canopy damage is markedly reduced in RIL relative to CL harvests (Howard, Rice, and Gullison, 1996; Johns, Barreto, and Uhl, 1996; Pereira et al., 2002). Figure 1-3. Significant relationship between increasing % canopy cover loss and increasing harvest intensity for planned logging. Data from sources cited in Appendix A. 1.3 Remote Sensing of Selective Logging Improving the ability to estimate the extent, and intensity, of selective logging from remote sensors is essential for accurate modeling of carbon sequestration and release (Schroeder and Winjom, 1995; Schroeder and Winjum, 1995; Potter, 1999), developing effective wild-fire control policies (Holdsworth and Uhl, 1997; Nepstad et al., 1999; Cochrane and Laurance, 2002; Cochrane, 2003), conservation of fauna and flora, and monitoring logging activities (Keller et al., 2002). Currently, estimates of deforestation rates in the tropics are based primarily on remote sensing analyses discriminating between forested and non-forested regions (Skole, 1993; Steininger et al., 2001a; Achard

PAGE 23

10 et al., 2002). Landsat-based measurements of total forest conversion, normally to agriculture or pasture, are convenient as rapid and cost efficient deforestation estimators (Skole, 1993; Skole and Tucker, 1993) but are incapable of detecting forest areas degraded due to selective logging or fire (Stone and Lefebvre, 1998; Nepstad et al., 1999). Acquiring accurate estimates of selective logging rates within the tropics has proven difficult (Stone and Lefebvre, 1998; Asner et al., 2002) because selective logging damage occurs on a fine spatial grain (Souza and Barreto, 2000; Pereira et al., 2002) compared with the spatial resolution of commonly available satellite imagery. Furthermore, the rapid regeneration of pioneer species in logged areas (Dickinson, Whigham, and Hermann, 2000; Fredericksen and Licona, 2000b; Fredericksen and Mostacedo, 2000) reduces indicators visible through optical remote sensing within 3 to 5 years (Stone and Lefebvre, 1998). Souza and Barreto (2000) were able to detect only 60% of field-verified logging patios (where logs are brought before being loaded on trucks for transport to a sawmill), and none that were > 3 years old due to rapid regeneration of vegetation. Prevalent remote sensing image analysis techniques used for monitoring of forest disturbances are active radar (Siegert and Hoffmann, 2000; Siegert et al., 2001), texture analysis (Stone and Lefebvre, 1998; Asner et al., 2002), and basic relationships between single radiometric bands and logging disturbances (Asner et al., 2002). Vegetation indices, such as NDVI, have had limited utility (Jasinski, 1990; Carlson and Ripley, 1997; Stone and Lefebvre, 1998). Among the more successful methods for identifying logging disturbances are deriving per-pixel % vegetation cover (Todd and Hoffer, 1998)

PAGE 24

11 from linear spectral mixture models (Hall, Shimabukuro, and Huemmrich, 1995; Garcia-Haro, Gilabert, and Melia, 1996; Cochrane and Souza, 1998; Shimabukuro et al., 1998; Souza and Barreto, 2000), or probabilistic Monte Carlo versions of linear spectral mixture models, such as AutoSWIR or AutoMCU (Asner and Lobell, 2000b; Lobell et al., 2001). 1.3.1 Textural and Single Band Analysis Textural analysis uses multi-pixel comparisons to enhance or diminish existing spatial variation (Asner et al., 2002; Debeir et al., 2002). Single band analysis compares the digital number, radiance, or reflectance from a single sensor band. Vegetation stress, or lack of typical vegetation spectral response, and the ability to discriminate vegetation from exposed soil and non-photosynthetic vegetation (residual logging slash) following selective logging may enhance the ability of remote sensors to identify these areas and will therefore be discussed in the following sections. Individual Landsat Thematic Mapper (T.M) band reflectances can correlate with logging disturbances. Landsat TM bands 1, 2, and 3 (covering the visible spectrum between 0.45 and 0.69 um) have been helpful in delineating areas of exposed soil following selective logging (Asner et al., 2002), possibly due to decreases in moisture content to which these bands are sensitive (Ripple, 1986; Bowman, 1989). Bands 1 and 2 are often avoided in such analyses because they are susceptible to atmospheric aerosol contamination (Krueger and Fischer, 1994; Asner et al., 2002). Landsat TM band 3 (red; centered at 0.67 nm) shows vegetation as very dark due to radiation absorption by foliar chlorophyll (Gaussman, 1977). TM band 4 (near-IR; centered at 0.83 nm) shows vegetation with high reflectance due to non-linear scattering of light by foliage, with soils and litter (NPV) having lower reflectance levels (Asner,

PAGE 25

12 1998). These bands has been shown to correlate both positively (Thomas et al., 1971) and negatively (Penuelas et al., 1993) with vegetation drought stress according to a complex suite of leaf physiological factors, including variations in leaf area index (LAI) and greater shadowing from leaves wilting or curling up when exposed to increasing levels of drought stress (Jackson and Ezra, 1985). The short wave infrared (SWIR) region of the spectrum, measured by Landsat TM bands 5 and 7, is a water absorption peak and thus decreasing SWIR reflectance has been found to correlate with increasing foliar water content (Tucker, 1980; Ripple, 1986; Bowman, 1989). Drought stress measurements using Landsat TM band 6 (thermal; centered at 11.45 nm) have focused on the increases in temperature (the thermal response) of plant foliage suffering water stress compared to the temperature of the surrounding air (Chuvieco et al., 1999). Asner et al. (2002) combined field measurements of canopy gap fraction along a time series with textural and band-by-band analysis of Landsat 7 Enhanced Thematic Mapper Plus (ETM+) data. Textural analysis was used to enhance post-logging variations between canopy and gap reflectance. These techniques, textural and single band analysis, were found sensitive only to high levels of canopy damage (>50% increase in canopy openness) and temporally limited to within 0.5 years post-harvest. These techniques may have some potential for broad delineation of very recently logged forests but are not useful for more detailed analyses of ecological or biogeochemical forest processes (Asner et al., 2002). 1.3.2 Band Indices Band combinations, ratios, and indices provide a powerful image analysis tool for the assessment of moisture content, vegetation stress, and related logging damages

PAGE 26

13 (Cibula, Zetka, and Rickman, 1992; Adegoke and Carleton, 2002; Aparicio et al., 2002). In the early 1980s it was found that leaf water content and photo-synthetically active biomass could be monitored through linear combinations of red and IR radiance changes (Tucker, 1979) and spectrum wavelengths between 1.55 and 1.75 um. Using these relationships Hunt et el. (1987 & 1989) developed a leaf water content index incorporating the wavelengths between .76-.90 um and 1.55-1.75 um. Band indices have been developed to measure vegetation stress and moisture content (Gilabert et al., 2002). They have been used to estimate chlorophyll content (Tucker, 1979) and photosynthesis rates (Choudhury, 1987), primary productivity (Curren, 1980), susceptibility to wild fire (Chuvieco et al., 1999; Chuvieco et al., 2002) and leaf aging, drop and stress (Bohlman, Adams, and Peterson, 1998) in the Amazon. Indices, such as the soil and atmosphere resistant vegetation index (SARVI) (Huete et al., 1997), water deficit index (WDI) (Moran et al., 1994) or equivalent water thickness (EWI) (Ceccato et al., 2001; Ceccato, Flasse, and Gregoire, 2002; Ceccato et al., 2002), have been developed to estimate vegetation stress and water content with good accuracy. The ratio of TM bands 4 to 5 has been found to be strongly indicative of changes in leaf water content (Hunt, Rock, and Nobel, 1987; Hunt and Rock, 1989). This is because decreases in leaf water content increase reflectance in the middle infrared spectrum while having little effect on reflectance in the near infrared spectra (TM band 5 and 4, respectively) (Knipling, 1970; Carter, 1991; Aldakheel and Danson, 1997). Rock et al. (1994) found that, for an area with 100% vegetation cover, TM band 5 reflectance increased with increasing water stress with no change in the reflectance of TM band 4. Bohlman and Adams (1998) used the TM band 4 to 5 ratio to determine leaf

PAGE 27

14 aging, leaf drop and water stress during the transition from wet season to dry season for forests in Maraba, Brazil. Several studies have noted the utility of the short wave infrared (SWIR 2 region, 2080 nm) for remote measurement of leaf moisture content and discrimination of vegetation from soil and non-photosynthetic vegetation (i.e. slash or litter) (Asner, 1998; Asner and Lobell, 2000a, b; Lobell et al., 2001) due to the dominance of water absorption by green plant spectra (Elvidge, 1990; Drake, Mackin, and Settle, 1999). 1.3.3 Linear Spectral Mixture Model (LSMM) Analysis Linear spectral un-mixing (Heinz, Chang, and Althouse, 1999; Heinz, 2001) or AutoSWIR (Asner and Lobell, 2000b) techniques decompose pixels to associated fractions of multiple selected materials of interest (MOI), termed endmembers. These materials of interest are chosen according to the ecological properties of the field location (as general as vegetation and soil or as specific as individual species if spectral seperability is adequate) and the desired resultant data product (Adams et al., 1995; Garcia-Haro, Gilabert, and Melia, 1999; Bateson, Asner, and Wessman, 2000). For successful linear un-mixing of pixel reflectance materials of interest must be chosen that exhibit purity or extremity within the dataset and are known to contribute to pixel reflectance over the entire landscape of study (Schanzer, 1993; Bateson, Asner, and Wessman, 1998; Bateson, Asner, and Wessman, 2000; Heinz, 2001). Sub-pixel fractions, with appropriate endmembers, will sum to 100% pixel reflectance (Asner, Hicke, and Lobell, 2002). The number of endmembers that can be unmixed from any given pixel is dependent on the dimensionality (a function of the number of bands and amount of random noise in the data) of the satellite imagery (Asner and Lobell, 2000b; Asner, Hicke, and Lobell, 2002).

PAGE 28

15 A least-squares based linear mixing model (constrained to 1) can be simplified to the following formula (Shimabukuro and Smith, 1991; Asner, Hicke, and Lobell, 2002): r 1 = (a ij x j ) + 1 where x j = 1 and, r i = spectral reflectance at the ith spectral band of the pixel; a ij = spectral reflectance known to the j th component at the i th spectral band; x j = value to be estimated from the proportion of the j th component within the pixel; e 1 = estimation error for the i th spectral band; i = number of spectral bands considered; j = number of components. The root mean square error fraction can serve as an indicator of how good the chosen endmembers are for the particular pixel. A classic study of forested ecosystem reflectance using hyper-spectral AVIRIS found that 98% of spectral variation was explained by linear mixtures of three endmembers: photosynthetic vegetation (PV), shade, and soil (Roberts, Smith, and Adams, 1993), with non-photosynthetic vegetation (NPV) not able to be directly distinguished from the soil endmember. NPV differences, however, discriminated through analysis of residual spectra were helpful in distinguishing distinct communities of green vegetation. The characteristic spectra of photosynthetic vegetation, dominated by foliar water absorbance across the spectrum (Elvidge, 1990) and C-H and O-H bonds in the SWIR region (Curran, 1989), as well as the presence of chlorophyll. NPV, lacking the water content of PV, and having distinct spectral features resulting from organic carbon bonds interacting with solar radiation (Curran, 1989), can be distinguished from the more similar spectra of soil partly through differences in the SWIR response attributable to the effects of cellulose and lignin in the vegetation (Roberts, Smith, and Adams, 1993) on the SWIR spectrum. Soil spectral properties vary according to mineralogy, clay content

PAGE 29

16 (Drake, Mackin, and Settle, 1999), and moisture content (Weidong et al., 2002) and roughness (Pinty, Verstraete, and Gobron, 1998). Shade spectra have been developed through both inversion of standard linear mixing models (Roberts, Smith, and Adams, 1993), and sampling of shaded areas in satellite imagery (Shimabukuro et al., 1998; Souza and Barreto, 2000), among other methods. Integration of a shade endmember can be helpful to compensate for the effect of topography, which is caused by differential illumination of the Earths surface and generally results in darker slopes facing away from the sun (Civco, 1989), and interand intra-canopy shadowing (Asner and Warner, 2003) which are both prevalent in satellite imagery. Shade can also be incorporated directly into the endmember bundles (Asner et al., 2004). Other methods for topographic correction are based on the Lambertian assumption that measured reflectance does not vary with view angle (Holben and Justice, 1980), which provide poor normalizations and often over-corrects for topography (Jensen, 1996), or on complicated non-Lambertian models that require development and validation of image based coefficients (Hodgson and Shelley, 1994). Linear spectral mixture analysis (SMA) has been used with success to locate areas of recent fire (Wessman, 1997; Cochrane and Souza, 1998), exposed soil from recent logging (Souza and Barreto, 2000), differentiate forested from non-forested areas through differences in shade fractions (Shimabukuro et al., 1998), estimate Amazonian transition forest biomass (Santos et al., 1999), and assess general land-cover change in tropical Amazonian forests (Adams et al., 1995). Souza and Barreto (2000) (Souza and Barreto, 2000) used LSMA to detect selectively logged Amazonian forest based upon sub pixel soil fractions. The study sites had 5 to 7 trees harvested per ha (Johns, Barreto, and Uhl,

PAGE 30

17 1996) for planned and unplanned selective logging. Souza and Barreto (2000) chose pixels having 20% or greater soil fraction as indicators of patios used for selective logging. This technique was temporally limited due to rapid regeneration of pioneer vegetation over exposed soil areas and was unable to locate sites five years post-logging. Adams et al. (1995) divided an Amazonian landscape into seven general categories based upon their LSMA endmember fractions. Primary forest areas visibly differed from those with large quantities of slash (i.e., areas recently selectively logged) due to higher fractions of NPV and soil, and reductions in the shade fraction. The temporal variations in the abundance of each endmember were used to assess landscape cover change. Asner et al. (2004) estimated per-pixel fractional cover of photosynthetic (PV), and non-photosynthetic (NPV) vegetation, and soil in Amazonian forests near Paragominas Brazil using an automated un-mixing model, termed AutoMCU incorporating endmember bundles. Endmember bundles are field derived reflectances (measured with a field spectroradiometer) of materials that encompass the full naturally occurring variability within the endmember class (for example, a specific soil at different moisture levels) (Bateson et al., 1998). Pixel un-mixing using endmember bundles allows for higher accuracy levels of sub-pixel fractional compositions and confidence interval estimates for those fractions (Asner and Lobell, 2000b). Asner et al. (2004) found significant differences between conventional (CL) and reduced-impact logging (RIL) PV, NPV, and soil endmember fractions that varied strongly with time since harvesting due to gap regeneration and canopy closure. Asner et al. (2002), employing the AutoMCU unmixing technique and endmember bundles was able to discriminate selectively logged areas in the eastern

PAGE 31

18 Amazonia for up to 3.5 years post logging. Canopy openness was found to be greater following conventional logging than reduced-impact logging immediately. Subsequent satellite and field-based measurements of canopy gap fraction were highly and inversely correlated. The 50% decrease in canopy openness derived from the unmixing process agreed well with ground based measurements. This technique seems the most promising but has yet to be studied for the gradient of harvest intensities and variety of forest types necessary for broader application in Amazon forests. In November 2003 a special issue of Remote Sensing of the Environment was published dedicated to remote sensing analyses of land use and land cover change, including selective logging, in the Brazilian Amazon. Five of the 13 papers in this issue featured linear spectral mixture model methodology. Numata et al. (2003) assessed sub-pixel fractional cover of green vegetation, shade, soil, and non-photosynthetic vegetation within pastures and found that they were dominated by NPV, whose dominance increased with increasing pasture age. Dengsheng, Lu et al. (2003) found that a LSMM approach was a promising method for distinguishing successional and mature forests, and that sub-pixel percentages of green vegetation and shade were the most sensitive to changes in forest structure. A study by Souza et al (2003) analyzing sub-pixel fractions PV, NPV, soil, and shade found that NPV was positively correlated with aboveground biomass and improved the ability to map selectively logged forests. A decision tree approach (dichotomous categorization) of the sub-pixel fractions was used to then successfully delineate between intact, logged, and regenerating forests (Souza et al., 2003). Asner et al. (2003) used the AutoMCU approach (Monte Carlo approach incorporating endmember bundles) to unmix Landsat Thematic Mapper pixels in areas bordering the

PAGE 32

19 Tapajos National Forest in the Central Amazon and found that PV and NPV fractions were useful for quantifying biophysical variability within and between pixels 1.3.4 Influences of Topography and Seasonality on Spectral Response The topographic effect is caused by differential solar illumination of the Earths surface, which generally results in darker slopes facing away from the sun and brighter slopes facing the sun (Civco, 1989). The topographic effect is a combination of Incident illumination defined as the orientation of the land surface to the suns rays Exitance reflectance defined as the energy reflected as a result of the slope, and Land topography and shadowing (ERDAS, 1999). Together these factors can cause identical land cover to be represented by different intensity values depending on the degree of shadowing. An ideal topographic normalization removes all intensity variation resulting from differential illumination, creating a pseudo flat reflectance surface. It is possible to correct for the topographic effect using either Lambertian or non-Lambertian reflectance models or through the use of mixture model techniques employing shade endmembers (Souza and Barreto, 2000). The Lambertian model normalizes imagery according to the cosine of the sun illumination angle (zenith) at the time of image acquisition and the slope/aspect information from the digital elevation model (DEM) of the area (Smith, Lin, and Ranson, 1980). The Lambertian assumption, that the measured reflectance does not vary with view angle (Holben and Justice, 1980), provides poor normalizations and often over-corrects images, with sun-facing slopes appearing darker than those facing away from the sun (Civco, 1989). This results from not including non-Lambertian scattered reflectance, such as diffuse skylight or light reflected from surrounding mountainsides (Jensen, 1996).

PAGE 33

20 Minneart and Sceicz (1961) proposed that all surfaces do not reflect incident radiation uniformly. The non-Lambertian reflectance model compensates for this by using image based correction factors within the algorithm (Hodgson and Shelley, 1994) and has been shown to have higher accuracy than Lambertian models (Smith, Lin, and Ranson, 1980) and fewer problems with over-correction (Civco, 1989). However, the development of a non-Lambertian model is time consuming and often requires field truthed data (ERDAS, 1999). Few studies have closely examined the abilities of a shade fraction or endmember bundle to remove the effect of topography. However, in general, topography is thought to have a minimal effect on the response of band indices, such as NDVI, as shade is largely photometric and results in a general decrease in spectral response which is independent of the bandwidth. A band ratio (such as NDVI) could therefore compensate for topographic differences and return topographically independent results. The effect of seasonality on the response of NDVI has been investigated in north-west Mexico (Salinas-Zavala, Douglas, and Diaz, 2002), where strong correlations were found between pluviometric data and atmospheric circulation and changes in NDVI. Other studies have found correlations between seasonality, including forest phenology, and the landscapes spectral response (Roberts et al., 1998; Asner and Lobell, 2000b; Ferreira et al., 2003; Siqueira, Chapman, and McGarragh, 2003).

PAGE 34

CHAPTER 2 POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA 2.1 Introduction Timber production within the Amazon basin has been estimated at 30 million cubic meters per year, based on regional sawmill production, but estimates of the areal extent and intensity of the selective logging practices that supply that timber are very poorly constrained (Nepstad et al., 1999; Lentini, Verissimo, and Sobral, 2003; Nepstad et al., 2004). Much of the selective logging in the region is clandestine, and in many cases, even legally registered forest management plans are extremely imprecise. Improving the ability to estimate the extent, and intensity, of selective logging is essential for monitoring of logging activities (Keller et al. 2002), and for accurate modeling of carbon sequestration and release (Schroeder and Winjom, 1995; Schroeder and Winjum, 1995; Potter, 1999), and developing effective wild-fire control policies (Holdsworth and Uhl, 1997; Nepstad et al., 1999; Cochrane and Laurance, 2002; Cochrane, 2003). Remote sensing technology may offer an objective means of determining the location, extent, and intensity of selective logging, but its use for those purposes is challenging because selective logging damage often occurs on a finer spatial grain than the spatial resolution of commonly available satellite imagery (Stone and Lefebvre, 1998; Souza and Barreto, 2000; Asner et al., 2002; Pereira et al., 2002), and forests rapidly regenerate after logging (Dickinson, Whigham, and Hermann, 2000; Fredericksen and 21

PAGE 35

22 Licona, 2000b; Fredericksen and Mostacedo, 2000) reducing indicators visible through optical remote sensing (Stone and Lefebvre, 1998). In Bolivia, selective logging is an important cause of degradation in the countrys lowland Amazon region (Cordero 2003). Previous timber extraction in Bolivia depleted forests of mahogany (Swietenia macrophylla), oak (Amburana cearensis), cedar (Cedrela sp.), morado (Macherium sp.), tarara (Centrolobium sp.), and tajibo (Tabebuia sp.) (CORDECRUZ, 1994). Although recent changes in the Bolivian Forestry Law provide an exemplary framework for good forest management (1996; Griffith, 1999), clandestine and poorly regulated logging activities continue, and the extent and intensity of ongoing selective logging in the Bolivian Amazon has not been quantified (Cordero, 2003). This study was designed to examine the potential applicability of remote sensing technology to the detection of selective logging in the Bolivian Amazon. In it I employ intensive spatial and temporal field measurements of structural changes associated with selective logging and then used these measurements to test the sensitivity and examine the temporal and spatial thresholds of a commonly used remote sensing vegetation index and an advanced linear spectral unmixing method. The unmixing method has previously been used for detection of selective logging in a limited number of locations in the Brazilian Amazon, where both standing and harvested volumes are substantially larger than at the study area I examined in Bolivia (Asner et al., 2002; Asner et al., 2004; Asner, Keller, and Silvas, 2004). The results of the analysis I present here can inform future efforts at monitoring the areal extent and spatial distribution of selective logging using remotely-sensed data.

PAGE 36

23 2.2 Site Description The study was conducted in the Agroindustria Forestal La Chonta Ltda. timber concession (15 47 S, 62 55 W) which encompasses 100,000 ha in the Guarayos forest preserve in the Department of Santa Cruz, Bolivia (Figure 2-1). The topography is slightly undulating and the vegetation is classified as Subtropical Humid Forest according to the Holdridge Life Zone System (Holdridge, 1971) and has an average biomass of 73 to 190 Mg/ha (Dauber, Teran, and Guzman, 2000). The elevation is 400 to 600 m above sea level, otherwise referred to as the Bolivian lowlands. Common canopy trees in the area, such as Hura crepitans, Ficus boliviana, and Pseudolmedia laevis, are typical of humid forests within Bolivia (Jackson, Fredericksen, and Malcolm, 2002). The average annual temperature is 15.3 C and the mean annual precipitation is 1,560 mm, though 77% of the annual precipitation falls between November and April (Appendix B). During the dry season temperatures often drop to 5 to 10 C due to Antarctic fronts (Gil, 1998). The soils are primarily moderately fertile inceptisols, though large areas of black anthrosols can be found throughout the concession (Calla, 2003; Paz, 2003). The region is vulnerable to wildfires (CAF, BOLFOR, and Geosystems, 2000), and 30% of the concession burned in 1995 (Gould et al., 2002).

PAGE 37

24 Figure 2-1. Location of research parcels in the La Chonta forestry concession, Department of Santa Cruz, Bolivia. The parcel boundaries are overlaid on a RGB composite image of ASTER bands 2, 3, and 1, respectively, from an image acquired on 30 June 2004. There are approximately 100 tree species with individuals >20 cm diameter at breast height (DBH) within La Chonta (Gil, 1997; Gil, 1998). The mean tree density is 88 trees/ha (Alvira, 2002). Eighteen timber species are currently harvested, including Ficus. sp., Pseudolmedia laevis, Hura crepitans, Ceiba pentandra, and Spondias mombin (BOLFOR, 2000). Average canopy height is 21 m (unpublished data) within a non-logged control parcel. The concession was previously high-graded for mahogany (Swietenia macrophylla) (Gil, 1998), but over 60% of the area is considered to be suitable for sustained-yield

PAGE 38

25 timber harvesting (Gil, 1997). The current annual cut is 2400 ha producing a wood volume of approximately 51,000 m 3 (Jackson, Fredericksen, and Malcolm, 2002). Average harvest intensity is 4.35 trees/ha (or 12.3 m 3 /ha of wood) (Jackson, Fredericksen, and Malcolm, 2002). The cutting cycle, as set by forestry law #1700 introduced in 1996, is 30 years (1996; Fredericksen, 2000), though several years are granted to complete extraction of a given annual block. La Chonta was certified by the Forest Stewardship Council (FSC) in early 1990 and abides by certification standards, including implementation of reduced impact logging (RIL) techniques (Johns, Barreto, and Uhl, 1996; Uhl, Barreto, and Verissimo, 1997; Nittler and Nash, 1999; Sist, 2000; Pereira et al., 2002) Inventory and mapping of trees to be harvested Planning of roads, log decks, and skid trails Vine cutting prior to harvest when necessary Directional felling, and Planning of extraction. Harvesting is based on a 50 cm minimum DBH cutting limit, with the exception of Hura crepitans and Ficus glabrata that have a minimum DBH of 70 cm (Jackson, Fredericksen, and Malcolm, 2002). Twenty percent of harvestable trees are left as seed trees. One year prior to harvesting, crop trees are selected, marked, and mapped, and some of the lianas in their crowns are cut (Alvira, 2002; Krueger, 2003). Prior to harvest, skid trails are built every 150 m intervals perpendicular to the main access road (Jackson, Fredericksen, and Malcolm, 2002). Directional felling of harvested trees minimizes ecological damage and improves ease of yarding (Krueger, 2003). Caterpillar 518C skidders equipped with rubber tires and winches with 15 m of steel cable are used to drag

PAGE 39

26 the logs to roadside log decks (Krueger, 2003), where they are loaded on trucks for transport to the concessions sawmill. 2.3 Methods Four logged parcels, ranging from 27 to 31 ha and two non-logged 27 hectare control parcels were used in this study. Two of the logged parcels, and both the control parcels, were previously established, measured, and mapped by the Instituto Boliviano de Investigaciones Forestals (IBIF). All four logged parcels were harvested using RIL harvesting techniques, with harvest intensities varying from 1 to 2 trees per ha (Table 2-1). Each parcel was logged at a different time, either <1, 6, 13 or 19 months prior to the collection of field data in July, 2003. Table 2-1. Characteristics of the selectively-logged parcels used in this study Parcel Parcel area (ha) Total trees harvested Harvest Intensity (trees/ha) <1 month post-harvest 29.7 56 1.8 6 months post-harvest 27.0 27 1.0 13 months post-harvest 32.0 64 2.0 19 months post-harvest 28.0 29 1.0 Within the constraints of the current study, it was not possible to measure replicate parcels for each stage of this selective logging chronosequence. As a result, in a formal sense, I am unable to extrapolate from the results I report below to all selectively-logged parcels in the region (Hurlbert, 1984). However, in this study, individual felling gaps and skid trail segments, rather than the parcels, are used as the units of analysis. The statistical tests I employ here are inferential and are used to provide an objective indication of whether significant differences between individual felling gaps, for

PAGE 40

27 example, were related to their location within a given parcel, and/or due to other factors, such as felling gap size class (Oksanen, 2001). Where parcel is a significant effect, I infer that the effect is largely a result of the differences in time post-harvest. I argue that this inference is justified because of the absence of plausible alternatives to explain such systematic between-parcel differences in individual felling gaps and skid trail segments. 2.3.1 Field Spatial Analyses Parcel boundaries, skid trails, and felling gaps were geo-located for the < 1 and 6 months post-harvest parcels using a global positioning system (GPS) unit (maintaining precision <10 m and with a minimum of 5 satellites visible) and entered into a geographic information system (GIS) (ArcGIS; ESRI, Redlands, California, USA). Skid trail and stump locations within the 13 and 19 months post-harvest parcels (450 m by 600 m) were mapped by IBIF. The maps were then geo-rectified using a minimum of 15 field GPS measurements per parcel. The root mean square error for the geo-located parcels was consistently < 5 m. The area of each felling gap was entered into the GIS using field measured azimuth of fall (adjusted for declination) from the stump and field length and width measurements. The length of the gap was measured as the longest axis. The width (minor) axis of the gap was measured perpendicular to the length (major) axis at the 50% gap length point, and operationally the gap was defined as an oval with these two axes. Although most felling gaps have more varied shapes, this assumption was sufficiently accurate for the questions addressed within this study and convenient for integration with a GIS. These oval polygons were geo-referenced to the previously geo-located stump locations. Gap edges were defined by 10 m tall vegetation surrounding the ground area disturbed by the fallen tree or yarding process.

PAGE 41

28 The definition of gap used in this study differs from ecological measurements of gaps (Brokaw, 1982; Uhl, 1988), which consider only areas with open canopy to be part of the gap. Because remote sensors are sensitive to ground disturbances occurring below forest canopies (Asner et al., 2004) I chose to define gaps with reference to the disturbed ground area, and separately estimate canopy openness within that area. The nature of the definition of gap used here means that the data I report should not be compared to measurements of gaps that follow the ecological convention (Brokaw, 1982; Uhl, 1988). Skid trail width was defined as the distance between the outer edges of the most widely separated wheel ruts, and the mean width of 172 measurements was used to buffer the geo-referenced skid trail centerlines to calculate per-parcel skid trail area. Skid trail area was calculated for a total of 10 parcels, including six additional parcels that had been mapped previously by IBIF. Relationships between the area of skid trail and harvest intensity were investigated using a Michaelis-Menten non-linear regression in JMP statistical software. The Michaelis-Menten non-linear regression (y = ((1*x) / (2+x))) was chosen to model the relationship as a previous study (Panfil and Gullison, 1998) showed that the total area of skid trails had a positive quadratic relationship with increasing harvest intensity. 2.3.2 Field Measurements and Analyses Within the logged parcels, field measurements were made in felling gaps and skid trails. Felling gaps were classified as: large (> 800 m 2 ), medium (400 to 800 m 2 ), or small (< 400 m 2 ), and were divided in half to form trunk and crown zones of equal size. All field measurements were made separately within the two zones. Skid trails were sampled in 100 m transects along straight sections of the trails. A separate set of measurements was made in each 10 m segment of the 100 m transects. Additionally, a 50 m X 50 m grid

PAGE 42

29 layout was used to establish measurement points within a 450m X 600 m unlogged control forest. Field measurements included cover estimates from 5 m above the ground surface for: photosynthetic vegetation (PV); non-photosynthetic vegetation (NPV), which includes trunks, branches and senesced leaves; exposed soil; and a separate estimate of lianas with green foliage (green foliage of lianas is also included in the PV estimation). Canopy openness was estimated using a scale of 0 to 1 defined as the proportion of a standard upward facing hemispherical mirror at 1.5 m height that has a clear view of the sky (no canopy obstruction). Previous studies have shown that a canopy densiometer has comparable accuracy to digital or film hemispherical photography (Englund, O'Brien, and Clark, 2000).Within the logged parcels, additional measurements included the maximum height of regeneration in felling gaps and skid trails, the height of residual non-photosynthetic vegetation in felling gaps (excluding the stump), and skid trail width. Photosynthetic and non-photosynthetic vegetation, exposed soil, and liana cover were estimated for the entire trunk and crown zones of the felling gaps, and every 10 m along the skid trails, within a 2-m band perpendicular to the direction of the trail. At the grid points in the unlogged control forest, these cover estimates were made within a 2 m diameter circle placed 1 m to the edge of the path that connected the grid points. In the felling gaps, canopy openness readings were taken in the middle of each zone along the length axis. For skid trails these readings were taken from the middle of the 2-m bands described for the cover estimates. Felling gaps that included more than one felled tree (defined as overlaid gaps and constituting < 5% of the total gap area) were identified in the GIS and removed prior to

PAGE 43

30 statistical analysis to avoid confounding relationships between field measurements taken in the trunk and crown felling gap zones. To analyze field data collected within the individual tree felling gaps, a mixed 3-way analysis of variance (ANOVA-SAS 2003) was used to test the main effects of parcel, size class (large, medium, and small), and gap zone (trunk vs. crown), and their interactions on canopy openness, vegetation height, PV, NPV, exposed soil, and NPV height in the individual tree felling gaps. For field data collected within the skid trail segments, one-way ANOVA was used to test for the effect of parcel on canopy openness, vegetation height, trail width, PV, NPV, and exposed soil. For both the felling gap and skid trail data, Tukeys and Dunnetts post-hoc tests were performed to identify significant, pair-wise differences between the four logged parcels, and between the individual logged parcels and the unlogged control forest parcel, respectively. 2.3.3 Remote Sensing Measurements and Analyses Fourteen ASTER (Advanced airborne thermal emission radiometer) satellite images were obtained of the study area during the summer of 2003. Of these images four were found to be sufficiently cloud-and error-free for use in this study. These images were acquired on 13 May 2003, 30 June 2003, 16 July 2003 and 17 August 2003. In addition to these images a pre-harvest image had been previously acquired on 11 August 2001. These images were obtained in universal transverse mercator (UTM), world geodetic system (WGS) 1984 datum, zone 20 south projection and preprocessed by NASA to L2B surface reflectance. The preprocessing compensated for differences in sun angle / image geometry and atmospheric differences between the images. ASTER surface reflectance data have been validated to provide surface reflectance within 1% for actual surface reflectance < 15% and within 7% of actual surface reflectance > 15% (Abrams

PAGE 44

31 and Hook, 2001). Field validation of ASTER imagery, however, indicates that the absolute radiometric correction are, in general, better than 4% (Thome et al., 1998; Yamaguchi et al., 2001). These corrections are performed using radiative transfer calculations with atmospheric aerosol content from outside sources, such as the MODIS satellite or climatology data (Abrams and Hook, 2001). The visible-infrared (15 m pixels) images were re-sized to 30 m using aggregate pixel mean values and co-registered to the short wave infrared (30 m pixels) image, then layer stacked using nearest neighbor to produce 9-band images. Band 9 was removed prior to imaging processing due to problems with atmospheric water vapor. The 30 June 2003 image was chosen as the base image as it had the least cloud interference. The remaining images were geo-referenced to the base image using a minimum of 80 image-to-image control points dispersed throughout the image. The RMS errors for each geo-referencing were < 15 m (or half a pixel). All images were then layer stacked using the nearest neighbor re-sampling to get absolute pixel overlay. The stacked multi-date image was then geo-referenced to 95 field GPS ground reference points (UTM, WGS 84, Zone 20 S) which were acquired during the summer of 2004. The RMS error was < 15 m in the final warp model. The image was warped using a 1 st order polynomial model with nearest neighbor re-sampling. Finally the image overlays were visually assessed by flickering between the May, July and both August images against the 30 June 2003 base image. Systematic off-sets were observed with the 16 July 2003 image and were corrected through direct adjustment to the image map reference coordinates. The 4 post-harvest and 1 pre-harvest control images of the parcels enable a multi-temporal assessment of the sensitivity of the remote

PAGE 45

32 sensing methodology to selective logging at < 1-4, 6-9, 13-16 and 19-22 months post harvest. A probabilistic spectral mixture model was used to decompose the ASTER image per-pixel surface reflectances into sub-pixel estimates of photosynthetic vegetation, non-photosynthetic vegetation, and exposed soil. Errors in the linear mixing assumption of the endmembers were shown in the per-pixel RMS error fraction. Development of this model was based on an automated probabilistic linear spectral unmixing procedure developed originally for woodland and shrubland ecosystems (Asner and Lobell, 2000a, b) and recently used for analysis of selective logging impacts in the Brazilian Amazon (Asner et al., 2004). I used a general database of photosynthetic and non-photosynthetic vegetation and soil spectra that had been collected over logged and unlogged sites in South America (G. Asner, Personal Communication) which were deconvulved to ASTER bandwidths using published ASTER band response coefficients. Endmember bundles of several hundred mean spectra were used in the unmixing procedure. The use of endmember bundles, rather than single endmembers, is a technique to incorporate naturally occurring endmember spectral variability into the unmixing model (Bateson, Asner, and Wessman, 2000). A separate shade endmember was not included, as shade levels of 0 to 30% were incorporated into the photosynthetic vegetation endmember bundle to account for topographic and intraand inter-crown shadowing which are prevalent within satellite imagery (Asner and Warner, 2003). The 4 post-harvest images were corrected for pre-existing differences in topography and forest structure among the study parcels by subtracting the NDVI and

PAGE 46

33 fractional values of the pre-harvest image from each of the post-harvest images. The variability between images associated with seasonality and atmospheric differences were removed by normalizing each logged parcel with the control parcel from the same image date. A digital elevation model (DEM) was obtained by request from NASAs Earth Observing System (EOS). The DEM was produced through stereoscopic comparison of nadir and side angle data from the 11 August 2001 pre-harvest control ASTER image, and has been validated to have <= 10 m relative accuracy (vertical) and < 50 m horizontal error (Abrams and Hook, 2001). The geo-location of the DEM was done through visually adjusting the DEM (through alterations to the map info reference coordinates) until shadows in a DEM based shaded relief model (based on the 2001 image from which it was created) matched up with the shadows in an RGB (bands 2, 3, and 1, respectively) composite of the 11 August 2001 ASTER image. Lambertian shaded relief images (on scale of 0 total reflectance) were modeled based on the sun elevation and azimuth. Separate 2-way repeated measures ANOVA s were used for each remote sensing variable (per-pixel NDVI, and PV, NPV and soil fractions) to test the main effects of parcel and image date and their interactions for large, medium, and small felling gaps, and for skid trails. Dunnetts post-hoc tests were performed to identify significant differences between the large, medium, and small felling gap, and skid trail pixels and pixels located in the unlogged control parcel. Within the logged parcels, residual forest pixels (defined as those > 10 m from a felling gap or a skid trail) were used to illustrate the size of the disturbance effects relative to between-parcel effects when comparing felling gap and control parcel reflectance. The August 11 2003 image data of the 6

PAGE 47

34 months post-harvest parcel was not used because the parcel had been re-entered for further extraction during that month. Separate linear regressions were run between NDVI, PV, NPV, and the soil fraction pixels within the control parcels for the four summer 2003 ASTER images (n = 530) and the Lambertian shaded relief values for those same pixels to estimate the influence of topographic shade. The July 16 2003 image was acquired closest to the date of field data collection, so I used this image to examine the strength of relationships between the field measurements within the felling gaps and the remote sensing responses of those same felling gaps using Pearson bivariate correlation analysis within the < 1 month and the 6 months post-harvest parcels. 2.4 Results 2.4.1 Field Spatial Analyses Higher harvest intensities within the logged parcels correlated with higher area in felling gaps. Felling gaps accounted for most of the disturbed area in the logged parcels, ranging from 4 to 11% of the total parcel area, while skid trails only accounted for a maximum of 5 % (Table 2-2). The spatial distribution of felling gaps and skid trails is illustrated in Figure 2-2. Gap size ranged from of 59 m 2 to 2200 m 2 and gaps > 800 m 2 were uncommon (Table 2-3). Table 2-2. Percent parcel area disturbed by tree fall gaps and skid trails Parcel Parcel area in ha Harvest intensity (trees/ha) % of parcel in felling gaps % of gap area in overlaid gaps % of parcel in skid trails <1 month post-harvest 30 1.9 8.7 3.4 4.0 6 months post-harvest 27 1.0 6.9 3.1 2.4 13 months post-harvest 32 2.0 10.5 6.7 5.1 19 months post-harvest 28 1.0 4.2 2.9 3.8

PAGE 48

35 Figure 2-2. Locations of tree fall gaps and skid trails are shown for the logged study parcels. Maps of the locations of felled trees in the 13and 19-month post harvest parcels were provided by IBIF, and were used as base maps for those parcels.

PAGE 49

36 Table 2-3. Sample size of large, medium, and small felling gaps within the logged parcels (Appendix C) Gap size Parcel <1 month post-harvest 6 months post-harvest 13 months post-harvest 19 months post-harvest Small 26 (29) 5 (5) 24 (30) 15 (19) Medium 27 (38) 13 (15) 14 (26) 5 (9) Large 3 (3) 8 (9) 6 (9) 1 (1) In parenthesis is the sample size before removing overlaid gaps The addition of data from eight other IBIF research parcels shows a clear quadratic relationship between harvest intensity and the percent of a parcel covered by skid trail working surfaces (Figure 2-3, root mean square error for the fit Michaelis-Menten model was 0.53. Estimates of 1 and 2 were 7.40 and 1.22, respectively). Figure 2-3. Michaelis Menten nonlinear model fit over harvest intensity versus % parcel area affected by skid trails. Data from Table 2-2 and Appendix D. 2.4.2 Field Measurements 2.4.2.1 Post-harvest recovery of forest structure in felling gaps Results of the 3-way ANOVA testing the main effects of parcel, gap size, and gap zone, and their interactions, on canopy openness, liana coverage, vegetation height,

PAGE 50

37 photosynthetic (PV) and non-photosynthetic vegetation (NPV), exposed soil, and non-photosynthetic vegetation (NPV) height are reported in Table 2-4. There were no significant 3-way interactions. Mean values of those variables for the four logged parcels and the control forest parcel are provided in Table 2-5. Table 2-6 lists mean values for the field measurements by felling gap size and Table 2-7 provides mean values of field variables for trunk and gap zones. Table 2-4. The F and P values for the main effects of parcel, gap size, and gap zone, and their interactions for mixed 3-way ANOVAs of variables measured in felling gaps Factors Parcel Gap size Gap zone Parcel gap size Parcel gap zone Gap size gap zone Canopy Openness 14.8*** 8.6** 7.0** 1.5 0.9 4.2* Liana coverage (%) 5.4** 0.6 31.6*** 0.6 18.8*** 2.3 Vegetation height (m) 11.0*** 3.4* 0.3 1.8 0.5 0.5 Photosynthetic vegetation (PV) coverage (%) 17.3*** 0.3 13.1** 3.2** 0.1 0.2 Non-photosynthetic vegetation (NPV) coverage (%) 7.3** 0.0 44.1*** 0.8 2.8* 0.7 Soil coverage (%) 5.2** 0.7 43.0*** 1.1 21.9*** 2.4 NPV height (m) 6.5*** 0.9 170.0*** 0.7 8.8*** 5.6*** Asterisks represent significance of main effects and interactions (* = P <0.05, ** = p <0.01 and *** = P <0.001). Canopy openness was significantly affected by parcel, size class, and gap zone, and there was a size class gap zone interaction. Canopy openness within felling gaps decreased significantly with time post harvest and was significantly greater for all logged parcels than for the control forest. Canopy openness within felling gaps also increased with increasing gap size, and trunk zones had a significantly less open canopy than in crown zones. The size class gap zone interaction reflects that canopy openness was greater in the crown zone than in trunk zone in large and medium gaps but not in small gaps (Figure 2-4).

PAGE 51

38 Table 2-5. Mean values of field measurement variables within felling gaps for <1-, 6-, 13-, and 19-months post-harvest parcels. Unlogged control forest values are provided for comparison. Factors Parcel mean (standard error) <1 month post-harvest 6 months post-harvest 13 months post-harvest 19 months post-harvest Control forest n = 56 n = 26 n = 44 n = 21 n = 130 Canopy Openness (%) 52.6 (6.2)***a 48.7 (3.3)***a 26.4 (2.7)***bc 18.1 (5.7)***c 3.7 (0.5) Liana Coverage (%) 12.8 (5.4)***ac 8.0 (3.6)***a 24.4 (2.8)c 27.1 (5.8)c 21.7 (2.4) Vegetation Height (m) 0.6 (0.4)***a 1.7 (0.3)***b 2.9 (0.2)***c 3.0 (0.5)***c 21.1 (1.0) Photosynthetic Vegetation Coverage (%) 31.4 (5.4)***a 58.9 (3.6)***b 70.8 (2.8)c 79.5 (5.8)c 71.2 (1.7) Non-Photosynthetic Vegetation Coverage (%) 49.6 (5.3)***a 36.6 (3.6)*b 26.1 (2.7)c 19.3 (5.7)c 28.2 (1.6) Exposed Soil Coverage (%) 13.6 (2.6)***a 3.2 (1.7)b 3.1 (1.3)b 1.3 (2.8)b 0.6 (0.3) NPV height (m) 2.6 (0.3) a 2.6 (0.2) a 1.8 (0.2) b 1.1 (0.4) b na Asterisks represent significant differences between treatment parcel felling gap and control forest values (* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant differences between felling gap values in the different treatment parcels (Tukey's test, P < 0.05). Table 2-6. Mean values (standard error) of field measurement variables for all large, medium, and small felling gaps. Factors Gap size and control (standard error) Large Medium Small n = 18 n = 59 n = 70 Canopy Openness (%) 46.8 (6.2)***b 36.8 (2.5)***b 25.5 (2.3)***a Liana Coverage (%) 22.5 (5.9) a 15.5 (2.4)***a 16.3 (2.5) a Vegetation Height (m) 2.4 (0.5)***ab 2.3 (0.2)***b 1.6 (0.2)***a Photosynthetic Vegetation Coverage (%) 59.2 (5.9) a 59.3 (2.4)***a 61.8 (2.5)***a Non-Photosynthetic Vegetation Coverage (%) 32.3 (5.8) a 33.0 (2.4)***a 33.4 (2.4)*a Soil Coverage (%) 6.1 (2.8) a 5.8 (1.2)***a 4.0 (1.2)***a NPV height (m) 2.3 (0.4) a 2.0 (0.2) a 1.8 (0.1) a Asterisks represent significant differences between treatment parcel felling gap and control forest values (* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant differences between felling gap values in the different treatment parcels (Tukey's test, P < 0.05).

PAGE 52

39 Table 2-7. Mean values (standard error) of measured variables for all trunk and canopy felling gap zones. Factors Gap zone (standard error) Trunk (n = 147) Canopy (n = 147) Canopy Openness (%) 33.5 (2.7)***a 39.3 (2.5)***b Liana Coverage (%) 10.4 (2.7)**a 25.7 (2.7)**b Vegetation Height (m) 2.0 (0.2)***a 2.1 (0.2)***a Photosynthetic Vegetation Coverage (%) 64.7 (2.6) a 55.6 (2.6)*b Non-Photosynthetic Vegetation Coverage (%) 24.1 (2.6)**a 41.8 (2.6)*b Soil Coverage (%) 10.0 (1.3)***a 0.6 (1.3) b NPV height (m) 0.8 (0.1) a 3.3 (0.2) b Asterisks represent significant differences between treatment parcel felling gap and control forest values (* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant differences between felling gap values in the different treatment parcels (Tukey's test, P < 0.05). Figure 2-4. Canopy openness of all felling gaps as affected by the interaction between gap size and gap zone. Error bars represent standard error of the mean. Liana coverage was affected by parcel and gap zone effects, as well as the parcel *gap zone interaction. Liana coverage dropped initially from the < 1 month post-harvest gaps to the 6 months old gaps, after which it increased dramatically. The < 1and 6months post-harvest parcels were significantly lower than the control forest mean but the

PAGE 53

40 13and 19-month post-harvest parcels were not (Table 2-5). Crown zones had significantly greater liana coverage than trunk zones (Table 2-7). The parcel gap zone interaction reflects that the gap zone differences are not strongly apparent until 13 months after logging when canopy zone liana % becomes much greater than that in the trunk zone (Figure 2-5a). The height of regenerating vegetation was greater in larger gaps, and in parcels that had more time to regrow following logging (Table 2-5, 2-6). Similarly, the coverage of PV was significantly affected by parcel and gap zone, as well as the parcel size class interaction. PV increased with time post-harvest and in the13and 19-months post-harvest plots, PV in the gaps was not significantly different from in the control forest (Table 2-5). The crown zone had significantly less PV than the trunk zone (Table 2-7). The interaction of parcel size class showed that small gaps had significantly higher PV only in the < 1-month post-harvest parcel (Figure 2-6). NPV was significantly affected by the main effects of parcel and gap zone, and the interaction of parcel gap zone. NPV decreased with time post-harvest and was indistinguishable from the control forest in the 13and 19-month post harvest gaps. The crown zone had significantly more NPV than the trunk zone (Table 2-7). The parcel gap zone interaction revealed consistently higher levels of NPV% in crown zones (versus trunk zones) that diminished with parcel (Figure 2-5b).

PAGE 54

41 Figure 2-5. The field factors of A) Liana, B) NPV, C) soil coverage, and D) NPV height as affected by the interaction between parcel and gap zone. Error bars represent standard error of the mean. See Table 2-3 for sample sizes. Soil exposure was also significantly affected by the main effects of parcel and gap zone, and the interaction of parcel gap zone. Soil exposure decreased with time post-harvest and only the < 1-month post-harvest gaps had significantly more exposed soil than the control forest. Although there was a trend towards increasing soil exposure with

PAGE 55

42 gap size it was not statistically significant. Soil exposure was almost twenty times greater in the trunk zone than in the crown zone. The parcel gap zone interaction reflected that the differences between gap zones diminished with increasing months post-harvest (Figure 2-5c). Figure 2-6. Photosynthetic vegetation coverage as affected by the interaction between parcel and gap size. Error bars represent standard error of the mean. See Table 2-3 for sample sizes. NPV height was significantly affected by the main effects of parcel and gap zone, as well as both parcel gap zone, and size class parcel interactions. NPV height decreased with increasing months post-harvest and was higher in the crown portion of the gap. The parcel gap zone interaction is a result of decreasing NPV height in the canopy zone with parcel but NPV height remaining the same in the trunk zone (Figure2-5d). The gap size gap zone interaction was a result of smaller gaps having decreased differences in NPV height between the canopy and trunk zones (data not shown).

PAGE 56

43 2.4.2.2 Skid trails Canopy openness, vegetation height, PV, NPV, and soil exposure were significantly affected by parcel (P < 0.05). Skid trail PV increased with time post-harvest whereas skid trail soil exposure decreased; patterns were less consistent for the other variables (Table 2-8). Table 2-8. Mean (standard error) for field factors within skid trails <1, 6, 13 and 19 months post-harvest. n Canopy openness n Veg. height (m) n Width (m) PV % NPV % Soil % <1 months post-harvest 63 17.3 (16.4)a*** 41 0.0 (.1)a*** 31 3.4 (0.4)a 5.8 (16.3)a*** 33.6 (22.8)a 59.5 (26.1)a*** 6 months post-harvest 45 10.1 (9.5)a** 45 0.7 (.3)a*** 10 3.3 (0.3)a 18.5 (14.3)b*** 47.5 (14.8)b** 32.0 (20.3)b*** 13 months post-harvest 15 5.8 (3.8)b 15 0.3 (.3)a*** na 3.4 (0.4)a na na na 19 months post-harvest 59 14.0 (14.9)a*** 59 1.75 (1.2)a*** 31 3.9 (0.4)a 51.5 (16.2)c*** 40.8 (17.3)b** 8.7 (14.6)c*** Asterisks represent significant differences between treatment and control parcels (* = P <0.05, ** = P <0.01, *** = P <0.001). 2.4.3 Remote Sensing Both NDVI and the soil fraction were significantly negatively correlated with increasing Lambertian shade levels (P < 0.001) while the NPV fraction was significantly positively correlated with increasing shade levels (P < 0.001). The PV fraction, however, was not correlated with shade intensity. Seasonality also affected NDVI, as well as the sub-pixel fractions (PV, NPV, soil), as illustrated in Figure 2-7. NDVI and PV fractional values within the control parcels declined steadily from May (early in the dry season) to mid-August (nearing the end of the dry season). Neither the NPV or soil fractions had strong correlations with seasonality.

PAGE 57

44 Figure 2-7. Control parcel mean values for NDVI and PV, NPV, and soil fractions versus image acquisition date. The dry season intensifies from May through August.

PAGE 58

45 2.4.3.1 Post-harvest recovery of spectral characteristics of felling gaps Felling gaps > 800 m 2 Figure 2-8 illustrates the evolution of spectral characteristics post-harvest for pixels in large felling gaps (Appendix E). Two-way repeated measures ANOVA revealed significant main effects of parcel for NDVI, PV, and NPV, and image date for PV and soil. The interaction effect was significant for NDVI and PV (Appendix E). NDVI was significantly lower than unlogged control pixels for up to 3 months after logging (P < 0.001), then higher at 6-8 months (P < 0.001) and at 16, and 20-22 months post-logging (P < 0.05). PV was lower for 2 and 3 months post-harvest (P < 0.001) and higher at 22 months post-logging (P < 0.05). NPV was higher 2 months post-logging (P < 0.05). The soil response had the highest variance. Felling gaps 400-800 m 2 Figure 2-9 illustrates the evolution of spectral characteristics post-harvest for pixels in medium felling gaps (Appendix E). Two-way repeated measures ANOVA revealed significant main effects of parcel for NDVI, PV, and NPV, and image date for NDVI, PV and soil. The interaction effect was significant for all the variables. NDVI was significantly lower than for unlogged pixels for up to 3 months after logging (P < 0.001), then higher at 15-16 months post-logging (P < 0.01). PV was lower for 1, 2 and 3 months post-logging (P < 0.01). NPV was higher 1-3 and 8 months post-logging (P < 0.05). Again the soil response had the highest variance. Felling gaps < 400 m 2 Figure 2-10 illustrates post-harvest spectral changes in small felling gaps (Appendix E). Two-way repeated measures ANOVA revealed significant main effects of parcel for NDVI, PV, and NPV, and image date for NDVI, PV and soil. The interaction effect was significant for all the variables. NDVI was significantly lower than for unlogged pixels only at 3 months post-logging (P < 0.001),

PAGE 59

46 but higher at 15 and 16 months post-logging (P < 0.05). PV was lower for 1, 2 and 3 months post-logging (P < 0.01). NPV was higher 1 and 2 months post-logging (P < 0.05). Figure 2-8. Difference between spectral characteristics of large felling gap and unlogged control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil. Error bars are standard errors for the large felling gap pixels; standard errors for the control pixels were < 0.001 on the y-axis. The differences between the treatment parcels residual forest and unlogged control pixels are shown to distinguish the disturbance effect from any potential effect of between-parcel differences.

PAGE 60

47 Figure 2-9. Difference between spectral characteristics of medium felling gap and unlogged control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil. Error bars are standard errors for the large felling gap pixels; standard errors for the control pixels were < 0.001 on the y-axis. The differences between the treatment parcels residual forest and unlogged control pixels are shown to distinguish the disturbance effect from any potential effect of between-parcel differences.

PAGE 61

48 Figure 2-10. Difference between spectral characteristics of small felling gap and unlogged control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil. Error bars are standard errors for the large felling gap pixels; standard errors for the control pixels were < 0.001 on the y-axis. The differences between the treatment parcels residual forest and unlogged control pixels are shown to distinguish the disturbance effect from any potential effect of between-parcel differences. 2.4.4 Linking Field and Remotely-Sensed Measurements Pearson bivariate correlations between field and remote sensing measurements of all felling gaps in the < 1 and 6 months post-harvest parcels are presented in Tables 2-12

PAGE 62

49 and 2-13, respectively. The significant positive correlations between NDVI and PV show they respond similarly to forest disturbances for both the < 1 and 6 months post-harvest parcels, and both are inversely correlated with NPV. Soil reflectance was also inversely correlated with NPV. Gap area was inversely correlated with NDVI and PV in the < 1month post-harvest parcel but only with NDVI in the 6-month post-harvest parcel. Canopy openness in the crown zone was also inversely correlated with NDVI and PV in the < 1-month post-harvest parcel. In the 6-month post-harvest parcel, crown zone PV coverage was inversely correlated with NPV reflectance, which was positively correlated with NPV coverage. PV coverage in the trunk zone of the < 1 month post-harvest parcel felling gaps was correlated with NDVI and PV reflectance, whereas NPV coverage was inversely correlated with NDVI. NPV coverage in the trunk zone of the 6 month post-harvest parcel was positively correlated with NPV and negatively correlated with soil reflectance. Table 2-12. Pearson bivariate correlations between field and remote sensing measurements of felling gaps in the < 1 month post-harvest parcel. NDVI PV NPV Soil NDVI 1 PV 0.736*** 1 NPV -0.450*** -0.612*** 1 Soil ns ns -0.835** 1 Gap area (m 2 ) -.322** -0.344** ns ns Gap canopy zone Canopy openness -0.382*** -0.311** ns ns Vegetation height (m) ns ns ns ns PV % coverage ns ns ns ns NPV % coverage ns ns ns ns Soil % coverage ns ns ns ns Gap trunk zone Canopy openness ns ns ns ns Vegetation height (m) ns ns ns ns PV % coverage 0.316** 0.265* ns ns NPV % coverage -0.248* ns ns ns Soil % coverage ns ns ns ns Asterisks represent significant correlations (* = P < 0.05, ** = P < 0.01, *** = P < 0.001).

PAGE 63

50 Table 2-13. Pearson bivariate correlations between field and remote sensing measurements of felling gaps in the 6 months post-harvest parcel. NDVI PV NPV Soil NDVI 1 PV 0.779*** 1 NPV -0.526** -0.589*** 1 Soil ns ns -0.748*** 1 Gap area (m 2 ) .338* ns ns ns Gap canopy zone Canopy openness ns ns ns ns Vegetation height (m) ns ns ns ns PV % coverage ns ns -0.428** ns NPV % coverage ns ns 0.444** ns Soil % coverage ns ns ns ns Gap trunk zone Canopy openness ns ns ns ns Vegetation height (m) ns ns ns ns PV % coverage ns ns ns ns NPV % coverage ns ns 0.459** -0.371* Soil % coverage ns ns ns ns Asterisks represent significant correlations (* = P < 0.05, ** = P < 0.01, *** = P < 0.001). 2.5 Discussion Development of effective remote sensing based programs to monitor selective logging requires an understanding of the spatial and temporal thresholds that constrain the applicability of remote sensing to the detection of selective logging. Although this study was conducted in the context of low harvest intensities the dynamics of recovery of structural and spectral characteristics following selective logging should be applicable to improving understanding of the signatures of selective logging (e.g. felling gaps and skid trails) throughout the tropics. I found that the NDVI and PV, NPV, and soil fractions were useful for identifying large and medium size tree fall gaps for between 3 and 6 months post-harvest. The PV fraction had the greatest response within felling gaps and, unlike NDVI, PV was not affected by topographic shade. The NPV and soil fractions were both highly correlated

PAGE 64

51 with topographic shade and were thus less useful for monitoring forest disturbances, especially in areas with more pronounced relief. Canopy openness defines the ability of remote sensors to view ground disturbances indicative of logging activities and, in general, as felling gaps age from < 1 to 19 months, canopy openness declines from 50 to 60% to less than 20%. Simultaneously, rapid vegetation growth, reaching nearly 2 to 5 m by 19 months covers over the originally exposed soil (primarily in the trunk zone), and NPV (primarily in the canopy zone) causing the relative percentages of PV, NPV and soil to change from 30, 50 and 10, respectively, immediately following harvest to 80, 20, and 0, respectively, after 19 months. Rapid liana growth, primarily in the crown zone, covers nearly 30 percent of the entire gap zone in an often dense mat of verdant lianas by 19 months post-harvest. This rapid reduction in overall canopy openness means that felling gaps become indistinguishable from the surrounding forest after around 6 months post-harvest. The process occurs faster for smaller gaps as they begin with less persistent residual NPV and are characterized by less initial canopy damage. Soil exposure within felling gaps, and therefore the utility of the soil fraction for identifying forest disturbances from selective logging, was limited primarily to the trunk zone. Though exposed soil in open areas is easily discerned from space, the gap trunk zone has little canopy damage, as compared with the crown zone, and by 6 months post-harvest canopy openness within the trunk zone is < 5%, due to canopy and vegetation regeneration. Although skid trails comprised 30 to 60% of the disturbed area, had the highest exposed soil levels, and had the slowest rates of vegetation recovery, they were not identifiable with remote sensing because they had little impact on canopy openness.

PAGE 65

52 Working in the Brazilian Amazon, (Asner et al., 2002) showed that single band and textural analysis techniques were not sensitive to canopy damages from selective logging that were < 50% of complete canopy coverage. The analytical techniques assessed in this study show a considerable improvement in sensitivity to lower levels of canopy damage. Asner et al. (2004), using AutoMCU derived per-pixel PV, NPV, and soil fractions of Landsat imagery, showed sensitivity to skid trails and felling gaps which diminished greatly from 0.5 to 1.5 year post-harvest, due to rapid regeneration of low-stature pioneer species. Remotely measured canopy openness values (derived from the PV fraction) of 12, 11, and 11 percent for felling gap pixels and 28, 11, and 12% for skid trail pixels were measured 0.5, 1.5, and 3.5 years post-harvest, respectively. Different from my results, Asner et al. (2004) found that felling gap and skid trail PV fraction remained consistently higher than in non-logged control forest. This may be a result of the higher harvest intensities in their study, leading to more prolonged canopy damage than was found in La Chonta. Few studies have linked remote sensing data directly to selective logging disturbances through the collection of extensive field data. The results of this study help to better understand the utility of currently available remote sensing technologies for monitoring selective logging, as well as identifying limitations that future remote sensors and image analysis technologies can address. Future efforts will seek to delineate logged areas based on the differences in reflectance that are apparent in felling gaps for several months, and on their spatial distribution.

PAGE 66

APPENDIX A GROUND, STAND, AND CANOPY DAMAGE AFTER SELECTIVE LOGGING Table A-1. Section 1 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha) 53

PAGE 67

54 Table A-2. Section 2 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha)

PAGE 68

55 Table A-3. Section 3 of ground, stand and canopy level forest damage after selective logging ordered according to level of harvest intensity (trees/ha)

PAGE 69

APPENDIX B MEAN MONTHLY PRECIPITATION IN LA CHONTA Figure B-1. Mean monthly precipitation (mm) measured in LThe pronounced dry season starting in April and lasting thvisible. Error bars represent standard errors around the m a Chonta from 1993. rough October is ean. 56

PAGE 70

APPENDIX C DISTRIBUTIOY PARCELS N OF FELLING GAP AREA SIZES WITHIN THE STUD Figure C-1. Gaps size classes for this study were small felling gaps < 400 m 2 medium felling gaps 400 m 2 to 800 m 2 and large felling gaps > 800 m 2 57

PAGE 71

APPENDIX D SKID TRAIL AREAS AND HARVEST INTENSITIES FOR ALL BOLFOR LONG TERM SILVICULTURAL RESEARCH PLOTS Table D-1. Skid trail areas and harvest intensities for all BOLFOR long term silvicultural research plots Parcel Skid trail area (ha) Parcel area (ha) # trees harvested Harvest intensity* % parcel area in skid trails 6 months post-harvest 0.652 27 27 1 2.41 19 months post-harvest 1.062 27.98 29 1.036 3.80 B1-M 1.12 27.23 31 1.13 4.11 < 1 month post-harvest 1.186 29.7 56 1.885 3.99 13 months post-harvest 1.63 32 64 2 5.09 B3-M 1.26 28.81 62 2.15 4.37 B2-M 1.52 29.19 78 2.67 5.21 B1-I 1.51 27.27 85 3.11 5.54 B3-I 1.83 31.47 116 3.68 5.82 M B2-I 1.48 27.23 123 4.52 5.44 ean trees/ha harvested 58

PAGE 72

APPENDIX E SIGNIFICANCE AND F VALUES OF 2-WAY REPEATED MEASURES ANOVAS TE SENSING VARIABLES OF FELLING GAP OF REMO PIXELS es of felling gap pixels g Gap Table E-1. Significance and F values of 2-way repeated measures ANOVAs of remote sensing variabl Fellin s > 800 m 2 Effects Soil NDVI PV NPV Parcel 12.08*** 6.85*** 2.83* NA Image Dat e 9* 44* *Image Date 3.31*** 3.22*** NA NA ap 800 m NA 2.7 NA .59** ParcelFelling G s 400 to 2 Ef fects I Soil NDV PV NPV Parcel *** 8*** N 9.18 15.9 9.62*** A Image D ate ** NA 19** Image e 15*** 7.*** 2.** 3.08Gaps < 400 m2 2.98 4.08 7.93* Parcel*g Dat 6. 67 31 *** Fellin Ef fects I Soil NDV PV NPV Parcel ** 6*** 4.73* 6.7 5.63*** 3. 69** Im age Date NA 12** *Image 2.s represent significcts anractions < 0.05, < 3.44* 3. 61 3.72* Parcel Date 3.02* 2.17 2.34** 59** Asterisk ant effe d inte (* = P ** = P 0.01, *** = P < 0.001) 59

PAGE 73

60 Table E-2. Mean differences and P Values for two-way repeated measures ANOVA post-hoc comparisons (Dunnetts Test) of NDVI, PV, NPV and soil fractions in large (> 800 m2) felling gaps versus unlogged control parcel pixels. Parcel Months post-harvest NDVI PV NPV Soil < 1 ns ns ns ns 1 -0.057*** ns ns ns 2 -0.052*** -0.079*** 0.082* ns < 1 month post-harvest 3 -0.050*** -0.072*** ns ns 6 0.018*** ns ns 0.058* 7 0.023*** n6 months post-harvest s ns ns 0.020*** ns ns ns ns ns ns a 8 13 ns 14 s 29* 15 est 16 13* s ns n ns -0.0 ns ns ns ns 13 months post-harv 0.0 ns n ns 19 ns ns ns ns 20 21 post-harvest 0.037* ns ns ns 0.045*** ns ns ns 19 months 22 .047* s represent significances ea co parcels ** = P < 0.01, < 0. 0.035 0 ns ns Asterisk t differen between tr tment and ntrol (* = P < 0.05, *** = P 001). a No August remote slable as a portion ensing d6 st-arceavaiof the pa re-lurin mont ata of the rcel was months poogged d harvest pg the last l was h of the study.

PAGE 74

61 Table E-3. Mean differences and P Values for two-way repeated measures ANOVA post-hoc comparisons (Dunnetts Test) of NDVI, PV, NPV and soil fractions in medium (400 to 800 m 2 ) felling gaps versus unlogged control parcel pixels. ParMp cel onths ost-harvest NDVI PV NPV Soil <071* 1 ns ns ns 0. 1 242*** .018* < 1postrvest 3 *** 17* 6 ns -0.016*** -0.031*** 0.041*** ns -0.015*** -0.028** 0.0 -0 month -ha -0.016*** -0.027 0.0 ns ns ns ns 7 ns 6 mharst 8 ns 6* 10*** ns ns ns onths post-a ve ns 0.02 ns 3 0.01 ns ns ns 14 Ns 1 13 postrvest 16 ns 19 ns ns ns ns 5 0.012** ns ns ns months -ha ns ns ns ns ns ns 20 ns 2 19 postrvest Astreatment and control parcels (* = P < 0 ns ns ns 1 ns ns ns ns months -ha 22 ns ns ns ns erisks represent significant differences between t .05, ** = P < 0.01, *** = P < 0.001). a N porthe parcel was re-logged during the last month of the study. o August remote sensing data of the 6 months post-harvest parcel was available as aion of t

PAGE 75

LIST REFERCES Abrams Ho. 2001. ASTha V ulsion boratory, Paena, CA. Achard, F., H. Eva, H. Stibig, P. Mayaux, J. Gallego, T. Richards, and J. Malingreau. 2002. Determinn of deforetes the worid tropical forests. 7: 9002. Adams, J. B., D. Sabo. Kapos, R.th, andGillespie. sificn of multisral imagased onctions of endmembers: to l-cover change in the Brazilian AmRemote Sensing of vironment 52: 137-154. Adegoktellite teorology 3: 395-405. Aldakheel, Y. Y., and F. M. Danson. 1997. Spectral reflectance of dehydrating leaves: Measurements and modelling. International Journal of Remote Sensing 18: 3683-3690. Alvira, D., E. F. Putz, and T. S. Fredericksen. 2004. Liana loads and post-logging liana densities after liana cutting in a lowland forest in Bolivia. Forest Ecology and Management 19: 73-86. Alvira, D. C. 2002. Liana loads and post-logging liana densities after liana cutting in a lowland forest in Bolivia, Masters Thesis, University of Florida, Gainesville, FL. Aparicio, N., D. Villegas, J. L. Araus, J. Casadesus, and C. Royo. 2002. Relationship between growth traits and spectral vegetation indices in durum wheat. Crop Science 42: 1547-1555. Appanah, S., and F. E. Putz. 1984. Climber abundance in virgin dipterocarp vorest and the effect of pre-felling climber cutting on logging damage. Malaysian Forester 47: 335-342. OF EN M., and S. ok ER users ndbook ersion 1. Jet Prop La sad atio station ra of ld's hum Science 29 99-1 l, V Filho, D. Roberts, M. S mi A. 1995. Clasapplication atio pect es b fra and azon. En e, J. O., and A. M. Carleton. 2002. Relations between soil moisture and savegetation indices in the US Corn Belt. Journal of Hydrome 62

PAGE 76

63 Armonia, J. 1995. Lista de las aves de Bolivia. BirdLife International, Santa Cruz, Bolivia. Asdak, C., P. G. Jarvis, P. van Gardingen, and A. Fraser. 1998. Rainfall interception loss in unlogged and logged forest areas of Central Kalimantan, Indonesia. Journal of Hydrology 206: 237-244Asner, G. P. 1998. Biophysical and biochemical sources of variability in canopy reflectance. Remote Sensing of Environment 64: 234-253. Asner, eforested areas bordering the Tapajos National Forest, Central Amazon. Remote Sensing of Environment 87: 507-520. Asner, In nklin [eds.], Remote sensing of forest environments: Concepts and case studies, 552. Kluwer Academic Publishers, New York, NY. Asner, canopy gaps following selective logging in the eastern Amazon. Global Change Asner, G., M. Keller, R. Pereira, and J. Zweede. 2002. Remote sensing of selective logging in Amazonia: Assessing limitations based on detailed field observations, Remote sensing of environmentAsner, G., M. Keller, R. Pereira, J. Zweede, and J. Silva. 2004. Canopy Damage and ______. 2000b. AutoSWIR: A general spectral unmixing algorithmm based on G. P., M. M. Bustamante, and A. Townsend. 2003. Scale dependence of biophysical structure in d G. P., J. A. Hicke, and D. Lobell. 2002. Per-pixel analysis of forest structure: Vegetation indices, spectral mixture analysis and canopy reflectance modeling.M. A. Wulder and S. E. Fra G., M. Keller, and J. N. M. Silvas. 2004. Spatial and temporal dynamics of forest Biology 10: 1-19. Landsat ETM+, and textural analysis. 80: 483-496. Recovery After Selective Logging in Amazonia: Field and Satellite Studies. Ecological Applications 14: 280-298. Asner, G., and D. Lobell. 2000a. A Biogeophysical Approach for Automated SWIR Unmixing of Soils and Vegetation. Remote Sensing of Environment 74: 99-112 2000-2400 nm endmember datasets and Monte Carlo analysis. Proceedings of the 9th Annual JPL Airborne Earth Science Workshop.

PAGE 77

64 Asner, G., and A. S. Warner. 2003. Canopy shadow in IKONOS satellite observatiotropical f ns of orests and savannas. Remote Sensing of Environment 87: 521-533. e and Remote Sensing 38: 1083-1094. ity in ceedings of the 7th Annual JPL Airborne Earth Science Workshop 1: 43-52. Bertault Kalimantan Indonesia. Forest Ecology and Management 94: 209-218. BohlmaBowman, W. D. 1989. The relationship between leaf water status, gas exchange, and -255. Brokaw, N. V. 1982. The definition of treefall gap and its effect on measures of forest dynamics. Biotropica 14: 158-160. r los Calla, Sl Bateson, A., G. Asner, and C. Wessman. 2000. Endmember bundles: A new approach to incorporating endmember variability into spectral mixture analysis. Ieee Transactions on Geoscienc Bateson, C., G. Asner, and C. Wessman. 1998. Incorporating endmember variabilspectral mixture analysis through endmember bundles. Pro t, J. G., and P. Sist. 1997. An experimental comparison of different harvesting intensities with reduced-impact and conventional logging in Eas n, S., J. Adams, and D. Peterson. 1998. Seasonal foliage changes in the eastern Amazon basin detected from Landsat thematic mapper satellite images. Biotropica 30: 376-391. BOLFOR. 2000. Study plan: Long-term silvicultural research project (LTSRP) in Bolivian tropical forests. BOLFOR, Santa Cruz, Bolivia. spectral reflectance in cotton leaves. Remote Sensing of Environment 30: 249 CAF, BOLFOR, and Geosystems. 2000. Bolivia: Determinacin del dao causado poincendios forestales ocurridos en los departamentos de Santa Cruz-Beni en los meses de Agosto y Septiembre de 1999. BOLFOR, Santa Cruz, Bolivia. A. 2003. Arqueloga de "La Chonta". BOLFOR, Santa Cruz, Bolivia. Camacho, O., W. Cordero, I. Martinez, and D. Rojas. 2001. Tasa de Deforestacion deDepartamento de Santa Cruz, Bolivia 1993-2000. BOLFOR and Superintendencia Forestal.

PAGE 78

65 Carlson, T., and D. Ripley. 1997. On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sensing of Environment 62: 241-255. Carter, G. A. 1991. Primary and secondary effects of water content on the spectral reflectance of leaves. American Journal of Botany 78: 916-924. Ceccato, P., S. Flasse, and J. Gregoire. 2002. Designing a spectral index to estimate Ceccato, P., S. Flasse, S. Tarantola, S. Jacquemoud, and J.-M. Gregoire. 2001. Detecting Ceccato, P., N. Gobron, S. Flasse, B. Pinty, and S. Tarantola. 2002. Designing a spectral Choudhury, B. J. 1987. Relationship between vegetation indices, radiation absorption, of : Remote sensing of large wildfires in the European Mediterranean Basin, 238. Chuvieco, E., D. Riano, I. Aguado, and D. Cocero. 2002. Estimation of fuel moisture 80. Cochrane, M. A. 2003. Fire science for rainforests. Nature 421: 913-919. vegetation water content from remote sensing data; Part 2. Validation and applications. Remote Sensing of Environment 82: 198-207. vegetation leaf water content using reflectance in the optical domain. Remote Sensing of Environment 77: 22-33. index to estimate vegetation water content from remote sensing data: Part 1; Theoretical approach. Remote Sensing of Environment 82: 188-197. and net photosynthesis evaluated by a sensitivity analysis. Remote Sensing Environment 22: 209-233. Chuvieco, E., M. Deshayes, N. Stach, D. Cocero, and D. Riano. 1999. Short-term riskfoliage moisture content estimation from satellite data. In E. Cchuvieco [ed.], Springer, Berlin, Germany. content from multitemporal analysis of Landsat Thematic Mapper reflectance data: applications in fire danger assessment. International Journal of Remote Sensing 23: 2145-2162. Cibula, W., E. Zetka, and D. Rickman. 1992. Response of Thematic Mapper bands to plant water stress. International Journal of Remote Sensing 13: 1869-18 Civco, D. 1989. Topographic normalization of Landsat Thematic Mapper digital imagery. Photogrammetric Engineering and Remote Sensing 55: 1303-1309.

PAGE 79

66 Cochrane, M. A., and W. F. Laurance. 2002. Fire as a large-scale edge effect in Amazonian forests. Journal of Tropical Ecology 18: 311-325. Cochrane, M. A., and C. Souza. 1998. Linear mixture model classification of burneforests in the Eastern Amazon. International Journal of Remote S d ensing 19: 3433-3440. CORDCORDECRUZ, Santa Cruz, Bolivia. Corder. FOR, Santa Cruz, Bolivia. CUMAT. 1992. Desbosque de la Amazonia Bolivia. Centro de Investigaciones de la Curren, P. J. 1980. Multispectral photographic remote sensing of vegetation amount and productivity. Proceedings of the Fourteenth International Symposium on Remote Curran 30: 271-278. DauberDebeir, O., I. Van den Steen, P. Latinne, P. Van Ham, and W. Elnore. 2002. Textural ECRUZ. 1994. Plan de uso del suelo (PLUS), una propuesta para el aprovechamiento sostenible de nuestros recursos naturales. o, W. 2003. Control de operaciones forestales con enfasis en la actividad ilegalDocumento Tecnico 120/2003. BOL Crome, F., L. Moore, and G. Richards. 1992. A study of logging damage in upland rainforest in north Queensland. Forest Ecology and Management 49: 1-29. Capacidad de Uso Mayor de la Tierra, La Paz, Bolivia. Sensing of the Environment, Ann Arbor, MI: 623-637. P. J. 1989. Remote sensing of foliar chemistry. Remote Sensing of Environment E., J. Teran, and R. Guzman. 2000. Estimaciones de biomasa y carbono en bosques naturales de Bolivia. Superintendencia Forestal, Santa Cruz, Bolivia. and contextual land-cover classification using single and multiple classifier systems. Photogrammetric Engineering and Remote Sensing 68: 597-605. Dengsheng, L., E. Moran, and M. Batistella. 2003. Linear mixture model applied to Amazonian vegetation classification. Remote Sensing of Environment 87: 456-469.

PAGE 80

67 Dickinson, M., D. Whigham, and S. Hermann. 2000. Tree regeneration in fellingnatural treefall disturbances in a semideciduous tropical forest and in Mexico. Forest Ecology and Management 134: 137-151. DIDF. mental de informacion y difusion forestal, Santa Cruz, Bolivia. logy in ry. Remote Sensing of Environment 68: 12-25. rical ents. Canadian Journal of Forest Resources 30: 1999-2005. ERDASorgia. Bolivia. ation index dynamics in the Brazilian Cerrado: An analysis within the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). Remote Sensing of Environment 87: 534-550. Fletcher, W. K., and J. Muda. 1999. Influence of selective logging and sedimentological Forestal. 2002. Informe anual gestion 2002. Superintendencia Forestal, Santa Cruz, Bolivia. Fox, J. 1968. Logging damage and the influence of climber cutting prior to logging in the lowland dipterocarp forest of Sabah. Malaysian Forestry 33: 326-347. 1996. Nueva ley forestal, No. 1700, 136. Proyecto de implementacion del sistema departa Drake, N., S. Mackin, and J. Settle. 1999. Mapping vegetation, soils and geosemiarid shrublands using spectral matching and mixture modeling of SWIR AVIRIS image Elvidge, C. 1990. Visible and infrared reflectance characteristics of dry plant materials. International Journal of Remote Sensing 12: 1775-1795. Englund, S., J. O'Brien, and D. Clark. 2000. Evaluation of digital and film hemisphephotography and spherical densiometry for measuring forest light environm ., 1999. ERDAS Field Guide. ERDAS, Inc, Atlanta, Ge Ergueta, P., and J. Sarmiento. 1992. Fauna silvestre de Bolivia: diversidad y conservacin. In M. Marconi [ed.], Conservacin de la diversidad biolgica en Bolivia. Centro de Datos para la Conservacin, La Paz, Ferreira, L. G., H. Yoshioka, A. Huete, AND E. E. Sano. 2003. Seasonal landscape and spectral veget process on geochemistry of tropical rainforest streams. Journal of Geochemical Exploration 67: 211-222.

PAGE 81

68 Fredericksen, T. 2000. Aprovechamiento forestal y conservacion de los bosques tropicales en Bolivia. BOLFOR, Santa Cruz, Bolivia. Fredericksen, N., and T. Fredericksen. 2002. Terrestrial wildlife response to logging and -ustainable Forestry 11. Fredericksen, T., and B. Mostacedo. 2000. Regeneration of timber species following gy and Management 131: 47-55. Frederiee Ecology and Management 171: 223-230.. Garcia-Gaussman, H. W. 1977. Reflectance of leaf components. Remote Sensing of Environment Gentry, A. H. 1995. Diversity and floristic composition of neotropical dry forests. In H. wildfire in a Bolivian tropical humid forest. Biodiversity and Conservation 11: 27-38. Fredericksen, N., T. Fredericksen, B. Flores, AND D. Rumiz. 1999. Wildlife use of different-sized logging gaps in a tropical dry forest. Tropical Ecology 40: 167175. Fredericksen, T., and J. Licona. 2000a. Invasion of non-commercial tree species after selection logging in a Bolivian tropical forest. Journal of S ______. 2000b. Encroachment of non-commercial tree species after selection logging in aBolivian tropical forest. Journal of Sustainable Forestry 11: 213-223. selection logging in a Bolivian tropical dry forest. Forest Ecolo cksen, T., and W. Pariona. 2002. Effect of skidder disturbance on commercial trregeneration in logging gaps in a Bolivian tropical forest. Forest Haro, F., M. Gilabert, and J. Melia. 1996. Linear spectral mixture modeling to estimate vegetation amount from optical spectral data. International Journal of Remote Sensing 17: 3373-3400. ______. 1999. Extraction of Endmembers from Spectral Mixtures. Remote Sensing of Environment 68: 237-253. 6: 1-9. A. M. S. H. Bullock., and E. Medina [ed.], Seasonally Dry Tropical Forests, 146-194. Cambridge University Press, Cambridge, UK.

PAGE 82

69 Gil, P. 1997. Plan de manejo forestal, Agroindustria Forestal La Chonta Ltda., SaCruz, Bolivia. Agroindustria Forestal La Chonta Ltda. nta Gilabert, M. A., J. Gonzalez-Piqueras, F. J. Garcia-Haro, and J. Melia. 2002. A 2: Gillman, G. P., D. F. Sinclair, R. Knowlton, and M. G. Keys. 1985. The effect of some st. ns for fire management. Forest Ecology and Management 165: 225-234. Griffithn de la ley B. BOLFOR, Santa Cruz, Bolivia. anes Bolivia. Forest Ecology and Management 59: 1-14. Hall, Fre decomposition and geometric reflectance methods. Ecological Applications 5: 993-1001. Heinz, D., C. Chang, and H. Althouse. 1999. Fully constrained least-squares based linear unmixing. IEEE Transactions on Geoscience and Remote Sensing: 1401. Heinz, D. C. 2001. Constrained least squares spectral unmixing for subpixel target detection, classification and quantification in hyperspectral and multispectral ______. 1998. Plan general de manejo forestal Empresa Agroindustrial La Chonta Ltda. La Chonta, Ltda, Santa Cruz, Bolivia. generalized soil-adjusted vegetation index. Remote Sensing of Environment 8303-310. soil chemical properties of the selective logging of a north Queensland rainforeForest Ecology and Management 12: 195-214. Gould, K., T. S. Fredericksen, F. Morales, D. Kennard, F. E. Putz, B. Mostacedo, and M. Toledo. 2002. Post-fire tree regeneration in lowland Bolivia: implicatio J. 1999. Resultados de los tres telleres regionales sobre la consolidacioforestal 1700. Doc. Tec. 76 Gullison, R., and J. Hardner. 1993. The effects of road design and harvest intensity on forest damage caused by selection logging: empirical results and a simulation model from the Bosque Chim ., Y. Shimabukuro, and K. Huemmrich. 1995. Remote sensing of forest biophysical structure using mixtu imagery. IEEE Transactions on Geoscience and Remote Sensing 62: 165.

PAGE 83

70 Hendrison, J. 1990. Damage-controlled logging in managed tropical rain forest in Suriname, Ecology and Management of Tropical Rain f orests in Suriname: 4. Wageningen Agricultural University, The Netherlands. Hodgso the topographic effect in remotely sensed imagery. The ERDAS Monitor 6: 4-6. Holben ting sensors. Photogrammetric Engineering and Remote Sensing 46: 1191-1200. HoldridHorne, R., and J. Gwalter. 1982. The recovery of rainforest overstory following logging. Howard, A., R. Rice, and R. Gullison. 1996. Simulated financial returns and selected n nd Huete, A. R., H. Q. Liu, K. Batchily, and L. W. Van. 1997. A comparison of vegetation of Hunt, E. R., B. N. Rock, and P. S. Nobel. 1987. Measurement of leaf relative water Hunt, E. R., and B. N. Rock. 1989. Detection of changes in leaf water content using near-. Hurlbert, S. H. 1984. Pseudoreplication and thents. Ecological Monographs 54: 187-211. n, M., and B. Shelley. 1994. Removing B. N., and C. O. Justice. 1980. The topographic effect on spectral response fromnadir-poin ge. 1971. Forest environments on tropical life zones: A pilot study. Pergamon Press, New York, NY. Holdsworth, A., and C. Uhl. 1997. Fire in Amazonian selectively logged rain forest and the potential for fire reduction. Ecological Applications 7: 713-725. Australian Forestry 13: 29-44. environmental impacts from four alternative silvicultural prescriptions applied ithe neotropics: a case study of the Chimanes Forest, Bolivia. Forest Ecology aManagement 89: 43-57. indices over a global set of TM image for EOS-MODIS. Remote SensingEnvironment 59: 440-451. content by infrared reflectance. Remote Sensing of Environment 22: 429-435. and middle-infrared reflectances. Remote Sensing of Environment 30: 43-54e design of ecological field experim

PAGE 84

71 Jackson, C. R., C. A. Sturm, and J. M. Ward. 2001. Timber harvest impacts on small headwater stream channels in the coast ranges of Washington. Journal of the American Water Resources Association 37: 1533-1549. Jackson, R. D., and C. E. Ezra. 1985. Spectral response of cotton to suddenly induced water stress. International Journal of R emote Sensing 6: 177-185. 166: 271-283. l o, and pixel scale. Remote Sensing of Environment 32: 169-187. Jensen,pective. Prentice-Hall, Inc., Upper Saddle River, NJ. Johns, A. 1992. Vertebrate responses to selective logging: implications for the design of logging systems. Transactions of the Royal Society of London 335: 437-442. Johns, Kaimowitz, D., P. Mendez, A. Puntodewo, and J. Vanclay. 2002. Spatial regression Agricultural technologies and tropical deforestation, 195-211. CABI Publishing Kaimowitz, D., G. Thiele, and P. Pacheco. 1999. The effects of structural adjustment on Jackson, S., T. Fredericksen, and J. Malcolm. 2002. Area disturbed and residual stand damage following logging in a Bolivian tropical forest. Forest Ecology and Management Jasinski, M. 1990. Sensitivity of the normalized difference vegetation index to subpixecanopy cover, soil albed J. 1996. Introductory digital image processing: a remote sensing pers J., P. Barreto, and C. Uhl. 1996. Logging damage during planned and unplanned logging operations in the eastern Amazon. Forest Ecology and Management 89. analysis of deforestation in Santa Cruz, Bolivia. In C. H. Wood and R. Porro [eds.], Deforestation and land use in the Amazon. University Press of FloridaGainesville. Kaimowitz, D., and J. Smith. 2001. Soybean technology and the loss of natural vegetation in Brazil and Bolivia. In A. Angelsen and D. Kaimowitz [eds.], and Center for International Forestry Research (CIFOR), Wallingford. deforestation and forest degradation in lowland Bolivia. World Development 27: 505-520.

PAGE 85

72 Keller, M., G. Asner, N. Silva, and M. Palace. 2002. Sustainability of selective logginupland forests in the Brazilian Amazon: Carbon budgets and remote sensing atools for evaluation of logging effects. In D. J. Z. e. al. [e g of s d.], Working forests in the tropics: Conservation through sustainable management? Columbia University Killeen, T., A. Jardim, F. Mamani, and R. Nelson. 1998. Diversity, composition and King, G. C., and W. S. Chapman. 1983. Floristic composition and structure of a rainforest Australian Journal of EcologyKnipling, E. B. 1970. Physical and physiological basis for the reflectance of visible and near-infrared radiation from vegetation. Remote Sensing of Environment 1: 155-Krueger, O., and J. Fischer. 1994. Correction of aerosol influence in Landsat 5 thematic GeoJournalKrueger, W. 2003. Efectos del marcado de rboles de futura cosecha y la planificacin de Lewis, O. T. 2001. Effect of experimental selective logging on tropical butterflies. Conservation Biology 15: 389-400. Lobell,over MDSMA. 1995. Memoria explicativa. Mapa forestal. Ministerio de Desarrollo Sostenible y Medio Ambiente., La Paz, Bolivia. Press, New York. structure of a tropical semideciduous forest in the Chiquitania region of Santa Cruz, Bolivia. Journal of Tropical Ecology 14: 803-827. area 25 yr after logging. 8: 415-423. 159. mapper data. 32: 61-70. pistas de arrastre en el aprovechamiento convencional con lmites diamtricos en un bosque tropical de Bolivia. BOLFOR, Santa Cruz, Bolivia. Lentini, M., A. Verissimo, and L. Sobral. 2003. Fatos florestais da Amazonia. Imazon, Belem, Brazil. D. B., G. P. Asner, B. E. Law, and R. N. Treuhaft. 2001. Subpixel canopy cestimation of coniferous forests in Oregon using SWIR imaging spectrometry. Journal of Geophysical Research-Atmospheres 106: 5151-5160. Malleux, J. 2000. Estado y cambios de la cobertura forestal en la republica de Bolivia para el FRA, 2000. FAO, Roma, Italia.

PAGE 86

73 Minneart, J. L., and G. Szeicz. 1961. The reciprocity principle in lunar photometry. Astrophysics Journal 93: 403-410. Wu. 1994. Integrating Amazonian vegetation, Moran, E., E. Brondizio, P. Mausel, and Y.land-use, and satellite data. BioScience 44. Myers,Alavalapati, F. E. Putz, and M. Schmink [eds.], Working forests in the Neotropics: conservation through sustainable management. Columbia University Nepstad, D., A. Verissimo, A. Alencar, C. Nobre, L. Eirivelthon, P. Lefebvre, P. Schlesinger, C. Potter, P. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks. Nittler, J., and D. Nash. 1999. The certification model for forestry in Bolivia. Journal of Numata, I., J. V. Soares, D. A. Roberts, F. C. Leonidas, A. C. Chadwick, and G. T. ed nces in Rondonia, Brazil. Remote Sensing of the Environment 87: 446-455. Oksane? and land use in the Amazon, 66-94. University Press of Florida, Gainesville. N., R. A. Mittermeier, C. G. Mittermeirer, G. A. B. da Fonseca, and J. Kent. 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853-858. Nepstad, D., A. Alencar, A. C. Barros, E. Lima, E. Mendoza, C. A. Ramos, and P. Lefebvre. 2004. Governing the Amazon timber industry. In D. J. Zarin, J. R. R. Press, New York, New York. 1999. Large-scale impoverishment of Amazonian forests by logging and fire. Nature 398: 505-508. Nicholson, D. I. 1958. An analysis of logging damage in tropical rainforest North Borneo. Malaysian Forester 21: 235-245. Forestry 97: 32-36. Batista. 2003. Relationships among soil fertility dynamics and remotely sensmeasures across pasture chronoseque n, L. 2001. Logic of experiments in ecology: is pseudoreplication a pseudoissueOIKOS 94: 27-38. Pacheco, P. 2002. Deforestation and forest degradation in lowland Bolivia. In C. H. Wood and R. Porro [eds.], Deforestation

PAGE 87

74 Panfil, S., and R. Gullison. 1998. Short term impacts of experimental timber harvest intensity on forest structure and com position in the Chimanes Forest, Bolivia. Forest Ecology and Management 102: 235-243. Paz, C.egetation of a lowland forest in Bolivia. Masters Thesis, University of Florida, Gainesville, FL. Penuela50-970 nm region as an indicator of plant water status. International Journal of Pereirand nagement 168: 77-89. ge. Pinard, M., F. Putz, and J. Tay. 2000. Lessons learned from the implementation of reduced-impact logging in hilly terrain in Sabah, Malaysia. International Forestry Pinard, M. A., F. Putz, and J. C. Licona. 1999. Tree mortality and vine proliferation ment 116: 247-252. Potter, C. S. 1999. Terrestrial biomass and the effects of deforestation on the global Prado, D. E., and R. J. Gibbs. 1993. Patterns of species distributions in the dry seasonal 27. Putz, F., and M. A. Pinard. 1993. Reducedethod. Conservation Biology 7: 755-757. 2003. Forest-use history and the soils and v s, J., I. Filella, C. Biel, L. Serrano, and R. Save. 1993. The reflectance at the 9 Remote Sensing 14: 1887-1905. R., J. Zweede, G. Asner, and M. Keller. 2002. Forest canopy damage arecovery in reduced-impact and conventional selective logging in eastern Para, Brazil. Forest Ecology and Ma Pinard, M., and F. Putz. 1996. Retaining forest biomass by reducing logging damaBiotropica 28: 278-295. Review 2: 33-39. following a wildfire in a subhumid tropical forest in eastern Bolivia. Forest Ecology and Manage Pinty, B., M. Verstraete, and N. Gobron. 1998. The effect of soil anisotropy on the radiance field emerging from vegetation canopies. Geophysical Research Letters25: 797-800. carbon cycle. BioScience 49: 769-778. forests of South America. Annals of the Missouri Botanical Garden 80: 902-9-impact logging as a carbon-offset m

PAGE 88

75 Putz, F. 1992. Silvicultural effects of lianas. In F. a. M. Putz, H [ed.], The biology of vines. Cambridge University Press, Cambridge. Reisinger, T. W., G. L. Simmons, and P. E. Pope. 1988. The impact of timber harvesting lied Roberts, D. A., B. W. Nelson, J. B. Adams, and F. Palmer. 1998. Spectral changes with Roberts, D. A., M. O. Smith, and J. Adams. 1993. Green vegetation, non-photosynthetic vegetation, and soils in AVIRIS data. Remote Sensing of Environment 44: 255-Rock, B. J., D. L. Williams, D. M. Moss, G. N. Lauten, and M. Kim. 1994. High-spectral e and Salinas-Zavala, C. A., A. V. Douglas, and H. F. Diaz. 2002. Interannual variability of Santos, J. R., M. Lacruz, M. Keil, and J. Kramer. 1999. A linear spectral mixture model e Sensing. Geosciences and RemoteSensing 3: 747. Schroesessing Brazil's carbon budget: I. Biotic carbon pools. Forest Ecology and Management 75: 77-86. on soil properties and seedling growth in the south. Southern Journal of AppForestry 12: 58-67. Ripple, W. J. 1986. Spectral reflectance relationship to leaf water stress. Photogrammetric Engineering and Remote Sensing 52: 1669-1675. leaf aging in Amazon caatinga. Trees Berlin 12: 315-325. 269. resolution field and laboratory optical reflectance measurements of red spruceastern hemlock needles and branches. Remote sensing of environment 47: 176-189. Rodriguez, T. M. 2001. Estado actual del manejo forestal en Bolivia. FAO, Santiago,Chile. NDVI in northwest Mexico. Associated climatic mechanisms and ecologicalimplications. Remote Sensing of Environment 82: 417-430. to estimate forest and savanna biomass at transition areas in Amazonia. IEEETransactions on Geoscience and Remot Schanzer, D. L. 1993. Comments on the least-squared mixing models to generate fractionimages derived for remote sensing multispectral data. IEEE Transactions on der, P., and J. K. Winjum. 1995. As

PAGE 89

76 Schroeder, P., and J. Winjum. 1995. Assessing Brazil's carbon budget: II. Biotic fluxeand net carbon balance. Forest Ecology and Man s agement 75: 87-99. sing shade matic Mapper images of the Amazon Region. International Journal of Remote Sensing Shimabukuro, Y., and J. Smith, T. Lin, and K. Ranson. 1991. The least-squares mixing models to generate fraction images derived from remote sensing multispectral Siegert, F., and A. A. Hoffmann. 2000. The 1998 forest fires in east Kalimantan Siegert, F., G. Ruecker, A. Hinrichs, and A. A. Hoffmann. 2001. Increased damage from Siqueira, P., B. Chapman, and G. McGarragh. 2003. The coregistration, calibration, and Remote Sist, P. 2000. Reduced-impact logging in the tropics: objectives, principles, and impacts. Sist, P.sting intensity versus sustainability in Indonesia. Forest Ecology and Management 108: 251-260. Sist, P. a 90-1996). Forest Ecology and Management 165: 85-103. Sekercioglu, C. H. 2002. Effects of forestry practices on vegetation structure and bird community of Kibale National Park, Uganda. Biological Conservation 107: 229-240. Shimabukuro, Y., G. Batista, E. Mello, J. Moreira, and V. Duarte. 1998. Ufraction image segmentation to evaluate deforestation in Landsat The 19: 535-541. data. IEEE Transactions on Geosciences and Remote Sensing 29: 747. (Indonesia): A quantitative evaluation using high resolution, multitemporal ERS-2 SAR images and NOAA-AVHRR hotspot data. Remote Sensing of Environment 72: 64-77. fires in logged forests during droughts caused by El Nino. Nature 414: 437-440. interpretation of multiseason JERS-1 SAR data over South America. Sensing of Environment 87: 389-403. International Forestry Review 2: 3-10. T. Nolan, J. Bertault, and D. Dykstra. 1998. Harve and N. Nguyen-The. 2002. Logging damage and the subsequent dynamics ofdipterocarp forest in East Kalimantan (19

PAGE 90

77 Skole, D. 1993. Measurement of deforestation in the Brazilian Amazon using satellite remote sensing. PhD, University of New Hampshire. Skole, D., and C. Tucker. 1993. Tropical deforestation and habitat fragmentation in theAmazon: satellite data from 1978 to 1988. Science 260: 1905-1906. Smith, J., T. Lin, and K. Ranson. 1980. The Lambertian assumption and Landsat data. Souza, C., and P. Barreto. 2000. An alternativazon. International Journal of Remote Sensing 21: 173-179. Souza, t models. Remote Sensing of Environment 87: 494-506. Steininical deforestation in the Bolivian Amazon. Environmental n 28: 127-134. Conservation, 127-134. Stone, temporal satellite data to evaluate selective logging in Para, Brazil. Remote Sensing. Sept. 19: 2517-2526. Stotz, D. F., J. Wbirds: ecology and conservation. University of Chicago Press, Chicago, USA. Tague, C., and L. Band. 2001. Simulating the impact of road construction and forest harvesting on hydrologic response. Earth Surface Processes and Landforms 26: Thomas, J. R., L. N. Naher, and R. G. Brown. 1971. Estimating leaf water content by reflectance measurements. Agronomy Journal 63: 845-847. Photogrammetric Engineering and Remote Sensing 46: 1183-1189. e approach for detecting and monitoring selectively logged forests in the Am C., L. Firestone, L. Moreira Silva, and D. Roberts. 2003. Mapping foresdegradation in the Eastern Amazon from SPOT 4 through spectral mixture ger, M., C. Tucker, J. Townshend, T. Killeen, A. Desch, V. Bell, and P. Ersts. 2001a. TropConservatio Steininger, M. K., C. J. Tucker, J. R. G. Townshend, T. J. Killeen, A. Desch, V. Bell, andP. Ersts. 2001b. Tropical deforestation in the Bolivian Amazon, Environmental T., and P. Lefebvre. 1998. Using multi-International Journal of Fitzpatrick, T. A. Parker, and D. K. Moskovits. 1996. Neotropical 135-151. mken, G. F. Oert

PAGE 91

78 Thome, K., K. Arai, H. Simon Hook, H. Kieffer, H. Lang, A. Tsuneo Matsunaga, A. OF. Palluconi, H. Sakuma, N. Slater, T. Takashima, H. no, Tonooka, S. Tsuchida, R. M. Welch, and E. Zalewski. 1998. ASTER preflight and inflight calibration and Todd, S., and R. Hoffer. 1998. Responses of spectral indices to variations in vegetation Tucker, C. J. 1980. Remote sensing of leaf water content in near infrared. Remote Uhl, C., P. Barreto, and A. Verissimo. 1997. Natural resource management in the Uhl, C., Clark, K., Dezzeo, N. and Magurrino, P., 1988. Vegetation dynamics in Amazonian treefall gaps. Ecology 69: 751-763. Uhl, C.l consequences of selective logging in an Amazon frontier: the case of Tailandia. Forest Ecology and Management 46: 243-273. Uhl, C.n: a case study from the Paragominas region of the state of Para. Biotropica 21: 98-106. Verissiral resource in Amazonia: the case of mahogany. Forestry Ecology and Management Vidal, E., J. Johns, J. Gerwing, P. Barreto, and C. Uhl. 1997. Vine management for reduced-impact logging in eastern Amazonia. Forest Ecology and Management the validation of level 2 products. IEEE Transactions on Geoscience and Remote Sensing 36: 1161-1172. cover and soil background. Photogrammetric Engineering and Remote Sensing 64: 915-921. Tucker, C. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment 8: 127-150. Sensing of Environment 10: 23-32. Brazilian Amazon: An integrated research approach. BioScience 47: 160-168. A. Verissimo, M. M. Mattos, Z. Brandino, and I. C. G. Vieira. 1991. Social, economic, and ecologica and I. Viera. 1989. Ecological impacts of selective logging in the Brazilian Amazo mo, A., P. Barreto, R. Tarifa, and C. Uhl. 1995. Extraction of a high-value natu 72: 39-60. 98: 105-114.

PAGE 92

79 Weaks, T. E., and R. C. Creekmore. 1981. A study of the hepatic flora of an infrequently timbered forest. Agriculture and Environment 6: 383-393. E. 1997. Canopy removal and residual stand damage during controlled selective Webb, logging in lowland swamp forest of northeast Costa Rica. Forest Ecology and Management 95. Weidong, L., F. Baret, G. Xingfa, T. QingxiRemote Sensing of Environment 81: 238-246. patterns in tall grass prairie using spectral mixture analysis. Ecological Applications 7: 493-511. White, L. J. T. 1994. The effects of commercial mechanized selective logging on a transect in lowland rainforest in Lope Reserve, Gabon. Journal of Tropical Whitman, A., N. Brokaw, and J. Hagan. 1997. Forest damage caused by selection logging and ce Z. Lanfen, and Z. Bing. 2002. Relating soil surface moisture to reflectance. Wessman, C. A., C.A. Bateson, and T.L. Benning. 1997. Detecting fire and grazing Ecology 10: 313-322. of mahogany (Swietenia macrophylla) in northern Belize. Forest Ecology Management 92: 87-96. Yamaguchi, Y., H. Fujisada, H. Tsu, I. Sato, H. Watanabe, M. Kato, M. Kudoh, A. B.Kahle, and M. Pniel. 2001. ASTER early image evaluation. Advances in SpaResearch 28: 69-76.

PAGE 93

BIOGRAPHICAL SKETCH Eben Broadbent, as a young child, lived with his family in Japan and traveled throughSciencezing in tropical areas, and with a minor in English. During this time, he worked in Costa Rica as d with bo Missouri Botanical Garden, looking for rare plant species studied congeneric epiphytes within the Monteverde area cloud forests. Aregeneration and butterfly diversity after natural and anthropogenic forest disturbances in the Parque Nacional Corcovado. Returning to the US, he began working with the environmental nonprofit firm of Hudsonia Ltd. mapping areas of biodiversity concern within Dutchess County, NY, for use by local towns in creating ecologically sensitive development plans. In 2001 he began an internship with BOLFOR, Proyecto de Manejo Forestal Sostenible de Bolivia, in Santa Cruz, Bolivia, identifying tree species within long term silvicultural research plots in the La Chonta forestry concession. For this thesis (part of his Master of Science degree in forestry) he returned to the La Chonta concession. out Eastern Asia. At the age of 12, he moved with his mother and sister to a solar-powered house, deep in the mountains of Vermont. He earned his Bachelor of degree from the University of Vermont with a major in botany, speciali an assistant teacher for a tropical ecology program based in Monteverde; and internetanists working for the within the Bosque Eterno de Los Ninos Rainforest Preserve. His undergraduate thesis niche partitioning among fter graduation, he returned to Costa Rica to conduct research on forest 80

PAGE 94

81 He is now working as a remote sensing and GIS technician for the Carnegie Institution of Washington at Stanford University. He is currently working on a project to identify selectively logged areas, deforestation, and new roads across the Brazilian Amazon. He is planning on be 2005, with the intention of studyre, ginning his doctoral studies in ing the effects of land-use change on feedbacks among selective logging, wildfiand climate change, for different aged deforestation frontiers in the Brazilian Amazon


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20110115_AAAAAL INGEST_TIME 2011-01-15T07:23:33Z PACKAGE UFE0007800_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 66494 DFID F20110115_AAAJUA ORIGIN DEPOSITOR PATH broadbent_e_Page_90.QC.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
c5ad3bcfbbec11202e7902cf36cef8fd
SHA-1
05d643656cfa91226180d2fa168adea437542768
66439 F20110115_AAAJUB broadbent_e_Page_39.QC.jpg
fd4a607944dff1be12cbd3d39d7e40ea
3fb039a25f0ebf1f8b9b9fc71c79dd103f034b83
25271604 F20110115_AAAJUC broadbent_e_Page_08.tif
9b27b00d8bf0d3696d50752049560859
426c72836a814ef597b8ca07b4fa034d29bce80d
110461 F20110115_AAAJUD broadbent_e_Page_64.jp2
56bee4b420696d5c95e15521b57ebfef
e2d8683390e31aa94c031e468451a21b4255a7df
177870 F20110115_AAAJUE broadbent_e_Page_55.jpg
b2295ef06f96cfea4f69e6c3fcecfd26
c80838257deac4df7462373c6cd927fd577c2fd7
1915 F20110115_AAAJUF broadbent_e_Page_25.txt
cc4565d268d03e66a1c9b6c4b893aa6d
1494b9bb32fd5f76071429c6d93ce664149cb18e
111396 F20110115_AAAKAA broadbent_e_Page_40.jp2
7165041e945a543b84e43503a62c5944
942105f4c83d225e581ac0075922ce3b38fcc161
8941 F20110115_AAAJUG broadbent_e_Page_01.pro
327a990d483d8813c281a1232e3d4789
173c2355be85439cfc5f5b3bccd04760e88ba141
114333 F20110115_AAAKAB broadbent_e_Page_41.jp2
33ec7b824d5223255432afeaa7314dcf
668d7fe0dc6a7455d4d2dffe67e31f8274932b06
2361 F20110115_AAAJUH broadbent_e_Page_39.txt
606609bae03dec06b93cfb73f92e245e
883f0cb15436d396e8418434b4fe8b95d2f9baab
107776 F20110115_AAAKAC broadbent_e_Page_42.jp2
8e1ec585ab86b4090212d0ca078c6747
18799d7d80fb603308f49a042a2ee5145943358a
31889 F20110115_AAAJUI broadbent_e_Page_92.pro
530c70eadb29d072aa3c85c13997fe34
12b8f0780c9a3b723766a2e498d33336f7ef5dfa
114696 F20110115_AAAKAD broadbent_e_Page_43.jp2
e625413ea55c76429da5e19b9afa43b5
5e96952498fbd4357587eb3a58f790d57a6aef82
48059 F20110115_AAAJUJ broadbent_e_Page_91.pro
9c69a76042afb20d3bc3c3324df62f06
65a66dc02fd239a359fb882a0599bfac5827b957
110365 F20110115_AAAKAE broadbent_e_Page_44.jp2
6ca8ba270a892f039dff4adb0d72aaf3
7ffbe4971320c26d90bb95b0b13ce63eba3becbc
211774 F20110115_AAAJUK broadbent_e_Page_46.jpg
508d3daa176af9c0504750d243da217a
a01c63a0748d5a13ee79ffe842921538b9b632ba
103814 F20110115_AAAKAF broadbent_e_Page_45.jp2
260d64a93546d3dfa5735101fce49459
b4675c4fe22457cf538c79fe15bd1ee18441d1ae
54932 F20110115_AAAJUL broadbent_e_Page_36.QC.jpg
980d5d90a2378f77e80c85b7dfa09c0f
21c96a8fedf79cfd6f078893fbaa7d6a8965007c
110446 F20110115_AAAKAG broadbent_e_Page_46.jp2
5dde84c19854605fff0a1544e31f6867
0c7f3d08cef6789972fffa8eb0e90a8e77af9700
F20110115_AAAJUM broadbent_e_Page_49.tif
32060f8d94d5e6a2506830ac3719022d
a2584422fba11512e82299240866f1f2da8fb63d
93415 F20110115_AAAKAH broadbent_e_Page_47.jp2
d123f8fa8878486dc4da6626085d47cc
fd7e337c67a7e65c34bd6f385d6fc955e0c34bbe
817365 F20110115_AAAJUN broadbent_e_Page_19.jp2
8fc666463a9980bf2866820597189c28
bdaf4d3d0904b72ce7951abdbd78769ba7a5cb46
1420 F20110115_AAAJTZ broadbent_e_Page_55.txt
3d1c15c587528ce7af80ae2685f02513
2efac669247b66b37be8dfe76d4e1180d629fdd5
1051985 F20110115_AAAKAI broadbent_e_Page_48.jp2
1b9d7f42e175f81c28e21fe46c18f604
4349890946c2a4b9f1490174bf55bc0ccf2b7a08
58433 F20110115_AAAJVC broadbent_e_Page_01.jpg
36f8fd130078444d06e27c5389717031
237337a330793525dfc4a665d18d777c71e55351
707301 F20110115_AAAJUO broadbent_e_Page_66.jp2
8a717daf055125ce77791e593a2574d0
2181be1850c2c5e472c3231074fe26283abaeb27
755629 F20110115_AAAKAJ broadbent_e_Page_49.jp2
732a52ea878c7146c9f1f7b294b37040
b8063b9405a357550861f93c553369ccab73e2a6
15441 F20110115_AAAJVD broadbent_e_Page_02.jpg
23a4dfb8031c0ec4e58d86ceccaae71c
b2b00c32a584341652bbd1303226aa827a7acb6e
220049 F20110115_AAAJUP broadbent_e_Page_76.jpg
24a05ed328ef310414ed96b7fdbf39d1
4a1cb56dee49360f760262277bf6bbf5e396b288
106240 F20110115_AAAKAK broadbent_e_Page_50.jp2
b851c582ace4b1f02cb88e59c9293058
644dd865c897cddf59c1e3b8274dbec8d0b0a0c7
14397 F20110115_AAAJVE broadbent_e_Page_03.jpg
becbb4792e790b381ee99a9d92af7661
4b2f6a957ae74168b30a68051943d31c334f66af
66749 F20110115_AAAJUQ broadbent_e_Page_82.QC.jpg
804d6406bbdb01dcc5c775480da2119a
ad86ed174bebd9fd995dcb6527397809a8011ba3
107598 F20110115_AAAKAL broadbent_e_Page_51.jp2
1e300238f71781629e6cb5899f4da586
d685ac93ed37df3990bc7047a431925f3cd4ed09
178454 F20110115_AAAJVF broadbent_e_Page_04.jpg
52e7088c0fb045eb95d4834e45443001
d46cccc71435ec850b35b638ec0ef4afe8995c34
208640 F20110115_AAAJUR broadbent_e_Page_15.jpg
abf17ed91e19d6defcc394babc7fe341
6c6de4e31717824a5a58c0c55fd29e609b3c6911
734150 F20110115_AAAKBA broadbent_e_Page_68.jp2
3065114bcf2df340b8c23de8078491ff
d40bd6396b9775d5ba44091b66d9ffc8045ff7ad
842673 F20110115_AAAKAM broadbent_e_Page_52.jp2
8923f1cf583614a8153be059fe679019
408471547e00eaba5f4e9fc92c5bd8057c39d123
153816 F20110115_AAAJVG broadbent_e_Page_05.jpg
da3dd0d38e0998e63fd08ab153137692
8f1ba6d7e2dd1e1621119734299f5f9db5bd5a35
59782 F20110115_AAAJUS broadbent_e_Page_75.QC.jpg
27b56888e1dff7eebecd90056cb29911
ede4f7b6c34791e1678d9e60ac4b3e258b165d56
87759 F20110115_AAAKAN broadbent_e_Page_53.jp2
5b898a5a02777351f1fe64760c5330b4
c34ce9d05f2514bba849e11692b6ce203f1b8769
235359 F20110115_AAAJVH broadbent_e_Page_06.jpg
115d3a92de2307159d60cddef2b617b6
c167187f1fbe4f81d8b6a6ea1cddbcc83a059adf
24317 F20110115_AAAJUT broadbent_e_Page_40thm.jpg
c67a264f6495e00670764ee7b7131211
b3dc632e622b7884969e8794b5d067aa45ac9878
335361 F20110115_AAAKBB broadbent_e_Page_69.jp2
af8d9ae224e8faf542ed95abba816e29
5c5e971d49ad2eecad8c2b8c3f29344276fe270f
729523 F20110115_AAAKAO broadbent_e_Page_54.jp2
64109f7eade4c0ed7fcd1c8f554136b1
ee253f0ed47b5b15125240fd17df0519fed0a6ef
227537 F20110115_AAAJVI broadbent_e_Page_07.jpg
5687bd70c03d1b3a9ce16f137607c939
91322cfddb9626318e095b77432442fc088ebda2
78758 F20110115_AAAJUU broadbent_e_Page_25.QC.jpg
12f436ded9619d6a8e17e259bb781b88
99804024390659c745b9cbbee59963cb6f1d169b
420907 F20110115_AAAKBC broadbent_e_Page_70.jp2
05ed82eed98fdced9862886864215d1c
4c3bf0770b3e8e3f99a1d9907877e40c4e8e1a73
817322 F20110115_AAAKAP broadbent_e_Page_55.jp2
e347e051672f1762161c006d23a7314d
f55654a0553c12ebb1464e35aaa8744e59aaaeae
290820 F20110115_AAAJVJ broadbent_e_Page_08.jpg
d261d953e38fe20b7d8c3a8d9920ed15
aa948d7b8d79830c86213c03f069fa31323695be
41562 F20110115_AAAKBD broadbent_e_Page_71.jp2
ea9f46c1d58a0b23ea5fb40a18160e26
8bfb708f64038482853b93d770623cd316671e2a
92157 F20110115_AAAKAQ broadbent_e_Page_56.jp2
06d32cd4de8654b8cf046d773be3d9b2
b50bbf449421f0339814704617f3cc5b8f273c71
173468 F20110115_AAAJVK broadbent_e_Page_09.jpg
2f26b127927fe5dd494d108db9c27ba5
d8f7644376de0debbcb05321780c1264d38a5a01
16539 F20110115_AAAJUV broadbent_e_Page_92thm.jpg
8f5c1ec35dd68ffd949c7662e81aa129
87a93b1499bbf07032837f365793eae901b170ab
50156 F20110115_AAAKBE broadbent_e_Page_72.jp2
3e4d336f0ef322d5177e5356083d9404
3f1ee675d306814779c92799942ae4d26a5326b0
388373 F20110115_AAAKAR broadbent_e_Page_57.jp2
827772bb030155d180389ee57ab9b9cc
928dfc1036af3421451b575aecee21ff1d90796b
280556 F20110115_AAAJVL broadbent_e_Page_10.jpg
be3348602f3baa49fd66868ee272b499
8d1669ae0ef774879101c881c707d43a4cb6b8f2
51161 F20110115_AAAJUW broadbent_e_Page_26.pro
0f2fa69c36c73454ba935e5014eb53e1
69152207b9daead0db7acbd679af19860f64d027
50414 F20110115_AAAKBF broadbent_e_Page_73.jp2
0adf0dda5435d787e7a953286fd86a79
2c29f9fdc9c46db7ec0d5f2f3ddea4732d9e0960
217919 F20110115_AAAJWA broadbent_e_Page_26.jpg
6a68c0103bbe76353a5268cb8a70386d
0006aaa213b03170a2bdcd1adc445ab03012ded1
110646 F20110115_AAAKAS broadbent_e_Page_58.jp2
73538facc69b322937ca7a387ec3c4d8
a9dba0b0a0b9b4fbc735b8365c807999f771fc0b
72281 F20110115_AAAJVM broadbent_e_Page_11.jpg
9415fc5f312fa95ce460a1425b3de3aa
b7721c36f8fd51ead3e86889d1e7a25ac47d54b4
2009 F20110115_AAAJUX broadbent_e_Page_31.txt
8185e65e4554f9c791dc5a3be78fb72c
7a760873cf63496d6397c19468fa33af715dfbcc
50394 F20110115_AAAKBG broadbent_e_Page_74.jp2
a77e3ce54d99cd64f41f031482fcf241
94bf6134290f6abf3aa36655103015a8efe4b58a
211119 F20110115_AAAJWB broadbent_e_Page_27.jpg
83cbe35c11e418ac7728891e56f7ea30
6fbd0ec9106ad7defea91616d55d55c1de02fb65
819828 F20110115_AAAKAT broadbent_e_Page_59.jp2
9b03382fa610b64efa8fd3e942eb535f
7cd5b8a52fa1a1502e95feaf16ecd683e4a33dab
176672 F20110115_AAAJVN broadbent_e_Page_12.jpg
c66b05e69436eef661e369dc71210a1a
8fb48dac558ceb3a8674fea6fc49d187981da16c
118954 F20110115_AAAJUY broadbent_e_Page_78.jp2
33f021c59a5bffba2ae71d2de93b973d
ca10ac4e459902530f03b2dbb9286b32e6d21c1c
95735 F20110115_AAAKBH broadbent_e_Page_75.jp2
884b4c998665bc39903f33d69babb74f
297e1a91ae4af44a003e8e50bcf111773f2d7f9a
207215 F20110115_AAAJWC broadbent_e_Page_28.jpg
2bbe21029d7df4ac1492c1e655f5a059
50de9ef92d6320394e1d94b2eb29ac61d6d8d99c
685021 F20110115_AAAKAU broadbent_e_Page_60.jp2
e3e5d1fc037dfb523af41a6d85d48c40
cd0aa3e58329595608a458a990649ed73afb1718
44215 F20110115_AAAJVO broadbent_e_Page_13.jpg
ad5019d189ac020d1b6fd90a86a2bb0f
411e086e41086fd12cf4ac556c5acac45590ba0b
110707 F20110115_AAAJUZ UFE0007800_00001.mets FULL
88b19e8fb3bdc0451e7afc79046bbb78
b8e8106139c466c7aded629b9e99cf3cbde9969b
113000 F20110115_AAAKBI broadbent_e_Page_76.jp2
334e9687024930090228873bd79d6f44
cc22556d9a1dfcbd32874b0b6231b9fe36f10a86
215650 F20110115_AAAJWD broadbent_e_Page_29.jpg
1064433c245f618a736405ce2d4827fd
86ff4b9303396d6ad8dc9012d7816f740e617cb8
821767 F20110115_AAAKAV broadbent_e_Page_61.jp2
204e9b2a3bc9d8307081fcbacfb9bb66
f828900ccbb6896b283ada22e018e2bd79b9a137
193506 F20110115_AAAJVP broadbent_e_Page_14.jpg
976983a0cf059f91eaf3ba2d38f5bb39
d03268b3e82d8caf619d7d579f5ae7602a8d4184
111087 F20110115_AAAKBJ broadbent_e_Page_77.jp2
ffaee6c43298060f0dfaa42941eec080
5cfaaef8c04807417b5784bd1f8c9cc416414c42
214731 F20110115_AAAJWE broadbent_e_Page_30.jpg
0c40d439c695b16fb0c567fe8fb98020
0d7c79ba799dc90ac60fbac42f1dd3176eedb274
98418 F20110115_AAAKAW broadbent_e_Page_62.jp2
aeb72305dfef20f721c9aa3bace379cc
9c18f8191066ab29dbf4d14e5b4c73c7fbd5e988
221435 F20110115_AAAJVQ broadbent_e_Page_16.jpg
ff3ab0665e248bb6901e2b5a8bd40369
436c23a151bf9c05a6a058b08c1715b230b499aa
104166 F20110115_AAAKBK broadbent_e_Page_79.jp2
e497088d1a08df18d87ec6dd979e15ec
43bd957ee9ee04d365d3c9f4a0d607715a401a5e
218100 F20110115_AAAJWF broadbent_e_Page_31.jpg
6e695775a0d0abb850dc2512810683bf
e89dd26a4a681122ae2bf0987321f33275ae2dc5
91283 F20110115_AAAKAX broadbent_e_Page_63.jp2
7087a466fc184d6c2fbd5efa952012aa
ade7a302f6af3e7a81b826202c46a063a59153d1
192796 F20110115_AAAJVR broadbent_e_Page_17.jpg
dc9acb1d6e0db9b654bc75d2f9d90704
a85cf87a00090beeb2d5b2498a00a3488ed7e681
1053954 F20110115_AAAKCA broadbent_e_Page_01.tif
06dcd05993ec2a214ae15fcc9f3d2c98
c0662bb7a6579d3ad52630d57b4cd423d457e70f
113161 F20110115_AAAKBL broadbent_e_Page_80.jp2
ff022cbee2863e836831ddd4cdb853a9
f666316dfe723acd16ff78cff61c9fc32845ab1d
210862 F20110115_AAAJWG broadbent_e_Page_32.jpg
6eed7cfef7895c6f39d77b2edfc3febd
be70a96bfb8c544dd42d196a31ed520b5bd7d416
99887 F20110115_AAAKAY broadbent_e_Page_65.jp2
5160c6c10a94aac961141e810ef50db8
9ce386b0aad20f7e0a97b75a0167fd2adc6b229a
207325 F20110115_AAAJVS broadbent_e_Page_18.jpg
800b562c7534937c4f06a61940dfa144
bd4766a56b0d6c56a9d71428be7a6b1c06dd49d7
F20110115_AAAKCB broadbent_e_Page_02.tif
81c835e9939b836d6d34faf58ea7d1f1
a35d644052dd98eac6555cf9d1a5c24a44147e5a
103907 F20110115_AAAKBM broadbent_e_Page_81.jp2
1d5f3bc2400941455ee1f5a348eba7ec
8ee754b1a35c7e9c06db6d1d0fc2fd4b67e90882
174713 F20110115_AAAJWH broadbent_e_Page_33.jpg
771c1e093053f29157bb2e00bcda1f61
62a3d1b01d1b8a2bb0bc5e3045ded66f5adbefb8
807313 F20110115_AAAKAZ broadbent_e_Page_67.jp2
9086883658169ac48601cae0187cdeb1
195df7970564ae6f0bbe7945a772807378cb4a09
186960 F20110115_AAAJVT broadbent_e_Page_19.jpg
016e2ced53ce9dd432f04c40ae4f7d57
213c3e53da9232df2b8a27c439cc2b52e3a80899
106396 F20110115_AAAKBN broadbent_e_Page_82.jp2
59b36b1f5ba4c3c2c101e4ace1d3d38c
cefc238d7621345bb12a5069fb6b491c22215915
185907 F20110115_AAAJWI broadbent_e_Page_34.jpg
70269ea13610f00bbf99df7cc665af04
4ed151f8aff2d2bbaa30249300631bdee07e6726
171484 F20110115_AAAJVU broadbent_e_Page_20.jpg
aeefd30cc0f0c2e9ede3ef8a2aa17d9f
bcfc403eba55d2800914770eb291534cc505d73b
F20110115_AAAKCC broadbent_e_Page_03.tif
7479ff50b8f435adeab4380760c86c22
7caa25e1b439a8550c3630d1067d76df2b282257
103762 F20110115_AAAKBO broadbent_e_Page_83.jp2
4fc0886f548b3abafab0dac1da9d024b
42f2a2978c1dfba569ec3b98b5ee9d781684941c
203313 F20110115_AAAJWJ broadbent_e_Page_35.jpg
f8a3d6232c7c3458997c82f58e6aa222
ecbd65751bb8b7146ce5f4cb974c176cb52113c4
208310 F20110115_AAAJVV broadbent_e_Page_21.jpg
c3858a2bed3c790c549aefdfd29728a3
7f85d82ac0a7480ae0468915037c34f8d51bc73f
F20110115_AAAKCD broadbent_e_Page_04.tif
e4473967777a6492ab476f5903e671e7
f8d78f3ce1facf035dff8ccc9e6754dc7869c9d1
110223 F20110115_AAAKBP broadbent_e_Page_84.jp2
6f3436bbc38ba56f93560eb09c1f1d19
28453d22acd91a8ca1e189841c986271007ed86c
158600 F20110115_AAAJWK broadbent_e_Page_36.jpg
c1eda892874172eea11922d5036178d0
5af4671218f38647e9611ea07fba1e1f89748ab0
F20110115_AAAKCE broadbent_e_Page_05.tif
847061cc02a361980ac953b17fe6316b
b32a3aba2ebf2bb0f8771c1adefdbb799c6c504b
110928 F20110115_AAAKBQ broadbent_e_Page_85.jp2
fd0055871f10f5f118a97d3da0f27644
33686d10dbc304249862f6993a9a4076a397a790
285721 F20110115_AAAJWL broadbent_e_Page_37.jpg
9abe615778e9603138937710a0a08230
9de961c086ea1bf56bcdea003d44253f166a7607
166585 F20110115_AAAJVW broadbent_e_Page_22.jpg
f3a0fcb18914ebbe3f89e35f544a5d73
9bbeb087aa187561b0acf41e3e0cf9566b5daa34
F20110115_AAAKCF broadbent_e_Page_06.tif
6f017e5c0eaf759b8daf76edec6d5927
4112f97277b1564aeb3b74b0a77489f0e84bea7c
102630 F20110115_AAAKBR broadbent_e_Page_86.jp2
46a6df8d855cc4dea8252f2e6fd52e84
c837cb1072fa9fb4c2bdedd181d065520836fae1
197617 F20110115_AAAJWM broadbent_e_Page_38.jpg
890dc1be85749fd1a1c74ee537249256
82b8f9b7b07a135ed2c6798e3e4ba78d7eb6d827
205876 F20110115_AAAJVX broadbent_e_Page_23.jpg
eb354a00f98df770dd05c41897328a78
817d9fd54d2511e450262284a084cd6a335155ec
F20110115_AAAKCG broadbent_e_Page_07.tif
18fb1087f68b10302fc572a65f039e77
1fc4fbe9c3632f69f816f4bbba6a79abbddf30ad
170583 F20110115_AAAJXA broadbent_e_Page_53.jpg
ba1dccae2e3fb57dda5d2ff7a21f3a73
e589226526918b27772004527fbd0bad8bca3623
106157 F20110115_AAAKBS broadbent_e_Page_87.jp2
5aed6bbfdf1173a97e9263846befddbf
212803aaea6962e43b2ec6d5e39b7ec69888d165
175578 F20110115_AAAJWN broadbent_e_Page_39.jpg
f1d1d0cda63625e6ee9a315b028f1757
1d39a96339a995333d0494c4848aa1b0079460a1
213129 F20110115_AAAJVY broadbent_e_Page_24.jpg
9928bf7deae260308326c2790d67da43
1b0709eae648786d0855562b3933f78b24ea83bb
F20110115_AAAKCH broadbent_e_Page_09.tif
ecfaafe0903d713318d63ec5e28b1206
38c00d56be18d1af8b85b67fa0bf1ddb23f44a70
161513 F20110115_AAAJXB broadbent_e_Page_54.jpg
eb297f826fb12a7d8c5e555cb74cf1a7
246b64357928b37d1b0965c6039e515300ccd296
108695 F20110115_AAAKBT broadbent_e_Page_88.jp2
5ff62c72a54f9b5824d8b9b24a0c0558
07cc73f2b02bfc5cb56e18416f823570151a5d04
210578 F20110115_AAAJWO broadbent_e_Page_40.jpg
893727ddc89dd2ed44624a16af3cb6d6
0f6959f7bc689619abadeebe84ed55d91b0214d8
209877 F20110115_AAAJVZ broadbent_e_Page_25.jpg
86c36d7ead6846c6e254ce60fe2f2f01
29b6d23570036c459e42100dcb69741247406bcc
F20110115_AAAKCI broadbent_e_Page_10.tif
87ccd9f42ff3818ba3245bde29e2465d
fe54a3a1b07d9e3fc2678279cfff96ed32bf74a4
174907 F20110115_AAAJXC broadbent_e_Page_56.jpg
f5f29ee0dbd267bf54e26ae25b7d8965
efc61cbada189ed39bbfd1732ca0769a0cf64cbd
109217 F20110115_AAAKBU broadbent_e_Page_89.jp2
7cceb444688f58283e9548e167987b41
c0fa1875d76804ab37b7ad9f5e1facd8d22ea196
218219 F20110115_AAAJWP broadbent_e_Page_41.jpg
ca5a2f224dd41b95b476a9732439ea06
a4e6c6158c74540edf7954312c7cb50c7500739f
F20110115_AAAKCJ broadbent_e_Page_11.tif
bdf0d705e250eb3122b966e21c409fdb
42b24ae240764cc91164ab0415e8807fb7e738ef
122920 F20110115_AAAJXD broadbent_e_Page_57.jpg
071bc0a76ea2816e46171af34e4c0723
016d86cacfc4db9d027add6042cddf51336534a5
107091 F20110115_AAAKBV broadbent_e_Page_90.jp2
6134b2b99b558fdae617545de8f756d8
f5f21091b4b0816c84662de0f9c6701a92808b63
206460 F20110115_AAAJWQ broadbent_e_Page_42.jpg
3b0c202bc7fc865280e023a558aed334
884932065d5e30a45d8d9e9e2ca631a4e4caa67c
F20110115_AAAKCK broadbent_e_Page_12.tif
3d1e2223d04710372c2d42f403a89e10
2e571231effdb2b5bbb2da4337f7307b7ec1785c
218161 F20110115_AAAJXE broadbent_e_Page_58.jpg
8b4cdbbcb442e57782d05b09dcfc08f9
769981c158f9037b75cdc765c9f54e7e77a63409
104702 F20110115_AAAKBW broadbent_e_Page_91.jp2
9bf7e923f576fdd5acd8ea54e99a1949
71aae441621458a6c765b89314e9dd07e9bdbee4
220783 F20110115_AAAJWR broadbent_e_Page_43.jpg
8ad06766550f8916926df16e4e39c7e2
7f27aaff7da4a3b0c2e9cf6207ced119dd6f2f10
F20110115_AAAKDA broadbent_e_Page_28.tif
fb49e233c792bb6d9ea7dee0a5c1b697
02ece7fbb02c2438bd21c19f7de70b2c2d792278
F20110115_AAAKCL broadbent_e_Page_13.tif
874830fb2c696416266e5eebc166a253
a79cb69d90612a2bbfb565123829b5b6e371ea15
197207 F20110115_AAAJXF broadbent_e_Page_59.jpg
fd9676a6dff9cf26668198a40b9105bd
749cec6173ce482d90408b824e0fe0181bf9156c
71938 F20110115_AAAKBX broadbent_e_Page_92.jp2
9996ef440995994b9d8dffe534c27e04
3a0bd88b6fbbb34e565b3e639debc47332359ef6
214527 F20110115_AAAJWS broadbent_e_Page_44.jpg
b2ef56b878b14b29ed307bd8ce4ae1ac
52b94cc0b80268d9dd1d5206020b4e100c3151a2
F20110115_AAAKDB broadbent_e_Page_29.tif
5279395a235abf48a4e00bf2b32886fc
d35bc8d8cde98014763fb673230d17adfe723ea4
F20110115_AAAKCM broadbent_e_Page_14.tif
d1838e8c40527a8d85b32d114fb8e942
b36b437ba645900e8d3820760b7e170da410e3ac
180503 F20110115_AAAJXG broadbent_e_Page_60.jpg
eae0e7ed8c6458a0c11766e600c8cbcf
048aecc09b55ace684d7011db8b4bccfe7172c44
92242 F20110115_AAAKBY broadbent_e_Page_93.jp2
709fc1528b6d95510b170e4836894478
331b56df3afeeb900bfa6b2bc4c265abc598fe29
198349 F20110115_AAAJWT broadbent_e_Page_45.jpg
ca5ff965b7cfa653feb36fa9bc87e0eb
2bce06d2ba6f192fd9f04c15eade64cb3377a577
F20110115_AAAKDC broadbent_e_Page_30.tif
eba36b38e3159415e9a4047b01a1513a
5605855bcf7c7aff14de7a971ebd33fcb6577c7b
F20110115_AAAKCN broadbent_e_Page_15.tif
f2ebe4052217c5e2bcb78f961f67c8fd
44d02a0c9d6b009e6d79f661d75936f6c7399797
204641 F20110115_AAAJXH broadbent_e_Page_61.jpg
489e0dd928cf59c76b106cce317506b1
b12c3baf812c651ead1538b2e8395424cd85c7b2
32921 F20110115_AAAKBZ broadbent_e_Page_94.jp2
6b600e055b4a52c259109a5057f010fa
73bb77f9bed42c86269ef55dc4933610d3b5e13e
179786 F20110115_AAAJWU broadbent_e_Page_47.jpg
ab7deac9855f09e6bd9864a7d02d9cfe
57b9286d73ac3ea2f8adb4dadcedb7ab698efa41
F20110115_AAAKCO broadbent_e_Page_16.tif
8a1ea995f9ecdf065fe6ec1b190c321f
f0c9ec6120b8aaa8521b354b0a59680ebcd52b8d
195698 F20110115_AAAJXI broadbent_e_Page_62.jpg
6d45b633f4293e782376502a8f59f082
a66503f5b7f680acb77bec1037c61b6aefbbdda2
231733 F20110115_AAAJWV broadbent_e_Page_48.jpg
1fa69bc775615dd4f3d586eda16853c8
49212587fe8f993a5df725079116983ee8d3dda7
F20110115_AAAKDD broadbent_e_Page_31.tif
d0e136508c1d28827b130832eacb102f
107db407fb6dd438e5146a5bb48ed3c1733b10b8
F20110115_AAAKCP broadbent_e_Page_17.tif
9ac25973105bca98a0a04b570527ff5f
8bddd70ed616c68543733624956e9b464d0f0e2d
181923 F20110115_AAAJXJ broadbent_e_Page_63.jpg
dd1fd9980072b4c0f5c6d42ade6e3585
e72555921df1c74eaf115623d56d720f74dbc801
F20110115_AAAKDE broadbent_e_Page_32.tif
2ba0d08d306ea5d25756fd69e2750fa0
61aab4d09e0cbfc78428d9b477774b415c9f8bf6
F20110115_AAAKCQ broadbent_e_Page_18.tif
bf9452fd4305e603c81f1742e78b0409
1abaa498e2243ae47cd5e1b1b2df3085fed6ab0b
209725 F20110115_AAAJXK broadbent_e_Page_64.jpg
138c369a345822b5dbd47b4f6ba28ed1
a752c8749c92aeb8588a9ef169e92d5e156815fa
181117 F20110115_AAAJWW broadbent_e_Page_49.jpg
0b22a4263fe35ff67180cbcd78dae3c4
d723dcb330fffd8d3ab9e2a824e63dafe653258e
F20110115_AAAKDF broadbent_e_Page_33.tif
d066f7ffa276c24a6c303cec1268173f
b19a0304b151c0672257b70640d13022fa3e0a40
F20110115_AAAKCR broadbent_e_Page_19.tif
fda05064150c582462e1d07500dce349
9bc66645ede009f865b46bd8032cf6132d91f2b2
193761 F20110115_AAAJXL broadbent_e_Page_65.jpg
f479f4b0444b6c626f70e0ad4b3a5359
2c97335914986f71bd7c11edbe7ba8a09eb4f3bb
F20110115_AAAKDG broadbent_e_Page_34.tif
cfcd41136457d6a628ac6cec79f699ab
8dca69791616fff6ecafdff998ca3a9be5ce3b7a
198985 F20110115_AAAJYA broadbent_e_Page_81.jpg
d074c1df1511fbd644484011e795f1bd
3abfff5a1f8c53eadafea8d180e1b7e3c4b1a75f
F20110115_AAAKCS broadbent_e_Page_20.tif
24463607b985aa9ed03a8d6b3cb4a4f8
d3df7b8c46ee6361e14899c262e0c4ddb6f8f4c0
132771 F20110115_AAAJXM broadbent_e_Page_66.jpg
6ef98bae139e7da22b947ebdd48ab39d
74dfdab80967109e45d2955112be92f3d384772c
199100 F20110115_AAAJWX broadbent_e_Page_50.jpg
084adb002f47198e00b5cf4a41fb3efe
9d1cae7a93bfaff1edd6088f3071604ead7d1975
F20110115_AAAKDH broadbent_e_Page_35.tif
a45f52d65fad2e72673f0a78973263ab
3a1389a9a8b3c7cb6a772ab32c2882b85909143b
204731 F20110115_AAAJYB broadbent_e_Page_82.jpg
20fb445c302bfeb69162b85935b40336
791c9b2ee47d7a60e3a2368e8fae0611b34d46a7
F20110115_AAAKCT broadbent_e_Page_21.tif
9da56e664e123dc63351aa609cb6c150
ffbbd4ac787024f36e72cb2302666143bb10137f
149191 F20110115_AAAJXN broadbent_e_Page_67.jpg
ccb739a38ccfb040af12d8369ee71e57
f494ee23096590be75e4547ed16e894ea13d5a28
198326 F20110115_AAAJWY broadbent_e_Page_51.jpg
7c064fd51e7c9a806a4c7884b8cacd64
8bfa7d9653ecd1b0fd9a6fa36ce9b0a4dc719b53
F20110115_AAAKDI broadbent_e_Page_36.tif
92f043ea67e4fb02528cffddac55a0c5
968421853288d7a9d8a1f572a10f924b2afbcab7
199287 F20110115_AAAJYC broadbent_e_Page_83.jpg
8922439db5c544c5b3c6530dfc97ac60
2e16e38157bedb0ebc16bc235dcd90d04817dfd6
F20110115_AAAKCU broadbent_e_Page_22.tif
35e31c96584ecfb2f202fe1e7f29682c
412d1fbc3e4fd58e7c9b3f469a5e99141a1e2200
138423 F20110115_AAAJXO broadbent_e_Page_68.jpg
57eecf389e83b8c488dcd337e7f90f75
f246dc23662d7c398aac0435b69e2d09d021afe8
188543 F20110115_AAAJWZ broadbent_e_Page_52.jpg
71da2076ae6f271001aef9a33c311f76
5b3073e9a6ac1910b5df2d2b32803cf3ebcc42e7
F20110115_AAAKDJ broadbent_e_Page_37.tif
d1ac78ece7b3fd7439b1f5de9b326684
c0a63777a28ee9612aeeb515527245af831a386a
210642 F20110115_AAAJYD broadbent_e_Page_84.jpg
18e5320d22e1375d7bc31110beff437e
0fd40172f045b56bfe4f9491e8a401e229e8f671
F20110115_AAAKCV broadbent_e_Page_23.tif
38b2f0ad076edcd2061904030131883c
5bd6538aac081cb55aa4ea0f6b60fe01c5a93913
99431 F20110115_AAAJXP broadbent_e_Page_69.jpg
fdedc4e3438b9e3471869dbfc94b8d4c
0ed077e9648e88dab64f44a3ad708b378f798ee1
F20110115_AAAKDK broadbent_e_Page_38.tif
54aecf04fbb4ae67af7ad0126318221c
63511434af6363740e8a7e2ca88c966d843c7b01
218580 F20110115_AAAJYE broadbent_e_Page_85.jpg
f637ed845c8b0c64b2ad351b64a596ce
bff8dac546414601812275b8e9eac9fb495b83cb
F20110115_AAAKCW broadbent_e_Page_24.tif
0ff5ef4ee6006490462215283eac8168
c9c9ed7b8174df8f137c0c1e1120589fc4009f0a
110527 F20110115_AAAJXQ broadbent_e_Page_70.jpg
76083059a1303877b00d093cdb6cd647
65aaa71d1e606448f172ade6bfbc909d2284b2bd
F20110115_AAAKDL broadbent_e_Page_39.tif
2c5725e2746078e8bc33cbb5bb55951b
5a8bb9f3043a0b153be3a8abd996c8bf393aefcf
202634 F20110115_AAAJYF broadbent_e_Page_86.jpg
d8c038c0fee1a3d7ec1efae14b4bad3e
e784be7ca7fd2caffed33bb3ec7f4566870645f7
F20110115_AAAKCX broadbent_e_Page_25.tif
f6627891ea70759014a8b4d92bfeb697
17a34fea5151eb9e1f1f220bad26f2a185660177
83078 F20110115_AAAJXR broadbent_e_Page_71.jpg
688fa97155253937adbf7bacab7b404f
826b663f79e970b8cfc143c25669f0f34048ab04
F20110115_AAAKEA broadbent_e_Page_55.tif
5d00ddf55b1c91680a2e1a522e0de3ea
a478edad7d860dff9fdcc1fd35b93f7ee5d5c124
F20110115_AAAKDM broadbent_e_Page_40.tif
1bcb9e1d856dd7b1c3a3bb4de3bf6a4d
2990562d1491dd11e4a181c496a971c8191b4883
200116 F20110115_AAAJYG broadbent_e_Page_87.jpg
e94716b10c905528cc5d6a88e6737299
dee8873a384cb0ae59ba025063a40af7481b4283
F20110115_AAAKCY broadbent_e_Page_26.tif
020f5e2cd46079f07b0c3e13e5250fe5
9352d397245f04e8a4400c9748d17888e9cbd40c
103005 F20110115_AAAJXS broadbent_e_Page_72.jpg
664aa4698e2778abbdf4f16744e3c7a7
bed299b1f79fb98e5e0b5fb21a5e3a2c25f9bad5
F20110115_AAAKEB broadbent_e_Page_56.tif
109bb6d6b822cf081faad9819eb63e4f
122afc97889d325f7f0b9727b55fc2187e7554a5
F20110115_AAAKDN broadbent_e_Page_41.tif
cad76b131af2cffdf0ac787b9e0f1ec0
8216182a96001fdb0de780050cdaec00df0b6ddf
210515 F20110115_AAAJYH broadbent_e_Page_88.jpg
b436a4d20fcb15851bacfce5de1e20fd
bfed8a97dcd6dcaa393c36b7a08e5bc34fedf395
F20110115_AAAKCZ broadbent_e_Page_27.tif
1a537d07baf171ea2661556f90821d8c
0eb2ab5b06f9567e015c28ad8f6b8ac57e677c26
100542 F20110115_AAAJXT broadbent_e_Page_73.jpg
72845f7ad2585fac653c7081e8a917d9
f5cbe258b3d657721eeee139328ffb9cb96f6e67
F20110115_AAAKEC broadbent_e_Page_57.tif
f5eb741b8d4fd0bca66cc2adc9faad93
7983d7510f0f7ada57044652c3a3f177c11051ae
F20110115_AAAKDO broadbent_e_Page_42.tif
905e46e6d8b6a189d3de258e44ec27cb
640d2b36075d4e0fb6861c634a4347188fdc7210
209449 F20110115_AAAJYI broadbent_e_Page_89.jpg
f33d92fc6c210362b70b162a634db5e1
1d0d7bf634eec571f2ddd1bfaab01a038884bc3d
103588 F20110115_AAAJXU broadbent_e_Page_74.jpg
48d279c1525b93bf3e0f2b96cc88912f
ea04cc4a45561aa90c822f1097112eaf7e6aacfe
F20110115_AAAKED broadbent_e_Page_58.tif
867de15aa9cc445414e1e473e182b700
cf16682887293d21e4d9a5b9037d7caf731f4c81
F20110115_AAAKDP broadbent_e_Page_43.tif
2c070105f16a044d22b7a98059fddaa4
db89239894c8d753456a2d4016cf4ed76479215e
209374 F20110115_AAAJYJ broadbent_e_Page_90.jpg
c91bc8f45fda65eba57f1745190b2ca8
d650a96deaffe3ad450187da0669cd70cd17a87c
181782 F20110115_AAAJXV broadbent_e_Page_75.jpg
58468bff72cc3540cbd26deb56908498
8018465e1db8a805a8bf557401323b395db90d74
F20110115_AAAKDQ broadbent_e_Page_44.tif
f28553cf61538fb57fe992037b3dfb02
8d72ff866e6b6d9791bdc2866bf13b9c00b3fb07
204114 F20110115_AAAJYK broadbent_e_Page_91.jpg
1db36cd30e62729d28fc83fd29533938
8adfc206e37e7587028b483eb920e7798e8df0be
213723 F20110115_AAAJXW broadbent_e_Page_77.jpg
ea901491d1039dd7c3a632c3f6fe3dca
ce71d534b63eff94cc4ddb16bcc56d79a7c5f404
F20110115_AAAKEE broadbent_e_Page_59.tif
a5732ae77c592ba505d0c1e4dd82686d
62c884b0fce85895a32835a54caeb4a49eb6913a
F20110115_AAAKDR broadbent_e_Page_45.tif
278449251259c23df06703d47ffd4255
7db9ac988d0f04dd8e92e7297d529122c08b3f17
143296 F20110115_AAAJYL broadbent_e_Page_92.jpg
f94763078c287a81435a7da7b08866a5
1b6d637f6108377d16769a59a1801898f5353fec
231517 F20110115_AAAJXX broadbent_e_Page_78.jpg
8f87c99ed59d005d098f5c9115106129
5bae06e3f4abd2581c4484eef1adfd075150e99a
F20110115_AAAKEF broadbent_e_Page_60.tif
787a09ff74e02a2b78f118ac63e1914a
5941c6f1b46a989af7d28984bcf914d1be2aec37
F20110115_AAAKDS broadbent_e_Page_46.tif
bce739a26b37d68587c6e2e9818250a6
97b5f9aa76d5a3bd9bc8c761db31251836d388c4
180439 F20110115_AAAJYM broadbent_e_Page_93.jpg
8857bfdb65b9a53706ea529cd8c67013
1814072889d24cfe7423ebd87294bdbe2193ac20
F20110115_AAAKEG broadbent_e_Page_61.tif
712c29108802d34d5008a12e35a297e6
ce2b7fe04fc5536a8b8fee63de7c9d0bf0c90384
22457 F20110115_AAAJZA broadbent_e_Page_13.jp2
906ae2fb5900ab7de528269246c8f080
ba6623f0623ff5033ec12b1f061c0083f9e202ee
64343 F20110115_AAAJYN broadbent_e_Page_94.jpg
d7ac08e404d3a89a0ff7956da3438b53
6f3e5e049c01b96c855b36f2c209f5933f20a6d2
200127 F20110115_AAAJXY broadbent_e_Page_79.jpg
96621b353cb851080d6ac9970ef1e587
a49e57954b879ca7b06193ff2d44219b68664018
F20110115_AAAKEH broadbent_e_Page_62.tif
11c71239a7c19bc1a2eb52f57955be4c
a43d61e688a5c3700b00706386c8a722365872e0
98404 F20110115_AAAJZB broadbent_e_Page_14.jp2
cc47e839d337cc7e0678c1346c28ce4c
b530c64bab2925813c2afcd908e7e513b7f599f5
F20110115_AAAKDT broadbent_e_Page_47.tif
e6c3bb3812492fa062465535e8a3bd1d
e0d526579e79f351ffd64a17b9556b9e88840cad
25645 F20110115_AAAJYO broadbent_e_Page_01.jp2
2084c33f0033e8bfddde85461ad97218
e582ce76b0e980e19696ad181e26cfd015aabf8c
219218 F20110115_AAAJXZ broadbent_e_Page_80.jpg
57aea924e3e4fc64b63bbd36b49526d8
fb34edb3d8ebd60f78b978314f39ed06dfa9fbd5
F20110115_AAAKEI broadbent_e_Page_63.tif
5acdfe4e09916906408fcc9023cd714e
8f11c833c3f510552fe29ffb8b2d9215a88ddec7
110415 F20110115_AAAJZC broadbent_e_Page_15.jp2
9bb4da3cf9fd04d2a73fd2265f4ba283
909ff4fc9a21d79b1d47e7acec1aa4c2362ffd08
F20110115_AAAKDU broadbent_e_Page_48.tif
6fe728ff8ec4d89a4d99dbe90306baa5
49eeeee75c72249233035989dad0d846adfb46b5
6009 F20110115_AAAJYP broadbent_e_Page_02.jp2
12c02c6cfd748a0a80179bfefbb7bf56
df2b172a5ce19c8caf70156ef3751fb5f364b8a0
F20110115_AAAKEJ broadbent_e_Page_64.tif
059f76fc8c3885d7d3eb5cc8a88a0869
f040d5d77cafd6f5e9406ab66d29be36036f15e3
116100 F20110115_AAAJZD broadbent_e_Page_16.jp2
c883e77171a524940c4413c1fee3ba59
4897d53c50a4a85171b94e07d36bc6f9374fc5cd
F20110115_AAAKDV broadbent_e_Page_50.tif
450f838a89280a924a7fbe8e585dffc0
2153bd475e1c36745767d952621311694c8838d6
5581 F20110115_AAAJYQ broadbent_e_Page_03.jp2
265520af5291cdca052c5848960651b7
2201b172e2ac1f67a86d7a80a1ba4319a648ddc8
F20110115_AAAKEK broadbent_e_Page_65.tif
8858f42254e83f79cf3e3859ee89f1c0
73ea9ab90d1b0a7acd2a9caa5514fff16ea61adb
101143 F20110115_AAAJZE broadbent_e_Page_17.jp2
84b3cce87d88b4230974c875c3ffe613
a85774d3fb6fad7ea006ee3c618c01826a44e970
F20110115_AAAKDW broadbent_e_Page_51.tif
0919399a8d5cc59337e508e3e62615b9
00fc375459e526e385862d2abb7b69031804db95
92400 F20110115_AAAJYR broadbent_e_Page_04.jp2
e9204a67ea4adfb365da8b9248367a87
4af92d4b5234cbc01ea1973d15af59187190892d
F20110115_AAAKFA broadbent_e_Page_81.tif
dde89699ece12bc483f08acff02ed66b
a57a708c54bc37ad1b81ff874b9a70907f30c251
F20110115_AAAKEL broadbent_e_Page_66.tif
45d114be0ef02ffaff5e664976105e8c
64085ef576df0731ae5bde491b5e6f29e05c9015
106432 F20110115_AAAJZF broadbent_e_Page_18.jp2
6a356dc8f6751102658e021baea9c748
622b445e2f317aa6b26484c73eae3bfc37142180
F20110115_AAAKDX broadbent_e_Page_52.tif
a4e04ccd58c3aa7dc4ea000aa116703d
32144560b03062eacaa54120ccd5176d6d03daa7
79203 F20110115_AAAJYS broadbent_e_Page_05.jp2
d66edd04c9877e20f060d4194c23988a
cc04fa6838bd8aff1ef5061766af78f95f2b0726
F20110115_AAAKFB broadbent_e_Page_82.tif
6b4ae2f6203a26876837cb98d5b50c3e
3bdfd57d90098434a9a50d18bc216753b7b8da93
F20110115_AAAKEM broadbent_e_Page_67.tif
1446049a36dcc1df41604dff82aed27e
c17d7a8470138ba9e410df68f9caa026696ca57f
758675 F20110115_AAAJZG broadbent_e_Page_20.jp2
ab8c46173681dd110fc11f77b81d271d
beed6f15d1461333f6ba010df3ddde991dc78127
F20110115_AAAKDY broadbent_e_Page_53.tif
1210bb32d10b67be4422146db6db6799
edc2c2885c09beb5d7ecba834adb21c85b8105b7
1051986 F20110115_AAAJYT broadbent_e_Page_06.jp2
2a497cba5af44a9d36b9f17775550742
24202ec2cd45d4284749373e50d351e260c18307
F20110115_AAAKFC broadbent_e_Page_83.tif
4fb1a21296d3d02a0b4c9bd17cf3e48d
23e14aaf378b5c84afc610eb6e84f05a35fc32d0
F20110115_AAAKEN broadbent_e_Page_68.tif
1a1eeb0b0a64b78e7d30fcff109087c7
0cf58d0ad08622fbb413bca5bf3dfc42e5c4bf41
108904 F20110115_AAAJZH broadbent_e_Page_21.jp2
cb774dbbced52670ad42b000536d4824
feea10872624332b09dc4c1aac355fe0e2ee1d48
F20110115_AAAKDZ broadbent_e_Page_54.tif
cf8a3960c0b9d03fe302f229890069e6
85b45d06dcde25f0085c7b4cc3bc3c47d6b83d7e
F20110115_AAAJYU broadbent_e_Page_07.jp2
ea39d8f0612681f1229d65428f6b0077
6b64fdf97d496522a09cd97c45fa78b3d7046392
F20110115_AAAKFD broadbent_e_Page_84.tif
d7cd609c2257534a8a85e3adc3bfb7b2
5794c64c6c67229f80fac812985c0322b3ffe017
F20110115_AAAKEO broadbent_e_Page_69.tif
b714d8e5f9820ab6d223158f81f47e29
0b3d61f2919ecda00bc7574b370c4ed2262c195e
730984 F20110115_AAAJZI broadbent_e_Page_22.jp2
ab68cd7bf013d66542d4f08d0051803c
92273d5771680c40257e3c46b59f590418540392
1051951 F20110115_AAAJYV broadbent_e_Page_08.jp2
d841b4507c63ab8f974396c55310a6ea
d145c66e5b97b32a5d0e3f2b20d258e7b46f9d42
F20110115_AAAKFE broadbent_e_Page_85.tif
4325d7bfb8a2c7cebc057f7160806b5f
98c305602f9b0f4bf0360363a0430ce5a97a0be4
F20110115_AAAKEP broadbent_e_Page_70.tif
c5b4aeaae95bf700af1a941e3f0a8ec2
d57957be3c96a033ae8a2c6d52866b0e72d992d6
108067 F20110115_AAAJZJ broadbent_e_Page_23.jp2
73683d711e30ba36025b16826d194f21
1d52b3cf3d841cbcbb4d32f57b362a6e35a1dba9
1051867 F20110115_AAAJYW broadbent_e_Page_09.jp2
3e5d7004b5beff1bf4df14c2d91119ba
d6d1a7338a6b47f520d78a71145348246b25aa55
F20110115_AAAKEQ broadbent_e_Page_71.tif
97bf8fecbb4bb09e51559c5db6dbc332
775338010dd71d221fff11b867b3e6cfdb4d64c3
110582 F20110115_AAAJZK broadbent_e_Page_24.jp2
1d6552571cba047ccadc6e69a1cb9ec9
a0be2d41e5dee672f039a777c5171b337ea47e25
1051953 F20110115_AAAJYX broadbent_e_Page_10.jp2
65591b1764b6c39ed09c8e1533590c52
6b7aea140bc879e80656142be6f398b62b3b62f8
F20110115_AAAKFF broadbent_e_Page_86.tif
2be3aae2b45bb80ca934f4ea32ddeeb3
ba31d26fedd3537b47f7f1ca2b69bf3c56d07a00
F20110115_AAAKER broadbent_e_Page_72.tif
6c8e4c501c81456d96a1e957ac696aff
d002a11bfafc595ec7fee2084a6e8c87cad0a5d5
107745 F20110115_AAAJZL broadbent_e_Page_25.jp2
5915b9507f999b1fad36561e0dfabf24
e7b76e7a00dbcbf66cdb261605a2afc5b7477dd7
356836 F20110115_AAAJYY broadbent_e_Page_11.jp2
82132827d4f899dc5d8efe204fea9998
804e185b43b69d87452417613084aa28a902ea36
F20110115_AAAKFG broadbent_e_Page_87.tif
2c9c375e4dfcc3301dfccf90998284ce
8c794a930f18102fe768250241ccfd3a7ef8a046
F20110115_AAAKES broadbent_e_Page_73.tif
9a56818b16403e4adb1b7bbcb4bb9f4e
e83072c8b43e8780f5f5cd6d4f2aa095569edf2e
111998 F20110115_AAAJZM broadbent_e_Page_26.jp2
822d2121bc234843957f44cf226f3563
e2feb1ed8bdaebc3be5c79d7e48d3ce60d81697d
F20110115_AAAKFH broadbent_e_Page_88.tif
119dd916142d01600510c773c808642e
21b45788071158d49035abf6c3a9b816f0e54502
F20110115_AAAKET broadbent_e_Page_74.tif
a3d2bdadef70d59f9c064ed9b9acc684
bea3914549ca9279613507918b50e5ea9dce54f2
109986 F20110115_AAAJZN broadbent_e_Page_27.jp2
a6d8a18bde1bf23e23079fd847213d5c
0cc412c86f07dd53fa460ad558d1da3f12927fff
87460 F20110115_AAAJYZ broadbent_e_Page_12.jp2
fa04f111299dfff2cb5975ea4a872db3
b9db34cface79bbfbeed125a9ab99de72cde8a0c
F20110115_AAAKFI broadbent_e_Page_89.tif
bc6eb68bca32c87cb92c47fd0f9c6156
9da2817413d7be0f226ac0905e03444cea9cd3b6
F20110115_AAAKEU broadbent_e_Page_75.tif
32ef202c709c7b6d97ffbfdc8f506272
89bb428e50dc11b9c721bc21fc1fbadfd472fe7f
106840 F20110115_AAAJZO broadbent_e_Page_28.jp2
7672124b96a6cc32ce728192d044b7c5
5c07a39f23516aed44cc83bea0fa08dce4f66997
F20110115_AAAKFJ broadbent_e_Page_90.tif
bbd2d168f5e0f1f1acac76f40afcfb43
0c9fd3c8dd037fae3997416d6b9a952a754e74e4
F20110115_AAAKEV broadbent_e_Page_76.tif
403c2f77786c47f29f8043bba935e03f
7f94d40d0ae2b2d176f664995fddb75f96ef3e50
112696 F20110115_AAAJZP broadbent_e_Page_29.jp2
b3c331e93d0dfcc557985ece0a6c0283
bbf7de40922ea2ba0df52e8acce53c7c7e589d19
F20110115_AAAKFK broadbent_e_Page_91.tif
d83c8f4bbd95e44da1451abc987af2ca
25cd80f8695d518964978513c36bc43ccb422b39
F20110115_AAAKEW broadbent_e_Page_77.tif
0fba58b1368287167174b50d01109da0
7f1853510d91a0813dee10f4bf4d7ee711d29dd2
111762 F20110115_AAAJZQ broadbent_e_Page_30.jp2
16346a02b8e5642ff91c385fd3e7788e
cd3bc41c41af9e4ee2a03123679e5f1ea59109e7
F20110115_AAAKFL broadbent_e_Page_92.tif
0e3ef4e3f88bd7e9bdab17bed505b9d7
1a87d9a1fc3952b67661358a6dec5535a97ef8d5
F20110115_AAAKEX broadbent_e_Page_78.tif
eaaa67c70a28d0414507d9810c68f3d4
1ba201ca8b722ce84c35e5eed4fdad59ad969777
114385 F20110115_AAAJZR broadbent_e_Page_31.jp2
66eec8c8dfe5a3a67c519d2623b92442
00398b8f4e147d85f985e72ef448c2ff6cf93eac
44814 F20110115_AAAKGA broadbent_e_Page_14.pro
0b96f791a60655dbb38448b61666264a
92b4d4bc23e30bb8a2026196f3a5fd9ca25a0db6
F20110115_AAAKFM broadbent_e_Page_93.tif
bd6c7dcfa95128e078951f453534b745
f5b835897318c978e6e8c4937c649f663c5630b4
F20110115_AAAKEY broadbent_e_Page_79.tif
12647d273b3c2969d59a2144a13f3c15
89c61f20aff67118e8d709da19fa72c1ff3d08f4
109589 F20110115_AAAJZS broadbent_e_Page_32.jp2
2604ecd818ea35b6a0f59dfbe0d74cf9
3976fde23bdc71dc2d03a0ee99e6dcf89c8aa42c
48868 F20110115_AAAKGB broadbent_e_Page_15.pro
a6dcd4ed3bd2457bbf83760cfa77b225
9fe00ea63367ddf21cb286f221ed5eb67f960264
F20110115_AAAKFN broadbent_e_Page_94.tif
15ec9fd92b857208a62955eb42b48702
64fc9f5d580cdea50d76468fc770e2cad4cb1e9e
F20110115_AAAKEZ broadbent_e_Page_80.tif
3472a84a8151a825faa44e99e3c2dc86
a9271eb084a18845d3b63afd9bc5d7779fe6b2f8
89674 F20110115_AAAJZT broadbent_e_Page_33.jp2
f1a866324a2954424f4edef4ec4c763e
95215a940bea85fa1a757ccbb719b80dbde6d38e
52122 F20110115_AAAKGC broadbent_e_Page_16.pro
1ed1b2f7dc8ec7eb8ae1bb75ce07199b
3ac809ae7c14eb06cabf6224b7aeaebeece16612
1306 F20110115_AAAKFO broadbent_e_Page_02.pro
dd68e9c07d57e540a5912ffaef02e11c
aed764490528befc846ac4cab663698b1f6d98aa
94760 F20110115_AAAJZU broadbent_e_Page_34.jp2
fad9ee434e24623fa720c51de4f2f68e
2ead53e8c5327dd16f5f18228aec897172eb1614
45746 F20110115_AAAKGD broadbent_e_Page_17.pro
76be708f02affc7f6ec0573950181ae1
e26bcb8b5fb98c7277427c08ca99e66761f40ad4
1198 F20110115_AAAKFP broadbent_e_Page_03.pro
1b7f9f445ef10e4204d0fb6ce2899f05
70f69ba8852743095ec2a6845f74561465253dc1
106607 F20110115_AAAJZV broadbent_e_Page_35.jp2
bbb32c55d79bf8db7e7bd49e1bc505d2
112ee778ead1d93dd2a661d99aac84fc78ad97b5
48115 F20110115_AAAKGE broadbent_e_Page_18.pro
83c65b0a58f1bab1a75064b28ff0f402
ce92d64742e61001af5082652872483693275077
41921 F20110115_AAAKFQ broadbent_e_Page_04.pro
310b35d540691cff6412c39f682c8402
2e08d281244344ff06f71e490dc5c98f92b72fcd
81091 F20110115_AAAJZW broadbent_e_Page_36.jp2
f43f69db28c1fe49bc0571d93f1f387c
531ec9cd5c63f450e36e03e5c74a9c4cedd1259e
38059 F20110115_AAAKGF broadbent_e_Page_19.pro
a4b64793675165d4db80acb677ca281c
bbfd2908d1922613f9946017a9694999c8fe11e2
35176 F20110115_AAAKFR broadbent_e_Page_05.pro
8f99503c17480c4b5a9ff095244e7bb4
6a87bdd3614828920253621250153571be50fd7e
F20110115_AAAJZX broadbent_e_Page_37.jp2
ae9f9cf213df67bb51f974a6af035656
4d5c5a03457c36c0a789e809730b1035016e0afd
57934 F20110115_AAAKFS broadbent_e_Page_06.pro
ebdad9a8a06cd0ad7781dcda7e073044
8c654909fce2238d88649140213753e3a0958082
101529 F20110115_AAAJZY broadbent_e_Page_38.jp2
d8b16a6170b58f80170411999ef9fb3e
276db91ad6ca64d7dec8159e94209501138ebe28
38658 F20110115_AAAKGG broadbent_e_Page_20.pro
b1f8974d902e3d72da0ce0af7ca14e2a
cc51b6b0c7e52db1e46fd676cfa8eb44cde18c71
48472 F20110115_AAAKFT broadbent_e_Page_07.pro
b73844c9aebd004426ab655bfbd1e725
f75976fcefd13a83f52b49c2ada29c714a4fae01
91059 F20110115_AAAJZZ broadbent_e_Page_39.jp2
850b491aaf5dad9a1c266b8438cd1752
f62ca50dabb42eeefc81a0669c74cc3840dfe0da
49612 F20110115_AAAKGH broadbent_e_Page_21.pro
d4a94011c2560860ee1064d53528cce2
0849a8188299d03c3c184a807dc90e4c3e83bbbc
61048 F20110115_AAAKFU broadbent_e_Page_08.pro
51f1f37fcdc5e39bd680c9c983abe926
372e6ff5b1ab4dfe12d0772f511175bdd5698044
32391 F20110115_AAAKGI broadbent_e_Page_22.pro
2c6e7b5cb0d5e84521e0b00904dbcb2d
8e6103e057e53466c82c6c7b2bb43946b4805193
31917 F20110115_AAAKFV broadbent_e_Page_09.pro
3b03c7ca0970d7b7ed99058a1b49a670
68fb8ac01caa97c98f5daccab9bf843913062ecb
48202 F20110115_AAAKGJ broadbent_e_Page_23.pro
5d29b83fd321a5777cb4c409aa9a4d37
4dc54e499497cefbbec30968567fca975e3715f4
60165 F20110115_AAAKFW broadbent_e_Page_10.pro
123589dcd91ac749917aec59e9008dce
14d7d5d40aecbfe85172b7fabd2bfc884a98601a
50136 F20110115_AAAKGK broadbent_e_Page_24.pro
7b20ab4ba3d8d841c84fb37db4733715
69b4fa389b2657210fc68af5cd5913ad43fa5076
8523 F20110115_AAAKFX broadbent_e_Page_11.pro
edfee7a3e5bd1f5bb473798d8182c962
ee94aea475b5a8c0716b12caf178e7d065861dcf
52853 F20110115_AAAKHA broadbent_e_Page_41.pro
0456f232eb149414f001fc04b14e6bc0
b27d5e34482f2d108424a784d94938037f5619bf
48300 F20110115_AAAKGL broadbent_e_Page_25.pro
fad7b2f8ea25f5c34cfaefe98ea0bb02
cd5823d4380a246173339d2944c71f37e9b1bdd4
39038 F20110115_AAAKFY broadbent_e_Page_12.pro
9a284c059d5d822fe29147a59de6675d
e53b3376f22046c8fb962ee2cf81455d3bc786bb
49465 F20110115_AAAKHB broadbent_e_Page_42.pro
2d1ccc56bfc12e08ea90c70f98623572
8d7fe65e3e84dfee7e5a0f4333f89c7f3bc21adf
50040 F20110115_AAAKGM broadbent_e_Page_27.pro
05a2fc5f1e7ccbc339eb8bc490592f41
95a8f67812cc6eaca9719f527a18bd418ef589f0
8705 F20110115_AAAKFZ broadbent_e_Page_13.pro
85f68118b4e4140a1238208de42e781d
037522d2c8ebebb1e152445b3a28bd02320986ce
51857 F20110115_AAAKHC broadbent_e_Page_43.pro
ea6869586d888c11e3d21e31e9375b7d
2d32a831cf4ba30c189e7b3383bba89548d93935
50074 F20110115_AAAKGN broadbent_e_Page_28.pro
9757765f1257fba35f965abf2c7de2e9
bd92be378afd662d02a4884b89e139babbb89d3e
49569 F20110115_AAAKHD broadbent_e_Page_44.pro
d1e6a6a34d950caf54b550c7780d33ad
7796eff8b5ef562112dbd6c1ba7d3ea3290e4a03
50667 F20110115_AAAKGO broadbent_e_Page_29.pro
b1b1e70fa03d6356bd6cb5802ade9346
87d725c0cfb9c2adf2ea16f108f39c75b6462894
46002 F20110115_AAAKHE broadbent_e_Page_45.pro
fa2542759f8d34cd5fa41be4205e8441
3638aae43df4abcb1d6799a5ffbf8faecf8c118b
50131 F20110115_AAAKGP broadbent_e_Page_30.pro
4cc9034cbc69e896ae81a1493601db18
b662641dd6ce89f1798105ab7232f629166048a8
49811 F20110115_AAAKHF broadbent_e_Page_46.pro
8c610530267c9a334214d07ff8eb00be
3a96e08af6d420200888c4326c7f532eace563c0
51181 F20110115_AAAKGQ broadbent_e_Page_31.pro
8f56fa2446e83668efdf10ea029b37dd
085738adf79b2df82740b3f240fab871b11ea095
46860 F20110115_AAAKHG broadbent_e_Page_47.pro
db3bd11bbc1df339858b150d4e59613a
d494a302d0fb28f73cc685bba163601bdc4b1a56
50488 F20110115_AAAKGR broadbent_e_Page_32.pro
851d4932a279a6d598d254d47ddbb35d
49234d1c808af50128fd3cbf8f87734a494d2541
39585 F20110115_AAAKGS broadbent_e_Page_33.pro
9e5e1c5b44fdfcd366000dbf6b2b45fc
f95f02cb2c4f5c08a1945326da280d4556aed6f0
17572 F20110115_AAAKHH broadbent_e_Page_48.pro
7aac7fc6970c9380528b962e2ae5d635
2a806bbd5b0a2e1cdf2d320cd2eeced086113355
42372 F20110115_AAAKGT broadbent_e_Page_34.pro
138b0f46e3fc0f2d1ad1c179243708ff
0ded9c9493989a7888ef0dadcc67cf7542240c3f
34367 F20110115_AAAKHI broadbent_e_Page_49.pro
cf48adca6e21e9e9e91066a8f4a8039e
a032bbc514b0a195c4c9b01f04bb25aa5f25def8
47520 F20110115_AAAKGU broadbent_e_Page_35.pro
483de52659db29abd94cee1512e1fe5a
ad665a90f2b0e1529741efc229124af235e2aa0a
53172 F20110115_AAAKHJ broadbent_e_Page_50.pro
e34f07308481bd1d01330aa855cfc687
4803c8f2c57216580f7f0961de39cd7c13bd4342
35398 F20110115_AAAKGV broadbent_e_Page_36.pro
5b51c1d7d2aab5aa7dba476f02d15a22
8452b7abfbc28bf536e198e7c9bc2194f7ab1b47
62357 F20110115_AAAKHK broadbent_e_Page_51.pro
5a936909c2c734e87ce2a7b37764fae7
94100aed4418be08e36876ffce9d17e411b0bfd1
26409 F20110115_AAAKGW broadbent_e_Page_37.pro
b83b66997fcb5345a2abaa256cac624c
2d65db488fe69ac556b3e31022a2cbf2a7b8eac8
14709 F20110115_AAAKIA broadbent_e_Page_67.pro
a8b95b9aa8637f377038bbc66f52b153
18e04a94305b31bdc7f0ab97bbd44e4c2afaaae1
39971 F20110115_AAAKHL broadbent_e_Page_52.pro
cb72354d419b0381cb812802173ad01a
f45885aba1dd8816853de271a23117c5b68e6617
46437 F20110115_AAAKGX broadbent_e_Page_38.pro
dbe9a0843f82b1880a5f80d49f085250
eb8cdc65dcdd777a1d9472b93b67b6bc9b3b5e7d
21916 F20110115_AAAKIB broadbent_e_Page_68.pro
a0d11559b37a9b28710583577861fcce
58bb2216ecbceea426dd61cb346b9bb7f017a130
39741 F20110115_AAAKHM broadbent_e_Page_53.pro
9743c6e52344fc3e05f155fbe921e4e6
42c19239f3be056bad9ee3a0cfc450da3a0c3fbd
45961 F20110115_AAAKGY broadbent_e_Page_39.pro
11b058355733b3590e157bbf5b373ab3
451b1ed1461e94b7d64c1055ce31da5ff08d296f
11550 F20110115_AAAKIC broadbent_e_Page_69.pro
f513f62aa6bb88972429de0776467b3a
8350e2a0053663de22fd15ca6b4470cec5a7c3c5
28103 F20110115_AAAKHN broadbent_e_Page_54.pro
3591b4f2355a66cc07846992b8c311a9
110bd6d1b32e11ecaa39b182d2b3b5f6cc2cecee
51133 F20110115_AAAKGZ broadbent_e_Page_40.pro
0bb89987e6e38ceeb5307871db1664c9
22a961073f8ac5bf6251a1550d9bbbb856397493
8170 F20110115_AAAKID broadbent_e_Page_70.pro
5e69138064f513207b788d09b28a65cb
8c1d62783f47da30f7d12fd93767f36dbd26e9f1
32278 F20110115_AAAKHO broadbent_e_Page_55.pro
3dc384be4da6cf954e66f3ef88e1b18b
01112bbd8e2e72c9ed4d868a6cd4548767daf665
19295 F20110115_AAAKIE broadbent_e_Page_71.pro
765d9e6477c263d584e83939a406f968
24325c77666a1466b555fb1441927a3a098754fd
47971 F20110115_AAAKHP broadbent_e_Page_56.pro
7b0c6887df454d62f9d0b88ee7351eca
21d44ed944a174aed3854d7c439f714313a06ec3
25558 F20110115_AAAKIF broadbent_e_Page_72.pro
30d02af871ca7402bfae6d157efe16f1
4b8860d53e1f217b47aba111d6b7efdd29eb6dc3
12027 F20110115_AAAKHQ broadbent_e_Page_57.pro
b887a803aabbe4d03ebc0ef4cb249ca7
9a847f364add0223d6238e8b5e3f245ab81932a6
32689 F20110115_AAAKIG broadbent_e_Page_73.pro
d275f3b673c76ae1ee14a62e35964995
61f26ec554aac068398ec749b52e4ffb1c228bdd
50523 F20110115_AAAKHR broadbent_e_Page_58.pro
a02d5657e8101b6b167edf5ebd1fd6c8
d3f4f98781f25a18f29591664de193ae5e979998
33064 F20110115_AAAKIH broadbent_e_Page_74.pro
8db3d7c09bbb58e9a30b68cd0ee336f4
7c8a8a510d6f3592b395efc9290a230bf9555283
38109 F20110115_AAAKHS broadbent_e_Page_59.pro
d1354b99fedf63601376ccb966d6c567
591150050c0ccc1f2354cb2adc356fb12ac44124
27756 F20110115_AAAKHT broadbent_e_Page_60.pro
c80e6cd4d59111a5d8c5ba0cdfdf81e9
b2ebd594f700d9cd209fe06ebe45b51f0b54d827
44224 F20110115_AAAKII broadbent_e_Page_75.pro
41e16b71357a42a1e3b28778d90dc788
bb5c6a11d2ae9018e2ed622cbdf4ec795951205f
37654 F20110115_AAAKHU broadbent_e_Page_61.pro
b81178f90af9de5104da4cff46e26dfb
2673a306ccb0c5ed829fd62dd670cabdc8ff6037
51600 F20110115_AAAKIJ broadbent_e_Page_76.pro
8d3561a512e8b9efe287c134f57a245a
9d30f0e07d9c5ca84a38b7c2ce80e23e635cdc96
53829 F20110115_AAAKHV broadbent_e_Page_62.pro
d99145692e271749aa4e0d60e3c38b73
b55a15b0f10fd923c6517d39eeb4adbc45ae5d48
49918 F20110115_AAAKIK broadbent_e_Page_77.pro
a5387fc83adf63f5dd1e52e8ed41c680
de176752fbf1d865427003dddeb0ad5093aca7b9
50002 F20110115_AAAKHW broadbent_e_Page_63.pro
0cf0d8db53c8b6889cd85b546b581e00
210abb21d66e481bfd70a81e691aa242c19185bd
54630 F20110115_AAAKIL broadbent_e_Page_78.pro
2fcfd8b0ba4d5e421b0e22ef5e90c93e
6b4c1c09c4eef45ee58f26318b64218341059f64
50596 F20110115_AAAKHX broadbent_e_Page_64.pro
f4df3d71dbd7c5d14450e824e4d07908
79d4c7dea7ebf65ab2a59f3384a2b3d58190e6bf
472 F20110115_AAAKJA broadbent_e_Page_01.txt
1b7f9e693ecabcea75dd61a2f44255df
bbfbb696de39ff3c2fb580797460f65c21a28ea6
47191 F20110115_AAAKIM broadbent_e_Page_79.pro
a3577a41f6a4fac69fcc9ad2783da0bb
a0dfa559be8934b137990101574d99d03c5c071c
45509 F20110115_AAAKHY broadbent_e_Page_65.pro
874667f80bd479110bb39dee87c855e6
bff8eec7d958d857e6c49516cb0b0e5b300d11c3
120 F20110115_AAAKJB broadbent_e_Page_02.txt
25db972b5e4b82701ba703bc0fca658f
79d250d091a8365b11e6d90363f836f36a3edb69
51561 F20110115_AAAKIN broadbent_e_Page_80.pro
cc5805e320c424a7c54017ea3adacf05
912fcdeb0b9dcf1a3237a74043bfa2ea4e5bc2c0
12887 F20110115_AAAKHZ broadbent_e_Page_66.pro
33ae3fb003140c8a6f4457c236d993b0
49629c11d17a9ef9e51539f3ea8b93fe1d2f333f
100 F20110115_AAAKJC broadbent_e_Page_03.txt
dd872ddf4691685968edd26c76a95e49
c76882b4a223305463d71093b72049499637a198
47519 F20110115_AAAKIO broadbent_e_Page_81.pro
c6725217da3065c3b18eab7822903c1e
bbca9bf6cfbe28926048a4d8204aff4facaf3a0a
1700 F20110115_AAAKJD broadbent_e_Page_04.txt
4ea46f2b11e6226879ddd84cd40f2987
c41e0d35ccf6954a19049ed9a673f37454322907
49218 F20110115_AAAKIP broadbent_e_Page_82.pro
85a05b8655553dc1ed6fc1571f7f90ff
5582c589e1eb14fbfdf04437aaeba1fe61365e8d
1422 F20110115_AAAKJE broadbent_e_Page_05.txt
e0b37a3162efe285ad84b8a336ede56f
0cec2c42f96c3022efbea0802d16d4150f2ff035
2459 F20110115_AAAKJF broadbent_e_Page_06.txt
2d8ac1ace48b257c7ef3856c3b7315f1
475ae8e840bd2468938707f6df55033e9f78c32b
47666 F20110115_AAAKIQ broadbent_e_Page_83.pro
e6e3b08d7d7ed7a5cddde9b6c4bf7578
4927965fd619f40991bdd1f139aa919ef773ad93
1983 F20110115_AAAKJG broadbent_e_Page_07.txt
e4e0a7d1ff0e7742d57cf695387ed5f9
8ac9c755328e3161b1de319f6da139bc23257c83
50738 F20110115_AAAKIR broadbent_e_Page_84.pro
02e71326df5f5e92b1b9752fc554ad73
cd0752fe7762b05789095659e1b2cf68cce534e0
2486 F20110115_AAAKJH broadbent_e_Page_08.txt
7d8d032357515ed20e2872238c7f2001
e73cd0115888fb51d0f73177141b6d9b8a1a7f57
51815 F20110115_AAAKIS broadbent_e_Page_85.pro
7a745a727cf1a0121471308d3daf80d6
a6885646cab9a39dce5db318a0694038e7ff691f
1301 F20110115_AAAKJI broadbent_e_Page_09.txt
4c4b4078d1d8dbda2a026c406445ccc7
391a421f62bfcdb01f53837458f159e3097376e7
48152 F20110115_AAAKIT broadbent_e_Page_86.pro
4f1e2a3708b22e3323ad4cabe62d4754
abe7a2b5b890b1a6adb2e35063a49c8ad6b676c3
49291 F20110115_AAAKIU broadbent_e_Page_87.pro
f9107d4d0b91597357ea6437835d1945
f23bf48da58ecbdbc56681688eb58b36ecc1ae1a
2445 F20110115_AAAKJJ broadbent_e_Page_10.txt
369824a126e1e6db6b5f62651ea9a596
57442b8f3cec229f5651f3e33673e2d7911cb161
50266 F20110115_AAAKIV broadbent_e_Page_88.pro
d08dd725dcdb91ab6fb06c1aa01b7c21
a65724b0dbb83570868b16241783bd32a0e836f7
349 F20110115_AAAKJK broadbent_e_Page_11.txt
196ae8b97b2afc2b8c9fe078d7df91cd
81e84f5357973d4a709634d8fec243320aa58921
49655 F20110115_AAAKIW broadbent_e_Page_89.pro
69d4942df56f809fd2748acf040a42d5
bf1e71fad6a163726d448524172e776899af9418
1976 F20110115_AAAKKA broadbent_e_Page_28.txt
75e0bf04063b146e1b819f3952d82d08
ca313fb54965dc5fe0192c5cc18ee68373167763
1710 F20110115_AAAKJL broadbent_e_Page_12.txt
58e52f4b8b549265285c942b3cb6fa28
a811f0ca704edbc5eabd916739cbb50fca46190f
F20110115_AAAKIX broadbent_e_Page_90.pro
42872de52cb132becae1cb653f719ef8
26ce1b4046bd713c29efc4bf07f5d139f06a0dd2
1993 F20110115_AAAKKB broadbent_e_Page_29.txt
156adfc9b4f8d04a16972b0833ad4adc
765849f1faa73fa6022c01476ef20ff5aec452eb
348 F20110115_AAAKJM broadbent_e_Page_13.txt
d6a35f345e2166163c76c98d60fe94d0
61245dc5d8120e237b8a0658f3eea29b8b56357d
41894 F20110115_AAAKIY broadbent_e_Page_93.pro
7947cbd0b0b79d38c994ab72618bfbcd
2b91dadc44ef6d8a5696552e78444f069f604e8c
1985 F20110115_AAAKKC broadbent_e_Page_30.txt
68c544ca3dd4b2dd0fcf73ba5557c18f
aaf11267cd987f31ead7f30d1fd8b30ed7023b31
1844 F20110115_AAAKJN broadbent_e_Page_14.txt
3e918dcd8f11c56625e48f6ef4ec157b
04e7bdea20957547dd653ac8ca38ed6d3e625636
13761 F20110115_AAAKIZ broadbent_e_Page_94.pro
09b78439096c6e2c8f1fa65c8502ab69
ba4b9807112fcbd53e6881bf0f836d5b90aa2f68
2001 F20110115_AAAKKD broadbent_e_Page_32.txt
e8b46a385cd0f296d2be40aacaf2857b
c2ccc0a26afb36263722871de8591bab8b4a71e7
1927 F20110115_AAAKJO broadbent_e_Page_15.txt
2d4341c34b45bad61b5ca3e100a7c647
5707b4cff37f337f06209361b12c414b6acd5bb7
1589 F20110115_AAAKKE broadbent_e_Page_33.txt
2c4a2a18727e750bab4ce182c89cde33
ce2986da5c0107912d27e2b93694e97cc7dfd25a
2061 F20110115_AAAKJP broadbent_e_Page_16.txt
f652966e1e175158ba7539f0948903f0
080a77a5b351ddd3ea65e78a9ce583d8029933ea
1757 F20110115_AAAKKF broadbent_e_Page_34.txt
8de2567c04db779958858aefb7c5ecc1
4792660f61e1388c2bdce295e2433bf6d9455f40
1820 F20110115_AAAKJQ broadbent_e_Page_17.txt
137da58de7494c393b9204e6ebd9e89a
453dda65a9a662945d1e0507abc9917a5b547952
1883 F20110115_AAAKKG broadbent_e_Page_35.txt
98b768eee90c51b5371b93dab9f33819
8cd8b4ec422eeb6abf034d4dd880abca4dbc2c01
1901 F20110115_AAAKJR broadbent_e_Page_18.txt
90534fc9916a4be9e2fc76d1c8c94340
112f5cff8e36c46630475f13eb9f3f86efdf8c28
1446 F20110115_AAAKKH broadbent_e_Page_36.txt
dc4663380263552444073f043cba3a87
0460a1fcb285496cb3423273472dd6b78d35c24e
1937 F20110115_AAAKJS broadbent_e_Page_19.txt
e8f8d1ac1f3576dde0dd9dbd7fb708cb
c593e9d46ccf4000f2aa129c35f9370e94b74c09
1108 F20110115_AAAKKI broadbent_e_Page_37.txt
965ae2d30fb3b876a614cd7b3efaa3e9
4e4d97dd875425ff0226e826dc7155b0bf86937d
1939 F20110115_AAAKJT broadbent_e_Page_20.txt
4c92977f5dbdf069df45d89f99dae73a
b6108bacd84cfaa69fbeedbd5acd4b0a48d10543
F20110115_AAAKKJ broadbent_e_Page_38.txt
afa3e7a9105a009b15bda111d24372be
b2d83482df4f64d825b6c086c2f39f3b5e6d6127
1953 F20110115_AAAKJU broadbent_e_Page_21.txt
e1e3ae61bd558a35ee7f49ed41283938
240862eb73c4b84ad2f771d9bdc911d9bbf151c7
1596 F20110115_AAAKJV broadbent_e_Page_22.txt
93632bebd243effaeb79af1a7107d2ac
fb39a89c0412bd96490c7d21348aea7bee36451d
2017 F20110115_AAAKKK broadbent_e_Page_40.txt
d17a300933e4901a336b69482822bbde
3917a69bd4107342d0cd183ab6569ea1389c3416
1900 F20110115_AAAKJW broadbent_e_Page_23.txt
5848e2bd560684c6d024b847b6a200ef
b6f8facbc19068599160abcd5a465112a92042b3
2080 F20110115_AAAKKL broadbent_e_Page_41.txt
9ddb3a9f8a9157f4d93bcbb495818bc3
5abe768af8ebfe44b50a04fb93c8d8717d4242d9
1977 F20110115_AAAKJX broadbent_e_Page_24.txt
fe7673da1507097b0154932f79da9243
6c1d8d75623e4fbc7102c54bb17aeaba8e01d426
637 F20110115_AAAKLA broadbent_e_Page_57.txt
203f5dc58c81e81db92b417fce9d973c
998ab6076100d4217cb5eef56173649ea445d038
1959 F20110115_AAAKKM broadbent_e_Page_42.txt
e8db2c603d150544d5ab109c4a5d0e72
5dd1dba2e66a79e54283cb12f955e535c26d5a9a
2020 F20110115_AAAKJY broadbent_e_Page_26.txt
fde1bcfd20f7fd55608a1f0b4767f7f6
7f9ec85fa4a837d245fc2e6cb93658c1b23ac237
F20110115_AAAKLB broadbent_e_Page_58.txt
f51e9e7c5337172a76be3eb54d01d6ce
7f316288f97213c19be96e9a21a37f7d8e818dfc
2031 F20110115_AAAKKN broadbent_e_Page_43.txt
1538527f9a5b2e540061eedfefa7f113
86ff035a9b5e486b11b646eabaae5d6de74854dd
F20110115_AAAKJZ broadbent_e_Page_27.txt
84604193dbd95625303c389afe287d8d
27d61c4ea3d43004e56f9f462c3490741fb2df78
1899 F20110115_AAAKLC broadbent_e_Page_59.txt
bc82a13c9cccace6c237b543fd7b5467
6224c9a43e3f4c9099dfeadd5dcc3a8cd1542d56
1955 F20110115_AAAKKO broadbent_e_Page_44.txt
dc3b98fd3b91d32372bac57c7ea1b7db
220fc10169f8b876a85040dbb304a28a84c1e314
1417 F20110115_AAAKLD broadbent_e_Page_60.txt
89159b5038316b543b27e44b2920c9d8
627981a6fa230d10e56c1dc38b3faba03d555c23
1823 F20110115_AAAKKP broadbent_e_Page_45.txt
38da418752e8c697ef0263eddf88d597
3bbbe324c546ddf91695af20258e4953b54d1042
1809 F20110115_AAAKLE broadbent_e_Page_61.txt
296fed5d351f896bc8275e77774e353b
a58d8ab19bddd2ff6f3e0e8e64c9e8580396b7a3
1961 F20110115_AAAKKQ broadbent_e_Page_46.txt
4e0baacf980bbf361013af259f2bb69e
924c5ec436529bde3c6e08812f97d6f78440e016
2461 F20110115_AAAKLF broadbent_e_Page_62.txt
b556c6b2d8579cb9141b793b6b544ce2
bd0397e9ba9dd78078247455e2b3c980168cb9a2
2056 F20110115_AAAKKR broadbent_e_Page_47.txt
23f826e26e47691641b6c0c54404e2df
e5f037420191ef6df00759aa3f89dbefe01ef24c
2292 F20110115_AAAKLG broadbent_e_Page_63.txt
cff93e9d80fb60e21a11395f5b905656
1d44598a5c026ab904d2af04401fc5e96f5b2dfd
815 F20110115_AAAKKS broadbent_e_Page_48.txt
9050299161f07a09882f64749f5c4e38
009db73be496ae99ffc984427fe2a65b49585f0a
1991 F20110115_AAAKLH broadbent_e_Page_64.txt
ccde189e92b480724ef01c318aa11f78
48ad2b6e6756d83cab2d98d4a82fc54f5956641f
1566 F20110115_AAAKKT broadbent_e_Page_49.txt
0abce02cc2624d904b50962ab3f78e7b
59ad3e95c0b9547e082193b1d1b0d3a116ddc9c9
1808 F20110115_AAAKLI broadbent_e_Page_65.txt
78cbf63e1546cee448bfaa4928f50012
6bed072d2d0061a6067f2567a18751a9b3f6078e
2368 F20110115_AAAKKU broadbent_e_Page_50.txt
07d3875533f180ef1444e12faab94d5e
fc34c0b13f88c34a769bae4db6bba03e25f417ff
950 F20110115_AAAKLJ broadbent_e_Page_66.txt
7bc402c7ecdc4c28d9ad6abb1034aae6
b54d17d2e883fea3fb2a4758d96d0cf5a9b65020
2569 F20110115_AAAKKV broadbent_e_Page_51.txt
f708f7ec4032284eb4d1cc2a3507b92e
dad11d13cb6f3aeb7a0692ec27e04295dcc4b5a3
1084 F20110115_AAAKLK broadbent_e_Page_67.txt
49c865b35ea353eaafcf53e0bde7e4ae
689cc190ab8931c7901fb65e30e0c1b0118199b3
1787 F20110115_AAAKKW broadbent_e_Page_52.txt
7e42be47c30ed926a10f948f0cfc2a25
dc13ea491ab8fdea5479c05bfde6769d097be4f5
1591 F20110115_AAAKKX broadbent_e_Page_53.txt
6ba581fc16125d58b51b41461a1c1a5b
755f795de75c80ce396ec0b56cf36dee557273c7
1980 F20110115_AAAKMA broadbent_e_Page_83.txt
2b5d3f6f55e3acdd18cc2beb02583144
1ad75f544f25807e8a7427cbaf1b9db9adad7f97
1062 F20110115_AAAKLL broadbent_e_Page_68.txt
4bfd6cf7cd875ad2d08e47b41382e035
f58e5dcbbdf8741421c95f050db2668dc6ca7f0e
1599 F20110115_AAAKKY broadbent_e_Page_54.txt
467c9b454195286035bd6e4bfee57a37
17dbd3ae9875793af141fb0fb65aae15c423af07
2113 F20110115_AAAKMB broadbent_e_Page_84.txt
3fd5e161fb7695aad239db44578130b4
f4963eb3346738a0dba581854688e93f6c4392b4
636 F20110115_AAAKLM broadbent_e_Page_69.txt
45cf0d3e1b1ca86f30a3ec18c2519831
378e73d9385a53e750d2863c579f56b7a98b3d88
2062 F20110115_AAAKKZ broadbent_e_Page_56.txt
78b078b8e85f7b15a7d9d00ec5a021ee
7f8284d547e1a6fc5a8962cfa6e735f1a3dcc3f3
2154 F20110115_AAAKMC broadbent_e_Page_85.txt
1a5c5304a8e7a601c29adc843f860ea4
a06bc2d8e74cb4d47cc9f246b9f1f066fedc971f
434 F20110115_AAAKLN broadbent_e_Page_70.txt
6959b80981d9e45e0bb9085ebb0d20d0
c28fd2520f8bcde2cdfe7145263f880e314e3149
2015 F20110115_AAAKMD broadbent_e_Page_86.txt
cd3591dbf8c03d6442f906e855abd450
f1c80c0dc2318bd9c2b03a06e2d202c550b8741b
937 F20110115_AAAKLO broadbent_e_Page_71.txt
6752a67d44adf5dfa49b318c776759cf
172d77992a2050900a02e19adf003a19f217276e
2050 F20110115_AAAKME broadbent_e_Page_87.txt
1fbfc74179dc7999b713c59888387e65
5ca0c3ad26e370988a98fadd52bc76d0fb4979cb
1369 F20110115_AAAKLP broadbent_e_Page_72.txt
c6757f5e8ec22ebd8d171637842cfc3b
3eebfb4ce64dd71a625cde1d8f1aaddfac378fee
2091 F20110115_AAAKMF broadbent_e_Page_88.txt
73a91a3763c173a3cf8bb972150d4da8
e991260d0f04e4ec826654999626c472f5f130e9
2011 F20110115_AAAKLQ broadbent_e_Page_73.txt
fb0363cbe81423bc8b312b08277bbd7a
8c633a50aa3c0a911a6104f1b402824c96b87457
2070 F20110115_AAAKMG broadbent_e_Page_89.txt
64ddb89adb63652b8461470e735645ed
52049ada97c87ac0759de42e06088701d6ce1259
1997 F20110115_AAAKLR broadbent_e_Page_74.txt
62e9fbe07c4850d5f69ab88251a7d39d
864231f91941511925b4adc6c2f7bdfd6d0b7215
F20110115_AAAKMH broadbent_e_Page_90.txt
769b18954034088766faa6d9437de288
d74e401174761a2c9447f4fb46c4a11999a10ebb
1849 F20110115_AAAKLS broadbent_e_Page_75.txt
976888d225a69d7a3539aba7a3b51a23
017a58d34ee6e5cac900a5e0da88f13f59133b51
2014 F20110115_AAAKMI broadbent_e_Page_91.txt
a9edb106c7a0b17d9782de331f232755
87d98c685f9ad647ac00272e76266c659dba940f
2153 F20110115_AAAKLT broadbent_e_Page_76.txt
ed093e937130683fdf07cd622ef8eea3
36b51b4b407bf21b7d96064116d24961a79be987
1348 F20110115_AAAKMJ broadbent_e_Page_92.txt
e447ee1d01dfcf97e1091b01d3f95397
b7d99c20dae7d50e0fd987d80cb97f18379fb72a
F20110115_AAAKLU broadbent_e_Page_77.txt
02dbd7ff409fbb5587e61c63c4253a58
58c854aaa466ac651b312b0acaaf1a72b4d7bb93
1697 F20110115_AAAKMK broadbent_e_Page_93.txt
45b32d9cf97631fa6db23b8522f7371f
729f06d588a7b8b4daa650ca4039a2faa37d5669
2257 F20110115_AAAKLV broadbent_e_Page_78.txt
9f2ff2a740d5bfcc6c6bedcaf21bf5f6
5526b0e1b63737b609113f89f6b3b5a3c35e86bb
593 F20110115_AAAKML broadbent_e_Page_94.txt
b9fa127c93d7739b854fbfef0e964eef
f379f289846dc3fc5e6e789407941fefe56a88b7
1963 F20110115_AAAKLW broadbent_e_Page_79.txt
a433b6761a10a54d0a0dd4cad506b867
a023c39a7a98e479b4773871efe7626be265e918
6128 F20110115_AAAKNA broadbent_e_Page_02.QC.jpg
23af5afdde41a5330f5908ee7a8d3faa
81a59ca56426f4f7866b5778591827fdb46b6ec0
2140 F20110115_AAAKLX broadbent_e_Page_80.txt
0cc6d137fb64a1ddb57009a4100ee83e
e09d1d97c0ef31573290844872a8881d94f87955
21669 F20110115_AAAKNB broadbent_e_Page_84thm.jpg
67f18cefb30053785d7d341cdaa01d5c
6d6cae93d1d166a869eee36b4932b70e4a0a2c8d
1722888 F20110115_AAAKMM broadbent_e.pdf
7d9d9889981e9364a9b2da771fe71650
38ca61db465676385eddee2f5abe85c6250b0a5c
1994 F20110115_AAAKLY broadbent_e_Page_81.txt
40f04fedadab2556773ab5bdcfd5d2ac
18238b6e54872800ead220e2283466c4f861bfb1
12503 F20110115_AAAKNC broadbent_e_Page_74thm.jpg
2655722854baad8c7f870166dd1aa224
7a5ad25205293ca8870329318fc4bec067642204
77071 F20110115_AAAKMN broadbent_e_Page_23.QC.jpg
821f6f5302fd64a0b911ecb3d6ffc6f7
044c54b5af6167c0177bd4e6e7a13308abcbe41b
2054 F20110115_AAAKLZ broadbent_e_Page_82.txt
b30688bdd2bc998a0df840e1448ed6b9
df97266c86842a0f155792b5720cd876034003a2
33297 F20110115_AAAKND broadbent_e_Page_73.QC.jpg
0bc7c8d3926e20bf60b77e7f031c16b9
6d135f85311638dc86a710d293162d4d5d82d24c
78457 F20110115_AAAKMO broadbent_e_Page_21.QC.jpg
d039aff8be40352ebcbeef6b4d11cf49
d2a5734bddfc3e9d00fffe1ed40dd6c8682bd7af
68279 F20110115_AAAKNE broadbent_e_Page_88.QC.jpg
06df9a55c898b1b2f84a167b51f33662
e70ff13523c95c82194d59d4173d8618f58741e1
44754 F20110115_AAAKMP broadbent_e_Page_59thm.jpg
26dd920989274955367731d281377661
162ec2ab9d4704afb9f89915cd4904e5b43ae6d7
79545 F20110115_AAAKNF broadbent_e_Page_58.QC.jpg
8e65b9110a45187fd7b6558e4c69c30e
25dc642cc25c54facbd0c1d9ae72120350680b44
73044 F20110115_AAAKMQ broadbent_e_Page_62.QC.jpg
5b70045614f439e439c47a87e2f149b3
1c2b1dac52c55906b960ddff74d087f70b10fa53
40648 F20110115_AAAKNG broadbent_e_Page_11.QC.jpg
02f8aba8659988e0be0b42ebead7daaa
45e2151eeea18fa51869fd893aff58eb9eb2610c
63365 F20110115_AAAKMR broadbent_e_Page_81.QC.jpg
47b2437a58d9c4616b29e7c0176ab618
73a7476cf7f29e5bf75a46b4142fa802a7feea33
83758 F20110115_AAAKNH broadbent_e_Page_19.QC.jpg
26ed84403e63632270202cf200ed2e09
845491fdfc13cf564cdbdc208ad02c4326980304
24273 F20110115_AAAKMS broadbent_e_Page_42thm.jpg
c639283900831d7dc28999fc6e154f13
239e2c50ad6b882450236832a0b14624c2f56ba8
68638 F20110115_AAAKNI broadbent_e_Page_84.QC.jpg
75a04ddf07ef455bee38db39c9597f93
2c3e9c69c1773c43ce2b2fb715aeddc46537d9aa
81575 F20110115_AAAKMT broadbent_e_Page_43.QC.jpg
4512caf271c72a4c6eb2532da6ea4797
aa59c57450ad9baeb537342b34a6e1fe9f9098c9
38056 F20110115_AAAKNJ broadbent_e_Page_66thm.jpg
c2d154e143993484392677a2e2155865
26409380f85db6382d307b1293efa546d52abe09
8254 F20110115_AAAKMU broadbent_e_Page_94thm.jpg
df2e6c1a949b52ad3f8742a6eef17df2
3b088ac6979a05340d470301ca08e52bc6f7c9dd
77167 F20110115_AAAKNK broadbent_e_Page_15.QC.jpg
3269a9266574c153bd80cd29b34d756a
ef6824d4979936598227d1ec46ac25d2f9a9daa4
70871 F20110115_AAAKMV broadbent_e_Page_14.QC.jpg
03ca15d569edeb06daf8eaaeefaa9ff8
e53db1a1b28365775d83333e5ca36c561308bb5b
48826 F20110115_AAAKNL broadbent_e_Page_08thm.jpg
049b8cb5cc82bf850cc8dde731bd22c7
f0413d4eb066c3d048f90d14b099b3408f0d7b2b
17760 F20110115_AAAKMW broadbent_e_Page_05thm.jpg
2e54f300c46b27ccdc1b30259dea1eb6
8098f545c18a466c0a3ef5796c32ac90839aee1f
23287 F20110115_AAAKNM broadbent_e_Page_28thm.jpg
c2d7b6904d4994b7c708661ad287b1cf
14f76c25182435482c373678722b7f9f5646f4b7
79818 F20110115_AAAKMX broadbent_e_Page_29.QC.jpg
caf9fd8a4c102e0365787ad7f6d6efa6
716e36b23216790b3cf8c90d3db59e75cdf86d4e
25243 F20110115_AAAKOA broadbent_e_Page_44thm.jpg
4915981a2ddf662665c064be72168595
ad0d4bbb792330a00b0bc4451f4022865f779376
19282 F20110115_AAAKMY broadbent_e_Page_75thm.jpg
c1ecdbabb5b05dcc9c6f52ab899b3f14
824db29ea81f0cc785732500638f929c3f6e4835
65101 F20110115_AAAKOB broadbent_e_Page_83.QC.jpg
246319ed328d9d0edd8287fd4cdf3b43
c83cdbe61cffc304190a1c08154d92952408b7ee
32121 F20110115_AAAKMZ broadbent_e_Page_71.QC.jpg
0acaf6bbb177bf2f70fd102903a4f60c
cdb69880a1c0e95dd87ad40519875e0dd8f6ba2f
25008 F20110115_AAAKOC broadbent_e_Page_58thm.jpg
35c619a26abcba722584150908edde6d
d088422fbed3259244b9388133e1a0a6d5fa64ac
21393 F20110115_AAAKNN broadbent_e_Page_90thm.jpg
6e471ef9c3b5fff564bf24649c8d2069
4f9e3fe9135922625648f6d29f9423cd9f4409c7
143552 F20110115_AAAKOD UFE0007800_00001.xml
10a8210e2d03eceb6d3de53de0908ebe
878590c7b5b93dbb36f4dc7bf62b5d88918279c6
65095 F20110115_AAAKNO broadbent_e_Page_53.QC.jpg
531a6d6a62216db00b9fa19dfee79036
aa83f3dbaeb420ca73938c006f27a63a5b01686d
7355 F20110115_AAAKOE broadbent_e_Page_01thm.jpg
353801e1866f50e59a20ce8b32b7fbd5
bf30c52ec5d04c6a894c34ed9c971a9f26100e35
44673 F20110115_AAAKNP broadbent_e_Page_06thm.jpg
9d40f4b200936e2443aa6fa4f257fd66
e85383e8fa40a0b2a016b6cf0d662593ff9908b5
20689 F20110115_AAAKOF broadbent_e_Page_01.QC.jpg
5ae91bf0608febe364051ac27d6abf4f
d63049e908b9bacd80381215165900d091dd7c5a
12713 F20110115_AAAKNQ broadbent_e_Page_72thm.jpg
243c3a8545af68b887a4aff0fd422a8b
511d0ec41598a4e5413253f48a0d28cee99036cb
3191 F20110115_AAAKOG broadbent_e_Page_02thm.jpg
f0361f057f8ab3092deb1c2f64725623
d6fe58c5289418225770803851c30133ad9ff50e
22548 F20110115_AAAKNR broadbent_e_Page_14thm.jpg
3667a8a0f2da7b8bfa0276ae1978fb84
9599961c6196e5087802fd6d8b6cfa592eda6fbe
5128 F20110115_AAAKOH broadbent_e_Page_03.QC.jpg
0b6d469cd7cd8c5a49457353210ce3e3
bd2ce11c737e384c0ac9212b423be3903fdb7b08
23451 F20110115_AAAKNS broadbent_e_Page_17thm.jpg
1d6babe3937705f0f48d536c59fe2aa3
f2c5a2f50f9890977a347588da871a9dc548b272
22081 F20110115_AAAKOI broadbent_e_Page_04thm.jpg
b163e44b67a83ae74c8b9206d0af5d02
febb38edd58883fcd473533f69270224f789dfe9
39149 F20110115_AAAKNT broadbent_e_Page_67thm.jpg
609a34635e4f38a0727ec5038feaf2e2
e223c8ce2e819560b00cbc8e8d0af18abbe9dbe6
69615 F20110115_AAAKOJ broadbent_e_Page_04.QC.jpg
fbbe40236fdf13350adae1fe708f6f13
66ac712bc6d0c2f6ce366f79cd40782848fd66be
2907 F20110115_AAAKNU broadbent_e_Page_03thm.jpg
19e755dc084939e8c73e82a1469d3e52
a1fc5fd1080284030c301d48bc65316de32c334c
59335 F20110115_AAAKOK broadbent_e_Page_05.QC.jpg
ea4118f3c2873b8e96e721c7edef8ba0
1639115df0c7231119f1ec22711c135b5969baf6
22077 F20110115_AAAKNV broadbent_e_Page_47thm.jpg
1090ce28819ecbcbaa8250b99c9b3af3
c1183e1d54055264e719a4c79e4ef57506c17eec
85292 F20110115_AAAKOL broadbent_e_Page_06.QC.jpg
8e4a8a2e51fcc048bed7a7bfc03776bf
81875a5833b4317ecb2ac780d01f9749259737cb
106145 F20110115_AAAKNW broadbent_e_Page_08.QC.jpg
0a1dfddae8153d01f94fc39328ade02f
386392f4cbed25da5c50c7343c4af17048c9e190
70010 F20110115_AAAKPA broadbent_e_Page_17.QC.jpg
2498ee92b09da37b70a92efd6baac86b
9fa75b79d9e40fa577e91d88260a7317dfd8c4be
44797 F20110115_AAAKOM broadbent_e_Page_07thm.jpg
5dedf4167f06df80aa14ced190327597
6ef3eee501512676cfeea32f9555d003e1dd9fce
39165 F20110115_AAAKNX broadbent_e_Page_68thm.jpg
774cbb2a8d3478de0fceaf12a74efe11
9d674f80d7dd6ed150219efa311ced5e710ba711
23888 F20110115_AAAKPB broadbent_e_Page_18thm.jpg
d1c86555420872cfb2e800beceb2a385
d290d9489c2a954e719c62f785407c7df7967f7a
87961 F20110115_AAAKON broadbent_e_Page_07.QC.jpg
44ec073d5d25d9122e4fe9a7f4625633
16fb17a902afdd722b34b566b79eff8f31d388a4
24445 F20110115_AAAKNY broadbent_e_Page_21thm.jpg
cad9b4f6a82fc71d5e9a7d259e374029
4e647499523c6673cab3115494b09e3dfead85b7
77606 F20110115_AAAKPC broadbent_e_Page_18.QC.jpg
2378c7bf8534e534106e8b8f6905e71a
89605cb3883aba8f84c1efe5138a1e29b5122c28
79362 F20110115_AAAKNZ broadbent_e_Page_44.QC.jpg
31d4fee9ac79a53fb1308a873e0c8107
a589bee6cb37cb4fe18fc671fbec2141ae612d6b
44843 F20110115_AAAKPD broadbent_e_Page_19thm.jpg
cba04ce6cedd4e0babcd9697e1b85791
7f4d1691a6076a0bc732cc015da48e75ad868700
38930 F20110115_AAAKOO broadbent_e_Page_09thm.jpg
88ec08b66b1e0821beeaaf0d0c443952
7fe1f36f50cb6056f78607210bca6aa634e7b73f
43852 F20110115_AAAKPE broadbent_e_Page_20thm.jpg
918dd687f8c6f2ff716b5951c2da6547
6bbdb39e5a092936d6618c19e4649b2e30045248
71290 F20110115_AAAKOP broadbent_e_Page_09.QC.jpg
328a963167f511758f0a3a15da981a88
53ccfe528a27fc9dd7478ec13e07f584c9cca435
80276 F20110115_AAAKPF broadbent_e_Page_20.QC.jpg
985cc8b7f03bd879107eb039e212d78a
7376a289d07186e546b2e946e6808b87a6210bb5
50086 F20110115_AAAKOQ broadbent_e_Page_10thm.jpg
d665908e9d3923aff50beaea9bba05c3
fd39d2b4b9019c300d9c4fc6eb29c0e459d0c0f9
43640 F20110115_AAAKPG broadbent_e_Page_22thm.jpg
4f3d4d32089f005659f370489125d49f
011ae4ad80e49f2482e7f23c176202b0304d6ec2
110065 F20110115_AAAKOR broadbent_e_Page_10.QC.jpg
377c6b08e2820b73b0ecabeb78bd67f4
94578aef4dd3654b5063da69baace86961f10d0a
79186 F20110115_AAAKPH broadbent_e_Page_22.QC.jpg
e942aae1bc984706f5dc7bcf4b8079c5
02efbb023df1c0265aa3306cca1ad57a98ca2b38
32017 F20110115_AAAKOS broadbent_e_Page_11thm.jpg
9d3e7c0669813cdd63f38e449f9e297f
076a9fa2e6ba5dec2f1dfdecb334253dc16f2f34
25208 F20110115_AAAKPI broadbent_e_Page_23thm.jpg
2693c54791953689464261f71b0ee5b8
a80531f05f20a28c545edbe9a09a08f9bcc9340f
19485 F20110115_AAAKOT broadbent_e_Page_12thm.jpg
a084831cbd009ce62192b5145a3bb561
0e8953f177ce9906800590dc88757354b6c100cc
24382 F20110115_AAAKPJ broadbent_e_Page_24thm.jpg
4009a1da237db834405797aefdd32f71
2c3b0a57f7c6f9af8fb6e690c033abf74c57f2ce
61277 F20110115_AAAKOU broadbent_e_Page_12.QC.jpg
644c2ebfb9163d55ee81e310e8d7e286
37ee2006bbd452feea43050985e618ebae63acaa
79239 F20110115_AAAKPK broadbent_e_Page_24.QC.jpg
c733ce503b30d9151321619f6fd67d9a
29f2c346450fa148107a0bc22c7e29c285170950
6586 F20110115_AAAKOV broadbent_e_Page_13thm.jpg
173c05b9a96e89c5d85bdb33738e378f
f5e97bba608bf759dc45ca6dcc9ffda8d544714f
24375 F20110115_AAAKPL broadbent_e_Page_25thm.jpg
0dd34d57ef891aac7079299512c8c9e9
b056ff2e065e85943c23e1f6bb4a286a3191f553
17339 F20110115_AAAKOW broadbent_e_Page_13.QC.jpg
1689f4c08412d73ed63c21b6284e156a
8b5e169fe666d2187fd94476e2d51c338c22c343
25342 F20110115_AAAKPM broadbent_e_Page_26thm.jpg
5dd422f87fbfeb986a1fbd24a99041a8
9a31b39c58e4106e26d45ad2d2d783c9870c00bf
25469 F20110115_AAAKOX broadbent_e_Page_15thm.jpg
169624afaba78121f5880fecc9b5bd08
6f2656ec43746cf6ef2595d13bf27fab7173b0d3
22064 F20110115_AAAKQA broadbent_e_Page_34thm.jpg
597dcada324dbd32de8b8e541fb92ff0
ba5ae3a2f2991c78d66a140902ad87664263080e
81992 F20110115_AAAKPN broadbent_e_Page_26.QC.jpg
13b727042d90d1ff9c432e87923ebfd1
26eac4fcfd42a3ffe9bc2c34cf6de85e849292bb
25675 F20110115_AAAKOY broadbent_e_Page_16thm.jpg
b4e49912d229449d4b47f58a62db66c1
1bec1ed887dabe880ab346ecaa550c627c31fd8f
68635 F20110115_AAAKQB broadbent_e_Page_34.QC.jpg
6738c574f3504428589e68189f12dcd4
47cbe7e381c17caf13b0e8cf071d97f1fe72b6e5
24271 F20110115_AAAKPO broadbent_e_Page_27thm.jpg
705c40be2006fbf97e309b74d3c6ca1a
4c25e7690001c603b55e59b8c35fa7a3ee306c92
83211 F20110115_AAAKOZ broadbent_e_Page_16.QC.jpg
f3cd367719184b38eb79af31f4fe8c12
62b9786f8e6f31c43cd2aa2f8fe1af75b2795f91
23707 F20110115_AAAKQC broadbent_e_Page_35thm.jpg
fb95ada8e90fe1f53e13b956a7a1d1ac
ace713d0f27e8186cb8b8632e4340e252b1815e8
75138 F20110115_AAAKQD broadbent_e_Page_35.QC.jpg
ea082afe9c2bcd3cb1a5e15d5e7acdf5
e26553cec9677d31a6dd7603aff88861e15dbf43
78626 F20110115_AAAKPP broadbent_e_Page_27.QC.jpg
75e3204eca08ad35f2dd778b9583b6ec
2cf50fa5a61dbc90139f02f163d0260ae425658c
18143 F20110115_AAAKQE broadbent_e_Page_36thm.jpg
3a8044ff975b6e7c22958ed61e572c38
8fc40d49842f659b7f9ad51a868166de91b1f481
75868 F20110115_AAAKPQ broadbent_e_Page_28.QC.jpg
92c94c07b9d1b9b00800a06e07b61efa
580beda34e68124100a50f9dec9060552006232e
50478 F20110115_AAAKQF broadbent_e_Page_37thm.jpg
a14ccadfca0d411b787129daa645a0b3
d1744b09046575d83c01b48f7ccc632145ca070e
25273 F20110115_AAAKPR broadbent_e_Page_29thm.jpg
4910320d2d6cf6f9dd3ae734554ae35d
ba0f6d54d52a284758d20e3150bae8c74f1e687a
109312 F20110115_AAAKQG broadbent_e_Page_37.QC.jpg
05a3a753a36f54790811bce3230b712a
fe16459b0aea004bc61cb19ae71a1608d929e7a6
25182 F20110115_AAAKPS broadbent_e_Page_30thm.jpg
a18e07486d3b56562beab16c3a6ad3dd
f6a4b5202c454576296af351305a83d4aeacb716
22480 F20110115_AAAKQH broadbent_e_Page_38thm.jpg
656e6a3909efe5cfa45295ccf15bb52c
a33d41bb48a3d6c665325bfd82123e9eb54486ca
80744 F20110115_AAAKPT broadbent_e_Page_30.QC.jpg
0795ba5ef2b1e93ebf3a7b4de103cdff
2476dd0eba21ef29660ffbbd6559619c4d54b565
71474 F20110115_AAAKQI broadbent_e_Page_38.QC.jpg
85e66c5625c6db106f2c9169bc2d02b3
53233cc642e61eb2540eea4de9e666e2a57cb30a
25398 F20110115_AAAKPU broadbent_e_Page_31thm.jpg
b8deb41d66449de6556025a1d64df272
2c403f1668066bb9c2e42a4bbde93948506c514d
21741 F20110115_AAAKQJ broadbent_e_Page_39thm.jpg
78f2870c323ecdb10fe89bd4ce4328e4
71bf1bc06e8128e87481abaec11632467903338a
81849 F20110115_AAAKPV broadbent_e_Page_31.QC.jpg
41fa0feee1868f80e4c94e1bc5b98523
70bc095d959e7f6a877a3feab4b58436d22f2ebb
78009 F20110115_AAAKQK broadbent_e_Page_40.QC.jpg
c75df13fa28ead10d10824558c5e4000
cbe344e3a34901f2d8b3787f698c16c15f4f94a7
24372 F20110115_AAAKPW broadbent_e_Page_32thm.jpg
3bf81b723b09766b09945b983d93a291
83e7c9134473442f2c1e09c2cd448ee0a3d01044
25310 F20110115_AAAKQL broadbent_e_Page_41thm.jpg
3fb8b453c4b1c84ec4af08ee58519ebb
ba794f5a014311efc92f52820253fa35f93b49e3
78877 F20110115_AAAKPX broadbent_e_Page_32.QC.jpg
43c4d76a2edb66024a3fbf07a5211004
22ee2afe1273cb59eff3510e6263c350c9b73a3d
21114 F20110115_AAAKRA broadbent_e_Page_51thm.jpg
2499aa90538652b4392d1121b752aa31
9a2b4eb8053cab3b6c24dc0e81deed3abfaf186d
83616 F20110115_AAAKQM broadbent_e_Page_41.QC.jpg
27dca6dd8af7b9057b3777362d90c7be
9896c7982de33c6c8156bddde7f372c0a09e178b
19701 F20110115_AAAKPY broadbent_e_Page_33thm.jpg
03989bb8d279c48bc5ad64b8321ac3a3
4c0e12e345bf8bdcd96f8ec86923dc3abba91565
69333 F20110115_AAAKRB broadbent_e_Page_51.QC.jpg
71583e02e92f8233d26345ce4e7051f5
634e8765bdb27b2ac5a089feff705db32affcfdb
77378 F20110115_AAAKQN broadbent_e_Page_42.QC.jpg
ab9c84865281614d50b9b125c921e461
ae676aacfc105a0e742c03f521a72e3e98afa2ab
66621 F20110115_AAAKPZ broadbent_e_Page_33.QC.jpg
b0df88c1d7a58d6fb8531ce36d8c5673
5f46762826f97ede7b3f68621ac61ab1053be57d
43778 F20110115_AAAKRC broadbent_e_Page_52thm.jpg
c46048442737caffba0b27f1703f5d25
ac89730fe76419b06f0a441ae79d6bfd46e20956
25348 F20110115_AAAKQO broadbent_e_Page_43thm.jpg
64db2890a84f3ef5e49622071e5fb588
9f3b14cb0e85129599b84edc72410ed116f18c5e
80298 F20110115_AAAKRD broadbent_e_Page_52.QC.jpg
01c4c9239dc02f2e10bbf95dea4025d2
937b2bc518ea8a9af60df1ce6e8d028267661df8
24617 F20110115_AAAKQP broadbent_e_Page_45thm.jpg
de1cbeb8b3a97df8764cf6f7bdcf187d
bcf116a613b5d291a87e8a8a12b997285704cd82
19958 F20110115_AAAKRE broadbent_e_Page_53thm.jpg
0b505ec24f3fa37913c868ebc0d98999
7e6493562be4a39e2f375a04e9a549a8fefe13b4
44420 F20110115_AAAKRF broadbent_e_Page_54thm.jpg
9555a128688ccd01f3fb901375cc9ac6
71297e9e4e75fc8d74c57787d19d33d24507801c
73937 F20110115_AAAKQQ broadbent_e_Page_45.QC.jpg
b6c71cfd280505a87d4d623db558aed6
00e184b577ef8d7eb4d5324bdc06185be9a9621e
77408 F20110115_AAAKRG broadbent_e_Page_54.QC.jpg
05a7605262421d2e16da00b1f09064c1
6e4cd98837718cc2ad54f67ba3fdf563d69be33d
25742 F20110115_AAAKQR broadbent_e_Page_46thm.jpg
cddec299a133171f2afec18f44d510ff
1a524e465ee5d12e966bdf9a514526402e43e9c4
45736 F20110115_AAAKRH broadbent_e_Page_55thm.jpg
9e510d11a57d10d14897f7681a81403b
f1953b9990b90acee45caa18ba248df4c8fcee22
80540 F20110115_AAAKQS broadbent_e_Page_46.QC.jpg
01838eecb0db380e64b77dacb37fe583
19c972056832f3c47177395042ae4531fb5d1b42
85115 F20110115_AAAKRI broadbent_e_Page_55.QC.jpg
89c9d54dda719a6964e5e31f9472b66b
8516b247bce90d047be82cd0938d9871ff9b4eec
66151 F20110115_AAAKQT broadbent_e_Page_47.QC.jpg
69a6ec028d037ace3201dd62e54ed696
fccb73c087d5a07c4dcd9396f0be7438880f97f3
21295 F20110115_AAAKRJ broadbent_e_Page_56thm.jpg
22cb75d1c6e53a398ae8dd4b63ca0aff
54e8d8cba05a49fb3aeb840915169dc20b968a2d
49339 F20110115_AAAKQU broadbent_e_Page_48thm.jpg
89724f63f1729d9b547c946adb52e964
81d4d0cc42a5da06882cd00d930ce82512fe8fec
64878 F20110115_AAAKRK broadbent_e_Page_56.QC.jpg
f251ca65064d9aef875a8e9f800af970
816c38bf1d285da03ab7bc32a59e8fdffad5ae49
99187 F20110115_AAAKQV broadbent_e_Page_48.QC.jpg
dcf6749ad25f1a23f670b82bde442a05
3c7188e203c7780b21a8cdc42571279eafa7dfab
38027 F20110115_AAAKRL broadbent_e_Page_57thm.jpg
a647879f14fae85fa1a74eb39bf730dc
cca6780db7e12473a2c20a4dfd9f9de5033f2706
43565 F20110115_AAAKQW broadbent_e_Page_49thm.jpg
7d3801637be79b3790a443d111a4d776
1a275369b179a00dd9e14ca5901cfa40ce95e1bf
67966 F20110115_AAAKSA broadbent_e_Page_67.QC.jpg
b2eac47933ea32ea84337367068f7b79
be4c4c778137fa107f205b7c98b09e296740678b
59508 F20110115_AAAKRM broadbent_e_Page_57.QC.jpg
8286761204c71ef35461e007d49202c1
71b5c95ad4d742d60c3cb327e730980384e71de8
78936 F20110115_AAAKQX broadbent_e_Page_49.QC.jpg
30421ee12fc2e07253c34b9789a45636
d31e5fa4d5cae4f7d16ea585b2e8a3162ba3d793
64443 F20110115_AAAKSB broadbent_e_Page_68.QC.jpg
8b0c83e8b7a8af72640e1975807e01fa
414ee249ea2d1f587f67a3af86d6f177642afeaa
84702 F20110115_AAAKRN broadbent_e_Page_59.QC.jpg
ab4b968d5f06b65c8cdb1e28b2df25f7
ac25f2ac4a32d3da0ed3008764ad861c9f3a20a4
23682 F20110115_AAAKQY broadbent_e_Page_50thm.jpg
bfef1b3d101d235e60fce55e95776406
cf78338c1f38b6d1ade9e17153a26e26cdc7e7c3
34994 F20110115_AAAKSC broadbent_e_Page_69thm.jpg
00f61a2b9baf5a34b48e3631fd54a376
dfb5ed2a6839d51f14cc1d5717f29e3a2d7c537c
41957 F20110115_AAAKRO broadbent_e_Page_60thm.jpg
0ee774b0a7c18b5c8f7aacb68a741423
f73c07e2ec5bee04106bb7ad78b2e673491ed494
75279 F20110115_AAAKQZ broadbent_e_Page_50.QC.jpg
b582c1873538015f82e49600463c202a
3070d3993c6a877237a764e49790bbee37d6fe89
52067 F20110115_AAAKSD broadbent_e_Page_69.QC.jpg
9ef2bc11fb8191d985a1585e470fa833
379d1e47eacbd83d911282fba4d9e038fad26360
77600 F20110115_AAAKRP broadbent_e_Page_60.QC.jpg
e70b6cbfea34eef232e25dfab15b818e
b153ceb3cd8329d8299d284f6444dcf116823677
34850 F20110115_AAAKSE broadbent_e_Page_70thm.jpg
0f486d500be9c289857455291386fb57
026f8b671cb20d8d412107276d54d073a89cab1e
45714 F20110115_AAAKRQ broadbent_e_Page_61thm.jpg
e2ea1bd838f5b442c820584b43258189
82475b0603a97508b748b766f8d20259689025e6
54090 F20110115_AAAKSF broadbent_e_Page_70.QC.jpg
2df181acb037c43619e9b0334eef6436
29940d9a9359bac44e7dad4e01439a805aae2e47
11736 F20110115_AAAKSG broadbent_e_Page_71thm.jpg
6705c73b3e025ac22cba5cf6196d9526
f086c6fe0c0cc267363ca8253b7c8392a7a12618
83938 F20110115_AAAKRR broadbent_e_Page_61.QC.jpg
2e52be5a7c98dd323dff68ded20a6658
ebf1e1671c8c51c32550eefcd5f5bb1f937520a9
37300 F20110115_AAAKSH broadbent_e_Page_72.QC.jpg
982bc76be7242e8f4f7aa34ba46c0e50
31c1e017f9b3fe52c20afa52cd11f8c441d01bf8
22223 F20110115_AAAKRS broadbent_e_Page_62thm.jpg
b8773a2e26ba229236ad4ca7dd64e0b2
a5ef7ba8377d316e1790b9a5b2514b7990ec49bf
10820 F20110115_AAAKSI broadbent_e_Page_73thm.jpg
048f26766bbb44d7d2ded089fbd139a5
3e6e81c8fff733e52f560eb7056c212ffbf34921
21024 F20110115_AAAKRT broadbent_e_Page_63thm.jpg
679780ae075fe5c829c82516525fe84b
609bb3f5335ac63bdc35302eb4d678f76d67efd3
36357 F20110115_AAAKSJ broadbent_e_Page_74.QC.jpg
d923e68b5b3cec5a8f17d4ee3bc6c0cd
9d3123ec6ed6d8e0f1801bd75621163c7628f3f7
64638 F20110115_AAAKRU broadbent_e_Page_63.QC.jpg
9d91028168caa858e8b719a2977a3015
5fd7bc01b7eabedede64a7b6bfc69b9a7dfde336
20888 F20110115_AAAKSK broadbent_e_Page_76thm.jpg
09163cd924651cdcd76a0fff6a670671
19c879259eea8dca68d1aacb05e954cdb598a9e2
24773 F20110115_AAAKRV broadbent_e_Page_64thm.jpg
38ce83c535bbd7423c18a9bdbabaa89f
887f213e0139542b97014159f3c44a73ab085b0e
78121 F20110115_AAAKRW broadbent_e_Page_64.QC.jpg
8ebf2c6d081b6d5173f7213f54ab2e37
d5d619b140f544d401c0f5ec981451d6943cbf65
67365 F20110115_AAAKSL broadbent_e_Page_76.QC.jpg
3523e55ebf89a6015b4e3c2a3cc4edaa
7b9b15ba0cc858e5d528ecb1244ccffd84608eaa
22042 F20110115_AAAKRX broadbent_e_Page_65thm.jpg
7286f67ee1d06c299032a8f334731eba
c7ba02dea8f535ac0f868df896ba8f01be04a8b6
65126 F20110115_AAAKTA broadbent_e_Page_86.QC.jpg
1c83d84956895b4dd3b94c068ef1d1e6
92e637d785100766a33282c61218d6a661701a24
21602 F20110115_AAAKSM broadbent_e_Page_77thm.jpg
30187a431f229ba51f65f0fd0f8403aa
079ea573711dc92a9cf5812398ff9c086f18072a
72962 F20110115_AAAKRY broadbent_e_Page_65.QC.jpg
590154b6d8acdc79ac050608ab943175
70e33e30b07c928ea6777b9e1f9fb60b9e6a59c8
22474 F20110115_AAAKTB broadbent_e_Page_87thm.jpg
377390d403e0eb070edc91442d4806eb
26505175e00c93e5089e0b531980574f15cf0b97
69940 F20110115_AAAKSN broadbent_e_Page_77.QC.jpg
c4877a466b72f34211f4c650fb0ebe6a
fa55ac14365dd24cb4f29497f0849019ee0b8dda
F20110115_AAAKRZ broadbent_e_Page_66.QC.jpg
d11bc01f3b1037999fc194bc80ce9abf
b4e3ad34e09f0a8473008b62e3df23411a13f6c5
66060 F20110115_AAAKTC broadbent_e_Page_87.QC.jpg
4a11684e7e4556847d4cdb526616951c
23023bbc5ec1358b310bd38545400045edcc15c7
23704 F20110115_AAAKSO broadbent_e_Page_78thm.jpg
8e0964e3c424358b9bb565129055b5c9
aa3798efe91c9bb9c453273e24844bbe0c62699f
21512 F20110115_AAAKTD broadbent_e_Page_88thm.jpg
9254cc1987d25a360ea3781779232226
e2679cbcd6d4d6db08a566ef66ef92af92412585
73419 F20110115_AAAKSP broadbent_e_Page_78.QC.jpg
619d5af29c9e654c0d2f52ba285a107b
dda05df257a85fb2f04b99e5a54503e1672f1cf5
21782 F20110115_AAAKTE broadbent_e_Page_89thm.jpg
ca9182ccb72ed45dd63d94a0d48306c7
b271c2e63b2654cd75edc48ee711239e5ab0c3ee
21801 F20110115_AAAKSQ broadbent_e_Page_79thm.jpg
410d6a58346bab1cc401c46cd70dee6d
1762ab81f446b2035e231988ff8634d0ee25fe04
67361 F20110115_AAAKTF broadbent_e_Page_89.QC.jpg
d6aa601e007a918fa85a248dc16295d6
15094c431de2eef25e64904c9863f149b5b16805
64288 F20110115_AAAKSR broadbent_e_Page_79.QC.jpg
13926b97583cd8ca84b98b6e831529f5
fa6fad7c5ed6f8b4e4a4b9ed600f6df8bdbdcaf4
21626 F20110115_AAAKTG broadbent_e_Page_91thm.jpg
13250a2a2fae3a090c2b02f660bd9eb0
3be77db1099c4de33bc9b5cb12c6e69f4adf89d8
64846 F20110115_AAAKTH broadbent_e_Page_91.QC.jpg
ad506d73f64fb767555c72ebf4bd907d
c749ad1c59c36cb17f921ce8b885920777c5651a
22412 F20110115_AAAKSS broadbent_e_Page_80thm.jpg
4ad357c85db3355c917bb48b158a28b8
f2d30c06ed6fb016e3bf935b2b04eaaabdf77da5
45099 F20110115_AAAKTI broadbent_e_Page_92.QC.jpg
40fe3e3651fe0b20ca949784358e41f4
d50c65750b6314ed630d179b07c94f9c13bb03c6
69803 F20110115_AAAKST broadbent_e_Page_80.QC.jpg
33a626007f6bb3f93585062ed0a5cc93
76119de36d16d7d26cc23b0ffe82cdebd4d86d7f
21450 F20110115_AAAKTJ broadbent_e_Page_93thm.jpg
4490afdccda963e13ff3ba993e2bcc2c
1318e114c763fba0dad7febea3844060c9e87f05
20594 F20110115_AAAKSU broadbent_e_Page_81thm.jpg
e54b6fd37845d40c3d635a000bfd1d54
c1aec7aeb8baf8231aac075ccb512d69a78da000
67488 F20110115_AAAKTK broadbent_e_Page_93.QC.jpg
000aabe9ac528cf4ef2ff00cb59c5941
cd9ea900a98fa8640833cafd8eb9f910690c31b0
23326 F20110115_AAAKSV broadbent_e_Page_82thm.jpg
4bcaa00e14fa1a95a3b53ce72a9da6ca
bec5b9f0e3647df7ad99904b5212ccd73319b6ba
26177 F20110115_AAAKTL broadbent_e_Page_94.QC.jpg
1fbcdcf61ea473f5e898e77be5226ff9
8f755e9ce3738e57b514dff6c22d882cf39d43fb
21727 F20110115_AAAKSW broadbent_e_Page_83thm.jpg
7e11ccf22594e24dd856842183012761
ce91e79b880141849cfdecbaaf3e68862ac7626b
22740 F20110115_AAAKSX broadbent_e_Page_85thm.jpg
b9970d30514e08bb8bc6025d9821de45
a2d8c7f8025363934af2df73c4c3bc31e3f4a77e
68342 F20110115_AAAKSY broadbent_e_Page_85.QC.jpg
1d56252f9dd1456034617ce918ce60fd
41ed330a623cb4105dfcf1aae150af41b1a9dd6e
22284 F20110115_AAAKSZ broadbent_e_Page_86thm.jpg
d2ab4e02dd7a3a47ad0747d8d7092661
2ea5b333279d8d8c95c0cd1d34bfd5b826acb1f1


Permanent Link: http://ufdc.ufl.edu/UFE0007800/00001

Material Information

Title: Post-Harvest Recovery of Forest Structure and Spectral Properties after Selective Logging in Lowland Bolivia
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0007800:00001

Permanent Link: http://ufdc.ufl.edu/UFE0007800/00001

Material Information

Title: Post-Harvest Recovery of Forest Structure and Spectral Properties after Selective Logging in Lowland Bolivia
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0007800:00001


This item has the following downloads:


Full Text












POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL
PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA
















By

EBEN NORTH BROADBENT


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Eben North Broadbent

































Dedicated to the forests of Bolivia















ACKNOWLEDGMENTS

I would like to thank my advisor (Daniel J. Zarin, School of Forest Resources and

Conservation). His unwavering support and guidance over these last 2.5 years made this

thesis possible. I would like to thank Michael W. Binford (Land Use and Environmental

Change Institute, Department of Geography) for helpful technical discussions, for my

formal introduction to remote sensing and geographic information systems, and for his

trust and generosity in allowing me use of the field spectrometry during my field work in

summer 2003. I would like to thank Francis "Jack" Putz, Department of Botany, for

taking the time to walk with me through the forests of the La Chonta concession; and for

his valuable critiques of my field methodology. Thanks go to Ramon Littell, Department

of Statistics, for insightful discussions regarding analysis of the field and remote sensing

measurements.

I would like to thank Gregory P. Asner (Carnegie Institution at Stanford

University) for inviting me to his lab and for helpful discussions about remote sensing of

selective logging. I also thank Amanda Cooper of the Carnegie Institution for processing

my imagery several times, and for helpful discussions regarding the AutoMCUC

procedure.

Great thanks go to Marielos Pefia-Claros (director of forestry investigation at

BOLFOR, and later IBIF) who made the field portion of my thesis possible. Our

discussions of research ideas (during my internship with BOLFOR in 2002 before

starting graduate school) and the many since then, helped make this thesis possible. Todd









Frederickson, Joaquin Justiniano, and everybody else at BOLFOR helped me extensively

through discussions, technical and logistical assistance, and basketball games. I thank the

employees of the Superintendencia Forestal for generous use of their space during my

presentation in summer 2004; and for their comments on my research.

My field assistant and good friend Victor Hugo-Lopez, one of the hardest workers I

know, motivated me to enter the densest liana tangles and provided insightful feedback

regarding the field methodology. His passion and dedication to forestry and forest

conservation in Bolivia will continue to inspire me.

Lucas Fortini, Roberta Veluci-Marlow, and Kelly Keefe made my time in the lab

and in Gainesville full of good memories. My friends Michael Burke, Michael McCarty,

and Brad Rosenheim gave me their support and friendship when I needed it over the

years.

I thank my mother, Taihaku; and my father and stepmother, Jeff and Jeylan, for

their love, support, and continual interest in my graduate studies over the years. Thanks

also go to my sister, Leafye, for her love and support.

I thank my wife, Angelica, for the uncountable nights spent laughing when we

could have been stressed, for making my life as wonderful as it is, and for her continual

help in making this thesis possible. Te amo siempre.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................................................... ............ ............. .. viii

LIST OF FIGURES ................................... ...... ... ................. .x

ABSTRACT ........ ........................... .. ...... .......... .......... xii

CHAPTER

1 INTRODUCTION AND LITERATURE REVIEW .............................................1

1.1 Deforestation, Forest Degradation, and Logging in Bolivia.............................. 1
1.2 Ecological Impacts of Selective Logging ............... ............................................ 3
1.2.1 G round A rea D isturbances ................................................ .............. 5
1.2.2 R esidual Stand D am age..................................................................... .... 6
1.2.3 Forest C anopy D am age ........................................ .......................... 7
1.3 Remote Sensing of Selective Logging......................................... ............... 9
1.3.1 Textural and Single Band Analysis ................................. ...................11
1.3.2 B and Indices ................................................................ ... .... ........ 12
1.3.3 Linear Spectral Mixture Model (LSMM) Analysis................................14
1.3.4 Influences of Topography and Seasonality on Spectral Response............19

2 POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL
PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA..........21

2.1 Introduction .....................................................................21
2.2 Site D description .......................... .............. ................. .... ....... 23
2 .3 M methods .........................................26
2.3.1 Field Spatial A nalyses ........................................ .......................... 27
2.3.2 Field M easurem ents and Analyses .................................. ............... 28
2.3.3 Remote Sensing Measurements and Analyses .......................................30









2.4 Results........................................... 34
2.4.1 Field Spatial A nalyses ........................................ .......................... 34
2.4.2 Field M easurem ents...................................... ..... ..........................36
2.4.2.1 Post-harvest recovery of forest structure in felling gaps..................36
2 .4 .2 .2 S k id trails.................................................................................... 4 3
2.4.3 R em ote Sensing ............... .... ........... .. ..... ........ .. ................. .. 43
2.4.3.1 Post-harvest recovery of spectral characteristics of felling gaps .....45
2.4.4 Linking Field and Remotely-Sensed Measurements..............................48
2 .5 D iscu ssio n ................................................... ................ 5 0

APPENDIX

A GROUND, STAND AND CANOPY DAMAGE AFTER SELECTIVE
L O G G IN G ............................................................... ..... ..... ........ 53

B MEAN MONTHLY PRECIPITATION IN LA CHONTA ........................................56

C DISTRIBUTION OF FELLING GAP AREA SIZES WITHIN THE STUDY
P A R C E L S ...................................... ....................................................57

D SKID TRAIL AREAS AND HARVEST INTENSITIES FOR ALL BOLFOR
LONG TERM SILVICULTURAL RESEARCH PLOTS .............................58

E SIGNIFICANCE AND F VALUES OF 2-WAY REPEATED MEASURES
ANOVAS OF REMOTE SENSING VARIABLES OF FELLING GAP
P IX E L S ...................................... ................................................... 5 9

L IST O F R EFE R E N C E S ............................................................................. ............. 62

B IO G R A PH IC A L SK E TCH ..................................................................... ..................80
















LIST OF TABLES


Table page

2-1. Characteristics of the selectively-logged parcels used in this study...........................26

2-2. Percent parcel area disturbed by tree fall gaps and skid trails.................................34

2-3. Sample size of large, medium, and small felling gaps within the logged
p a rc e ls ........................................................................ 3 6

2-4. The F and P values for the main effects of parcel, gap size, and gap zone,
and their interactions for mixed 3-way ANOVAs of variables measured
in fellin g g ap s ...................................... ............................. ................ 3 7

2-5. Mean values of field measurement variables within felling gaps for < 1-,
6-, 13-, and 19-months post-harvest parcels. Unlogged control forest
values are provided for com prison. ............................................. ............... 38

2-6. Mean values (+standard error) of field measurement variables for all
large, m medium and sm all felling gaps........................................... ............... 38

2-7. Mean values standardd error) of measured variables for all trunk and
canopy felling gap zones. ..... ........................... ........................................39

2-8. Mean standardd error) for field factors within skid trails < 1, 6, 13 and
19 m months post-harvest ............................................... ... ..... .. ....... .... 43

2-12. Pearson bivariate correlations between field and remote sensing
measurements of felling gaps in the < 1 month post-harvest parcel ......................49

2-13. Pearson bivariate correlations between field and remote sensing
measurements of felling gaps in the 6 months post-harvest parcel........................50

A-1. Section 1 of ground, stand and canopy level forest damage after
selective logging ordered according to level of harvest intensity
(trees/ha) ..................................... .................. ............... ........... 53

A-2. Section 2 of ground, stand and canopy level forest damage after
selective logging ordered according to level of harvest intensity
(trees/ha) ..................................... .................. ............... ........... 54









A-3. Section 3 of ground, stand and canopy level forest damage after
selective logging ordered according to level of harvest intensity
(trees/ha) ..................................... .................. ............... ........... 55

D-1. Skid trail areas and harvest intensities for all BOLFOR long term
silvicultural research plots......................................................... ............... 58

E-1. Significance and F values of 2-way repeated measures ANOVAs of
remote sensing variables of felling gap pixels .......................................................59

E-2. Mean differences and P Values for two-way repeated measures ANOVA
post-hoc comparisons (Dunnett's Test) of NDVI, PV, NPV, and soil
fractions in large (> 800 m2) felling gaps versus unlogged control
parcel pixels ......... ... ................................................. ................. ... 60

E-3. Mean differences and P Values for two-way repeated measures ANOVA
post-hoc comparisons (Dunnett's Test) of NDVI, PV, NPV, and soil
fractions in medium (400 to 800 m2) felling gaps versus unlogged control
parcel pixels ......... ... ......................................... ........ .......... .......... 61
















LIST OF FIGURES


Figure p

1-1. Ground area disturbed for varying levels of harvest intensity in planned
and unplanned logging. .................................. .. ............ .......... .... ....

1-2. Residual stand damage for various levels and harvest intensity in planned
and unplanned logging. .................................. .. ............ .......... .... ....

1-3. Significant relationship between increasing % canopy cover loss and
increasing harvest intensity for planned logging.....................................................9

2-1. Location of research parcels in the La Chonta forestry concession. ..........................24

2-2. Locations of tree fall gaps and skid trails are shown for the logged
study parcels. ........................................................................ 35

2-3. Michaelis-Menten nonlinear model fit over harvest intensity versus
% parcel area affected by skid trails.................................. .......................... 36

2-4. Canopy openness of all felling gaps as affected by the interaction
betw een gap size and gap zone. ........................................ .......................... 39

2-5. The field factors of Liana, NPV, soil coverage, and NPV height as
affected by the interaction between parcel and gap zone ......................................41

2-6. Photosynthetic vegetation coverage as affected by the interaction
betw een parcel and gap size. ........................................ .......................................42

2-7. Control parcel mean values for NDVI and PV, NPV, and soil fractions
versus im age acquisition date ............................................................ ..... ........44

2-8. Difference between spectral characteristics of large felling gap and
unlogged control parcel pixels for NDVI, PV, NPV, and Soil.. ............................46

2-9. Difference between spectral characteristics of medium felling gap and
unlogged control parcel pixels for NDVI, PV, NPV, and Soil. ............................47

2-10. Difference between spectral characteristics of small felling gap and
unlogged control parcel pixels for NDVI, PV, NPV, and Soil. ............................48









B-1. Mean monthly precipitation (mm) measured in La Chonta from
1993-2001 ......... ......... ............................... 56

C-1. Gaps size classes for this study were small felling gaps < 400 m2
medium felling gaps 400 m2 to 800 m2, and large felling gaps > 800 m2...............57















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL
PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA

By

Eben North Broadbent

May 2005

Chair: Daniel J. Zarin
Major Department: School of Forest Resources and Conservation

Our study combined extensive field measurements of the spatial and temporal

dynamics of felling gaps and skid trails < 1-19 months post-harvest in a forest in lowland

Bolivia with remote sensing measurements through simultaneous ASTER satellite

overflights during the summer of 2003. An advanced probabilistic spectral mixture model

(referred to as AutoMCUc) was used to derive per-pixel fractional cover estimates of

photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and soil. These

results were compared with the normalized difference in vegetation index (NDVI) and

field-derived GIS maps of felling gaps and skid trails.

We found that NDVI, PV, NPV, and soil fractions were useful for identifying

felling gaps > 400 m2 for up to 6 months after logging, and for identifying felling gaps

< 400 m2 for up to 3 months after logging; but they were not useful for identifying skid

trails. The PV fraction was most sensitive to felling gaps. The NPV and soil fractions

were both highly correlated with topographic shade and were thus less useful for









monitoring forest disturbances, especially in areas with more pronounced relief. These

results identify important spatial and temporal thresholds relevant to monitoring selective

logging with remote sensing; and may be used in the development of automated

programs for identifying selectively logged forests in the region.














CHAPTER 1
INTRODUCTION AND LITERATURE REVIEW

We examined post-harvest recovery of forest structure and spectral characteristics

in felling gaps and skid trails after selective logging in a forest concession located in the

Department of Santa Cruz, Bolivia. Here, is a brief overview of the forest sector in

Bolivia; followed by a description of the ecological impacts of selective logging and

recent approaches for monitoring selective logging in the tropics using analysis of

remotely sensed imagery.

1.1 Deforestation, Forest Degradation, and Logging in Bolivia

There are about 40 million ha of Amazonian lowland tropical forest in Bolivia

(Steininger et al., 2001a) and over 53 million hectares (nearly 48% of the national

territory) of forest cover in the country as a whole (Rodriguez, 2001). Recently, Bolivian

lowland forests were placed among the top global conservation priorities (Myers et al.,

2000; Steininger et al., 2001b) because of their high diversity of flora (Gentry, 1995) and

fauna (Armonia., 1995; Stotz et al., 1996), and their abundance of habitat types (Prado

and Gibbs, 1993; Killeen et al., 1998). Over 316 species of mammals, including 16

species of primates (Ergueta and Sarmiento, 1992); 1274 species of birds (Armonia.,

1995); and over 20,000 species of flowering plants, including more than 2000 species of

trees and shrubs (Malleux, 2000), have been discovered so far in Bolivian forests.

Historically, deforestation rates in the Bolivian lowlands were low, with only 2.4

million ha (or 5.6%) of the original forested area cleared by 1990 (CUMAT, 1992).

Recently, deforestation has accelerated from 80,000 ha (or 0.2% of the forested area)









per year in the late-1980s, to more than 270,000 ha per year through the mid-1990s

(MDSMA, 1995; Rodriguez, 2001; Pacheco, 2002). Deforestation in the Amazon portion

of the Department of Santa Cruz alone (which covers approximately 61% of the entire

department) has increased dramatically from 38,000 ha per year cleared annually between

1986 to 1990 (CUMAT, 1992) to more than 200,000 ha per year in the mid-1990s

(Camacho et al., 2001; Kaimowitz et al., 2002; Pacheco, 2002), partly as a result of

increased colonization and expansion of cattle ranching (Pacheco, 2002) and soybean

production (Kaimowitz, Thiele, and Pacheco, 1999; Kaimowitz and Smith, 2001).

Analysis of the spatial patterns of deforestation in the Department of Santa Cruz, Bolivia,

shows that areas near roads and population centers are most likely to become deforested

(Kaimowitz et al., 2002).

Selective logging, defined as the extraction of timber species (Verissimo et al.,

1995) having the greatest economic value (Uhl, Barreto, and Verissimo, 1997), is also an

important factor in degradation in Bolivia. Previous timber extraction in Bolivia depleted

forests of mahogany (Swietenia macrophylla), oak (Amburana cearensis), cedar (Cedrela

sp.), morado (Macherium sp.), tarara (Centrolobium sp.), and tajibo (Tabebuia sp.)

(CORDECRUZ, 1994).

In 2001, there were 40 million ha of forest designated for permanent forestry

production (Rodriguez, 2001); equal to approximately 77% of the total national forest

cover. Of this area only 8.5 million ha (or approximately 16 % of the total forested area)

have active logging rights (Forestal, 2002). Illegal logging in Bolivia comprises much of

the total selectively logged area. Cordero (2003) reports that of 133 inspections of

logging operations by the Bolivian Superintendencia Forestal, 39% were found to be









illegal with a further 19% legal, but not in compliance with regulations. In addition to

causing extensive forest degradation, illegal logging makes legal operations less

financially competitive; and illegally logged areas are often subsequently converted to

pasture or agriculture, or bum in wildfires.

1.2 Ecological Impacts of Selective Logging

Forest damage resulting from selective logging operations can be divided into the

general categories of: (1) ground area disturbances; (2) residual stand damages; and, (3)

canopy cover reductions (Uhl and Viera, 1989; White, 1994; Johns, Barreto, and Uhl,

1996; Jackson, Fredericksen, and Malcolm, 2002). Ground-area disturbance results from

the construction and use of skid trails, logging roads, and log landings (Nicholson, 1958;

Fox, 1968; Gillman et al., 1985; Jackson, Fredericksen, and Malcolm, 2002; Pereira et

al., 2002), which may result in soil compaction and damage hydrological functions

(Reisinger, Simmons, and Pope, 1988; Jackson, Sturm, and Ward, 2001). Residual stand

damage results from harvesting of trees; which can damage or kill surrounding trees and

vegetation and disturb regeneration. Stand-level disturbances can be inferred from

changes in composition of forest regeneration after logging (Weaks and Creekmore,

1981; King and Chapman, 1983; Uhl and Viera, 1989; Panfil and Gullison, 1998;

Fredericksen and Licona, 2000a; Fredericksen and Licona, 2000b; Fredericksen and

Mostacedo, 2000; Jackson, Fredericksen, and Malcolm, 2002); from residual stand

damage (Nicholson, 1958; Whitman, Brokaw, and Hagan, 1997); or simply from the

overall reduction in basal area (Webb, 1997). Canopy-cover reduction results from felling

of trees, which then causes further damage as they fall and remove other neighbor trees'

canopies through direct impact or liana intercanopy connections. Canopy disturbances,

such as increases in canopy openness (Home and Gwalter, 1982; Crome, Moore, and









Richards, 1992; Pereira et al., 2002) due to selective logging have been insufficiently

studied (Pereira et al., 2002). Though often difficult to distinguish disturbance category

boundaries, they still serve as a simple foundation for comparisons between varying

harvest intensities and silvicultural treatments.

Reduced impact logging (RIL) techniques can significantly minimize forest

damage as compared with conventional logging (CL). The main components of RIL

logging (Putz and Pinard, 1993; Pinard and Putz, 1996; Bertault and Sist, 1997; Uhl,

Barreto, and Verissimo, 1997; Pinard, Putz, and Tay, 2000; Sist, 2000; Pereira et al.,

2002) are

* Inventory and mapping to reduce waste during logging
* Planning of roads, log decks, and skid trails
* Vine cutting prior to harvest
* Planning of extraction, and
* Directional felling and bucking of trunks.

Many of these RIL components were mandated in 1996 when the Bolivian

government implemented Forestry Law #1700, which instituted new legal and regulatory

frameworks for control and monitoring of forestry operations (Griffith, 1999; Alvira,

Putz, and Fredericksen, 2004).

The forest damage resulting from CL and RIL techniques varies widely, and

depends on factors such as basal area removed, minimum cutting diameters, and forest

type (Gullison and Hardner, 1993; Pinard and Putz, 1996; Panfil and Gullison, 1998).

Conventional logging has been shown to damage 10-40% of the living forest biomass

(Uhl et al., 1991); may disrupt ecological processes (Uhl and Viera, 1989), including the

regeneration of commercially valuable species (Fredericksen and Licona, 2000a;

Fredericksen and Mostacedo, 2000); may alter species composition (Johns, 1992;









Fredericksen et al., 1999; Lewis, 2001; Sekercioglu, 2002; Fredericksen and

Fredericksen, 2002); and may affect forest biogeochemical processes (Asner, Keller, and

Silvas, 2004).

1.2.1 Ground Area Disturbances

Ground area disturbances (GAD) commonly result from the construction and use of

log landings, logging roads, skid trails, or impacts from tree felling. Effects of GAD

include increased levels of soil compaction (Whitman, Brokaw, and Hagan, 1997;

Jackson, Fredericksen, and Malcolm, 2002), and altered site hydrology (Asdak et al.,

1998; Fletcher and Muda, 1999; Tague and Band, 2001).

Relationships between harvest intensity and GAD are difficult to establish due to

the large variety of harvesting practices (Gullison and Hardner, 1993). Figure 1-1

illustrates the relationship for data obtained from 25 published articles, of which 14 were

classified as having planned and 11 unplanned logging operations. Ground area

disturbances were less common after reduced-impact logging (RIL) compared to

conventional logging (CL) in Paragominas, Brazil (Johns, Barreto, and Uhl, 1996; Pereira

et al., 2002). Panfil & Gullison (1998) found a strong relationship between increasing

harvest intensities and increasing GAD within the Chimanes Forest, Bolivia, at even the

relatively low harvest intensities of 1 to 6 trees/ha.

Asner et al. (2004) found that between 4.8 and 11.2 % of the ground area was

disturbed after RIL and CL, respectively, logging in an Amazonian forest. Harvest

intensities for the RIL and CL logging in their study were 3 and 6.4 trees/ha, respectively.

Skid trails comprised 2.9 to 8.8% of the parcel following harvest. Jackson et al. (2002)

found that selective logging at 4.35 trees/ha in a tropical forest in Bolivia damaged

approximately 50% of the total area studied.










50
a a Planned Logging
Unplanned Logging
40
"o
a)

S30 -



r 20 A
P O A
(C
10 A
AA At


0

0 5 10 15 20

Harvest intensity
Figure 1-1. Ground area disturbed for varying levels of harvest intensity in planned and
unplanned logging. Data from sources cited in Appendix A.

Of the disturbed area half was in the form of skid trails, roads, and log landings, and half

was in the form of felling gaps (Jackson, Fredericksen, and Malcolm, 2002). Pereira et al.

(2002) studied differences in disturbed area and canopy openness for RIL and CL

techniques between 100 ha plots that were harvested at similar intensities (-3

individuals/ha). The total ground area disturbed was twice as great for CL (8.9 to 11.2%

vs. 4.6 to 4.8% for CL and RIL, respectively) as it was for RIL (Pereira et al., 2002).

1.2.2 Residual Stand Damage

Aspects of residual stand damage, such as tree mortality (Johns, Barreto, and Uhl,

1996; Bertault and Sist, 1997; Webb, 1997; Panfil and Gullison, 1998; Sist et al., 1998;

Sist and Nguyen-The, 2002), alterations in subsequent species composition (Panfil and

Gullison, 1998), and regeneration (Fredericksen and Mostacedo, 2000) have been widely

studied in the tropics. A significant linear relationship (P < 0.05, R2 = 0.96) was found

between increasing percentages of residual stand damage and increasing harvest intensity










for unplanned logging operations for the literature reviewed in Appendix A, although the

sample size (n = 4) was limited.

70
a Planned Logging
60 Unplanned logging


50

E 40 -
(U
10









O 3 6 9 12 15

Harvest intensity
Figure 1-2. Residual stand damage for various levels and harvest intensity in planned and
unplanned logging. Data from sources cited in Appendix A.
-10


A


0 3 6 9 12 15

Harvest intensity
Figure 1-2. Residual stand damage for various levels and harvest intensity in planned and
unplanned logging. Data from sources cited in Appendix A.

1.2.3 Forest Canopy Damage

Reductions in forest canopy cover are strongly related to silvicultural treatments,

such as pre-harvest liana cutting (Putz, 1992; Vidal et al., 1997; Pinard, Putz, and Licona,

1999), as well as to increased harvesting intensity (Gullison and Hardner, 1993; Bertault

and Sist, 1997). Silvicultural interventions can increase or decrease canopy coverage. For

example, tree girdling or poisoning, and the cutting of unmarketable species, can lead to

larger canopy reductions. Other techniques, such as vine cutting to minimize inter-canopy

connections, can reduce the canopy damage per tree harvested for some forest types

(Appanah and Putz, 1984; Vidal et al., 1997).

Panfil & Gullison (1998) found a correlation between increasing harvest intensities

and increased canopy damage, which reached an asymptote at greater harvesting









intensities due to the re-use of previously constructed skid trails and logging roads.

Pereira et al (2002), found canopy openness of 16.5% and 21.9% for two CL blocks as

compared with 4.9% and 10.9% for two blocks harvested with a RIL approach that

included extensive pre-planning of roads, log decks and skid trails, as well as planned

directional felling and vine cutting prior to harvest (Uhl, Barreto, and Verissimo, 1997).

Other studies (Hendrison, 1990; Johns, Barreto, and Uhl, 1996), have found similar

results.

Temporal patterns of recovery following logging operations have been studied

extensively at the residual stand level (Dickinson, Whigham, and Hermann, 2000;

Fredericksen and Pariona, 2002), including alterations in timber regeneration

(Fredericksen and Mostacedo, 2000) and residual stand mortality rates (Sist and Nguyen-

The, 2002). Canopy level damage, however, and its subsequent recovery (i.e. canopy

closure), have been little studied (Pereira et al., 2002). Cannon et al. (1994) assessed

canopy damage and closure for blocks in West Kalimantan, Indonesia. Three sites were

selected that had undergone similar harvest intensities and had been harvested 0.5, 1 and

8 years prior to their study. The overall canopy openness for the sites decreased from

63% to 49% to 21% with increasing time since harvesting.

An analysis of planned and unplanned logging in Appendix A showed a significant

linear relationship (P < 0.05, R2 = 0.72, n = 7) between increasing harvest intensity and

increasing levels of canopy loss (Figure 1-3). No relationship was found for unplanned

logging, though it was obvious that much greater levels of canopy loss occurred at low

harvest intensities. The utility of the relationship between increasing harvest intensity and

increasing canopy loss for remotely monitoring logging operations is discussed in the










following section. Forest canopy damage is markedly reduced in RIL relative to CL

harvests (Howard, Rice, and Gullison, 1996; Johns, Barreto, and Uhl, 1996; Pereira et al.,

2002).

60
A Planned logging
50 Unplaned logging



S30
0

c 20

10





0 5 10 15 20

Harvest intensity
Figure 1-3. Significant relationship between increasing % canopy cover loss and
increasing harvest intensity for planned logging. Data from sources cited in
Appendix A.

1.3 Remote Sensing of Selective Logging

Improving the ability to estimate the extent, and intensity, of selective logging from

remote sensors is essential for accurate modeling of carbon sequestration and release

(Schroeder and Winjom, 1995; Schroeder and Winjum, 1995; Potter, 1999), developing

effective wild-fire control policies (Holdsworth and Uhl, 1997; Nepstad et al., 1999;

Cochrane and Laurance, 2002; Cochrane, 2003), conservation of fauna and flora, and

monitoring logging activities (Keller et al., 2002). Currently, estimates of deforestation

rates in the tropics are based primarily on remote sensing analyses discriminating

between forested and non-forested regions (Skole, 1993; Steininger et al., 2001a; Achard









et al., 2002). Landsat-based measurements of total forest conversion, normally to

agriculture or pasture, are convenient as rapid and cost efficient deforestation estimators

(Skole, 1993; Skole and Tucker, 1993) but are incapable of detecting forest areas

degraded due to selective logging or fire (Stone and Lefebvre, 1998; Nepstad et al.,

1999).

Acquiring accurate estimates of selective logging rates within the tropics has

proven difficult (Stone and Lefebvre, 1998; Asner et al., 2002) because selective logging

damage occurs on a fine spatial grain (Souza and Barreto, 2000; Pereira et al., 2002)

compared with the spatial resolution of commonly available satellite imagery.

Furthermore, the rapid regeneration of pioneer species in logged areas (Dickinson,

Whigham, and Hermann, 2000; Fredericksen and Licona, 2000b; Fredericksen and

Mostacedo, 2000) reduces indicators visible through optical remote sensing within 3 to 5

years (Stone and Lefebvre, 1998). Souza and Barreto (2000) were able to detect only

60% of field-verified logging patios (where logs are brought before being loaded on

trucks for transport to a sawmill), and none that were > 3 years old due to rapid

regeneration of vegetation.

Prevalent remote sensing image analysis techniques used for monitoring of forest

disturbances are active radar (Siegert and Hoffmann, 2000; Siegert et al., 2001), texture

analysis (Stone and Lefebvre, 1998; Asner et al., 2002), and basic relationships between

single radiometric bands and logging disturbances (Asner et al., 2002). Vegetation

indices, such as NDVI, have had limited utility (Jasinski, 1990; Carlson and Ripley,

1997; Stone and Lefebvre, 1998). Among the more successful methods for identifying

logging disturbances are deriving per-pixel % vegetation cover (Todd and Hoffer, 1998)









from linear spectral mixture models (Hall, Shimabukuro, and Huemmrich, 1995; Garcia-

Haro, Gilabert, and Melia, 1996; Cochrane and Souza, 1998; Shimabukuro et al., 1998;

Souza and Barreto, 2000), or probabilistic Monte Carlo versions of linear spectral

mixture models, such as AutoSWIR or AutoMCU (Asner and Lobell, 2000b; Lobell et

al., 2001).

1.3.1 Textural and Single Band Analysis

Textural analysis uses multi-pixel comparisons to enhance or diminish existing

spatial variation (Asner et al., 2002; Debeir et al., 2002). Single band analysis compares

the digital number, radiance, or reflectance from a single sensor band. Vegetation stress,

or lack of typical vegetation spectral response, and the ability to discriminate vegetation

from exposed soil and non-photosynthetic vegetation (residual logging slash) following

selective logging may enhance the ability of remote sensors to identify these areas and

will therefore be discussed in the following sections.

Individual Landsat Thematic Mapper (T.M) band reflectances can correlate with

logging disturbances. Landsat TM bands 1, 2, and 3 (covering the visible spectrum

between 0.45 and 0.69 um) have been helpful in delineating areas of exposed soil

following selective logging (Asner et al., 2002), possibly due to decreases in moisture

content to which these bands are sensitive (Ripple, 1986; Bowman, 1989). Bands 1 and 2

are often avoided in such analyses because they are susceptible to atmospheric aerosol

contamination (Krueger and Fischer, 1994; Asner et al., 2002).

Landsat TM band 3 (red; centered at 0.67 nm) shows vegetation as very dark due to

radiation absorption by foliar chlorophyll (Gaussman, 1977). TM band 4 (near-IR;

centered at 0.83 nm) shows vegetation with high reflectance due to non-linear scattering

of light by foliage, with soils and litter (NPV) having lower reflectance levels (Asner,









1998). These bands has been shown to correlate both positively (Thomas et al., 1971) and

negatively (Penuelas et al., 1993) with vegetation drought stress according to a complex

suite of leaf physiological factors, including variations in leaf area index (LAI) and

greater shadowing from leaves wilting or curling up when exposed to increasing levels of

drought stress (Jackson and Ezra, 1985).

The short wave infrared (SWIR) region of the spectrum, measured by Landsat TM

bands 5 and 7, is a water absorption peak and thus decreasing SWIR reflectance has been

found to correlate with increasing foliar water content (Tucker, 1980; Ripple, 1986;

Bowman, 1989). Drought stress measurements using Landsat TM band 6 (thermal;

centered at 11.45 nm) have focused on the increases in temperature (the thermal

response) of plant foliage suffering water stress compared to the temperature of the

surrounding air (Chuvieco et al., 1999).

Asner et al. (2002) combined field measurements of canopy gap fraction along a

time series with textural and band-by-band analysis of Landsat 7 Enhanced Thematic

Mapper Plus (ETM+) data. Textural analysis was used to enhance post-logging variations

between canopy and gap reflectance. These techniques, textural and single band analysis,

were found sensitive only to high levels of canopy damage (>50% increase in canopy

openness) and temporally limited to within 0.5 years post-harvest. These techniques may

have some potential for broad delineation of very recently logged forests but are not

useful for more detailed analyses of ecological or biogeochemical forest processes (Asner

et al., 2002).

1.3.2 Band Indices

Band combinations, ratios, and indices provide a powerful image analysis tool for

the assessment of moisture content, vegetation stress, and related logging damages









(Cibula, Zetka, and Rickman, 1992; Adegoke and Carleton, 2002; Aparicio et al., 2002).

In the early 1980s it was found that leaf water content and photo-synthetically active

biomass could be monitored through linear combinations of red and IR radiance changes

(Tucker, 1979) and spectrum wavelengths between 1.55 and 1.75 um. Using these

relationships Hunt et el. (1987 & 1989) developed a leaf water content index

incorporating the wavelengths between .76-.90 um and 1.55-1.75 um.

Band indices have been developed to measure vegetation stress and moisture

content (Gilabert et al., 2002). They have been used to estimate chlorophyll content

(Tucker, 1979) and photosynthesis rates (Choudhury, 1987), primary productivity

(Curren, 1980), susceptibility to wild fire (Chuvieco et al., 1999; Chuvieco et al., 2002)

and leaf aging, drop and stress (Bohlman, Adams, and Peterson, 1998) in the Amazon.

Indices, such as the soil and atmosphere resistant vegetation index (SARVI) (Huete et al.,

1997), water deficit index (WDI) (Moran et al., 1994) or equivalent water thickness

(EWI) (Ceccato et al., 2001; Ceccato, Flasse, and Gregoire, 2002; Ceccato et al., 2002),

have been developed to estimate vegetation stress and water content with good accuracy.

The ratio of TM bands 4 to 5 has been found to be strongly indicative of changes in

leaf water content (Hunt, Rock, and Nobel, 1987; Hunt and Rock, 1989). This is because

decreases in leaf water content increase reflectance in the middle infrared spectrum while

having little effect on reflectance in the near infrared spectra (TM band 5 and 4,

respectively) (Knipling, 1970; Carter, 1991; Aldakheel and Danson, 1997).

Rock et al. (1994) found that, for an area with 100% vegetation cover, TM band 5

reflectance increased with increasing water stress with no change in the reflectance of

TM band 4. Bohlman and Adams (1998) used the TM band 4 to 5 ratio to determine leaf









aging, leaf drop and water stress during the transition from wet season to dry season for

forests in Maraba, Brazil.

Several studies have noted the utility of the short wave infrared (SWIR 2 region,

2080-2280 nm) for remote measurement of leaf moisture content and discrimination of

vegetation from soil and non-photosynthetic vegetation (i.e. slash or litter) (Asner, 1998;

Asner and Lobell, 2000a, b; Lobell et al., 2001) due to the dominance of water absorption

by green plant spectra (Elvidge, 1990; Drake, Mackin, and Settle, 1999).

1.3.3 Linear Spectral Mixture Model (LSMM) Analysis

Linear spectral un-mixing (Heinz, Chang, and Althouse, 1999; Heinz, 2001) or

AutoSWIR (Asner and Lobell, 2000b) techniques decompose pixels to associated

fractions of multiple selected materials of interest (MOI), termed endmembers. These

materials of interest are chosen according to the ecological properties of the field location

(as general as vegetation and soil or as specific as individual species if spectral

seperability is adequate) and the desired resultant data product (Adams et al., 1995;

Garcia-Haro, Gilabert, and Melia, 1999; Bateson, Asner, and Wessman, 2000).

For successful linear un-mixing of pixel reflectance materials of interest must be

chosen that exhibit purity or extremity within the dataset and are known to contribute to

pixel reflectance over the entire landscape of study (Schanzer, 1993; Bateson, Asner, and

Wessman, 1998; Bateson, Asner, and Wessman, 2000; Heinz, 2001). Sub-pixel fractions,

with appropriate endmembers, will sum to 100% pixel reflectance (Asner, Hicke, and

Lobell, 2002). The number of endmembers that can be unmixed from any given pixel is

dependent on the dimensionality (a function of the number of bands and amount of

random noise in the data) of the satellite imagery (Asner and Lobell, 2000b; Asner,

Hicke, and Lobell, 2002).









A least-squares based linear mixing model (constrained to 1) can be simplified to

the following formula (Shimabukuro and Smith, 1991; Asner, Hicke, and Lobell, 2002):

r = (aij xj) + el, where xj = 1 and,
ri= spectral reflectance at the ith spectral band of the pixel;
aij = spectral reflectance known to the jth component at the ith spectral band;
xj = value to be estimated from the proportion of the jth component within the pixel;
el = estimation error for the ith spectral band;
i = number of spectral bands considered;
j = number of components.


The root mean square error fraction can serve as an indicator of how good the chosen

endmembers are for the particular pixel.

A classic study of forested ecosystem reflectance using hyper-spectral AVIRIS

found that 98% of spectral variation was explained by linear mixtures of three

endmembers: photosynthetic vegetation (PV), shade, and soil (Roberts, Smith, and

Adams, 1993), with non-photosynthetic vegetation (NPV) not able to be directly

distinguished from the soil endmember. NPV differences, however, discriminated

through analysis of residual spectra were helpful in distinguishing distinct communities

of green vegetation.

The characteristic spectra of photosynthetic vegetation, dominated by foliar water

absorbance across the spectrum (Elvidge, 1990) and C-H and O-H bonds in the SWIR

region (Curran, 1989), as well as the presence of chlorophyll. NPV, lacking the water

content of PV, and having distinct spectral features resulting from organic carbon bonds

interacting with solar radiation (Curran, 1989), can be distinguished from the more

similar spectra of soil partly through differences in the SWIR response attributable to the

effects of cellulose and lignin in the vegetation (Roberts, Smith, and Adams, 1993) on the

SWIR spectrum. Soil spectral properties vary according to mineralogy, clay content









(Drake, Mackin, and Settle, 1999), and moisture content (Weidong et al., 2002) and

roughness (Pinty, Verstraete, and Gobron, 1998).

Shade spectra have been developed through both inversion of standard linear

mixing models (Roberts, Smith, and Adams, 1993), and sampling of shaded areas in

satellite imagery (Shimabukuro et al., 1998; Souza and Barreto, 2000), among other

methods. Integration of a shade endmember can be helpful to compensate for the effect of

topography, which is caused by differential illumination of the Earth's surface and

generally results in darker slopes facing away from the sun (Civco, 1989), and inter- and

intra-canopy shadowing (Asner and Warner, 2003) which are both prevalent in satellite

imagery. Shade can also be incorporated directly into the endmember bundles (Asner et

al., 2004). Other methods for topographic correction are based on the Lambertian

assumption that measured reflectance does not vary with view angle (Holben and Justice,

1980), which provide poor normalizations and often over-corrects for topography

(Jensen, 1996), or on complicated non-Lambertian models that require development and

validation of image based coefficients (Hodgson and Shelley, 1994).

Linear spectral mixture analysis (SMA) has been used with success to locate areas

of recent fire (Wessman, 1997; Cochrane and Souza, 1998), exposed soil from recent

logging (Souza and Barreto, 2000), differentiate forested from non-forested areas through

differences in shade fractions (Shimabukuro et al., 1998), estimate Amazonian transition

forest biomass (Santos et al., 1999), and assess general land-cover change in tropical

Amazonian forests (Adams et al., 1995). Souza and Barreto (2000) (Souza and Barreto,

2000) used LSMA to detect selectively logged Amazonian forest based upon sub pixel

soil fractions. The study sites had 5 to 7 trees harvested per ha (Johns, Barreto, and Uhl,









1996) for planned and unplanned selective logging. Souza and Barreto (2000) chose

pixels having 20% or greater soil fraction as indicators of patios used for selective

logging. This technique was temporally limited due to rapid regeneration of pioneer

vegetation over exposed soil areas and was unable to locate sites five years post-logging.

Adams et al. (1995) divided an Amazonian landscape into seven general categories

based upon their LSMA endmember fractions. Primary forest areas visibly differed from

those with large quantities of slash (i.e., areas recently selectively logged) due to higher

fractions of NPV and soil, and reductions in the shade fraction. The temporal variations

in the abundance of each endmember were used to assess landscape cover change.

Asner et al. (2004) estimated per-pixel fractional cover of photosynthetic (PV), and

non-photosynthetic (NPV) vegetation, and soil in Amazonian forests near Paragominas

Brazil using an automated un-mixing model, termed AutoMCUC, incorporating

endmember bundles. Endmember bundles are field derived reflectances (measured with a

field spectroradiometer) of materials that encompass the full naturally occurring

variability within the endmember class (for example, a specific soil at different moisture

levels) (Bateson et al., 1998). Pixel un-mixing using endmember bundles allows for

higher accuracy levels of sub-pixel fractional compositions and confidence interval

estimates for those fractions (Asner and Lobell, 2000b). Asner et al. (2004) found

significant differences between conventional (CL) and reduced-impact logging (RIL) PV,

NPV, and soil endmember fractions that varied strongly with time since harvesting due to

gap regeneration and canopy closure.

Asner et al. (2002), employing the AutoMCUc unmixing technique and

endmember bundles was able to discriminate selectively logged areas in the eastern









Amazonia for up to 3.5 years post logging. Canopy openness was found to be greater

following conventional logging than reduced-impact logging immediately. Subsequent

satellite and field-based measurements of canopy gap fraction were highly and inversely

correlated. The 50% decrease in canopy openness derived from the unmixing process

agreed well with ground based measurements. This technique seems the most promising

but has yet to be studied for the gradient of harvest intensities and variety of forest types

necessary for broader application in Amazon forests.

In November 2003 a special issue of Remote Sensing of the Environment was

published dedicated to remote sensing analyses of land use and land cover change,

including selective logging, in the Brazilian Amazon. Five of the 13 papers in this issue

featured linear spectral mixture model methodology. Numata et al. (2003) assessed sub-

pixel fractional cover of green vegetation, shade, soil, and non-photosynthetic vegetation

within pastures and found that they were dominated by NPV, whose dominance increased

with increasing pasture age. Dengsheng, Lu et al. (2003) found that a LSMM approach

was a promising method for distinguishing successional and mature forests, and that sub-

pixel percentages of green vegetation and shade were the most sensitive to changes in

forest structure. A study by Souza et al (2003) analyzing sub-pixel fractions PV, NPV,

soil, and shade found that NPV was positively correlated with aboveground biomass and

improved the ability to map selectively logged forests. A decision tree approach

dichotomouss categorization) of the sub-pixel fractions was used to then successfully

delineate between intact, logged, and regenerating forests (Souza et al., 2003). Asner et

al. (2003) used the AutoMCUc approach (Monte Carlo approach incorporating

endmember bundles) to unmix Landsat Thematic Mapper pixels in areas bordering the









Tapajos National Forest in the Central Amazon and found that PV and NPV fractions

were useful for quantifying biophysical variability within and between pixels.

1.3.4 Influences of Topography and Seasonality on Spectral Response

The topographic effect is caused by differential solar illumination of the Earth's

surface, which generally results in darker slopes facing away from the sun and brighter

slopes facing the sun (Civco, 1989). The topographic effect is a combination of

* Incident illumination defined as the orientation of the land surface to the sun's rays
* Exitance reflectance defined as the energy reflected as a result of the slope, and
* Land topography and shadowing (ERDAS, 1999).

Together these factors can cause identical land cover to be represented by different

intensity values depending on the degree of shadowing. An ideal topographic

normalization removes all intensity variation resulting from differential illumination,

creating a pseudo flat reflectance surface.

It is possible to correct for the topographic effect using either Lambertian or non-

Lambertian reflectance models or through the use of mixture model techniques

employing shade endmembers (Souza and Barreto, 2000). The Lambertian model

normalizes imagery according to the cosine of the sun illumination angle (zenith) at the

time of image acquisition and the slope/aspect information from the digital elevation

model (DEM) of the area (Smith, Lin, and Ranson, 1980). The Lambertian assumption,

that the measured reflectance does not vary with view angle (Holben and Justice, 1980),

provides poor normalizations and often over-corrects images, with sun-facing slopes

appearing darker than those facing away from the sun (Civco, 1989). This results from

not including non-Lambertian scattered reflectance, such as diffuse skylight or light

reflected from surrounding mountainsides (Jensen, 1996).









Minneart and Sceicz (1961) proposed that all surfaces do not reflect incident

radiation uniformly. The non-Lambertian reflectance model compensates for this by

using image based correction factors within the algorithm (Hodgson and Shelley, 1994)

and has been shown to have higher accuracy than Lambertian models (Smith, Lin, and

Ranson, 1980) and fewer problems with over-correction (Civco, 1989). However, the

development of a non-Lambertian model is time consuming and often requires field

truthed data (ERDAS, 1999).

Few studies have closely examined the abilities of a shade fraction or endmember

bundle to remove the effect of topography. However, in general, topography is thought to

have a minimal effect on the response of band indices, such as NDVI, as shade is largely

photometric and results in a general decrease in spectral response which is independent of

the bandwidth. A band ratio (such as NDVI) could therefore compensate for topographic

differences and return topographically independent results.

The effect of seasonality on the response of NDVI has been investigated in north-

west Mexico (Salinas-Zavala, Douglas, and Diaz, 2002), where strong correlations were

found between pluviometric data and atmospheric circulation and changes in NDVI.

Other studies have found correlations between seasonality, including forest phenology,

and the landscape's spectral response (Roberts et al., 1998; Asner and Lobell, 2000b;

Ferreira et al., 2003; Siqueira, Chapman, and McGarragh, 2003).














CHAPTER 2
POST-HARVEST RECOVERY OF FOREST STRUCTURE AND SPECTRAL
PROPERTIES AFTER SELECTIVE LOGGING IN LOWLAND BOLIVIA

2.1 Introduction

Timber production within the Amazon basin has been estimated at 30 million cubic

meters per year, based on regional sawmill production, but estimates of the areal extent

and intensity of the selective logging practices that supply that timber are very poorly

constrained (Nepstad et al., 1999; Lentini, Verissimo, and Sobral, 2003; Nepstad et al.,

2004). Much of the selective logging in the region is clandestine, and in many cases, even

legally registered forest management plans are extremely imprecise. Improving the

ability to estimate the extent, and intensity, of selective logging is essential for

monitoring of logging activities (Keller et al. 2002), and for accurate modeling of carbon

sequestration and release (Schroeder and Winjom, 1995; Schroeder and Winjum, 1995;

Potter, 1999), and developing effective wild-fire control policies (Holdsworth and Uhl,

1997; Nepstad et al., 1999; Cochrane and Laurance, 2002; Cochrane, 2003).

Remote sensing technology may offer an objective means of determining the

location, extent, and intensity of selective logging, but its use for those purposes is

challenging because selective logging damage often occurs on a finer spatial grain than

the spatial resolution of commonly available satellite imagery (Stone and Lefebvre, 1998;

Souza and Barreto, 2000; Asner et al., 2002; Pereira et al., 2002), and forests rapidly

regenerate after logging (Dickinson, Whigham, and Hermann, 2000; Fredericksen and









Licona, 2000b; Fredericksen and Mostacedo, 2000) reducing indicators visible through

optical remote sensing (Stone and Lefebvre, 1998).

In Bolivia, selective logging is an important cause of degradation in the country's

lowland Amazon region (Cordero 2003). Previous timber extraction in Bolivia depleted

forests of mahogany (Swietenia macrophylla), oak (Amburana cearensis), cedar (Cedrela

sp.), morado (Macherium sp.), tarara (Centrolobium sp.), and tajibo (Tabebuia sp.)

(CORDECRUZ, 1994). Although recent changes in the Bolivian Forestry Law provide

an exemplary framework for good forest management (1996; Griffith, 1999), clandestine

and poorly regulated logging activities continue, and the extent and intensity of ongoing

selective logging in the Bolivian Amazon has not been quantified (Cordero, 2003).

This study was designed to examine the potential applicability of remote sensing

technology to the detection of selective logging in the Bolivian Amazon. In it I employ

intensive spatial and temporal field measurements of structural changes associated with

selective logging and then used these measurements to test the sensitivity and examine

the temporal and spatial thresholds of a commonly used remote sensing vegetation index

and an advanced linear spectral unmixing method. The unmixing method has previously

been used for detection of selective logging in a limited number of locations in the

Brazilian Amazon, where both standing and harvested volumes are substantially larger

than at the study area I examined in Bolivia (Asner et al., 2002; Asner et al., 2004; Asner,

Keller, and Silvas, 2004). The results of the analysis I present here can inform future

efforts at monitoring the areal extent and spatial distribution of selective logging using

remotely-sensed data.









2.2 Site Description

The study was conducted in the Agroindustria Forestal La Chonta Ltda. timber

concession (150 47' S, 620 55' W) which encompasses 100,000 ha in the Guarayos forest

preserve in the Department of Santa Cruz, Bolivia (Figure 2-1). The topography is

slightly undulating and the vegetation is classified as Subtropical Humid Forest according

to the Holdridge Life Zone System (Holdridge, 1971) and has an average biomass of 73

to 190 Mg/ha (Dauber, Teran, and Guzman, 2000). The elevation is 400 to 600 m above

sea level, otherwise referred to as the Bolivian lowlands. Common canopy trees in the

area, such as Hura crepitans, Ficus boliviana, and Pseudolmedia laevis, are typical of

humid forests within Bolivia (Jackson, Fredericksen, and Malcolm, 2002). The average

annual temperature is 15.3 C and the mean annual precipitation is 1,560 mm, though

77% of the annual precipitation falls between November and April (Appendix B). During

the dry season temperatures often drop to 5 to 10 OC due to Antarctic fronts (Gil, 1998).

The soils are primarily moderately fertile inceptisols, though large areas of black

anthrosols can be found throughout the concession (Calla, 2003; Paz, 2003). The region

is vulnerable to wildfires (CAF, BOLFOR, and Geosystems, 2000), and 30% of the

concession burned in 1995 (Gould et al., 2002).









Research Parels in the La Chonta Forestr) Concession
SO 5 12 M 1W

a i 2 3 4 S


Figure 2-1. Location of research parcels in the La Chonta forestry concession,
Department of Santa Cruz, Bolivia. The parcel boundaries are overlaid on a
RGB composite image of ASTER bands 2, 3, and 1, respectively, from an
image acquired on 30 June 2004.

There are approximately 100 tree species with individuals >20 cm diameter at

breast height (DBH) within La Chonta (Gil, 1997; Gil, 1998). The mean tree density is 88

trees/ha (Alvira, 2002). Eighteen timber species are currently harvested, including Ficus.

sp., Pseudolmedia laevis, Hura crepitans, Ceibapentandra, and Spondias mombin

(BOLFOR, 2000). Average canopy height is 21+1 m (unpublished data) within a non-

logged control parcel.

The concession was previously high-graded for mahogany (Swietenia macrophylla)

(Gil, 1998), but over 60% of the area is considered to be suitable for sustained-yield


104M









timber harvesting (Gil, 1997). The current annual cut is 2400 ha producing a wood

volume of approximately 51,000 m3 (Jackson, Fredericksen, and Malcolm, 2002).

Average harvest intensity is 4.35 trees/ha (or 12.3 m3/ha of wood) (Jackson,

Fredericksen, and Malcolm, 2002). The cutting cycle, as set by forestry law #1700

introduced in 1996, is 30 years (1996; Fredericksen, 2000), though several years are

granted to complete extraction of a given annual block.

La Chonta was certified by the Forest Stewardship Council (FSC) in early 1990 and

abides by certification standards, including implementation of reduced impact logging

(RIL) techniques (Johns, Barreto, and Uhl, 1996; Uhl, Barreto, and Verissimo, 1997;

Nittler and Nash, 1999; Sist, 2000; Pereira et al., 2002)

* Inventory and mapping of trees to be harvested
* Planning of roads, log decks, and skid trails
* Vine cutting prior to harvest when necessary
* Directional felling, and
* Planning of extraction.

Harvesting is based on a 50 cm minimum DBH cutting limit, with the exception of

Hura crepitans and Ficus glabrata that have a minimum DBH of 70 cm (Jackson,

Fredericksen, and Malcolm, 2002). Twenty percent of harvestable trees are left as seed

trees. One year prior to harvesting, crop trees are selected, marked, and mapped, and

some of the lianas in their crowns are cut (Alvira, 2002; Krueger, 2003). Prior to harvest,

skid trails are built every 150 m intervals perpendicular to the main access road (Jackson,

Fredericksen, and Malcolm, 2002). Directional felling of harvested trees minimizes

ecological damage and improves ease of yarding (Krueger, 2003). Caterpillar 518C

skidders equipped with rubber tires and winches with 15 m of steel cable are used to drag










the logs to roadside log decks (Krueger, 2003), where they are loaded on trucks for

transport to the concession's sawmill.

2.3 Methods

Four logged parcels, ranging from 27 to 31 ha and two non-logged 27 hectare

control parcels were used in this study. Two of the logged parcels, and both the control

parcels, were previously established, measured, and mapped by the Instituto Boliviano de

Investigaciones Forestals (IBIF). All four logged parcels were harvested using RIL

harvesting techniques, with harvest intensities varying from 1 to 2 trees per ha (Table

2-1). Each parcel was logged at a different time, either <1, 6, 13 or 19 months prior to the

collection of field data in July, 2003.

Table 2-1. Characteristics of the selectively-logged parcels used in this study
Harvest
Parcel area Total trees Itest
Parcel Intensity
Parce(ha) harvested (tees
(trees/ha)
<1 month
month 29.7 56 1.8
post-harvest

6 months
27.0 27 1.0
post-harvest

13 months
32.0 64 2.0
post-harvest

19 months
28.0 29 1.0
post-harvest

Within the constraints of the current study, it was not possible to measure

replicate parcels for each stage of this selective logging chronosequence. As a result, in a

formal sense, I am unable to extrapolate from the results I report below to all selectively-

logged parcels in the region (Hurlbert, 1984). However, in this study, individual felling

gaps and skid trail segments, rather than the parcels, are used as the units of analysis.

The statistical tests I employ here are inferential and are used to provide an objective

indication of whether significant differences between individual felling gaps, for









example, were related to their location within a given parcel, and/or due to other factors,

such as felling gap size class (Oksanen, 2001). Where parcel is a significant effect, I infer

that the effect is largely a result of the differences in time post-harvest. I argue that this

inference is justified because of the absence of plausible alternatives to explain such

systematic between-parcel differences in individual felling gaps and skid trail segments.

2.3.1 Field Spatial Analyses

Parcel boundaries, skid trails, and felling gaps were geo-located for the < 1 and 6

months post-harvest parcels using a global positioning system (GPS) unit (maintaining

precision <10 m and with a minimum of 5 satellites visible) and entered into a geographic

information system (GIS) (ArcGIS; ESRI, Redlands, California, USA).

Skid trail and stump locations within the 13 and 19 months post-harvest parcels

(450 m by 600 m) were mapped by IBIF. The maps were then geo-rectified using a

minimum of 15 field GPS measurements per parcel. The root mean square error for the

geo-located parcels was consistently < 5 m.

The area of each felling gap was entered into the GIS using field measured azimuth

of fall (adjusted for declination) from the stump and field length and width

measurements. The length of the gap was measured as the longest axis. The width

(minor) axis of the gap was measured perpendicular to the length (major) axis at the 50%

gap length point, and operationally the gap was defined as an oval with these two axes.

Although most felling gaps have more varied shapes, this assumption was sufficiently

accurate for the questions addressed within this study and convenient for integration with

a GIS. These oval polygons were geo-referenced to the previously geo-located stump

locations. Gap edges were defined by 10 m tall vegetation surrounding the ground area

disturbed by the fallen tree or yarding process.









The definition of gap used in this study differs from ecological measurements of

gaps (Brokaw, 1982; Uhl, 1988), which consider only areas with open canopy to be part

of the gap. Because remote sensors are sensitive to ground disturbances occurring below

forest canopies (Asner et al., 2004) I chose to define gaps with reference to the disturbed

ground area, and separately estimate canopy openness within that area. The nature of the

definition of gap used here means that the data I report should not be compared to

measurements of gaps that follow the ecological convention (Brokaw, 1982; Uhl, 1988).

Skid trail width was defined as the distance between the outer edges of the most

widely separated wheel ruts, and the mean width of 172 measurements was used to buffer

the geo-referenced skid trail centerlines to calculate per-parcel skid trail area. Skid trail

area was calculated for a total of 10 parcels, including six additional parcels that had been

mapped previously by IBIF. Relationships between the area of skid trail and harvest

intensity were investigated using a Michaelis-Menten non-linear regression in JMPo

statistical software. The Michaelis-Menten non-linear regression (y = ((1 *x) / (02+x)))

was chosen to model the relationship as a previous study (Panfil and Gullison, 1998)

showed that the total area of skid trails had a positive quadratic relationship with

increasing harvest intensity.

2.3.2 Field Measurements and Analyses

Within the logged parcels, field measurements were made in felling gaps and skid

trails. Felling gaps were classified as: large (> 800 m2), medium (400 to 800 m2), or small

(< 400 m2), and were divided in half to form trunk and crown zones of equal size. All

field measurements were made separately within the two zones. Skid trails were sampled

in 100 m transects along straight sections of the trails. A separate set of measurements

was made in each 10 m segment of the 100 m transects. Additionally, a 50 m X 50 m grid









layout was used to establish measurement points within a 450m X 600 m unlogged

control forest.

Field measurements included cover estimates from 5 m above the ground surface

for: photosynthetic vegetation (PV); non-photosynthetic vegetation (NPV), which

includes trunks, branches and senesced leaves; exposed soil; and a separate estimate of

lianas with green foliage (green foliage of lianas is also included in the PV estimation).

Canopy openness was estimated using a scale of 0 to 1 defined as the proportion of a

standard upward facing hemispherical mirror at 1.5 m height that has a clear view of the

sky (no canopy obstruction). Previous studies have shown that a canopy densiometer has

comparable accuracy to digital or film hemispherical photography (Englund, O'Brien,

and Clark, 2000).Within the logged parcels, additional measurements included the

maximum height of regeneration in felling gaps and skid trails, the height of residual

non-photosynthetic vegetation in felling gaps (excluding the stump), and skid trail width.

Photosynthetic and non-photosynthetic vegetation, exposed soil, and liana cover

were estimated for the entire trunk and crown zones of the felling gaps, and every 10 m

along the skid trails, within a 2-m band perpendicular to the direction of the trail. At the

grid points in the unlogged control forest, these cover estimates were made within a 2 m

diameter circle placed 1 m to the edge of the path that connected the grid points. In the

felling gaps, canopy openness readings were taken in the middle of each zone along the

length axis. For skid trails these readings were taken from the middle of the 2-m bands

described for the cover estimates.

Felling gaps that included more than one felled tree (defined as overlaid gaps and

constituting < 5% of the total gap area) were identified in the GIS and removed prior to









statistical analysis to avoid confounding relationships between field measurements taken

in the trunk and crown felling gap zones. To analyze field data collected within the

individual tree felling gaps, a mixed 3-way analysis of variance (ANOVA-SASC, 2003)

was used to test the main effects of parcel, size class (large, medium, and small), and gap

zone (trunk vs. crown), and their interactions on canopy openness, vegetation height, PV,

NPV, exposed soil, and NPV height in the individual tree felling gaps. For field data

collected within the skid trail segments, one-way ANOVA was used to test for the effect

of parcel on canopy openness, vegetation height, trail width, PV, NPV, and exposed soil.

For both the felling gap and skid trail data, Tukey's and Dunnett's post-hoc tests were

performed to identify significant, pair-wise differences between the four logged parcels,

and between the individual logged parcels and the unlogged control forest parcel,

respectively.

2.3.3 Remote Sensing Measurements and Analyses

Fourteen ASTER (Advanced airborne thermal emission radiometer) satellite

images were obtained of the study area during the summer of 2003. Of these images four

were found to be sufficiently cloud-and error-free for use in this study. These images

were acquired on 13 May 2003, 30 June 2003, 16 July 2003 and 17 August 2003. In

addition to these images a pre-harvest image had been previously acquired on 11 August

2001. These images were obtained in universal transverse mercator (UTM), world

geodetic system (WGS) 1984 datum, zone 20 south projection and preprocessed by

NASA to L2B surface reflectance. The preprocessing compensated for differences in sun

angle / image geometry and atmospheric differences between the images. ASTER surface

reflectance data have been validated to provide surface reflectance within 1% for actual

surface reflectance < 15% and within 7% of actual surface reflectance > 15% (Abrams









and Hook, 2001). Field validation of ASTER imagery, however, indicates that the

absolute radiometric correction are, in general, better than 4% (Thome et al., 1998;

Yamaguchi et al., 2001). These corrections are performed using radiative transfer

calculations with atmospheric aerosol content from outside sources, such as the MODIS

satellite or climatology data (Abrams and Hook, 2001).

The visible-infrared (15 m pixels) images were re-sized to 30 m using aggregate

pixel mean values and co-registered to the short wave infrared (30 m pixels) image, then

layer stacked using nearest neighbor to produce 9-band images. Band 9 was removed

prior to imaging processing due to problems with atmospheric water vapor.

The 30 June 2003 image was chosen as the base image as it had the least cloud

interference. The remaining images were geo-referenced to the base image using a

minimum of 80 image-to-image control points dispersed throughout the image. The RMS

errors for each geo-referencing were < 15 m (or half a pixel). All images were then layer

stacked using the nearest neighbor re-sampling to get absolute pixel overlay. The stacked

multi-date image was then geo-referenced to 95 field GPS ground reference points

(UTM, WGS 84, Zone 20 S) which were acquired during the summer of 2004. The RMS

error was < 15 m in the final warp model. The image was warped using a 1st order

polynomial model with nearest neighbor re-sampling.

Finally the image overlays were visually assessed by flickering between the May,

July and both August images against the 30 June 2003 base image. Systematic off-sets

were observed with the 16 July 2003 image and were corrected through direct adjustment

to the image map reference coordinates. The 4 post-harvest and 1 pre-harvest control

images of the parcels enable a multi-temporal assessment of the sensitivity of the remote









sensing methodology to selective logging at < 1-4, 6-9, 13-16 and 19-22 months post

harvest.

A probabilistic spectral mixture model was used to decompose the ASTER image

per-pixel surface reflectances into sub-pixel estimates of photosynthetic vegetation, non-

photosynthetic vegetation, and exposed soil. Errors in the linear mixing assumption of the

endmembers were shown in the per-pixel RMS error fraction. Development of this model

was based on an automated probabilistic linear spectral unmixing procedure developed

originally for woodland and shrubland ecosystems (Asner and Lobell, 2000a, b) and

recently used for analysis of selective logging impacts in the Brazilian Amazon (Asner et

al., 2004).

I used a general database of photosynthetic and non-photosynthetic vegetation and

soil spectra that had been collected over logged and unlogged sites in South America (G.

Asner, Personal Communication) which were deconvulved to ASTER bandwidths using

published ASTER band response coefficients. Endmember bundles of several hundred

mean spectra were used in the unmixing procedure. The use of endmember bundles,

rather than single endmembers, is a technique to incorporate naturally occurring

endmember spectral variability into the unmixing model (Bateson, Asner, and Wessman,

2000). A separate shade endmember was not included, as shade levels of 0 to 30% were

incorporated into the photosynthetic vegetation endmember bundle to account for

topographic and intra- and inter-crown shadowing which are prevalent within satellite

imagery (Asner and Warner, 2003).

The 4 post-harvest images were corrected for pre-existing differences in

topography and forest structure among the study parcels by subtracting the NDVI and









fractional values of the pre-harvest image from each of the post-harvest images. The

variability between images associated with seasonality and atmospheric differences were

removed by normalizing each logged parcel with the control parcel from the same image

date.

A digital elevation model (DEM) was obtained by request from NASA's Earth

Observing System (EOS). The DEM was produced through stereoscopic comparison of

nadir and side angle data from the 11 August 2001 pre-harvest control ASTER image,

and has been validated to have <= 10 m relative accuracy (vertical) and < 50 m horizontal

error (Abrams and Hook, 2001). The geo-location of the DEM was done through visually

adjusting the DEM (through alterations to the map info reference coordinates) until

shadows in a DEM based shaded relief model (based on the 2001 image from which it

was created) matched up with the shadows in an RGB (bands 2, 3, and 1, respectively)

composite of the 11 August 2001 ASTER image. Lambertian shaded relief images (on

scale of 0-1 total reflectance) were modeled based on the sun elevation and azimuth.

Separate 2-way repeated measures ANOVA s were used for each remote sensing

variable (per-pixel NDVI, and PV, NPV and soil fractions) to test the main effects of

parcel and image date and their interactions for large, medium, and small felling gaps,

and for skid trails. Dunnett's post-hoc tests were performed to identify significant

differences between the large, medium, and small felling gap, and skid trail pixels and

pixels located in the unlogged control parcel. Within the logged parcels, residual forest

pixels (defined as those > 10 m from a felling gap or a skid trail) were used to illustrate

the size of the disturbance effects relative to between-parcel effects when comparing

felling gap and control parcel reflectance. The August 11 2003 image data of the 6










months post-harvest parcel was not used because the parcel had been re-entered for

further extraction during that month. Separate linear regressions were run between NDVI,

PV, NPV, and the soil fraction pixels within the control parcels for the four summer 2003

ASTER images (n = 530) and the Lambertian shaded relief values for those same pixels

to estimate the influence of topographic shade.

The July 16 2003 image was acquired closest to the date of field data collection, so

I used this image to examine the strength of relationships between the field measurements

within the felling gaps and the remote sensing responses of those same felling gaps using

Pearson bivariate correlation analysis within the < 1 month and the 6 months post-harvest

parcels.

2.4 Results

2.4.1 Field Spatial Analyses

Higher harvest intensities within the logged parcels correlated with higher area in

felling gaps. Felling gaps accounted for most of the disturbed area in the logged parcels,

ranging from 4 to 11% of the total parcel area, while skid trails only accounted for a

maximum of 5 % (Table 2-2). The spatial distribution of felling gaps and skid trails is

illustrated in Figure 2-2. Gap size ranged from of 59 m2 to 2200 m2, and gaps > 800 m2

were uncommon (Table 2-3).

Table 2-2. Percent parcel area disturbed by tree fall gaps and skid trails
ParParcel Harvest intensity % of parcel in % of gap area in % of parcel
Parcel
area in ha (trees/ha) felling gaps overlaid gaps in skid trails
<1 month
month 30 1.9 8.7 3.4 4.0
post-harvest
6 months
27 1.0 6.9 3.1 2.4
post-harvest
13 months
32 2.0 10.5 6.7 5.1
post-harvest
19 months
s 28 1.0 4.2 2.9 3.8
post-harvest














0


b o
0


hOc


, 0
QIN


C

t


<1 month post-harvest parcel


13 months post-harvest parcel


Projection UTM
Datum: WGS 84
Zone: 20 South


o 40 80 160 240 32Ri
1 1Me[9[


U


Legend
Stumps
O Gap areas
Skid trails
I Boundary
O~ V


0o








6 months post-harvest parcel












o9 c


19 moths post-harvest parcel
19 months post-harvest parcel


Figure 2-2. Locations of tree fall gaps and skid trails are shown for the logged study
parcels. Maps of the locations of felled trees in the 13- and 19-month post
harvest parcels were provided by IBIF, and were used as base maps for those
parcels.










Table 2-3. Sample size of large, medium, and small felling gaps within the logged parcels
(Appendix C)
Gap size Parcel
<1 month post- 6 months post- 13 months post- 19 months post-
harvest harvest harvest harvest
Small 26 (29) 5 (5) 24 (30) 15 (19)
Medium 27 (38) 13 (15) 14 (26) 5 (9)
Large 3 (3) 8 (9) 6 (9) 1 (1)
In parenthesis is the sample size before removing overlaid gaps


The addition of data from eight other IBIF research parcels shows a clear quadratic

relationship between harvest intensity and the percent of a parcel covered by skid trail

working surfaces (Figure 2-3, root mean square error for the fit Michaelis-Menten model

was 0.53. Estimates of 01 and 02 were 7.40 and 1.22, respectively).


S1 3 4, 5
Hanvct lateflity (Trea per hectare)

Figure 2-3. Michaelis Menten nonlinear model fit over harvest intensity versus % parcel
area affected by skid trails. Data from Table 2-2 and Appendix D.

2.4.2 Field Measurements

2.4.2.1 Post-harvest recovery of forest structure in felling gaps

Results of the 3-way ANOVA testing the main effects of parcel, gap size, and gap

zone, and their interactions, on canopy openness, liana coverage, vegetation height,










photosynthetic (PV) and non-photosynthetic vegetation (NPV), exposed soil, and non-

photosynthetic vegetation (NPV) height are reported in Table 2-4. There were no

significant 3-way interactions. Mean values of those variables for the four logged parcels

and the control forest parcel are provided in Table 2-5. Table 2-6 lists mean values for the

field measurements by felling gap size and Table 2-7 provides mean values of field

variables for trunk and gap zones.

Table 2-4. The F and P values for the main effects of parcel, gap size, and gap zone, and
their interactions for mixed 3-way ANOVAs of variables measured in felling
gaps
Parcel Parcel gap Gap size *
Factors Parcel Gap size Gap zone
gap size zone gap zone
Canopy Openness 14.8*** 8.6** 7.0** 1.5 0.9 4.2*
Liana coverage (%) 5.4** 0.6 31.6*** 0.6 18.8*** 2.3
Vegetation height (m) 11.0*** 3.4* 0.3 1.8 0.5 0.5
Photosynthetic
vegetation (PV) 17.3*** 0.3 13.1** 3.2** 0.1 0.2
coverage (%)
Non-photosynthetic
vegetation (NPV) 7.3** 0.0 44.1*** 0.8 2.8* 0.7
coverage (%)
Soil coverage (%) 5.2** 0.7 43.0*** 1.1 21.9*** 2.4
NPV height (m) 6.5*** 0.9 170.0*** 0.7 8.8*** 5.6***
Asterisks represent significance of main effects and interactions (* = P <0.05, ** =p <0.01 and *** = P
<0.001).


Canopy openness was significantly affected by parcel, size class, and gap zone,

and there was a size class gap zone interaction. Canopy openness within felling gaps

decreased significantly with time post harvest and was significantly greater for all logged

parcels than for the control forest. Canopy openness within felling gaps also increased

with increasing gap size, and trunk zones had a significantly less open canopy than in

crown zones. The size class gap zone interaction reflects that canopy openness was

greater in the crown zone than in trunk zone in large and medium gaps but not in small

gaps (Figure 2-4).











Table 2-5. Mean values of field measurement variables within felling gaps for <1-, 6-,
13-, and 19-months post-harvest parcels. Unlogged control forest values are provided for
comparison.


Factors


Canopy Openness (%)
Liana Coverage (%)
Vegetation Height (m)
Photosynthetic
Vegetation Coverage
(%)
Non-Photosynthetic
Vegetation Coverage
(%)
Exposed Soil Coverage
(%)
NPV height (m)


Parcel mean standardd error)
<1 month post- 6 months post-
harvest harvest
n = 56 n = 26

52.6 (6.2)***a 48.7 (3.3)***a


12.8 (5.4)***ac
0.6 (0.4)***a


13 months post-
harvest
n =44

26.4 (2.7)***bc


19 months
post-harvest
n =21

18.1 (5.7)***c


Control forest

n= 130


3.7 (0.5)


8.0 (3.6)***a 24.4 (2.8)c 27.1 (5.8)c 21.7(2.4)
1.7 (0.3)***b 2.9 (0.2)***c 3.0 (0.5)***c 21.1 (1.0)


31.4 (5.4)***a 58.9 (3.6)***b


49.6 (5.3)***a


36.6 (3.6)*b


13.6 (2.6)***a 3.2 (1.7)b


2.6 (0.3) a


2.6 (0.2) a


70.8 (2.8)c


26.1 (2.7)c

3.1 (1.3)b
1.8 (0.2)b


79.5 (5.8)c


71.2 (1.7)


19.3 (5.7)c 28.2 (1.6)


1.3 (2.8)b
1.1 (0.4)b


0.6 (0.3)
na


Asterisks represent significant differences between treatment parcel felling gap and control forest values
(* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant differences between
felling gap values in the different treatment parcels (Tukey's test, P < 0.05).



Table 2-6. Mean values standardd error) of field measurement variables for all large,
medium, and small felling gaps.


Factors


Gap size and control (+standard error)
Large Medium


n= 18


n =59


Small

n = 70


Canopy Openness (o%)
Liana Coverage (%)
Vegetation Height (m)
Photosynthetic Vegetation
Coverage (%)
Non-Photosynthetic
Vegetation Coverage (%)
Soil Coverage (%)
NPV height (m)


46.8 (6.2)***b
22.5 (5.9) a
2.4 (0.5)***ab

59.2 (5.9) a

32.3 (5.8) a

6.1 (2.8) a
2.3 (0.4) a


36.8 (2.5)***b
15.5 (2.4)***a
2.3 (0.2)***b

59.3 (2.4)***a

33.0 (2.4)***a

5.8 (1.2)***a
2.0 (0.2) a


25.5 (2.3)***a
16.3 (2.5) a
1.6 (0.2)***a

61.8 (2.5)***a

33.4 (2.4)*a

4.0 (1.2)***a
1.8 (0.1) a


Asterisks represent significant differences between treatment parcel felling gap and control forest
values (* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant
differences between felling gap values in the different treatment parcels (Tukey's test, P < 0.05).










Table 2-7. Mean values standardd error) of measured variables for all trunk and canopy
felling gap zones.
Factors Gap zone (+standard error)
Trunk (n = 147) Canopy (n = 147)
Canopy Openness (%) 33.5 (2.7)***a 39.3 (2.5)***b
Liana Coverage (%) 10.4 (2.7)**a 25.7 (2.7)**b
Vegetation Height (m) 2.0 (0.2)***a 2.1 (0.2)***a
Photosynthetic Vegetation Coverage (%) 64.7 (2.6) a 55.6 (2.6)*b
Non-Photosynthetic Vegetation Coverage 24.1 (2.6)**a 41.8 (2.6)*b
(%)
Soil Coverage (%) 10.0 (1.3)***a 0.6 (1.3) b
NPV height (m) 0.8 (0.1) a 3.3 (0.2) b
Asterisks represent significant differences between treatment parcel felling gap and control forest
values (* = P <0.05, ** = P <0.01, *** = P <0.001). Different letters represent significant
differences between felling gap values in the different treatment parcels (Tukey's test, P < 0.05).


Small Medium Large

Gap size class

Figure 2-4. Canopy openness of all felling gaps as affected by the interaction between
gap size and gap zone. Error bars represent standard error of the mean.

Liana coverage was affected by parcel and gap zone effects, as well as the parcel

*gap zone interaction. Liana coverage dropped initially from the < 1 month post-harvest

gaps to the 6 months old gaps, after which it increased dramatically. The < 1- and 6-

months post-harvest parcels were significantly lower than the control forest mean but the









13- and 19-month post-harvest parcels were not (Table 2-5). Crown zones had

significantly greater liana coverage than trunk zones (Table 2-7). The parcel gap zone

interaction reflects that the gap zone differences are not strongly apparent until 13 months

after logging when canopy zone liana % becomes much greater than that in the trunk

zone (Figure 2-5a).

The height of regenerating vegetation was greater in larger gaps, and in parcels

that had more time to regrow following logging (Table 2-5, 2-6). Similarly, the coverage

of PV was significantly affected by parcel and gap zone, as well as the parcel size class

interaction. PV increased with time post-harvest and in the 13- and 19-months post-

harvest plots, PV in the gaps was not significantly different from in the control forest

(Table 2-5). The crown zone had significantly less PV than the trunk zone (Table 2-7).

The interaction of parcel size class showed that small gaps had significantly higher PV

only in the < 1-month post-harvest parcel (Figure 2-6).

NPV was significantly affected by the main effects of parcel and gap zone, and the

interaction of parcel gap zone. NPV decreased with time post-harvest and was

indistinguishable from the control forest in the 13- and 19-month post harvest gaps. The

crown zone had significantly more NPV than the trunk zone (Table 2-7). The parcel *

gap zone interaction revealed consistently higher levels of NPV% in crown zones (versus

trunk zones) that diminished with parcel (Figure 2-5b).







41


50 80
A B

40
6o

30 (
S30
o 8 40




10


0 0




30 6
C D

25 5

a)
20 4




S10 > 2
25
g 20








Time post-harvest Time post-harvest
Trunk zone
Canopy zone

Figure 2-5. The field factors of A) Liana, B) NPV, C) soil coverage, and D) NPV height
as affected by the interaction between parcel and gap zone. Error bars
represent standard error of the mean. See Table 2-3 for sample sizes.

Soil exposure was also significantly affected by the main effects of parcel and gap


zone, and the interaction of parcel gap zone. Soil exposure decreased with time post-

harvest and only the < 1-month post-harvest gaps had significantly more exposed soil

than the control forest. Although there was a trend towards increasing soil exposure with
than the control forest. Although there was a trend towards increasing soil exposure with










gap size it was not statistically significant. Soil exposure was almost twenty times greater

in the trunk zone than in the crown zone. The parcel gap zone interaction reflected that

the differences between gap zones diminished with increasing months post-harvest

(Figure 2-5c).

120
Large gaps
Medium gaps
100- Small gaps


80


.2 60
e-I

U 40-
0
0

S20 -



S oth 6 -oth oh g oh

Time post-harvest

Figure 2-6. Photosynthetic vegetation coverage as affected by the interaction between
parcel and gap size. Error bars represent standard error of the mean. See Table
2-3 for sample sizes.

NPV height was significantly affected by the main effects of parcel and gap zone,

as well as both parcel gap zone, and size class parcel interactions. NPV height

decreased with increasing months post-harvest and was higher in the crown portion of the

gap. The parcel gap zone interaction is a result of decreasing NPV height in the canopy

zone with parcel but NPV height remaining the same in the trunk zone (Figure2-5d). The

gap size gap zone interaction was a result of smaller gaps having decreased differences

in NPV height between the canopy and trunk zones (data not shown).










2.4.2.2 Skid trails

Canopy openness, vegetation height, PV, NPV, and soil exposure were

significantly affected by parcel (P < 0.05). Skid trail PV increased with time post-harvest

whereas skid trail soil exposure decreased; patterns were less consistent for the other

variables (Table 2-8).

Table 2-8. Mean (+standard error) for field factors within skid trails <1, 6, 13 and 19
months post-harvest.
Canopy Veg. height
n openness n (m) n Width (m) PV % NPV % Soil %
<1 months 17.3 )a*** 31 34 ( 5.8 33.6 59.5
post-harvest (16.4)a*** 0.0 (.)a*** 31 3.4 (0a (16.3)a*** (22.8)a (26.1)a***

6 months 10.1 18.5 47.5 32.0
post-harvest (9.5)a** 45 0.7 (.3)a*** 10 3.3 (0.3)a (14.3)b*** (14.8)b** (20.3)b***

13 months
month 15 5.8 (3.8)b 15 0.3 (.3)a*** na 3.4 (0.4)a na na na
post-harvest

19 months 14.0 1.75 51.5 40.8 8.7
post-harvest (14.9)a***** 31 3.9 (0.4)a (16.2)c*** (17.3)b** (14.6)c***

Asterisks represent significant differences between treatment and control parcels (* = P <0.05, ** = P <0.01, *** = P
<0.001).



2.4.3 Remote Sensing

Both NDVI and the soil fraction were significantly negatively correlated with

increasing Lambertian shade levels (P < 0.001) while the NPV fraction was significantly

positively correlated with increasing shade levels (P < 0.001). The PV fraction, however,

was not correlated with shade intensity. Seasonality also affected NDVI, as well as the

sub-pixel fractions (PV, NPV, soil), as illustrated in Figure 2-7. NDVI and PV fractional

values within the control parcels declined steadily from May (early in the dry season) to

mid-August (nearing the end of the dry season). Neither the NPV or soil fractions had

strong correlations with seasonality.




























I I







13 10 311A 1S 3TE\ Vmag cq uIsti1o d ta te l10

ASTER image acquisition date


-.- PV
-0- N PV
---V Soil


ASTER image acquisition date
NSTER image acquisition date


Figure 2-7. Control parcel mean values for NDVI and PV, NPV, and soil fractions versus
image acquisition date. The dry season intensifies from May through August.


-*- ND\I


1.00


0.95


0.90


0.85


0.80


0.75


0.70





1.2




1.0


0.8 )
0.4


^^3b Ju^3


-









2.4.3.1 Post-harvest recovery of spectral characteristics of felling gaps

Felling gaps > 800 m2. Figure 2-8 illustrates the evolution of spectral

characteristics post-harvest for pixels in large felling gaps (Appendix E). Two-way

repeated measures ANOVA revealed significant main effects of parcel for NDVI, PV,

and NPV, and image date for PV and soil. The interaction effect was significant for

NDVI and PV (Appendix E). NDVI was significantly lower than unlogged control pixels

for up to 3 months after logging (P < 0.001), then higher at 6-8 months (P < 0.001) and at

16, and 20-22 months post-logging (P < 0.05). PV was lower for 2 and 3 months post-

harvest (P < 0.001) and higher at 22 months post-logging (P < 0.05). NPV was higher 2

months post-logging (P < 0.05). The soil response had the highest variance.

Felling gaps 400-800 m2. Figure 2-9 illustrates the evolution of spectral

characteristics post-harvest for pixels in medium felling gaps (Appendix E). Two-way

repeated measures ANOVA revealed significant main effects of parcel for NDVI, PV,

and NPV, and image date for NDVI, PV and soil. The interaction effect was significant

for all the variables. NDVI was significantly lower than for unlogged pixels for up to 3

months after logging (P < 0.001), then higher at 15-16 months post-logging (P < 0.01).

PV was lower for 1, 2 and 3 months post-logging (P < 0.01). NPV was higher 1-3 and 8

months post-logging (P < 0.05). Again the soil response had the highest variance.

Felling gaps < 400 m2. Figure 2-10 illustrates post-harvest spectral changes in

small felling gaps (Appendix E). Two-way repeated measures ANOVA revealed

significant main effects of parcel for NDVI, PV, and NPV, and image date for NDVI, PV

and soil. The interaction effect was significant for all the variables. NDVI was

significantly lower than for unlogged pixels only at 3 months post-logging (P < 0.001),








46



but higher at 15 and 16 months post-logging (P < 0.05). PV was lower for 1, 2 and 3


months post-logging (P < 0.01). NPV was higher 1 and 2 months post-logging (P < 0.05).


0.12
0.10
0.08


0.04
0.02
0.00
4-002
-0.04
-0.06
-0.08
-0.10
-0.12




0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12


A)














-- Large gap NDVI
-o- Residual forest NDVI
--------T-- -- I
0 5 10 15 20 25

Months post harvest



C)






-_ -iyo I-J -r^ p






-- Large gap NPV
-o- Residual forest NPV

0 5 10 15 20 25

Months post harvest


0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12




0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12


Figure 2-8. Difference between spectral characteristics of large felling gap and unlogged
control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil. Error bars are
standard errors for the large felling gap pixels; standard errors for the control
pixels were < 0.001 on the y-axis. The differences between the treatment
parcel's residual forest and unlogged control pixels are shown to distinguish
the disturbance effect from any potential effect of between-parcel differences.


B)














--- Large gap PV
-- Residual forest PV

0 5 10 15 20 2Z
Months post harvest



D)














-*- Large gap soil
--- Residual forest soil

0 5 10 15 20 2i
Months post harvest













A)













--- Medium gap NDVI
-o- Residual forest NDVI

0 5 10 15 20 25

Months post harvest



C)













--- Medium gap NPV
-o- Residual forest NPV

0 5 10 15 20 25
Months post harvest


0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12




0.12
0.10
0.08
0.06
0.04
0.02
0-00
-0,02
-0.04
-0.06
-0.08
-0.10
-0.12


B)













--- Medium gap PV
-o- Residual forest PV

0 5 10 15 20 25
Months post harvest



D)













-.- Medium gap soil
--Residual forest soil

0 5 10 15 20 25
Months post harvest


Figure 2-9. Difference between spectral characteristics of medium felling gap and
unlogged control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil.
Error bars are standard errors for the large felling gap pixels; standard errors
for the control pixels were < 0.001 on the y-axis. The differences between the
treatment parcel's residual forest and unlogged control pixels are shown to
distinguish the disturbance effect from any potential effect of between-parcel
differences.












0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12




0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.12


0 5 10 15 20 2!
Months post harvest


C)














-- Small gap NPV
-o- Residual forest NPV

0 5 10 15 20
Months post harvest


0.12
0.10
0.08
0.06
0.04
0.04



-002
-0.04


-0.08
-0.10
-012


Figure 2-10. Difference between spectral characteristics of small felling gap and
unlogged control parcel pixels for A) NDVI, B) PV, C) NPV and D) Soil.
Error bars are standard errors for the large felling gap pixels; standard errors
for the control pixels were < 0.001 on the y-axis. The differences between the
treatment parcel's residual forest and unlogged control pixels are shown to
distinguish the disturbance effect from any potential effect of between-parcel
differences.


2.4.4 Linking Field and Remotely-Sensed Measurements


Pearson bivariate correlations between field and remote sensing measurements of


all felling gaps in the < 1 and 6 months post-harvest parcels are presented in Tables 2-12


A)














--- Small gap NDVI
-o- Residual forest NDVI


0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-012


B)














Small gap PV
-- Residual forest PV

0 5 10 15 20 25
Months post harvest



D)













- Small gap soil
-o- Residual forest soil

0 5 10 15 20 25
Months post harvest










and 2-13, respectively. The significant positive correlations between NDVI and PV show

they respond similarly to forest disturbances for both the < 1 and 6 months post-harvest

parcels, and both are inversely correlated with NPV. Soil reflectance was also inversely

correlated with NPV. Gap area was inversely correlated with NDVI and PV in the < 1-

month post-harvest parcel but only with NDVI in the 6-month post-harvest parcel.

Canopy openness in the crown zone was also inversely correlated with NDVI and PV in

the < 1-month post-harvest parcel. In the 6-month post-harvest parcel, crown zone PV

coverage was inversely correlated with NPV reflectance, which was positively correlated

with NPV coverage. PV coverage in the trunk zone of the < 1 month post-harvest parcel

felling gaps was correlated with NDVI and PV reflectance, whereas NPV coverage was

inversely correlated with NDVI. NPV coverage in the trunk zone of the 6 month post-

harvest parcel was positively correlated with NPV and negatively correlated with soil

reflectance.

Table 2-12. Pearson bivariate correlations between field and remote sensing
measurements of felling gaps in the < 1 month post-harvest parcel.
NDVI PV NPV Soil
NDVI 1
PV 0.736*** 1
NPV -0.450*** -0.612*** 1
Soil ns ns -0.835** 1
Gap area (m2) -.322** -0.344** ns ns
Gap canopy zone
Canopy openness -0.382*** -0.311** ns ns
Vegetation height (m) ns ns ns ns
PV % coverage ns ns ns ns
NPV % coverage ns ns ns ns
Soil % coverage ns ns ns ns
Gap trunk zone
Canopy openness ns ns ns ns
Vegetation height (m) ns ns ns ns
PV% coverage 0.316** 0.265* ns ns
NPV % coverage -0.248* ns ns ns
Soil % coverage ns ns ns ns
Asterisks represent significant correlations (* = P < 0.05, ** = P < 0.01, *** = P < 0.001).










Table 2-13. Pearson bivariate correlations between field and remote sensing
measurements of felling gaps in the 6 months post-harvest parcel.
NDVI PV NPV Soil
NDVI 1
PV 0.779*** 1
NPV -0.526** -0.589*** 1
Soil ns ns -0.748*** 1
Gn area (m2r .338* ns ns ns


Gap canopy zone
Canopy openness ns
Vegetation height (m) ns
PV % coverage ns
NPV % coverage ns
Soil % coverage ns
Gap trunk zone
Canopy openness ns
Vegetation height (m) ns
PV % coverage ns
NPV % coverage ns
Soil % coverage ns
Asterisks represent significant correlations (*


ns ns ns
ns ns ns
ns -0.428** ns
ns 0.444** ns
ns ns ns

ns ns ns
ns ns ns
ns ns ns
ns 0.459** -0.371*
ns ns ns
P < 0.05, ** = P < 0.01, *** = P < 0.001).


2.5 Discussion

Development of effective remote sensing based programs to monitor selective

logging requires an understanding of the spatial and temporal thresholds that constrain

the applicability of remote sensing to the detection of selective logging. Although this

study was conducted in the context of low harvest intensities the dynamics of recovery of

structural and spectral characteristics following selective logging should be applicable to

improving understanding of the signatures of selective logging (e.g. felling gaps and skid

trails) throughout the tropics.

I found that the NDVI and PV, NPV, and soil fractions were useful for identifying

large and medium size tree fall gaps for between 3 and 6 months post-harvest. The PV

fraction had the greatest response within felling gaps and, unlike NDVI, PV was not

affected by topographic shade. The NPV and soil fractions were both highly correlated


*









with topographic shade and were thus less useful for monitoring forest disturbances,

especially in areas with more pronounced relief.

Canopy openness defines the ability of remote sensors to view ground disturbances

indicative of logging activities and, in general, as felling gaps age from < 1 to 19 months,

canopy openness declines from 50 to 60% to less than 20%. Simultaneously, rapid

vegetation growth, reaching nearly 2 to 5 m by 19 months covers over the originally

exposed soil (primarily in the trunk zone), and NPV (primarily in the canopy zone)

causing the relative percentages of PV, NPV and soil to change from 30, 50 and 10,

respectively, immediately following harvest to 80, 20, and 0, respectively, after 19

months. Rapid liana growth, primarily in the crown zone, covers nearly 30 percent of the

entire gap zone in an often dense mat of verdant lianas by 19 months post-harvest. This

rapid reduction in overall canopy openness means that felling gaps become

indistinguishable from the surrounding forest after around 6 months post-harvest. The

process occurs faster for smaller gaps as they begin with less persistent residual NPV and

are characterized by less initial canopy damage.

Soil exposure within felling gaps, and therefore the utility of the soil fraction for

identifying forest disturbances from selective logging, was limited primarily to the trunk

zone. Though exposed soil in open areas is easily discerned from space, the gap trunk

zone has little canopy damage, as compared with the crown zone, and by 6 months post-

harvest canopy openness within the trunk zone is < 5%, due to canopy and vegetation

regeneration. Although skid trails comprised 30 to 60% of the disturbed area, had the

highest exposed soil levels, and had the slowest rates of vegetation recovery, they were

not identifiable with remote sensing because they had little impact on canopy openness.









Working in the Brazilian Amazon, (Asner et al., 2002) showed that single band

and textural analysis techniques were not sensitive to canopy damages from selective

logging that were < 50% of complete canopy coverage. The analytical techniques

assessed in this study show a considerable improvement in sensitivity to lower levels of

canopy damage. Asner et al. (2004), using AutoMCUc derived per-pixel PV, NPV, and

soil fractions of Landsat imagery, showed sensitivity to skid trails and felling gaps which

diminished greatly from 0.5 to 1.5 year post-harvest, due to rapid regeneration of low-

stature pioneer species. Remotely measured canopy openness values (derived from the

PV fraction) of 12, 11, and 11 percent for felling gap pixels and 28, 11, and 12% for skid

trail pixels were measured 0.5, 1.5, and 3.5 years post-harvest, respectively. Different

from my results, Asner et al. (2004) found that felling gap and skid trail PV fraction

remained consistently higher than in non-logged control forest. This may be a result of

the higher harvest intensities in their study, leading to more prolonged canopy damage

than was found in La Chonta.

Few studies have linked remote sensing data directly to selective logging

disturbances through the collection of extensive field data. The results of this study help

to better understand the utility of currently available remote sensing technologies for

monitoring selective logging, as well as identifying limitations that future remote sensors

and image analysis technologies can address. Future efforts will seek to delineate logged

areas based on the differences in reflectance that are apparent in felling gaps for several

months, and on their spatial distribution.




















APPENDIX A
GROUND, STAND, AND CANOPY DAMAGE AFTER SELECTIVE LOGGING


.1,
u "-





relf
ti
1-I



' 0



,I-
.1







4-I
-,i 9 o;







,2-,
'aIt



_1 1 -




vi

'j i d


- S
,71




-i
ji











9: --
-I
-I a

S-a



?Ss



- la

;I"

'!'"


*1
(1

'0

01


a 3




-5 I
S li











I ,
-' 3 criZ 0'













c
9
a $^Is'









^111




t e In










a

c s



p^ M r S *- <


-


- "


-u


s I
e-
g,-2
, a
- -C






-2<
Val


ts.I1














o "




- V




||
-1 .lS
J|^



oa 'p t3









Biliij
1^1'

iij *


03 'bt
e a


s 1
t-

























71 1, SSr-
. ^3 3 .
-1' C- L-V HZ 0m





g c 21a sq-:
5r^ '1^ -a ^ ^


"
en 9T


oNw


C Ci


aj
~s R




E





9 5 B




Ca ~ C
cf 5 J-?



13
w


cc to c
V

-- ,Lg j (






C 2U o' i S
C F.l S


2
C T
C m


Sen
00


Cl


m
00 CS cf '0
- C C= "


5-



r*1
00
C-A CA





C -4





o rt


sC
o a


-** C


2

2 a


U '



tr g3 & Q S


*u" S1










C)
o,


CeS




"3

a5
S o 3i;





t43






P
O fi bT xd

I8s-s

















2 t:
-^ ot" d

fl30-I 05
















o* a o






.S 43





i- K

j R -&
C Si'


-^^" g


h
C3
n

|j








E"


'U4
.r9"

















06


Q! s ^ |
IS' t *f^ ^







go ii M~"
"C0















edx
Si |5 t S
I ~\ IS' d vrI2 ^ 'f j


















x M 0

Izaen (n m EY
00
c -oa












22




2.
r r ES ^, n. '. s ca




















> 04


1A1
'to-p
|>(g d Q!5 2 ?! 8l l

~.

-: C:


"6I b I

3 S.
' ~ *
)s Q *- 2 2-'




B Cl E S 2
~-E U
0 -e



tU | o 0Cl.8 C 0l
-; 09 B 0
i^" c a3" a i i

4i I-l~ S.I a15 ^^sS,
Oui f sa ^ a aig. -,, s


0 Cf


. I



tU ag, g^^' ^J-^
-T a l1 |




C ^ C^ -n?3
g a a a~? gl
u uS 3
001

Cr o T a
i g ipi
IS rJ-fc ~ L CQ OtI COM t ~ '


('C *- F> -o


C I 1 I -s


s ~ ~ II ll ll, li 'E^ 1
0C C f


F-r FC0t a U




















APPENDIX B
MEAN MONTHLY PRECIPITATION IN LA CHONTA


January Feburary March April May June July August September October November December
Month
Precipitation (mm)



Figure B-1. Mean monthly precipitation (mm) measured in La Chonta from 1993-2001.
The pronounced dry season starting in April and lasting through October is
visible. Error bars represent standard errors around the mean.


400


350


300


250

S
200


a. 150


100




















APPENDIX C
DISTRIBUTION OF FELLING GAP AREA SIZES WITHIN THE STUDY PARCELS


dOn


Li
u
S
I 25
0*
= 25


* 2o

01
- 15
0)
e


Gap Area Classes (m2)
Figure C-1. Gaps size classes for this study were small felling gaps < 400 m2, medium
felling gaps 400 m2 to 800 m2, and large felling gaps > 800 m2

















APPENDIX D
SKID TRAIL AREAS AND HARVEST INTENSITIES FOR ALL BOLFOR LONG
TERM SILVICULTURAL RESEARCH PLOTS

Table D-1. Skid trail areas and harvest intensities for all BOLFOR long term silvicultural
research plots


Parcel


6 months post-
harvest
19 months
post-harvest
B1-M
< 1 month
post-harvest
13 months
post-harvest
B3-M
B2-M
B1-I

B3-I

B2-I


Skid trail Parcel # trees Harvest % parcel area
area (ha) area (ha) harvested intensity* in skid trails


0.652

1.062

1.12

1.186

1.63

1.26
1.52
1.51

1.83

1.48


27

27.98

27.23

29.7

32

28.81
29.19
27.27

31.47

27.23


1

1.036

1.13

1.885

2

2.15
2.67
3.11

3.68

4.52


2.41

3.80

4.11

3.99

5.09

4.37
5.21
5.54

5.82

5.44


* Mean trees/ha harvested
















APPENDIX E
SIGNIFICANCE AND F VALUES OF 2-WAY REPEATED MEASURES ANOVAS
OF REMOTE SENSING VARIABLES OF FELLING GAP PIXELS

Table E-1. Significance and F values of 2-way repeated measures ANOVAs of remote
sensing variables of felling gap pixels

Felling Gaps > 800 m2

Effects NDVI PV NPV Soil
Parcel 12.08*** 6.85*** 2.83* NA
Image Date NA 2.79* NA 44.59***
Parcel*Image Date 3.31*** 3.22*** NA NA

Felling Gaps 400 to 800 m2

Effects NDVI PV NPV Soil
Parcel 9.18*** 15.98*** 9.62*** NA
Image Date 2.98* 4.08** NA 197.93***
Parcel*Image Date 6.15*** 7.67*** 2.31** 3.08***

Felling Gaps < 400 m2

Effects NDVI PV NPV Soil
Parcel 4.73*** 6.76*** 5.63*** 3.69**
Image Date 3.44* 3.61* NA 123.72***
Parcel*Image Date 3.02* 2.17* 2.34** 2.59**
Asterisks represent significant effects and interactions (* = P < 0.05, ** = P <
0.01, *** = P < 0.001).







60


Table E-2. Mean differences and P Values for two-way repeated measures ANOVA post-
hoc comparisons (Dunnett's Test) of NDVI, PV, NPV and soil fractions in
large (> 800 m2) felling gaps versus unlogged control parcel pixels.
Months
Parcel Months NDVI PV NPV Soil
post-harvest
< 1 ns ns ns ns
< 1 month
Month 1 -0.057*** ns ns ns
post-harvest 2 -0.052*** -0.079*** 0.082* ns

3 -0.050*** -0.072*** ns ns
6 months 6 0.018*** ns ns 0.058*
post-harvesta 7 0.023*** ns ns ns
8 0.020*** ns ns ns
13 ns ns ns ns
13 months
14 ns ns ns -0.029*
post-harvest
post-harvest 15 ns ns ns ns
16 0.013* ns ns ns
19 ns ns ns ns
19 months
19 months 20 0.037* ns ns ns
post-harvest
post-harvest 21 0.045*** ns ns ns
22 0.035* 0.047* ns ns
Asterisks represent significant differences between treatment and control parcels
(* = P < 0.05, ** = P < 0.01, *** = P < 0.001).
a No August remote sensing data of the 6 months post-harvest parcel was
available as a portion of the parcel was re-logged during the last month of the
study.







61


Table E-3. Mean differences and P Values for two-way repeated measures ANOVA post-
hoc comparisons (Dunnett's Test) of NDVI, PV, NPV and soil fractions in
medium (400 to 800 m2) felling gaps versus unlogged control parcel pixels.

Months
Parcel NDVI PV NPV Soil
post-harvest

< 1 ns ns ns 0.071*
< month 1 -0.016*** -0.031*** 0.041*** ns
post-harvest 2 -0.015*** -0.028** 0.042*** -0.018*

3 -0.016*** -0.027*** 0.017* ns
6 months post- 6 ns ns ns ns
harvest 7 ns ns ns ns
8 ns ns 0.026* ns
13 0.010*** ns ns ns
13 months 14 Ns
14 Ns ns ns ns
post-harvest
post-harvest 15 0.012** ns ns ns

16 ns ns ns ns
19 ns ns ns ns
19 months
Sm 20 ns ns ns ns
post-harvest
post-harvest 21 ns ns ns ns
22 ns ns ns ns
Asterisks represent significant differences between treatment and control parcels (* = P
< 0.05, ** = P < 0.01, *** = P < 0.001).
a No August remote sensing data of the 6 months post-harvest parcel was available as a
portion of the parcel was re-logged during the last month of the study.
















LIST OF REFERENCES


Abrams, M., and S. Hook. 2001. ASTER users handbook Version 1. Jet Propulsion
Laboratory, Pasadena, CA.


Achard, F., H. Eva, H. Stibig, P. Mayaux, J. Gallego, T. Richards, and J. Malingreau.
2002. Determination of deforestation rates of the world's humid tropical forests.
Science 297: 999-1002.


Adams, J. B., D. Sabol, V. Kapos, R. Filho, D. Roberts, M. Smith, and A. Gillespie.
1995. Classification of multispectral images based on fractions of endmembers:
application to land-cover change in the Brazilian Amazon. Remote Sensing of
Environment 52: 137-154.


Adegoke, J. O., and A. M. Carleton. 2002. Relations between soil moisture and satellite
vegetation indices in the US Corn Belt. Journal ofHydrometeorology 3: 395-405.


Aldakheel, Y. Y., and F. M. Danson. 1997. Spectral reflectance of dehydrating leaves:
Measurements and modelling. International Journal of Remote Sensing 18: 3683-
3690.


Alvira, D., E. F. Putz, and T. S. Fredericksen. 2004. Liana loads and post-logging liana
densities after liana cutting in a lowland forest in Bolivia. Forest Ecology and
Management 19: 73-86.


Alvira, D. C. 2002. Liana loads and post-logging liana densities after liana cutting in a
lowland forest in Bolivia, Master's Thesis, University of Florida, Gainesville, FL.


Aparicio, N., D. Villegas, J. L. Araus, J. Casadesus, and C. Royo. 2002. Relationship
between growth traits and spectral vegetation indices in durum wheat. Crop
Science 42: 1547-1555.


Appanah, S., and F. E. Putz. 1984. Climber abundance in virgin dipterocarp vorest and
the effect of pre-felling climber cutting on logging damage. Malaysian Forester
47: 335-342.









Armonia, J. 1995. Lista de las aves de Bolivia. BirdLife International, Santa Cruz,
Bolivia.


Asdak, C., P. G. Jarvis, P. van Gardingen, and A. Fraser. 1998. Rainfall interception loss
in unlogged and logged forest areas of Central Kalimantan, Indonesia. Journal of
Hydrology 206: 237-244.


Asner, G. P. 1998. Biophysical and biochemical sources of variability in canopy
reflectance. Remote Sensing of Environment 64: 234-253.


Asner, G. P., M. M. Bustamante, and A. Townsend. 2003. Scale dependence of
biophysical structure in deforested areas bordering the Tapajos National Forest,
Central Amazon. Remote Sensing of Environment 87: 507-520.


Asner, G. P., J. A. Hicke, and D. Lobell. 2002. Per-pixel analysis of forest structure:
Vegetation indices, spectral mixture analysis and canopy reflectance modeling. In
M. A. Wulder and S. E. Franklin [eds.], Remote sensing of forest environments:
Concepts and case studies, 552. Kluwer Academic Publishers, New York, NY.


Asner, G., M. Keller, and J. N. M. Silvas. 2004. Spatial and temporal dynamics of forest
canopy gaps following selective logging in the eastern Amazon. Global Change
Biology 10: 1-19.


Asner, G., M. Keller, R. Pereira, and J. Zweede. 2002. Remote sensing of selective
logging in Amazonia: Assessing limitations based on detailed field observations,
Landsat ETM+, and textural analysis. Remote sensing of environment 80: 483-
496.


Asner, G., M. Keller, R. Pereira, J. Zweede, and J. Silva. 2004. Canopy Damage and
Recovery After Selective Logging in Amazonia: Field and Satellite Studies.
Ecological Applications 14: 280-298.


Asner, G., and D. Lobell. 2000a. A Biogeophysical Approach for Automated SWIR
Unmixing of Soils and Vegetation. Remote Sensing of Environment 74: 99-112.


2000b. AutoSWIR: A general spectral unmixing algorithmm based on
2000-2400 nm endmember datasets and Monte Carlo analysis. Proceedings of the
9th Annual JPL Airborne Earth Science Workshop.









Asner, G., and A. S. Warner. 2003. Canopy shadow in IKONOS satellite observations of
tropical forests and savannas. Remote Sensing of Environment 87: 521-533.


Bateson, A., G. Asner, and C. Wessman. 2000. Endmember bundles: A new approach to
incorporating endmember variability into spectral mixture analysis. leee
Transactions on Geoscience and Remote Sensing 38: 1083-1094.


Bateson, C., G. Asner, and C. Wessman. 1998. Incorporating endmember variability in
spectral mixture analysis through endmember bundles. Proceedings of the 7th
Annual JPL Airborne Earth Science Workshop 1: 43-52.


Bertault, J. G., and P. Sist. 1997. An experimental comparison of different harvesting
intensities with reduced-impact and conventional logging in East Kalimantan
Indonesia. Forest Ecology and Management 94: 209-218.


Bohlman, S., J. Adams, and D. Peterson. 1998. Seasonal foliage changes in the eastern
Amazon basin detected from Landsat thematic mapper satellite images.
Biotropica 30: 376-391.


BOLFOR. 2000. Study plan: Long-term silvicultural research project (LTSRP) in
Bolivian tropical forests. BOLFOR, Santa Cruz, Bolivia.


Bowman, W. D. 1989. The relationship between leaf water status, gas exchange, and
spectral reflectance in cotton leaves. Remote Sensing of Environment 30: 249-255.


Brokaw, N. V. 1982. The definition of treefall gap and its effect on measures of forest
dynamics. Biotropica 14: 158-160.


CAF, BOLFOR, and Geosystems. 2000. Bolivia: Determinaci6n del daho causado por los
incendios forestales ocurridos en los departamentos de Santa Cruz-Beni en los
meses de Agosto y Septiembre de 1999. BOLFOR, Santa Cruz, Bolivia.


Calla, S. A. 2003. Arquelogia de "La Chonta". BOLFOR, Santa Cruz, Bolivia.


Camacho, O., W. Cordero, I. Martinez, and D. Rojas. 2001. Tasa de Deforestacion del
Departamento de Santa Cruz, Bolivia 1993-2000. BOLFOR and Superintendencia
Forestal.









Carlson, T., and D. Ripley. 1997. On the relation between NDVI, fractional vegetation
cover, and leaf area index. Remote Sensing of Environment 62: 241-255.


Carter, G. A. 1991. Primary and secondary effects of water content on the spectral
reflectance of leaves. American Journal of Botany 78: 916-924.


Ceccato, P., S. Flasse, and J. Gregoire. 2002. Designing a spectral index to estimate
vegetation water content from remote sensing data; Part 2. Validation and
applications. Remote Sensing of Environment 82: 198-207.


Ceccato, P., S. Flasse, S. Tarantola, S. Jacquemoud, and J.-M. Gregoire. 2001. Detecting
vegetation leaf water content using reflectance in the optical domain. Remote
Sensing of Environment 77: 22-33.


Ceccato, P., N. Gobron, S. Flasse, B. Pinty, and S. Tarantola. 2002. Designing a spectral
index to estimate vegetation water content from remote sensing data: Part 1;
Theoretical approach. Remote Sensing of Environment 82: 188-197.


Choudhury, B. J. 1987. Relationship between vegetation indices, radiation absorption,
and net photosynthesis evaluated by a sensitivity analysis. Remote Sensing of
Environment 22: 209-233.


Chuvieco, E., M. Deshayes, N. Stach, D. Cocero, and D. Riano. 1999. Short-term risk:
foliage moisture content estimation from satellite data. In E. Cchuvieco [ed.],
Remote sensing of large wildfires in the European Mediterranean Basin, 238.
Springer, Berlin, Germany.


Chuvieco, E., D. Riano, I. Aguado, and D. Cocero. 2002. Estimation of fuel moisture
content from multitemporal analysis of Landsat Thematic Mapper reflectance
data: applications in fire danger assessment. International Journal of Remote
Sensing 23: 2145-2162.


Cibula, W., E. Zetka, and D. Rickman. 1992. Response of Thematic Mapper bands to
plant water stress. International Journal of Remote Sensing 13: 1869-1880.


Civco, D. 1989. Topographic normalization of Landsat Thematic Mapper digital imagery.
Photogrammetric Engineering and Remote Sensing 55: 1303-1309.


Cochrane, M. A. 2003. Fire science for rainforests. Nature 421: 913-919.









Cochrane, M. A., and W. F. Laurance. 2002. Fire as a large-scale edge effect in
Amazonian forests. Journal of Tropical Ecology 18: 311-325.


Cochrane, M. A., and C. Souza. 1998. Linear mixture model classification of burned
forests in the Eastern Amazon. International Journal of Remote Sensing 19: 3433-
3440.


CORDECRUZ. 1994. Plan de uso del suelo (PLUS), una propuesta para el
aprovechamiento sostenible de nuestros recursos naturales. CORDECRUZ, Santa
Cruz, Bolivia.


Cordero, W. 2003. Control de operaciones forestales con enfasis en la actividad illegal.
Documento Tecnico 120/2003. BOLFOR, Santa Cruz, Bolivia.


Crome, F., L. Moore, and G. Richards. 1992. A study of logging damage in upland
rainforest in north Queensland. Forest Ecology and Management 49: 1-29.


CUMAT. 1992. Desbosque de la Amazonia Bolivia. Centro de Investigaciones de la
Capacidad de Uso Mayor de la Tierra, La Paz, Bolivia.


Curren, P. J. 1980. Multispectral photographic remote sensing of vegetation amount and
productivity. Proceedings of the Fourteenth International Symposium on Remote
Sensing of the Environment, Ann Arbor, MI: 623-637.


Curran, P. J. 1989. Remote sensing of foliar chemistry. Remote Sensing of Environment
30: 271-278.


Dauber, E., J. Teran, and R. Guzman. 2000. Estimaciones de biomasa y carbon en
bosques naturales de Bolivia. Superintendencia Forestal, Santa Cruz, Bolivia.


Debeir, O., I. Van den Steen, P. Latinne, P. Van Ham, and W. Elenore. 2002. Textural
and contextual land-cover classification using single and multiple classifier
systems. Photogrammetric Engineering and Remote Sensing 68: 597-605.


Dengsheng, L., E. Moran, and M. Batistella. 2003. Linear mixture model applied to
Amazonian vegetation classification. Remote Sensing of Environment 87: 456-
469.









Dickinson, M., D. Whigham, and S. Hermann. 2000. Tree regeneration in felling and
natural treefall disturbances in a semideciduous tropical forest in Mexico. Forest
Ecology andManagement 134: 137-151.


DIDF. 1996. Nueva ley forestal, No. 1700, 136. Proyecto de implementation del sistema
departmental de information y difusion forestal, Santa Cruz, Bolivia.


Drake, N., S. Mackin, and J. Settle. 1999. Mapping vegetation, soils and geology in
semiarid shrublands using spectral matching and mixture modeling of SWIR
AVIRIS imagery. Remote Sensing of Environment 68: 12-25.


Elvidge, C. 1990. Visible and infrared reflectance characteristics of dry plant materials.
International Journal of Remote Sensing 12: 1775-1795.


Englund, S., J. O'Brien, and D. Clark. 2000. Evaluation of digital and film hemispherical
photography and spherical densiometry for measuring forest light environments.
Canadian Journal ofForest Resources 30: 1999-2005.


ERDAS., 1999. ERDAS Field Guide. ERDAS, Inc, Atlanta, Georgia.


Ergueta, P., and J. Sarmiento. 1992. Fauna silvestre de Bolivia: diversidad y
conservaci6n. In M. Marconi [ed.], Conservaci6n de la diversidad biol6gica en
Bolivia. Centro de Datos para la Conservaci6n, La Paz, Bolivia.


Ferreira, L. G., H. Yoshioka, A. Huete, AND E. E. Sano. 2003. Seasonal landscape and
spectral vegetation index dynamics in the Brazilian Cerrado: An analysis within
the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). Remote
Sensing ofEnvironment 87: 534-550.


Fletcher, W. K., and J. Muda. 1999. Influence of selective logging and sedimentological
process on geochemistry of tropical rainforest streams. Journal of Geochemical
Exploration 67: 211-222.


Forestal. 2002. Informe annual gestion 2002. Superintendencia Forestal, Santa Cruz,
Bolivia.


Fox, J. 1968. Logging damage and the influence of climber cutting prior to logging in the
lowland dipterocarp forest of Sabah. Malaysian Forestry 33: 326-347.









Fredericksen, T. 2000. Aprovechamiento forestal y conservation de los bosques
tropicales en Bolivia. BOLFOR, Santa Cruz, Bolivia.


Fredericksen, N., and T. Fredericksen. 2002. Terrestrial wildlife response to logging and
wildfire in a Bolivian tropical humid forest. Biodiversity and Conservation 11:
27-38.


Fredericksen, N., T. Fredericksen, B. Flores, AND D. Rumiz. 1999. Wildlife use of
different-sized logging gaps in a tropical dry forest. Tropical Ecology 40: 167-
175.


Fredericksen, T., and J. Licona. 2000a. Invasion of non-commercial tree species after
selection logging in a Bolivian tropical forest. Journal of Sustainable Forestry 11.


2000b. Encroachment of non-commercial tree species after selection logging in a
Bolivian tropical forest. Journal of Sustainable Forestry 11: 213-223.


Fredericksen, T., and B. Mostacedo. 2000. Regeneration of timber species following
selection logging in a Bolivian tropical dry forest. Forest Ecology and
Management 131: 47-55.


Fredericksen, T., and W. Pariona. 2002. Effect of skidder disturbance on commercial tree
regeneration in logging gaps in a Bolivian tropical forest. Forest Ecology and
Management 171: 223-230..


Garcia-Haro, F., M. Gilabert, and J. Melia. 1996. Linear spectral mixture modeling to
estimate vegetation amount from optical spectral data. International Journal of
Remote Sensing 17: 3373-3400.


1999. Extraction of Endmembers from Spectral Mixtures. Remote Sensing of
Environment 68: 237-253.


Gaussman, H. W. 1977. Reflectance of leaf components. Remote Sensing of Environment
6: 1-9.


Gentry, A. H. 1995. Diversity and floristic composition of neotropical dry forests. In H.
A. M. S. H. Bullock., and E. Medina [ed.], Seasonally Dry Tropical Forests, 146-
194. Cambridge University Press, Cambridge, UK.









Gil, P. 1997. Plan de manejo forestal, Agroindustria Forestal La Chonta Ltda., Santa
Cruz, Bolivia. Agroindustria Forestal La Chonta Ltda.


1998. Plan general de manejo forestal Empresa Agroindustrial La Chonta Ltda.
La Chonta, Ltda, Santa Cruz, Bolivia.


Gilabert, M. A., J. Gonzalez-Piqueras, F. J. Garcia-Haro, and J. Melia. 2002. A
generalized soil-adjusted vegetation index. Remote Sensing of Environment 82:
303-310.


Gillman, G. P., D. F. Sinclair, R. Knowlton, and M. G. Keys. 1985. The effect of some
soil chemical properties of the selective logging of a north Queensland rainforest.
Forest Ecology and Management 12: 195-214.


Gould, K., T. S. Fredericksen, F. Morales, D. Kennard, F. E. Putz, B. Mostacedo, and M.
Toledo. 2002. Post-fire tree regeneration in lowland Bolivia: implications for fire
management. Forest Ecology and Management 165: 225-234.


Griffith, J. 1999. Resultados de los tres telleres regionales sobre la consolidation de la ley
forestal 1700. Doc. Tec. 76B. BOLFOR, Santa Cruz, Bolivia.


Gullison, R., and J. Hardner. 1993. The effects of road design and harvest intensity on
forest damage caused by selection logging: empirical results and a simulation
model from the Bosque Chimanes Bolivia. Forest Ecology and Management 59:
1-14.


Hall, F., Y. Shimabukuro, and K. Huemmrich. 1995. Remote sensing of forest
biophysical structure using mixture decomposition and geometric reflectance
methods. Ecological Applications 5: 993-1001.


Heinz, D., C. Chang, and H. Althouse. 1999. Fully constrained least-squares based linear
unmixing. IEEE Transactions on Geoscience and Remote Sensing: 1401.


Heinz, D. C. 2001. Constrained least squares spectral unmixing for subpixel target
detection, classification and quantification in hyperspectral and multispectral
imagery. IEEE Transactions on Geoscience and Remote Sensing 62: 165.









Hendrison, J. 1990. Damage-controlled logging in managed tropical rain forest in
Suriname, Ecology and Management of Tropical Rain forests in Suriname: 4.
Wageningen Agricultural University, The Netherlands.


Hodgson, M., and B. Shelley. 1994. Removing the topographic effect in remotely sensed
imagery. The ERDAS Monitor 6: 4-6.


Holben, B. N., and C. O. Justice. 1980. The topographic effect on spectral response from
nadir-pointing sensors. Photogrammetric Engineering and Remote Sensing 46:
1191-1200.


Holdridge. 1971. Forest environments on tropical life zones: A pilot study. Pergamon
Press, New York, NY.


Holdsworth, A., and C. Uhl. 1997. Fire in Amazonian selectively logged rain forest and
the potential for fire reduction. Ecological Applications 7: 713-725.


Home, R., and J. Gwalter. 1982. The recovery of rainforest overstory following logging.
Australian Forestry 13: 29-44.


Howard, A., R. Rice, and R. Gullison. 1996. Simulated financial returns and selected
environmental impacts from four alternative silvicultural prescriptions applied in
the neotropics: a case study of the Chimanes Forest, Bolivia. Forest Ecology and
Management 89: 43-57.


Huete, A. R., H. Q. Liu, K. Batchily, and L. W. Van. 1997. A comparison of vegetation
indices over a global set of TM image for EOS-MODIS. Remote Sensing of
Environment 59: 440-451.


Hunt, E. R., B. N. Rock, and P. S. Nobel. 1987. Measurement of leaf relative water
content by infrared reflectance. Remote Sensing of Environment 22: 429-435.


Hunt, E. R., and B. N. Rock. 1989. Detection of changes in leaf water content using near-
and middle-infrared reflectances. Remote Sensing of Environment 30: 43-54.


Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments.
Ecological Monographs 54: 187-211.









Jackson, C. R., C. A. Sturm, and J. M. Ward. 2001. Timber harvest impacts on small
headwater stream channels in the coast ranges of Washington. Journal of the
American Water Resources Association 37: 1533-1549.


Jackson, R. D., and C. E. Ezra. 1985. Spectral response of cotton to suddenly induced
water stress. International Journal of Remote Sensing 6: 177-185.


Jackson, S., T. Fredericksen, and J. Malcolm. 2002. Area disturbed and residual stand
damage following logging in a Bolivian tropical forest. Forest Ecology and
Management 166: 271-283.


Jasinski, M. 1990. Sensitivity of the normalized difference vegetation index to subpixel
canopy cover, soil albedo, and pixel scale. Remote Sensing of Environment 32:
169-187.


Jensen, J. 1996. Introductory digital image processing: a remote sensing perspective.
Prentice-Hall, Inc., Upper Saddle River, NJ.


Johns, A. 1992. Vertebrate responses to selective logging: implications for the design of
logging systems. Transactions of the Royal Society of London 335: 437-442.


Johns, J., P. Barreto, and C. Uhl. 1996. Logging damage during planned and unplanned
logging operations in the eastern Amazon. Forest Ecology and Management 89.


Kaimowitz, D., P. Mendez, A. Puntodewo, and J. Vanclay. 2002. Spatial regression
analysis of deforestation in Santa Cruz, Bolivia. In C. H. Wood and R. Porro
[eds.], Deforestation and land use in the Amazon. University Press of Florida,
Gainesville.


Kaimowitz, D., and J. Smith. 2001. Soybean technology and the loss of natural
vegetation in Brazil and Bolivia. In A. Angelsen and D. Kaimowitz [eds.],
Agricultural technologies and tropical deforestation, 195-211. CABI Publishing
and Center for International Forestry Research (CIFOR), Wallingford.


Kaimowitz, D., G. Thiele, and P. Pacheco. 1999. The effects of structural adjustment on
deforestation and forest degradation in lowland Bolivia. World Development 27:
505-520.









Keller, M., G. Asner, N. Silva, and M. Palace. 2002. Sustainability of selective logging of
upland forests in the Brazilian Amazon: Carbon budgets and remote sensing as
tools for evaluation of logging effects. In D. J. Z. e. al. [ed.], Working forests in
the tropics: Conservation through sustainable management? Columbia University
Press, New York.


Killeen, T., A. Jardim, F. Mamani, and R. Nelson. 1998. Diversity, composition and
structure of a tropical semideciduous forest in the Chiquitania region of Santa
Cruz, Bolivia. Journal of Tropical Ecology 14: 803-827.


King, G. C., and W. S. Chapman. 1983. Floristic composition and structure of a rainforest
area 25 yr after logging. Australian Journal of Ecology 8: 415-423.


Knipling, E. B. 1970. Physical and physiological basis for the reflectance of visible and
near-infrared radiation from vegetation. Remote Sensing of Environment 1: 155-
159.


Krueger, O., and J. Fischer. 1994. Correction of aerosol influence in Landsat 5 thematic
mapper data. GeoJournal 32: 61-70.


Krueger, W. 2003. Efectos del marcado de arboles de future cosecha y la planificaci6n de
pistas de arrastre en el aprovechamiento conventional con limits diametricos en
un bosque tropical de Bolivia. BOLFOR, Santa Cruz, Bolivia.


Lentini, M., A. Verissimo, and L. Sobral. 2003. Fatos florestais da Amazonia. Imazon,
Belem, Brazil.


Lewis, O. T. 2001. Effect of experimental selective logging on tropical butterflies.
Conservation Biology 15: 389-400.


Lobell, D. B., G. P. Asner, B. E. Law, and R. N. Treuhaft. 2001. Subpixel canopy cover
estimation of coniferous forests in Oregon using SWIR imaging spectrometry.
Journal of Geophysical Research-Atmospheres 106: 5151-5160.


Malleux, J. 2000. Estado y cambios de la cobertura forestal en la republican de Bolivia
para el FRA, 2000. FAO, Roma, Italia.


MDSMA. 1995. Memoria explicativa. Mapa forestal. Ministerio de Desarrollo Sostenible
y Medio Ambiente., La Paz, Bolivia.









Minneart, J. L., and G. Szeicz. 1961. The reciprocity principle in lunar photometry.
Astrophysics Journal 93: 403-410.


Moran, E., E. Brondizio, P. Mausel, and Y. Wu. 1994. Integrating Amazonian vegetation,
land-use, and satellite data. BioScience 44.


Myers, N., R. A. Mittermeier, C. G. Mittermeirer, G. A. B. da Fonseca, and J. Kent.
2000. Biodiversity hotspots for conservation priorities. Nature 403: 853-858.


Nepstad, D., A. Alencar, A. C. Barros, E. Lima, E. Mendoza, C. A. Ramos, and P.
Lefebvre. 2004. Governing the Amazon timber industry. In D. J. Zarin, J. R. R.
Alavalapati, F. E. Putz, and M. Schmink [eds.], Working forests in the
Neotropics: conservation through sustainable management. Columbia University
Press, New York, New York.


Nepstad, D., A. Verissimo, A. Alencar, C. Nobre, L. Eirivelthon, P. Lefebvre, P.
Schlesinger, C. Potter, P. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks.
1999. Large-scale impoverishment of Amazonian forests by logging and fire.
Nature 398: 505-508.


Nicholson, D. I. 1958. An analysis of logging damage in tropical rainforest North
Borneo. Malaysian Forester 21: 235-245.


Nittler, J., and D. Nash. 1999. The certification model for forestry in Bolivia. Journal of
Forestry 97: 32-36.


Numata, I., J. V. Soares, D. A. Roberts, F. C. Leonidas, A. C. Chadwick, and G. T.
Batista. 2003. Relationships among soil fertility dynamics and remotely sensed
measures across pasture chronosequences in Rondonia, Brazil. Remote Sensing of
the Environment 87: 446-455.


Oksanen, L. 2001. Logic of experiments in ecology: is pseudoreplication a pseudoissue?
OIKOS 94: 27-38.


Pacheco, P. 2002. Deforestation and forest degradation in lowland Bolivia. In C. H.
Wood and R. Porro [eds.], Deforestation and land use in the Amazon, 66-94.
University Press of Florida, Gainesville.









Panfil, S., and R. Gullison. 1998. Short term impacts of experimental timber harvest
intensity on forest structure and composition in the Chimanes Forest, Bolivia.
Forest Ecology and Management 102: 235-243.


Paz, C. 2003. Forest-use history and the soils and vegetation of a lowland forest in
Bolivia. Master's Thesis, University of Florida, Gainesville, FL.


Penuelas, J., I. Filella, C. Biel, L. Serrano, and R. Save. 1993. The reflectance at the 950-
970 nm region as an indicator of plant water status. International Journal of
Remote Sensing 14: 1887-1905.


Pereira, R., J. Zweede, G. Asner, and M. Keller. 2002. Forest canopy damage and
recovery in reduced-impact and conventional selective logging in eastern Para,
Brazil. Forest Ecology andManagement 168: 77-89.


Pinard, M., and F. Putz. 1996. Retaining forest biomass by reducing logging damage.
Biotropica 28: 278-295.


Pinard, M., F. Putz, and J. Tay. 2000. Lessons learned from the implementation of
reduced-impact logging in hilly terrain in Sabah, Malaysia. International Forestry
Review 2: 33-39.


Pinard, M. A., F. Putz, and J. C. Licona. 1999. Tree mortality and vine proliferation
following a wildfire in a subhumid tropical forest in eastern Bolivia. Forest
Ecology and Management 116: 247-252.


Pinty, B., M. Verstraete, and N. Gobron. 1998. The effect of soil anisotropy on the
radiance field emerging from vegetation canopies. Geophysical Research Letters
25: 797-800.


Potter, C. S. 1999. Terrestrial biomass and the effects of deforestation on the global
carbon cycle. BioScience 49: 769-778.


Prado, D. E., and R. J. Gibbs. 1993. Patterns of species distributions in the dry seasonal
forests of South America. Annals of the Missouri Botanical Garden 80: 902-927.


Putz, F., and M. A. Pinard. 1993. Reduced-impact logging as a carbon-offset method.
Conservation Biology 7: 755-757.









Putz, F. 1992. Silvicultural effects of lianas. In F. a. M. Putz, H [ed.], The biology of
vines. Cambridge University Press, Cambridge.


Reisinger, T. W., G. L. Simmons, and P. E. Pope. 1988. The impact of timber harvesting
on soil properties and seedling growth in the south. S.nlhuiw Journal ofApplied
Forestry 12: 58-67.


Ripple, W. J. 1986. Spectral reflectance relationship to leaf water stress.
Photogrammetric Engineering and Remote Sensing 52: 1669-1675.


Roberts, D. A., B. W. Nelson, J. B. Adams, and F. Palmer. 1998. Spectral changes with
leaf aging in Amazon caatinga. Trees Berlin 12: 315-325.


Roberts, D. A., M. O. Smith, and J. Adams. 1993. Green vegetation, non-photosynthetic
vegetation, and soils in AVIRIS data. Remote Sensing of Environment 44: 255-
269.


Rock, B. J., D. L. Williams, D. M. Moss, G. N. Lauten, and M. Kim. 1994. High-spectral
resolution field and laboratory optical reflectance measurements of red spruce and
eastern hemlock needles and branches. Remote sensing of environment 47: 176-
189.


Rodriguez, T. M. 2001. Estado actual del manejo forestal en Bolivia. FAO, Santiago,
Chile.


Salinas-Zavala, C. A., A. V. Douglas, and H. F. Diaz. 2002. Interannual variability of
NDVI in northwest Mexico. Associated climatic mechanisms and ecological
implications. Remote Sensing of Environment 82: 417-430.


Santos, J. R., M. Lacruz, M. Keil, and J. Kramer. 1999. A linear spectral mixture model
to estimate forest and savanna biomass at transition areas in Amazonia. IEEE
Transactions on Geoscience and Remote Sensing.


Schanzer, D. L. 1993. Comments on the least-squared mixing models to generate fraction
images derived for remote sensing multispectral data. IEEE Transactions on
Geosciences and RemoteSensing 3: 747.


Schroeder, P., and J. K. Winjum. 1995. Assessing Brazil's carbon budget: I. Biotic carbon
pools. Forest Ecology and Management 75: 77-86.









Schroeder, P., and J. Winjum. 1995. Assessing Brazil's carbon budget: II. Biotic fluxes
and net carbon balance. Forest Ecology and Management 75: 87-99.


Sekercioglu, C. H. 2002. Effects of forestry practices on vegetation structure and bird
community of Kibale National Park, Uganda. Biological Conservation 107: 229-
240.


Shimabukuro, Y., G. Batista, E. Mello, J. Moreira, and V. Duarte. 1998. Using shade
fraction image segmentation to evaluate deforestation in Landsat Thematic
Mapper images of the Amazon Region. International Journal of Remote Sensing
19: 535-541.


Shimabukuro, Y., and J. Smith, T. Lin, and K. Ranson. 1991. The least-squares mixing
models to generate fraction images derived from remote sensing multispectral
data. IEEE Transactions on Geosciences and Remote Sensing 29: 747.


Siegert, F., and A. A. Hoffmann. 2000. The 1998 forest fires in east Kalimantan
(Indonesia): A quantitative evaluation using high resolution, multitemporal ERS-2
SAR images and NOAA-AVHRR hotspot data. Remote Sensing of Environment
72: 64-77.


Siegert, F., G. Ruecker, A. Hinrichs, and A. A. Hoffmann. 2001. Increased damage from
fires in logged forests during droughts caused by El Nino. Nature 414: 437-440.


Siqueira, P., B. Chapman, and G. McGarragh. 2003. The coregistration, calibration, and
interpretation of multiseason JERS-1 SAR data over South America. Remote
Sensing ofEnvironment 87: 389-403.


Sist, P. 2000. Reduced-impact logging in the tropics: objectives, principles, and impacts.
International Forestry Review 2: 3-10.


Sist, P., T. Nolan, J. Bertault, and D. Dykstra. 1998. Harvesting intensity versus
sustainability in Indonesia. Forest Ecology and Management 108: 251-260.


Sist, P., and N. Nguyen-The. 2002. Logging damage and the subsequent dynamics of a
dipterocarp forest in East Kalimantan (1990-1996). Forest Ecology and
Management 165: 85-103.









Skole, D. 1993. Measurement of deforestation in the Brazilian Amazon using satellite
remote sensing. PhD, University of New Hampshire.


Skole, D., and C. Tucker. 1993. Tropical deforestation and habitat fragmentation in the
Amazon: satellite data from 1978 to 1988. Science 260: 1905-1906.


Smith, J., T. Lin, and K. Ranson. 1980. The Lambertian assumption and Landsat data.
Photogrammetric Engineering and Remote Sensing 46: 1183-1189.


Souza, C., and P. Barreto. 2000. An alternative approach for detecting and monitoring
selectively logged forests in the Amazon. International Journal of Remote
Sensing 21: 173-179.


Souza, C., L. Firestone, L. Moreira Silva, and D. Roberts. 2003. Mapping forest
degradation in the Eastern Amazon from SPOT 4 through spectral mixture
models. Remote Sensing of Environment 87: 494-506.


Steininger, M., C. Tucker, J. Townshend, T. Killeen, A. Desch, V. Bell, and P. Ersts.
2001 a. Tropical deforestation in the Bolivian Amazon. Environmental
Conservation 28: 127-134.


Steininger, M. K., C. J. Tucker, J. R. G. Townshend, T. J. Killeen, A. Desch, V. Bell, and
P. Ersts. 2001b. Tropical deforestation in the Bolivian Amazon, Environmental
Conservation, 127-134.


Stone, T., and P. Lefebvre. 1998. Using multi-temporal satellite data to evaluate selective
logging in Para, Brazil. International Journal of Remote Sensing. Sept. 19: 2517-
2526.


Stotz, D. F., J. W. Fitzpatrick, T. A. Parker, and D. K. Moskovits. 1996. Neotropical
birds: ecology and conservation. University of Chicago Press, Chicago, USA.


Tague, C., and L. Band. 2001. Simulating the impact of road construction and forest
harvesting on hydrologic response. Earth Surface Processes and Landforms 26:
135-151.


Thomas, J. R., L. N. Namken, G. F. Oerther, and R. G. Brown. 1971. Estimating leaf
water content by reflectance measurements. Agronomy Journal 63: 845-847.









Thome, K., K. Arai, H. Simon Hook, H. Kieffer, H. Lang, A. Tsuneo Matsunaga, A. Ono,
F. Palluconi, H. Sakuma, N. Slater, T. Takashima, H. Tonooka, S. Tsuchida, R.
M. Welch, and E. Zalewski. 1998. ASTER preflight and inflight calibration and
the validation of level 2 products. IEEE Transactions on Geoscience andRemote
Sensing 36: 1161-1172.


Todd, S., and R. Hoffer. 1998. Responses of spectral indices to variations in vegetation
cover and soil background. Photogrammetric Engineering and Remote Sensing
64: 915-921.


Tucker, C. 1979. Red and photographic infrared linear combinations for monitoring
vegetation. Remote Sensing of Environment 8: 127-150.


Tucker, C. J. 1980. Remote sensing of leaf water content in near infrared. Remote
Sensing ofEnvironment 10: 23-32.


Uhl, C., P. Barreto, and A. Verissimo. 1997. Natural resource management in the
Brazilian Amazon: An integrated research approach. BioScience 47: 160-168.


Uhl, C., Clark, K., Dezzeo, N. and Magurrino, P., 1988. Vegetation dynamics in
Amazonian treefall gaps. Ecology 69: 751-763.


Uhl, C., A. Verissimo, M. M. Mattos, Z. Brandino, and I. C. G. Vieira. 1991. Social,
economic, and ecological consequences of selective logging in an Amazon
frontier: the case of Tailandia. Forest Ecology and Management 46: 243-273.


Uhl, C., and I. Viera. 1989. Ecological impacts of selective logging in the Brazilian
Amazon: a case study from the Paragominas region of the state of Para.
Biotropica 21: 98-106.


Verissimo, A., P. Barreto, R. Tarifa, and C. Uhl. 1995. Extraction of a high-value natural
resource in Amazonia: the case of mahogany. Forestry Ecology andManagement
72: 39-60.


Vidal, E., J. Johns, J. Gerwing, P. Barreto, and C. Uhl. 1997. Vine management for
reduced-impact logging in eastern Amazonia. Forest Ecology and Management
98: 105-114.









Weaks, T. E., and R. C. Creekmore. 1981. A study of the hepatic flora of an infrequently
timbered forest. Agriculture and Environment 6: 383-393.


Webb, E. 1997. Canopy removal and residual stand damage during controlled selective
logging in lowland swamp forest of northeast Costa Rica. Forest Ecology and
Management 95.


Weidong, L., F. Baret, G. Xingfa, T. Qingxi, Z. Lanfen, and Z. Bing. 2002. Relating soil
surface moisture to reflectance. Remote Sensing of Environment 81: 238-246.


Wessman, C. A., C.A. Bateson, and T.L. Benning. 1997. Detecting fire and grazing
patterns in tall grass prairie using spectral mixture analysis. Ecological
Applications 7: 493-511.


White, L. J. T. 1994. The effects of commercial mechanized selective logging on a
transect in lowland rainforest in Lope Reserve, Gabon. Journal of Tropical
Ecology 10: 313-322.


Whitman, A., N. Brokaw, and J. Hagan. 1997. Forest damage caused by selection logging
of mahogany (Swietenia macrophylla) in northern Belize. Forest Ecology and
Management 92: 87-96.


Yamaguchi, Y., H. Fujisada, H. Tsu, I. Sato, H. Watanabe, M. Kato, M. Kudoh, A. B.
Kahle, and M. Pniel. 2001. ASTER early image evaluation. Advances in Space
Research 28: 69-76.















BIOGRAPHICAL SKETCH

Eben Broadbent, as a young child, lived with his family in Japan and traveled

throughout Eastern Asia. At the age of 12, he moved with his mother and sister to a

solar-powered house, deep in the mountains of Vermont. He earned his Bachelor of

Science degree from the University of Vermont with a major in botany, specializing in

tropical areas, and with a minor in English. During this time, he worked in Costa Rica as

an assistant teacher for a tropical ecology program based in Monteverde; and interned

with botanists working for the Missouri Botanical Garden, looking for rare plant species

within the Bosque Eterno de Los Ninos Rainforest Preserve. His undergraduate thesis

studied niche partitioning among congeneric epiphytes within the Monteverde area cloud

forests.

After graduation, he returned to Costa Rica to conduct research on forest

regeneration and butterfly diversity after natural and anthropogenic forest disturbances in

the Parque Nacional Corcovado. Returning to the US, he began working with the

environmental nonprofit firm of Hudsonia Ltd. mapping areas of biodiversity concern

within Dutchess County, NY, for use by local towns in creating ecologically sensitive

development plans.

In 2001 he began an internship with BOLFOR, Proyecto de Manejo Forestal

Sostenible de Bolivia, in Santa Cruz, Bolivia, identifying tree species within long term

silvicultural research plots in the La Chonta forestry concession. For this thesis (part of

his Master of Science degree in forestry) he returned to the La Chonta concession.






81


He is now working as a remote sensing and GIS technician for the Carnegie

Institution of Washington at Stanford University. He is currently working on a project to

identify selectively logged areas, deforestation, and new roads across the Brazilian

Amazon. He is planning on beginning his doctoral studies in 2005, with the intention of

studying the effects of land-use change on feedbacks among selective logging, wildfire,

and climate change, for different aged deforestation frontiers in the Brazilian Amazon.