1 THE NATURE OF SAVANNAS THAT REPLACE DRY FORESTS IN CHIQUITANIA, BOLIVIA By JOSEPH WILLIAM VELDMAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FO R THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 J oseph W illiam V eldman
3 To Aaron, Ava, and Helen
4 ACKNOWLEDGMENTS development as a scient ist. I will be forever grateful to Hope College Professors Greg Murray, Kathy Winnett Murray, William Mungall, and Mark Northuis who encouraged me to attend graduate school. I also thank my committee members, Emilio Bruna, Michelle Mack, and George Tanner for their suggestions and guidance in both the developmental and final stages of my dissertation. Thanks also to Botany faculty Kaoru Kitajima, Stuart McDaniel, Chris Baraloto, and Jack Ewel for their support. Departmental staff Kathy Jones, Kim Williams, Denise Caswell, Karen Patterson and Tangelyn Mitchell helped me navigate the University of Florida system and kept research funds flowing to me in Bolivia. My parents Jim and Lorraine Veldman, my sister Katie, and dear friend Nathan Geisler gave me the love and support without which this research would mean little. Lin Cassidy and Camila Pizano are my Gainesville family and will forever be my neighbors even if living on different continents. I spent many weekends running at San Felasco with John Engels, Christine Archer, Alfredo Rios, Santiago Espinosa, Franklin Paniagua, Amy Duchelle, Gaby Stocks, Wendy Lin Bartels, and Peter Derby. Since returning to Florida I have been blessed with new friends Jared Westbrook, Lucas Majure, Meribeth Latvis, Nick Miles and Alejandro Berrio, as well as friends from Bolivia now at the University of Florida : Marlene Soriano and Martijn Slot. I have been privileged to share offices with a number of graduate students in the Departments of Botany and Biology. My first years at the University of Florida were spent at McCarty Hall in the office of Fire and Chemical Ecology with Morgan Varner and Matt Palumbo. On the 3 rd floor of McCarty I also had the camaraderie of Geoff Blate, Bill Grauel, Claudia Romero, Skya Murphy, Ana Ele uterio, Paulo Brando, Camila Pizano, Julia Reiskind and
5 Ann Wagner. In Bartram Hall I am fortunate to work with Vincent Mejibe, Drew Silver, Yuan Zhou, Alexander Shenkin, Xavier Haro, and Mathew McConnachie. During time I spent in Ann Arbor, University of Michigan Professor John Vandermeer generously gave me space in his lab; I sincerely appreciate that David Allen, Jahi Chappell, Heidi Lie re, Zach Miller, Doug Jackson, Shalene Jha, Aley Joseph, and Senay Yitbarek shared their office, friendship, and idea s. I am also grateful to University of Michigan Professor Robyn Burnham for sharing her excitement for ecology and inviting me to the LOTS discussion group that includes David Marvin and Becca Sonday. This research was possible because of the hospitality of the Bolivian people I will forever feel that Bolivia is my second home. I owe a huge debt of gratitude to Bonifacio Mostacedo who encouraged me to pursue research in Bolivia. In Bolivia the people of the Instituto Boliviano de Investigacin Forestal (IBIF) gave me friendship, guidance, and support: Bonifacio Mostacedo, Marielos Pea Claros, Zulma Villegas, Marisol Toledo, Lourens Poorter, Geovana Carreo, Juan Carlos Licona, Alfredo Alarcon, Carlos Pinto, Emma Nuez, Laly Dominguez, Bety Flores, Israe l Ibaez, Miguel ngel Chaves, Mayra Maldonado, Jose Luis Santivaez, Vincent Vroomans, Saskia Grootemaat, Ari Martinez, Skya Murphy, and Margo Stoddard. I additionally wish to acknowledge the friendship of Elibeth Paredo, Remedios Calle, Eva Maria Rivero, Gabriela Gutierrez and Javier Chambi I s pent much of my time in Bolivia with Lars Markesteijn, fellow doctoral student and biologist, who shared his friendship and ideas about life and ecology. Field work would not have been impossible without the hel p of field assistants including: Juan Alvarez, Juan Carlos Alvarez, Israel Melgar, the late Juan Pesoa, Juan Pasavare, and Jose Iraipi. In particular I am honored to have worked with Don Juan Alvarez of Concepcion who
6 accompanied me tirelessly for months of field work. Throughout Chiquitania, ranchers and rural villagers gave me a place to camp, food to eat, permission to study their land, and directions. The Colombara f amily deserves special thanks for the use of Estancia Cacarachi and for providing hors es to ride to field sites. Thanks to Don Juan of San Francisco, the residents of San Jose, Don Negro of Concepcion and many others. The portions of this research conducted in selectively logged forest were facilitated by the managers and employees of INPA Parket LTD; in par ticular I thank Paul Rozenboom, William Pariona, and Daniel Velasquez. Don Wilfredo and Don Raul kept my pick up truck functioning between trips to the field. I was fortunate to use the plant collection at the Herbario del Oriente Bolivia no ( Museo de Historia Natural Noel Kempff Mercado ) and thank botanists Fabiana Mamani, Daniel Soto, Moises Mendoza, and Daniel Villarroel. Nelson Joaquin of the Centro de Investigacin Agricola Tropical (CIAT Santa Cruz) helped me learn about tropical fora ge grasses. Financial support for this research came from the Garden Club of America Award in Tropical Botany, a National Science Foundation Doctoral Dissertation Improvement Grant, and a University of Florida Alumni Doctoral Fellowship.
7 TABLE OF CONTENT S page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 9 LIST OF FIGURES ................................ ................................ ................................ ....................... 10 ABSTRACT ................................ ................................ ................................ ................................ ... 12 CHAPTER 1 TROPICAL DRY FOREST REPLACEMENT BY SAVAN NAS ON THE SOUTHERN EDGE OF THE AMAZON BASIN ................................ ................................ 14 Introduction ................................ ................................ ................................ ............................. 15 Materials and Methods ................................ ................................ ................................ ........... 19 Study Area ................................ ................................ ................................ ....................... 19 Satellite Image Analysis ................................ ................................ ................................ .. 19 Savanna Floristic, Fuels, and Soil Analyses ................................ ................................ .... 21 Results ................................ ................................ ................................ ................................ ..... 24 Satellite Image Change Analysis ................................ ................................ ..................... 24 Savanna Floristics ................................ ................................ ................................ ............ 25 Fuel Loads and Fire History ................................ ................................ ............................ 29 Soil Fertility and Grass Species Composition ................................ ................................ 29 Discussion ................................ ................................ ................................ ............................... 30 Conclusion ................................ ................................ ................................ .............................. 34 2 SELECTIVE LOGGING AN D FIRE AS DRIVERS OF ALIEN GRASS INVASION IN A BOLIVIAN TROPIC AL DRY FOREST ................................ ................................ ...... 45 Introduction ................................ ................................ ................................ ............................. 46 Methods ................................ ................................ ................................ ................................ .. 48 Study Site ................................ ................................ ................................ ......................... 48 Grass Species Distributions ................................ ................................ ............................. 49 Alien Grass Responses to Fire ................................ ................................ ......................... 50 Data Analysis ................................ ................................ ................................ ................... 52 Results ................................ ................................ ................................ ................................ ..... 53 Grass Species Distributions ................................ ................................ ............................. 53 Alien Grass Responses to Fire ................................ ................................ ......................... 55 Discussion ................................ ................................ ................................ ............................... 56 Grass Species Distributions ................................ ................................ ............................. 56 Alien Grass Responses to Fire ................................ ................................ ......................... 59 Conclusion s ................................ ................................ ................................ ............................. 60
8 3 LONG DISTANCE DISPERSAL O F INVASIVE GRASSES B Y LOGGING VEHICLES IN A TROPIC AL DRY FOREST ................................ ................................ ...... 66 Introduction ................................ ................................ ................................ ............................. 67 Methods ................................ ................................ ................................ ................................ .. 69 Study Site ................................ ................................ ................................ ......................... 69 Shade House and Seed Addition Experiments ................................ ................................ 70 Seed Rain from Logging Vehicles ................................ ................................ .................. 72 Results ................................ ................................ ................................ ................................ ..... 75 Conditions Required for G rass Establishment ................................ ................................ 75 Seed Dispersal by Logging Vehicles ................................ ................................ ............... 76 Discussion ................................ ................................ ................................ ............................... 77 4 GUADUA PANICULATA (BAMBUSOIDEAE) IN TH E BOLIVIAN CHIQUITAN IA: FIRE ECOLOGY AND A P OTENTIAL NATIVE FORA GE GRASS ................................ 82 Introduction ................................ ................................ ................................ ............................. 83 Methods ................................ ................................ ................................ ................................ .. 85 Biophysical Characteristics of Bamboo Stands ................................ ............................... 86 Forage Production and Nutrient Content ................................ ................................ ......... 86 Data and Analysis ................................ ................................ ................................ ............ 87 Results and Discussion ................................ ................................ ................................ ........... 87 Guadua and Fire ................................ ................................ ................................ .............. 87 Guadua as a Native Forage ................................ ................................ ............................. 9 0 APPENDIX A GRASS SPECIES IN NAT IVE AND DERIVED SAVA NNAS ................................ ........... 99 B TREE SPECIES IN NATI VE AND DERIVED SAVAN NAS ................................ ............ 102 LIST OF REFERENCES ................................ ................................ ................................ ............. 106 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 114
9 LIST OF TABLES Table page 1 1 Tree canopy cover and graminoid cover of training samples (used for the satellite image classification) and native and derived savannas i ncluded in the savanna floristic study.. ................................ ................................ ................................ ................... 35 1 2 Accuracy assessment of classified images used in the 1986 to 2005 change trajectory analysis. ................................ ................................ ................................ .............................. 35 1 3 Area of change trajectory classes from 1986 and 2005. ................................ .................... 36 1 4 Soil variables measured in native and derived savannas listed in increasing order of CCA 1 axis scores relatin g savanna grass species and soil variables. ............................... 37 2 1 Grass cover in five forest microhabitats 1 5 yr after selective logging. ............................ 62 3 1 Alien grass seeds collected from motor vehicles arriving at log landings during selective logging of a seasonally dry Bolivian forest ................................ ........................ 80 4 1 Locations and number of transects for eight site s used to study Guadua in Chiquitania, Bolivia ................................ ................................ ................................ .......... 92 4 2 Biophysical variables measured in Guadua stands in Chiquitania ................................ .... 93 4 3 Com parison of forage (leaf) production and total biomass of Guadua paniculata and cultivated forage grasses in the genus Urochloa (including the former Brachiaria brizantha, B. decumbens, and Panicum maxima) ................................ ............................. 94 4 4 Concentration of protein, phosporous (P), and calcium (Ca) in forage of Guadua paniculata harvested at Estancia Cacarachi, 23 km east of San Ignacio de Velasco, Santa Cruz Bolivia. ................................ ................................ ................................ ............ 95 A 1 Frequency, mean cover and maximum cover for grass species in native and derived savannas. ................................ ................................ ................................ .......................... 100 B 1 Frequency, mean stem density, and mean basal area for tree species in native and derived savannas. ................................ ................................ ................................ ............. 103
10 LIST OF FIGURES Figure page 1 1 Study location in eastern lowland Bolivia. ................................ ................................ ........ 38 1 2 Land cover class distributions based on error corrected area totals for the 1986 (Landsat TM) and 2005 (CBERS 2) classifications. ................................ ......................... 38 1 3 Modified change trajectory map created from classified 1986 (Landsat TM) and 2005 (CBERS 2) images used to identify forest replacing (derived) savan nas and native (old) savannas. ................................ ................................ ................................ ......... 39 1 4 Detrended correspondence analysis of grass cover by native and exotic species in native and derived savannas. ................................ ................................ .............................. 40 1 5 Detrended correspondence analysis of tree species by basal area in native and derived savannas. ................................ ................................ ................................ ............... 41 1 6 Frequency distributions of numbers of native grass and tree species in native and derived savannas. ................................ ................................ ................................ ............... 42 1 7 Proportion of savanna tree spe cies, forests tree species, and palms in native and derived savannas. ................................ ................................ ................................ ............... 43 1 8 Fuel loads produced by graminoids (Poaceae and Cyperaceae), herbs, and shrubs in native and derived savannas. ................................ ................................ .............................. 43 1 9 Canonical correspondence analysis of grass cover by (native and exotic )species and soil variables measured in native and derived savannas. ................................ ................... 44 2 1 Canopy openness, grass cover density and species richness in five microhabitats arranged in order of increasing disturbance by selective logging, 1 5 yr after harvest .... 63 2 2 Grass cover and species richness in relation to tree canopy openness 1 5 yr after selective logging for all habitats combined ................................ ................................ ...... 64 2 3 Responses of alien grass populations and tree canopy c over to an experimental fire in a selectively logged Bolivian tropical dry forest. ................................ .............................. 65 3 1 Survival and total biomass at 1 mo of Urochloa maxima seedlings emerging from 50 seeds sown and grown under 0, 1, 10 or 70% sunlight for 1 mo. ................................ ...... 80 3 2 Emergence of Urochloa maxima seedlings (mean+1SE) 1 mo after planting 200 seeds in 0.2 m 2 plots in a 2 x 2 factorial design combining two habita t treatments (unlogged forests and felling gaps) and two soil treatments (scarified soil and untreated soil ). ................................ ................................ ................................ .................... 81
11 4 1 Relationship between Guadua abundance and tree canopy cover ................................ .... 95 4 2 Relationships between trees, bamboo fuel loads, and fire intensities, that suggest the combination of Guadua and fire influences tree abundance in the Chiquitano dry forest. ................................ ................................ ................................ ................................ 96 4 3 Diagrams that incorporate fire to explain abundances of Guadua and trees in the Chiquitano dry forest. ................................ ................................ ................................ ........ 97 4 4 Relationship between the total stand ing biomass of Guadua and the mass of lea ves suitable as cattle forage. ................................ ................................ ................................ ..... 98
12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements f or the Degree of Doctor of Philosophy THE NATURE OF SAVANNAS THAT REPLACE DRY FORESTS IN CHIQUITANIA, BOLIVIA By J oseph William Veldman May 2010 Chair: Francis E. Putz Major: Botany Climate change, deforestation, agriculture, logging, drought, and f ire are predicted to interact in ways that will render forests of the Amazon Basin susceptible to conversion to savanna but it remains unclear what sort of savanna will replace forest and the identities of the grasses and tree species that will dominate I studied savannas that replace d dry forests in the region of Chiquitania in eastern lowland Bolivia w hich is on the southern rim of the Amazon Basin in a transition zone between forest and savanna biomes. Using satellit e image analysis of a 22,500 km 2 a rea, I identified 1420 km 2 that was transformed from forest to savanna between 1986 and 2005. The savannas that replaced forest were dominated by one or two native or exotic grasses, and were floristically distinct from native savannas. The trees in recent ly derived savannas were primarily thin barked forest species and not fire tolerant species typical of fire maintained savannas. Derived savannas had lower native grass and tree diversity than native savannas, occurred on more fertile soils and supported f uel loads that were twice as large. The majority of derived savannas were dominated by either exotic forage grasses in the genus Urochloa or by Guadua paniculata a native bamboo. Guadua stands were characterized by frequent fires, charred dead trees and large fuel loads (up to 16 Mg/ha) ; clearly fire played a role
13 in their creation O n a positive note, foliar analysis of Guadua indicates its potential use as cattle forage. To better understand the role of land use practices in promoting savanna expansio n into forested areas I studied grasses in a selectively logged forest. Forest invasion by alien grasses was promoted by selecti ve logging and fire through open ing the canopy and creating sites suitable for C 4 grasses, the most abundant of which was U. ma xima ; no native savanna species were found to colonize logged forest. To determine how grass propagules arrive d in the forest, I collected seeds falling from logging trucks and demonstrate d that grass seeds are carried far into the forest by motor vehicle s. This research improves our understanding of how the floristic compositions of forest replacing savannas are influenced by human land use, soil fertility, and seed dispersal. S avanna expansion models could be improved by clarifying the sort of savanna th at will replace forests and the ecosystem level consequences ( e.g., fuel loads) of derived compar ed to native savannas
14 CHAPTER 1 TROPICAL DRY FOREST REPLACEMENT BY SAVAN NAS ON THE SOUTHERN EDGE OF THE AMAZON BASIN Current trends in land use, fire reg imes, and climate change may promote savanna expansion in the Amazon Basin, but what type of savanna s will replace forests and which grass species will be involved remains unclear. I used Landsat TM and CBERS 2 satellite image s to quantify forest replace me nt by savanna between 1986 and 2005 i n 22,500 km 2 of eastern lowland Bolivia. I then sampled savannas derived from forest and native s avannas to compare their grass and tree species compositions flammability, and soil properties Savannas replaced 1420 k m 2 of forest (8.4% of forest in 1986) a rate comparable to deforest ation for agriculture ( 1200 km 2 ). D etrended and canonical correspondence analyses (DCA and CCA, respectively ) showed n ative and derived savannas differed both floristically and in soil fer tility ; native savannas occurred on sandy acidic soils and derived savannas on relatively fertile soils. Whereas native savannas contained a diversity of grass species the derived sav annas were usually dominated by either Guadua paniculata ( a native bambo o) Urochloa brizantha or U. maxima (exotic forage grasses), Imperata brasiliensis (a nat ive invader of swidden plots), Digitaria insularis (a native ruderal), or the native fire adapted herb Hyptis suaveolens (Lamiaceae ) J ust one of 17 derived savanna s r esembled native savannas in grass species composition The trees in derived savannas were not thick barked species characteristic of savannas environments (e.g., Curatella americana ) but forest trees ( e.g., Anadenanthera colombrina ) and fire tolerant palms ( Attalea spp. ). Grasses and herbs in derived savannas on fertile soils produced fuel loads twice as large compared to native savannas. The relationships between soil fertility, grass species composition, and fuel loads could be used to improve savanna exp ansion models, clarify the sorts of savannas that will expand, and emphasize the differences between native and derived savannas.
15 Introduction Vast areas of Amazon forest are expected, or assumed, to be vulnerable to savanna expansion when subjected to th e combined effects of climate change, intensification of land use, and fire. Predictions of savanna expansion are based on climate models that project the Amazon B asin to become too dry to support forests (e.g., Cox et al. 2004) as well as on the effects of logging, road building, and pasture creation on forest flammability and susceptibility to grass invasions (Uhl & Buschbacher 1985; Cochrane et al. 1999 ; Laurance 2003). These factors do not act in isolation but interact as positive feedbacks between de forestation, drought, tree mortality, fire, and atmospheric carbon dioxide concentrations (Nepstad et al. 2004; Golding & Betts 2008). It is these drivers of deforestation in combination with the grass f ire cycle (D'Antonio & Vitousek 1992), and recognitio n of savanna as an alternative stable state (sensu Holling 1973 ; Suding et al 2004) that have le d to expectations of forest replacement by savannas. There remains a great deal of uncertainty about what is meant by savanna expansion, whether savanna will be the ultimate fate of tropical forests subjected to climate change and fire (Mahli et al. 2009) and if so, whether we should expect the expansion of species rich native savannas or dominance by a few invasive grass species. Savanna definitions are compl icated in the Neotropics in part because of the existence of both native savannas and extensive areas dominated by African forage grasses (Daubenmire 1972) Furthermore, the term savanna can encompass a huge range of vegetation types that may have little more in common than graminoid dominated understories and sparse t ree cover (e.g., Putz & Redford 2010). Native savannas, such as those of the cerrado biome of South America are amongst the most species diverse and threatened ecosystems in the world (Oliv iera et al. 2002). C limate shifts have influenced the distribution and extent of native savannas for millennia (Mayle et al. 2004) In contrast, modified fire regimes coupled with the introduction of
16 exotic forage grasses in the Neotropics resulted in the creation of vast areas of physiognomic or in areas that were formerly forests (Parsons 1972). A strictly physiognomic definition of savanna ignores the identities of the graminoids and forbs involved, and whether th ey are native or exotic, naturally regenerated or planted. Nonetheless, there are reasons to consider grass dominated systems collectively, particularly with respect to the increasing frequency of fires and the grass fire cycle (Cochrane et al. 1999). I u broadly on a physiognomic basis to describe areas with a graminoid or herb dominated understory and <65% tree canopy cover in full recognition of the fact that areas fitting this definition can be floristically and functionally very different, even if grasses, fire, and sparse tree cover are common features forest savannas. Studies on the vulnerability of Amazon forests to environmental change have not adequately addressed what sort of savanna w ill replace forest. In describing the possible outcomes of forest degradation in the Amazon Basin authors use terms such as savanna like scrub (Nepstad et al. 1999) or savanna like vegetation (Nepstad et al. 2008), old fields or scrub sa vanna (Cochrane & Schulze 1999), and savanna (Hutyra et al. 2005). Another ill defined term, savannization, has been used to describe both the spread of native savanna (Cavelier et al. 1998) as well as a range of processes that degrade forest (Borhidi, 1988; Nepstad et al. 1999). Complicating matters, studies that assume that the climate currently associated with savannas can predict future savanna distribution imply that the native savanna biome of South America will shift northward as the Amazon B asin dries (e.g., Hutyra et al. 2005 ; Mahli et al. 2009) ; despite this implication, to date there is little evidence that large scale native savanna expansion is occurring
17 Determining what sorts of savannas are expanding and what species are involved is criti cal to understanding ecosystem level consequences of vegetation shifts. Differences between native and exotic grass species can result in substantial differences in their ability to co exist with trees, invade forest, or drive savanna expansion (e.g., Lipp incott 2000 ; Platt & Gottschalk 2001 ; Hoffmann et al. 2004). Tree species identity and natural history are important given that thick barked savanna trees are more fire tolerant than thin barked forest trees (Hoffmann et al. 2009), and that savanna expansi on in areas that lack fire adapted tree species (i.e., much of the Amazon Basin), are unlikely to maintain tree cover important to ecosystem services (i.e. precipitation ; Hoffman et al. 2000). A goal of this study is to determine where savanna has already replaced forest in lowland Bolivia, to identify the grass and tree species involved, and determine the consequences for biodiversity and ecosystem flammability. The seasonally dry forests of eastern lowland Bolivia represent an appropriate study system fo r advancing our understanding of the process of forest conversion to savanna. The region known as Chiquitania contains a mosaic of vegetation types dominated by seasonally dry tropical deciduous forest i nterspersed with native savannas that are western dis juncts of Brazilian C errado s (Killeen et al. 1990). The forests of Chiquitania are transitional between the more humid Amazonian forest to the north and the drier scrub woodlands to the south (Killeen et al. 2005). O n the perip hery of the Amazon Basin t hes e forests and are expected to be particularly vulnerable to replacement by savanna due to seasonal drought and high flammability (Hutyra et al. 2005), as well as documented sensitivity to climate change (Mayle et al. 2000). These forests are also subject t o the anthropogenic drivers of deforestation and savanna expansion that operate throughout the Amazon. In particular, agricultural conversion, road building, and logging have all contributed to the fragmentation, degradation, and increased fire susceptibil ity of Chiquitano
18 dry forest (Steininger et al. 2001). Importantly, the diverse native grass flora of Chiquitania, plus naturalized African forage grasses, permits investigation of the savanna forming potential of a variety of grass species Variability in soil fertility makes the region suitable for studying the interaction between soils and savanna grass species composition. Soil fertility is thought to determine native savanna forest distributions in South America (Hoffman et al. 2009); in Chiquitania, n ative savannas generally occur on shallow sandy soils or well drained oxisols, whereas forests occur on younger, relatively fertile soils developed from the granite rock of the Precambrium Shield that underlies the region (Killeen et al. 1990). Giv en the huge range in soil characteristics that occur across the Amazon Basin ( Quesada 2009 ), a better understanding of associations between soil type and grass species assemblages will be important for predicting savanna expansion. The objective of this study wa s to determine how much and what kind of savanna has already replaced forest on the southern rim of the Amazon Basin in eastern lowland Bolivia. I use d satellite imagery to quantify the extent of forest replacement by savanna between 1986 and 2005. Based o n this analysis I identified and sampled native (old) savannas and newly derived savannas (i.e. areas that changed from forest to savanna between 1986 and 2005 ) to determine their grass and tree species compositions to measure fine fuel loads and to co llect soil samples I use th ese data to describe forest replacing savannas, to determine the degree to which they floristically resemble native savanna s to relate soil chemical and physical properties to grass species assemblages, and to suggest ecosystem consequences of native as opposed to derived savanna expansion.
19 Materials and Methods Study A rea This research covers 22,500 km 2 of northern Chiquitania bounded by the limits of CBERS 2 satellite scenes and the San Ramon Trinidad Highway on the southwest ( Figure 1 1). a mean annual temperature of 24C with extreme temp eratures of 4 37C; mean annual precipitation is 1160 mm but ranges 799 1859 mm (data from Administracin de Aeropuertos y Servicios Auxiliares a la Navegacin Area collected in Concepcin, 1943 2005). Seven months each year receive < 100 mm precipit ation (April through October) and during four months (June through September), potential evapotranspiration exceeds rainfall (M. Pea Claros, unpublished data ). Satellite Image A nalysis To quantify the extent of forest replacement by savan na during 1986 2 005, I digitally processed satellite images to create a change trajectory map that identified areas transformed from forest to savanna. Due to the seasonality of the region and the predominance of deciduous trees, I chose images from the early to mid dry s eason, a time of year before complete leaf senescence but with many cloud free days. For the 1986 classification, I used adjacent 1986 Landsat Thematic Mapper (TM) scenes for Path 230 Row 071 and Path 229 Row 071 from 2 July 1986 and 27 July 1986, respecti vely, available from the Global Land Cover Facility, University of Maryland, USA. The Path 230 Row 071 scene was already orthogeorectified and I obtained an orthogeorectified scene for TM Path 229 Row 071 from 16 July 1988. Due to technical failures in La ndsat 7 and limited availability of current Landsat TM scenes, I chose two CBERS 2 (China Brazil Earth Resource Satellite) scenes from INPE (National Institute of Space
20 Research), Brazil for Path 172 Row 118 from 24 July 2005 and Path 171 Row 118 from 27 J uly 2005. The data were projected to WGS84, UTM Zone 20 S, resampled to 30 m x 30 m spatial resolution, and geo referenced via image to image rectification using the orthorectified TM images with a root mean square (RMS) error of < 0.5 pixels. During June July 2007 I visited 190 sites distributed across the study region to create a database of Geographic Positioning System (GPS) coordinates and corresponding vegetation descriptions needed to interpret current and historic satellite images. At each 30 x 30 m site (i.e., the size of one Landsat pixel), I measured tree canopy cover at three random points with a spher ical densitometer at (Lemmon 1956) and visually estimated the percent ground cover of gr aminoids (of any height) I also took digital photographs of each site and made notes on current and past land use practices. Based on this information I defined each site as forest, savanna, large scale agriculture, or wetland for classification of the current image (Table 1 1). A point was only assigned to a different class for 1986 than in 2005 when there was convincing evidence that class shift had occurred. For example, a present day pasture that contained standing dead forest trees 50 80 cm in DBH and that looked like forest in the image would be classifie d as forest for interpretation of the 1986 image. I used a random selection of half of the training samples to create spectral signatures based on histogram equalized CBERS 2 B ands 2, 3, 4 and histogram equalized TM B ands 2,3,4, 5 and then assigned pixe ls to the four classes using the maximum likelihood model in Erdas Imagine 9.0. (Leica Geosystems, Norcross, Georgia USA) Improved pastures and savannas were the only classes that were not consistently separable based on spectral signatures alone. Large s cale agriculture, including improved pastures and industrial monocrops, were easily distinguished from savannas because ranchers usually clear forests in a manner that
21 creates regularly shaped borders (Steininger et al. 2001; Martinez M. et al. 2003). I ma nually edited the classified images to correct areas of large scale agriculture that would have been classified as savanna based on spectral signature alone. Once each of the images was classified, I used the remaining half of the training samples to perfo rm accuracy assessments. The accuracy results I report are the mean values for the two TM 1986 scenes and the two CBERS 2005 scenes weighted by their proportional contribution to the 1986 2005 change trajectory analysis. I used image subtraction to detect changes in pixel classifications between 1986 and 2005 to show the areas where savanna replaced forest. Prior to overlaying classified images to produce the change trajectory map I created a mosaic of the two adjacent scenes for each year. I used the error matrix produced by the accuracy assessment to correct the total area estimates for each cover class and each trajectory class (van Deusen 1996); both raw (map values) and corrected (unbiased) values for cover classifications are presented. Savanna F loris tic Fuels, and Soil Analyse s During the mid to late rainy season of 2008 (February and May), I conducted detailed field sampling of areas determined by satellite image analysis to have shifted from forest to savanna between 1986 and 2005. I used a strat ified random sample of pixels from acros s the study region to identify savanna patches > 9 ha that included 20 derived savanna s and 12 old (native) savannas (i.e., areas classified as savanna in both 1986 and 2005). I excluded some savannas (~10% of all d erived savannas) in roadless portions of the study region. In each of the selected sites, I established five 20 m variable width transects at random locations. Along each transect, I measured ground cover density and vegetation height up to 2 m at 41 point s spaced at 0.5 m intervals of the following classes: 1) grass cover by species; 2) other herbaceous plants; 3) woody shrubs; 4) leaf litter; and, 5) bare soil. At the ends and midpoint of each transect, I clipped all standing biomass < 25 mm diameter from 0.5 m x 0.5 m plots in the following classes:
22 1) graminoids (i.e., Poaceae, Cyperaceae, Juncaceae); 2) other herbs; and, 3) shrubs. I also collected litter from these three plots (0.75 m 2 /transect) and sorted the material collected as graminoid litter or litter from all other sources. G rass species present in the these plots but not at the 41 ground cover density points were assigned a cover value of 1%. To obtain more accurate estimates of standing biomass of the native bamboo Guadua paniculata three 1 m 2 clip plots were measured. After weighing biomass in the field, subsamples were dried at 65C for 48 hr to correct for moisture. At the ends and midpoint of each transect I estimated tree canopy cover with a spherical densiometer (Lemmon 1956) held at 1 .5 m. I identified to species and measured the DBH and height of all trees >5 cm DBH that fell within 4 m of the transect ( i.e., 20 m x 8 m). As an indication of the intensity of past fires, I measured the maximum char height on the trunk of each tree (Fry 2008). Although, s ome grasses ( Guadua paniculata Hyparrenhia rufa and Urochloa maxima ) can exceed 2 m in height, I included grass cover >2 m tall in estimates of ground cover but not for tree canopy cover. At the midpoints of each of the five transects per site, I measured soil depth to 2 m with a bucket auger, collected soil samples for bulk density determination (0 4 cm) in 6.8 cm diameter tins, and extracted soil cores (0 20 cm, 6.8 cm diameter) for laboratory analysis. When soil depth was < 20 cm, I collected an equivalent volume of soil from multiple shallower cores. I mixed the five soil cores from each site before sending samples for analysis by the soils laboratory of the Center fo r Tropical Agriculture Research (CIAT Santa Cruz ) Soils were anal yzed for : pH (1:5 in water); electrical conductivity (1:5 in water); soluble cations Ca, Mg Na, K (atomic absorption (AA) following soil digestion in ammonium acetate; for Ca and Mg soil was additionally treated with lanthanum chloride); interchangeab le acidity (titration with potassium hydroxide of soil diluted in potassium chloride); Al (floride titration); available P (modified
23 Olsen method for pH<5.4, modified Bray and Kurt z method for pH > 5.4 ) ; total N (Micro Khejldaljh method) organic material ( Walkley Blach method); and texture (sedimentation analysis). Cation exchange capacity was calculated as the sum of interchangeable acidy plus the concentration of soluble cations ( Ca, Mg Na, K). B ulk density samples were oven dried at 105C for 24 hr pri or to weighing. I identified grass species using the nomenclature of Renvoize (1998), with assistance of keys provided therein and by Killeen (1990). Nomenclature departs from Renvoize (1998) for the former Brachiaria brizantha B. decumbens and Panicum maxima that molecular data place within Urochloa (Aliscioni et al. 20 03). Tree nomenclature follows Jardim et al. ( 2003) and Killeen et al. (1993). Identifications were confirmed at the Noel Kempff Mercado Museum of Natural History Herbarium in Santa Cruz, Bolivia (SCZ) where voucher specimens were deposited. Note that after first mention, species are referred to by their generic name (ex c e p t where ambiguous). I used detrended correspondence analys i s (DCA) to ordinate savannas based on their gra ss and tree species compositions I performed two ordinations, one using percent grass cover by species and a second using tree basal area by species. Only species present in at least two sites were included in the ordinations; the grass and tree DCAs included 52 and 64 species, respectively. I performed both ordinations in R 2.9.0 (the R Foundation for Statistical Computing) using the decorana command from the vegan 1.15 2 package. Three of the derived savannas sampled did not fit within my definition of savanna in t erms of tree canopy cover and herbaceous ground cover. After excluding these wooded areas that were misclassified as derived savanna in the change trajectory map, I analyzed data from n = 12 native and n = 17 derived savannas. To compare how grass and tr ee species belonging to particular vegetation associations
24 (forests or savannas) or functional groups (native savanna grasses or exotic forage grasses) were distributed in derived versus native savannas, I classified grasses and trees a priori based on spe cies descriptions by Killeen (1990) and Jardim et al. (2003), respectively. I compared native and derived savanna diversity on the basis of numbers of species of native grasses, savanna trees, and forest trees; the proportions of all trees that were sava nna species, forest species, or palms were also compared. To control for the effect of area sampled, I counted the species present on a per transect basis. As such, values for each savanna are the average from five 20 m transects. The number of native gra sses includes all species recorded in cover counts or present in three 0.5 m x 0.5 m sub plots per transect. The number of tree species (>5 cm DBH) were those present in each transect (160 m 2 sampled per transect). To compare numbers of species in native a nd derived savannas I tests and Mann Whitney U tests. I used canonical correspondence analysis (CCA) to relate grass species composition to soil variables measured in native and derived savannas. CCA was performed in R 2.9.0 using the cca command from the vegan 1.15 2 package I tested for differences in soil characteristics between native and derived savannas using t tests. I compared native and derived savanna fuel loads and individual biomass components (i.e., graminoid litter, stand her bs, etc.) using Mann Whitney U tests. Results Satellite Image Change Analysis The 1986 Landsat TM and 2005 CBERS 2 classifications had overall accuracies of 90% and 89%, respectively, with overall Kappa values of 0.84 and 0.84 (Table 1 2). The 1986 and 200 5 classifications differed in accuracy for forests and savannas such that there was high confidence in pixels classified as forest in 1986 and high confidence in pixels classified as savanna in 2005 (Table 1 (or consumer s risk, the
25 probability that a classified pixel is indeed that cover class; v an Deusen 1996) for forest (100%) forest (84%) was less than savanna (91% ). These values signify high confidence that areas classified as derived savanna in the change trajectory map are areas of former forest. The trade off for confidence in derived savanna detection is that the analysis provides a conservative estimate of sa vanna formation and overestimates succession from savanna to forest. After correcting for this trade off, I estimate that between 1986 and 2005, total forest cover declined from 83% of the study area to 68%, savanna increased from 14% to 22%, and large sca le agriculture increased from 2% to 9% ( Figure 1 3). Of the 17,700 km 2 classified as forest in 1986 in the change trajectory analysis, 4.5% (800 km 2 ) changed to savanna by 2005 and 5.4% (959 km 2 ) was converted to large scale agriculture (Table 1 3); both of these estimates are conservative and error corrected values are 7.6% (1420 km 2 ) and 6.4% (1200 km 2 ), respectively. Some 2000 km 2 was classified as savanna in both 1986 and 2005 and was defined as native (old) savanna for the floristic studies that foll ow. Savanna Floristics Field sampling and subsequent DCA performed on grass species data showed two kinds of native savannas: typical cerrado and areas dominated by Guadua Native savannas were clustered together on DCA axis 1 (Fig 4a); 10 of 12 native savannas were floristically distinct from the derived savannas, a position in ordination space that reflects the DCA scores of native cerrado grasses compared to the scores of other species ( Figure 1 4b,c). The two native savannas that were not closely gr ouped with the cerrados contained the native bamboo Guadua paniculata either as the dominant grass species or in lower abundance along with cerrado grasses. Derived savannas were considerably more dispersed in ordination space than native savannas ( Figure 1 4a). The positions of most derived savannas corresponded with the species
26 scores of the following dominant grasses: Guadua U. brizantha and U. maxima Imperata brasiliensis and Digitaria insularis ( Figure 1 4c). With the exception of a single derived savanna that had a grass flora floristically similar to the cerrados the only derived savannas that clustered near native savannas were Guadua dominated. The tree DCA revealed a distinct cluster of native savannas but little grouping of derived sites ( Figure 1 5a). The site scores for native savannas corresponded to the species scores of Curatella americana (Dilleniaceae), Qualea grandiflora (Vochysiaceae), Caryocar brasilensis (Caryocaraceae), Diptychandra aurantiaca (Caesalpinoideae), and other tree s considered to be characteristic of savanna by Jardim et al. (2003; Figure 1 5b,c). In contrast, the derived savannas typically contained a few remnant forest tree species such as Anadenathera columbrina (Mimisoideae; a dominant forest species), palms in the genus Attalea ( A. phalerata A. speciosa Arecaceae), or woodland forest species ( e.g., Astronium urundeuva, Anacardiaceae) that occur across a range of habitat types. Whereas the grass ordination revealed two types of native savannas ( cerrados and tho se with Guadua ), the analysis based on tree species composition showed all native savannas as part of one distinct cluster. None of the derived savannas clustered amongst the native savannas, but four sites had DCA 1 scores similar to native savanna. These sites clearly differed from native savannas along DCA 2 based on the presence of Anadenathera Astronium Attalea speciosa and Magonia pubescens (Sapindaceae). Native savannas had more native grass species and a greater number of savanna tree species t han derived savannas, whereas derived savanna contained more forest tree species ( Figure 1 6). The average numbers of native grass species per transect were 5.7 0.6 and 1.7 0.3 for native and derived savannas, respectively (mean1SE; Figure 1 6a). Me dian savanna tree species richness values per transect (160 m 2 ) were 2.5 (range 1.2 5.8) and 0.4 (range 0 7.2)
27 for native and derived savannas, respectively ( Figure 1 6b). Mean forest tree species richness values per 160 m 2 were 0.6 0.1 and 1.3 0.3 for native and derived savannas, respectively (mean1SE; Figure 1 6c). The relative proportions of savanna trees, forest trees, and palms differed between native and derived savannas with native savannas composed primarily of savanna tree species and deri ved savannas containing higher proportions of forest trees. In native savannas, 83% and 14% of trees were savanna and forest species, respectively, compared to derived savannas where 22% and 72% of trees were savanna and forest species, respectively ( Figur e 1 7). Relative abundance of palms did not differ between savanna types, but maximum abundance of palms was greater in derived savanna. In addition to these comparisons, grass and tree species listed in order of their re spective DCA 1 scores (Figure A 1; Figure B 1 ) shows the limited number of species shared by derived and native savannas. There were exceptions to the trend of high native grass diversity in native savannas and low savanna tree diversity in derived savannas. One exception was a single deri ved savanna located in the transition zone between forest and native savanna. This savanna had the highest mean tree species richness of any site sampled (7.2 savanna and 4.4 forest tree species/transect) and was the only area identified as recently deriv ed savanna in the satellite analysis that was similar to diverse native savanna in grass and tree species composition; it is possible that this site was misclassified as having shifted from forest to savanna. The exception to the relatively high grass spec ies richness in native savannas was a site where Guadua appeared to exclude most other grasses and that contained few trees (only 1.4 savanna and 0.6 forest tree species per transect). The majority of the native savannas sampled (11 of 12) and one of the newly derived savannas (1 of 17) contained a diversity of grass species typical of cerrado vegetation (Killeen et
28 al. 1990; Killeen & Hinz 1992 a ). The most abundant species were the native grasses Elionurus muticus Trachypogon spicatus Thrasya petrosa, Paspalum stellatum, and the naturalized African forage grass Hyparrhenia rufa. In lower abundances but frequently present in cerrado were the native grasses Andropogon selloanus Axonopus barbigerus A. marginatus P. erianthum and Schizachyrium sanguin eum Of the 77 grass species encountered in this study 36 occurred exclusively in cerrados The most common forest replacing savannas (7 of 17 derived savannas) were dominated by Guadua On the basis of their nearly arboreal habit, bamboos are commonly l umped by ecologists with woody plants or are considered a vegetation class of their own. In this study I included bamboo dominated vegetation within the broad definition of savanna. Guadua the most common bamboo of this region, produces large quantities o f fine fuel and resprouts readily following fire s (Veldman 2008). Guadua stands, which dominate ten of thousands of h ectares in the region (Killeen 1990), can be characterized by frequent fires and severely charred or dead trees (Veldman 2008). Although Gu adua can grow up to 7 m tall, most stands form grass canopies of only 2 5 m. Given its interactions with fire in Chiquitania, it seems appropriate to include Guadua within my conceptual frame work of grass fire feedbacks influencing forest to savanna conve rsion. Four exotic forage grasses of African origin were among the most common species encountered in this study. Urochloa dominated savannas ( U. brizantha U. decumbens and/or U. maxima occurring alone or in mixed swards) accounted for six of 17 derived savannas. Another African pasture grass, Hyparrhenia, is widely naturalized in cerrados but rarely occurred in derived savannas. Two of the derived savannas located on abandoned swidden plots contained monodominant stands of Imperata brasiliensis Impe rata is a rhizomatous native grass that was
29 a frequent component but never abundant in native savannas. A third area of long fallowed swidden agriculture was dominated by the native annual herb Hyptis suaveolens (Lamiaceae). Hyptis formed a canopy of 1.5 to 2.5 m tall and appeared to exclude most other plants including grasses; the only trees interspersed with Hyptis were palms with charred trunks indicative of intense fires. Digitaria insularis a native ruderal, was abundant on a single site where it gre w on the ash of burnt slash along the border of a Urochloa pasture; Digitaria was also present in the fallowed swidden plots dominated by Imperata Fuel Loads and Fire History Compared to native savannas, derived savannas had higher biomass values for sta nding graminoids, graminoid litter, and other litter, but the two savanna types did not differ in standing herb or shrub biomass (Figure 1 8). Total fuel loads (biomass <25 mm diameter) in derived savannas (median 8.0 Mg/ha; range 3.5 20) were twice that o f native savannas (median 3.6 Mg/ha; range 2.7 9.2; Figure 1 9). Median fuel loads of standing graminoids, standing herbs, and graminoid litter (woody shrubs and other litter excluded) were 6.2 Mg/ha (range 0.8 20) and 2.4 Mg/ha (range 0.7 7.5) in nati ve and derived savannas, respectively (Figure 1 8). Maximum char heights in native savannas (2.0 1.4 m) exceeded those measured in derived savannas (0.8 0.9 m; mean 1SD, t test, p = 0.009, n = 12, 17, respectively). All native savannas and 16 of 17 de rived savannas contained charred trees. Both native and derived savannas contained similar densities of standing dead trees ~26 trees/ha. Soil Fertility and Grass Species Composition Concentrations of plant essential nutrients (N, K, Ca, and Mg), cation ex change capacities, electrical conductivities, organic matter, pH, and silt were greater in derived savannas than native savanna (Table 1 4). Native savanna soils were more often sandy, with high Al, acidity, and bulk density (Table 1 4). Savanna types did not differ in clay content, Na concentration, or soil
30 depth. CCA of savanna grasses and soils showed very similar clustering of sites and species compared to DCA of grass species only (Figures 1 9, Figure 1 4 respectively). CCA 1 appears to represent a so il fertility gradient with most native savannas ( cerrados ) occurring on infertile soils and derived savannas occurring on fertile soils (Figure 1 9). Discussion Over a twenty year period (1986 2005) in eastern lowland Bolivia, savannas replaced forests at approximately the same rate as deforestation for large scale agriculture. These new savannas were dominated by a native bamboo, a few exotic forage grasses, or native invaders of fallowed swidden plots. In rare cases, savanna expansion resulted in high na tive grass and tree diversity, but these high diversity transitions zones along native savanna forest margins accounted for a small fraction of total forest to savanna conversion. The large difference in species richness and floristic composition between n ative and newly derived savannas underscores the contrasting biodiversity implications of native savanna expansion compared to forest replacement by derived savannas. Of the species that dominate d forest replacing savannas, Guadua Imperata Urochloa spp. and Hyptis appear to represent alternative stable states maintained by positive feedbacks with fire, or in the case of Hyptis a combination of fire and allelopathy. Many bamboo species interacted with fire (Keeley & Bond 1999) and bamboo stands are thou ght to be an alternative stable state to broadleaf forest in both temperate ( e.g., Arundinaria gigantea ; Gagnon & Platt 2008) and Neotropical ( e.g., Guadua sarcocarpa and G. weberii ; Griscom & Ashton 2003, 2006) regions. P asture grasses in the genus Urochl oa introduced from Africa as cattle forage are widespread across the N eotropics (Parsons 1972). In the study region, U. brizantha and U. maxima are the most common planted pasture species, are roadside weeds, and invade disturbed forest in areas that rec eive motor vehicle traffic (Veldman & Putz 2010). Both U. maxima and U.
31 brizantha are bunch grasses whereas culms of U. decumbens spread laterally and root at the nodes. All three species form dense canopies, with U. maxima (2 3 m tall) typically exceeding the height s of U. brizantha and U. decumbens (1 1.5 m tall). Because these forage grasses are both intentionally planted and accidentally introduced ( Veldman et al. 2009), it is difficult to ascertain the genesis of all Urochloa stands in this study. Uro chloa dominated areas included pasture margins, where annual fires probably contribute to pasture expansion, as well as roadsides where fires and grazing are common but grasses may have been planted or simply colonized. Imperata a native rhizomatous grass is described by Imperata stands as one type of anthropogenic savanna in lowland Bolivia. In northern Bolivia, Imperata abundance in small scale agriculture p lots was reported to increase with repeated slash and burn cycles (Sools 2007), and in Peru Imperata invasion of swidden plots followed by repeated fires was considered critical step in the conversion of forest to savanna (Scott 1977). An ecologically sim ilar congener of Asian origin, I. cylindrica, has changed fire regimes, replaced native vegetation from southeast Asia to the southeastern United States ( e.g., Lippincott 2000), and is considered one of the most invasive plants in the world (MacDonald 2004 ). Hypti s, a n herb native to the Neotropics, is invasive in India were it colonizes shifting agriculture sites and was observed to enhance dry season fires (Murthy et al. 2007). Additionally, Hyptis is reputedly allelopathic and limit s the germination and growth of both grasses and trees (Totey et al. 1994; Mokat et al. 2005). Soil fertility helps to explain differences between native and derived savanna grass species composition (Table 1 4 ; Figure 1 9) Consistent with previous studies in the region (Kille en et al. 1990 ; Killeen & Hinz 1992a), I found native savannas to occurr on nutrient poor acidic soils In
32 contrast, derived savanna s occurred on relatively fertile soils normally associate d with forests not surprising for sites that were forests within the last two decades. The absence of native savanna grasses in most derived savannas may be due to dispersal limitation, a competitive disadvantage on fertile soils, or both. Native savanna grasses adapted to infertile soils are probably unable to take adv antage of nutrients available in more fertile sites, whereas grasses with high potential growth rates (e.g., Guadua and Urochloa spp. ) are physiologically ill equipped for infertile soils but exploit disturbed forest soils that tend to be richer in nutrie nts (Grime et al. 1997) A n apparent consequence of productive grasses and h erbs growing on fertile soils, derived savannas had fuel loads that were twice as large compared to native savannas. cteristics are human land uses. As such, w e should expect the determinants of derived savanna grass species compositions to be more complicated than simply soil fertility and fire history factors that define native savanna forest boundaries (Hoffmann et al. 2009). When people clear forests for agriculture they prefer the most fertile soils. When people establish pastures they intentionally introduce exotic forage species. In some cases savanna formation may be driven by complex socio political factors: in one derived savanna U. brizantha and U. maxima were clearly planted, but never grazed, pr esumably as a way to lay claim to the land or otherwise show that it was being put to productive use a case of savanna formation resulting from issues related to land tenure T he use of climate data from to predict savanna expansion into the Amazon Basin (e.g., Mahli et al. 2009) ma y be inappropriate because d erived savanna species are not functional equivalents of the species foun d in old growth native savannas (i.e., derived savanna grasses produce greater fuel loads). Additionally, underlying environmental gradients (i.e., soil fertility) and human activities favor a few grass species over
33 cerrado grasses. Native savannas may exp and in areas with low soil fertility, but when savannas expand onto fertile forest soils we should expect them to be composed of a few grasses and herbs species that produce large fuel loads. Differences in fuel loads are relevant given the differences in fire tolerance between tree species characteristic of forest compared to savanna (Hoffmann 2009); in lowland Bolivia and throughout much of the Amazon Basin forest trees are susceptible to even low intensity fires (Pinard & Huffman 1997). Given that most of the trees I found in derived savannas were forest species unlikely to survive repeated grass fueled fires, maintenance of tree cover may depend on whether savanna tree species are able to colonize. Forest replacement by species poor savanna is just on e sort of vegetation shift in lowland Bolivia. Although this study focused on places transformed from forest to savanna, selective logging, fires, and swidden agriculture can result in a variety of derived vegetation states (Putz & Redford 2010). One such outcome is liana dominance, a state transition that may be favored by the increased competitiveness of lianas under increasing atmospheric concentrations of CO 2 (Foster et al. 2008) or in response to forest fires (Pinard et al. 1999). Although my data sup port other reports that swidden fallows are invaded by fire promoting grasses and herbs (Scott 1977; Killeen, 1990; Renvoize 1998; Sools 2007), savanna is not the only fate of abandoned agricultural land. Indeed, the remote sensing analysis showed 10% of t he area classified as large scale agriculture in 1986 had regenerated to forest by 2005. This study was conducted in a region that contains native savannas, but in forests that are far from native savannas, the likelihood of native savanna expansion seems less likely, at least over decadal time scales. Although it is clear that repeated fires can push forests into a severely degraded state (e.g., Barlow & Peres 2008), the presence of one or more forest invading and fire promoting grasses is likely a trigger that promotes forest to savanna conversion. Data collected
34 in forests that are currently far from propagule sources of invasive grass species may be quite different than those subjected to the synergistic effect of disturbance plus invasion by savanna for ming species. Clearly m ore needs to be known about the biology of the grasses, herbs, and trees that dominate recently derived savannas as compared to those characteristic of native savanna. Species differences in fine fuel production, height s soil nutri ent requirements, seed production, capacity for vegetative expansion, and drought resistance could result very different vegetation climate fire interactions in the Amazon Basin. Conclusion The derived savannas that replaced tropical forest in this study had low grass and tree species diversity, were floristically distinct from native savannas, were usually dominated by one or two species of grasses and contained remnant forest tree species. The distinction between diverse native savannas and low diversit y derived savannas is important given the threatened status of the native savannas of South America and predictions that Amazonian forests are increasingly susceptible to savanna expansion. Soil fertility should be considered when predicting which grass sp ecies are likely to replace forests; fertile soils are more likely to be colonized by a few highly productive native or exotic species rather than native savanna grasses Species identities are important because grass species with high growth rates are cap able of producing large fuel loads on fertile soils. As research on large scale tropical forest savanna dynamics progresses, we ought to be explicit about w hat sort of savanna is involved.
35 Table 1 1. Tree canopy cover and graminoid cover of training samp les (used for the satellite im age classification) and native and derived savannas included in the savanna floristic study. Values are mean percent1SD. n Tree C anopy C over (%) Graminoid C over (%) Training Sample C lasses Forests 81 8 9 10 6 1 2 Savannas 63 18 1 8 79 1 6 Large S cale A griculture 34 4 6 85 1 8 Wetland s /Water 1 2 Floristic S tudy S ites Native S avannas 12 22 14 51 16 Derived S avannas 17 30 19 50 23 Table 1 2. Accuracy assessment of classified images used in the 1986 to 2005 change trajectory analysis. TM 1986 CBERS 2005 Class A ccuracy (%) A ccuracy (%) A ccuracy (%) A ccuracy (%) Forest 87 100 95 84 Savanna 96 76 78 91 Large S cale A gricultu re 83 100 92 97 Water 100 86 83 100 Overall A ccuracy 90 89 Overall Kappa S tatistic 0.84 0.84
36 Table 1 3. Area of change trajectory classes from 1986 and 2005. Raw totals are the sum of all pixels as indicated by the change trajectory analysis that was used to identify native and derived savannas for field sampling. Error corrected totals were adjusted based on the results of the accuracy assessment to more accurately reflect true transition class values. Map Totals Error Correct ed Estimates Trajectory (1986 2005) Trajectory C lass S ynonyms Color Figure 1 3 km 2 % of S tudy A rea km 2 % of S tudy A rea Forest forest old growth forest gray 15963 71.1 14182 64.9 Forest savanna derived savanna yellow 800 3.6 1418 6.5 F orest large scale agriculture agricultural conversion white 959 4.3 1196 5.5 Savanna forest forest succession gray 1713 7.6 1115 5.1 Savanna savanna native (old) savannas brown 1986 8.9 2579 11.8 Savanna large s cale agriculture agricultural conversion white 481 2.1 440 2.0 Large scale agriculture forest forest succession gray 42 0.2 62 0.3 Large scale agriculture savanna persistent agriculture white 50 0.2 148 0. 7 Large scale agriculture large scale agriculture persistent agriculture white 176 0.8 366 1.7 Water (in any year) black 268 1.2 340 1.6
37 Table 1 4 Soil variables measured in native ( n = 12) and derived ( n = 17) savannas listed in increasing order of CCA 1 axis scores relating savanna grass species and soil variables. S ignificant differences noted for p<0.05 (t tests) Soil Variable Native Savannas (mean 1SD) Derived Savannas (mean 1SD) p CCA 1 Bulk Density (g/cm 2 ) 1.25 0.11 1.03 0.10 *** 0.84 Acidity (cmol/Kg) 0.9 0.7 0.1 0.1 ** 0.69 Al (cmol/Kg) 0.5 0.7 0.0 0.0 0.61 Sand (%) 63 10 51 17 0.40 Clay (%) 17 8 17 4 ns 0.00 Na (cmol/Kg) 0.10 0.02 0.12 0.05 ns 0.12 Depth (m) 1.3 0.7 1.5 0.6 ns 0.26 P (mg/kg) 1 0 5 6 0.28 K (cmol/Kg) 0.16 0.07 0.30 0.15 ** 0.34 Silt (%) 21 11 31 16 0.39 pH 5.2 0.5 6.4 0.6 *** 0.44 19 9 75 52 *** 0.48 Cation Exchange Capacity (cmol/Kg) 2.5 0.7 6.6 3.4 *** 0.51 Ca (cmol/Kg) 0.9 0.4 4.5 2.6 *** 0.56 Mg (cmol/Kg) 0.5 0.3 1.5 0.8 *** 0.61 N (%) 0.09 0.04 0.17 0.05 *** 0.64 Organic Matter (%) 1.0 0.6 2.3 0 .4 *** 0.81
38 Figure 1 1. Study location in eastern lowland Bolivia. The shaded area corresponds to the intersection of the Landsat TM (1986) and CBERS 2 (2005) satellite images analyzed to detect forest replacement by savanna. Figure 1 2. Land co ver class distributions based on error corrected area totals for the 1986 (Landsat TM) and 2005 (CBERS 2) classifications.
39 Figure 1 3. Modified change trajectory map created from classified 1986 (Landsat TM) and 2005 (CBERS 2) images used to identify fo rest replacing (derived) savannas and native (old) savannas. Trajectory classes other than forest savanna and savanna savanna have been grouped in the color coding for visual simplification; see Table 1 3 for clarification of color groupings. Axes are lab eled with UTM coordinates for WGS 84 Zone 20 south. The four unclassified strips that transect the map correspond to pixels of low CBERS 2 image quality that were excluded from the analysis.
40 Figure 1 4. Detrended correspondence analysis of grass cover by native and exotic species in native and derived savannas. Ordination plots display the scores for sites and species along the first two DCA axes. (a) Ordination of 52 grass species present in at least two site s (b) Ordination of sites; n = 12 native sava nnas, and n = 17 derived savannas. (c) Composite ordination of sites and species with the ten most abundant grasses species noted.
41 Figure 1 5. Detrended correspondence analysis of tree species by basal area in native and derived savannas. Ordination pl ots display the DCA scores for sites and species along the first two DCA axes. (a) Ordination of sites; n = 12 native savannas, and n = 17 derived savannas. (b) Ordination of 64 tree species present in at least two site s ; shading of species characteristic of savanna or forest is based on Jardim et al. (2003). (c) Composite ordination of species and sites with the ten most abundant tree species noted.
42 Figure 1 6 Frequency distributions of numbers of native grass and tree species in native ( n = 12) and d erived ( n = 17) savannas. (a b) mean number of native grass species present per transect; native and derived savannas differ at p < 0.001 (t test). (c d) mean number of savanna tree species present per transect; native and derived savannas differ at p < 0 .001 (Mann Whitney U test). (e f) mean number of forest tree species present per transect; native and derived savannas differ at p < 0.05 (t test). Trees >5 cm DBH were counted in 20 m x 8 m transects and are classified as savanna or forest species; palms were excluded from the analysis.
43 Figure 1 7. Proportion of savanna tree species, forests tree species, and palms in native ( n = 12) and derived ( n = 17) savannas. Native and derived savannas differed in proportions (% of total stems >5 cm DBH) of sava nna and forest species (t tests, p<0.001). Figure 1 8. Fuel loads produced by graminoids (Poaceae and Cyperaceae), herbs, and shrubs in native and derived savannas. Fuel loads are dry weights of biomass <25 mm diameter. Graminoid + Herbs includes stand ing graminoid fuel, standing herb fuel, and graminoid litter. Asterisks deno te differences between native ( n = 12) and derived ( n = 17) savannas at p <0.05 (Mann Whitney U tests).
44 Figure 1 9. Canonical correspondence analysis of grass cover by (native and exotic )species and soil variables measured in native ( n = 12) and derived ( n = 17) savannas. Ordination pl ots display the scores for sites, species, and soil variables along the first two C CA axes All grasses were included in the CCA but only the 10 most abundant species are displayed.
45 CHAPTER 2 SELECTIVE LOGGING AN D FIRE AS DRIVERS OF ALIEN GRASS INVASION IN A BOLIVIAN TROPICAL DR Y FOREST 1 Logging is an integral component of most conceptual models that relate human land use and climate change to tr opical deforestation via positive feedbacks involving fire. Given that grass invasions can substantially alter fire regimes, I studied grass distributions in a tropical dry forest 1 5 yr after selective logging, and experimentally tested the effect of fore st fire on populations of invasive grasses. In unlogged forests and in microhabitats created by selective logging I found a total of four alien and 16 native grass species. Grasses covered 2% of unlogged and 4% of logged forest, with grass cover in logged forest concentrated in areas directly disturbed by logging; log landings and roads had relatively greater grass cover (37% and 17%, respectively) than did skid trails (10%) and felling gaps (8%). Total grass cover and grass species richness increased with canopy openness and were greatest in sites most severely disturbed by logging. The grass flora of these disturbed areas was composed mostly of native ruderal species (e.g., Digitaria insularis, Leptochloa virgata ), a native bamboo ( Guadua paniculata ), and Urochloa (Panicum) maxima a caespitose C 4 pasture grass introduced from Africa. Urochloa maxima formed monodominant stands (up to 91% cover and 2 3 m tall) and grew on 69% of log landings and 38% of roads. To better understand the potentially synergistic effects of logging and fire on the early stages of grass invasion, I tested the effect of a 12 ha experimental fire on U. maxima populations in a selectively logged forest. Three years after the fire, the area covered by alien grass in burned forest incre ased five fold from 400 m 2 (pre fire) to 1 Reprinted with permission from: Veldman, J. W., B. Mostacedo, M. Pea Claros, and F. E. Putz. 2009. Selective logging and fire as drivers of alien grass invasion in a Boli vian tropical dry forest. Forest Ecology and Management 258 : 1643 1649.
46 1660 m 2 ; over the same period in a logged but unburned (control) area, U. maxima cover decreased from 398 to 276 m 2 Increased canopy openness due to fire induced tree mortality corresponded with the greater magnitu de of grass invasion following fire. Selective logging of this dry forest on the southern edge of the Amazon Basin promotes alien grass invasion; when coupled with fire, the rate of invasion substantially increased. Recognition of the grass promoting pote ntial of selective logging is important for understanding the possible fates of tropical forests in fire prone regions. Introduction The combined effects of road building, logging, increased intensity of land use, climate change, and fire in the Amazon Ba sin are predicted to result in widespread replacement of forest by savanna (e.g., Nepstad et al. 1999 ; Cox et al. 2004 ; Hutyra et al. 2005). Such projections are based on positive feedbacks between deforestation and climate change (Shukla et al. 1990; Bett s et al. 2004), vegetation disturbance and fire (Cochrane et al. 1999), and drought and fire (Nepstad et al. 2004). A critical step in the process of forest conversion to savanna (i.e., to an open canopied ecosystem with grass dominance of the understory), is invasion of forest by savanna grasses. Unfortunately, the factors that control grass invasions and their consequences for tropical forests are not well understood. Given the increasing susceptibility of tropical forests to fire (e.g., Hutyra et al. 20 05 ; Golding & Betts 2008), it is important to recognize land use practices the promote grasses. Forests generally recover, albeit slowly at times, from severe degradation and even from outright destruction (Uhl et al. 1988 ) In contrast, a forested area pushed into the alternative stable state of grass dominance (i.e., another stability domain sensu Holling 1973) faces huge obstacles to recovery of its forested condition primarily due to the effects of grasses on fire regi mes (D'Antonio & Vitousek 1992) Most grasses produce large amounts of fine fuel, resprout readily
47 from basal meristems, and thrive in frequently burned areas (e.g., Daubenm ire 1972). As such, we may view tropical forest conversion to savanna in light of disturbance regime shifts from one of very rare fires to one of frequent fires that favor grasses and limit forest regeneration [e.g., Mack & D'Antonio 1998; see also Bond (2 008) for a comprehensive review of factors influencing the balance between grasses and trees]. Logging figures prominently among land uses that degrade tropical forests in ways that might contribute to grass invasions and subsequent forest conversion to sa vanna. Logging opens canopies, disturbs soil, and otherwise renders forests more prone to fire (Uhl & Buschbacher 1985 ; Laurance 2003; but see Blate 2005). Tropical dry forests may be at particular risk of grass invasion following logging due to their natu rally open canopies, severe seasonal limitations on water availability, and slow tree growth rates. Tropical dry forest susceptibility to grass invasion is evident in Guanacaste, Costa Rica where forests have largely been converted to derived savannas of t he African grass Hyparrhenia rufa (Daubenmire 1972), and in Hawaii where invasion of dry forest by another African grass, Melinis minutiflora drastically altered ecosystem functions (Mack & D'Antonio 2003). In this study I investigated forest invasion b y grasses in eastern lowland Bolivia where agricultural conversion, road building, and logging all contribute to the fragmentation, My primary objectives we re to determine if disturbance by logging promotes forest colonization by grasses, and to identify which grass species invade logged forests. To this end, I present data on the identities and distributions of native and alien grasses in a tropical dry fore st 1 5 yrs after selective logging. In light of the grass fire cycle, I hypothesized that the rate of alien grass invasion is amplified by forest fires. To test this hypothesis I evaluated the effect of a 12 ha
48 controlled burn on the population growth of t he introduced African pasture grass, Urochloa maxima (formerly Panicum maximum and Megathyrsus maximus ) in a partially grass invaded area of selectively logged, seasonally dry tropical forest on the southern rim of the Amazon Basin. Methods Study Site This research was carried out in the Bolivian region of Chiquitania in selectively logged areas of the 30,000 ha private property of INPA Parket Ltd. (hereafter INPA). The region contains a mosaic of vegetation types dominated by tropical deciduous dry forest (bosque seco Chiquitano) interspersed with native savannas floristically related to the Brazilian cerrados (Killeen et al. 1990 ; Jardim et al. 2003). These dry forests, in the transition zone between more humid Amazonian forest to the north and drier savan nas and scrub woodlands (Gran Chaco) to the south, are thought to be particularly vulnerable to replacement by savanna due to seasonal drought, high flammability (Hutyra et al. 2005), and sensitivity to climate change (Mayle et al. 2000). Although a histor (Killeen et al. 2005), most trees do not develop bark thick enough to withstand even low intensity fires (Pinard & Huffman 1997). Indeed, anthropogenic forest fires are considered among the greatest challenges to forest conservation in the region (Steininger et al. 2001). (extreme temperatures range from 4C to 37C) and receives an average of 1160 mm of p recipitation annually (range 799 1859 mm, data from Administracin de Aeropuertos y Servicios Auxiliares a la Navegacin Area at Concepcin collected 1943 2005), with seven months each receiving < 100 mm (April through October) and four months (June thr ough September) during which potential evapotranspiration exceeds rainfall (M. Pea Claros,
49 unpublished data). Between 2002 2006, 900 ha yr 1 of INPA were selectively logged at an intensity of 4 trees ha 1. A single road accesses the property from the t own of Concepcin, 40 km to the west, and passes through a matrix of forest fragments, native savannas, cattle pastures, and small farms before entering the continuous forests of INPA. These nearby agricultural lands are frequently burned and are sources o f fires that threaten the property annually. Due to fire suppression efforts by forest managers, no wildfires occurred on the property for the last 20 years (1987 2007, M. Pea, pers. obs.). Grass S pecies D istributions To determine the effect of forest dis turbance by logging on grass colonization, I studied the identities and distributions of grasses within a 2,600 ha area of INPA that was selectively harvested between 2002 2006 using reduced impact logging techniques. Using a map of the property that showe d the locations of roads and log landings as well as year of harvest, I randomly selected 16 log landings to sample during the mid rainy season (February and March 2007). From these landings I located one example of each of the following microhabitats: pri mary skid trails; felling gaps; logging roads; and, forest patches with no signs of logging activity [ n = 16 per microhabitat; N = 80 (16 x 5 microhabitat types)]. Sampling included areas 1, 2, 3, 4, and 5 yr post logging ( n = 4, 3, 4, 2, and 3, respecti vely). In each microhabitat I randomly located a single 10 x 3 m plot to sample grass cover, grass species presence/absence, and forest canopy openness. I selected microhabitats such that all plots were a minimum of 100 m from one another but were never >2 50 m from the landing. To estimate ground cover density within each plot, I classified 33 evenly spaced sample points in a 3 x 11 point grid as grass by species, other vegetation, or no plant cover. Canopy openness was measured at the center and
50 endpoints of the central axis of each plot using a spherical canopy densiometer (Lemmon 1956) held 1.3 m from the ground. I identified all grass species according to Killeen (1990) and comparisons with herbarium specimens at the Museo Nacional de Historia Natural No el Kempff Mercado in Santa Cruz, Bolivia (USZ) where voucher specimens were deposited. I photosynthetic pathway as either C 3 or C 4 based on Killeen & Hinz (1992b) and Guissani et al. (2001). Nomenclature departs from Killeen (1990 ) for U. maxima and U. brizantha (formerly Panicum maximum and Brachiaria brizantha respectively). Molecular data show both species should be together in Urochloa (Guissani et al. 2001; but see Simon & Jacobs (2003) who classify U. maxima as Megathyrsus m aximus ). Note that after first mention, species are referred to by their generic names except where ambiguous. Alien G rass R esponses to F ire I used a 12 ha controlled burn to test the effect of fire on the population growth of U. maxima in a partially gras s invaded area of selectively logged forest. Although U. maxima was generally limited to roads and log landings, a portion of INPA (one harvest unit of 100 ha) had substantial populations growing on skid trails probably because skidders re used grass invad ed log landings and entered the forest along grass invaded roads. In 2005 I mapped two separate grass invaded skid trail networks that were selectively logged the previous year; both networks supported similar U. maxima populations at the start of the exp eriment (see description below). One skid trail network fell within an area assigned to be burned in a 12 ha prescribed fire and the other network served as a control for comparison; as such, I established two unreplicated treatments: logged+burned and log ged only. The implementation and replication of large scale fire experiments that simulate wild fire events are difficult and often cost prohibitive in remote tropical forests. Most studies of tropical
51 forest responses to fires are from areas that burned as a result of wild fires and do not include pre fire data (e.g., Cochrane & Schulze 1999). Well designed, albeit unreplicated, studies on tropical forest fires that combine pre and post fire data from experimentally burned and control plots can yield imp ortant results on vegetation responses to fire (e.g., Balch et al. 2008). I used a BACI (before after control impact) designed experiment (Smith 2002), to study the population growth of U. maxima in two areas of selectively logged forest, one experimentall y burned and the other an unburned control. Before the fire, I estimated the extent of U. maxima cover along skid trails and in felling gaps by measuring the width and length of each patch of U. maxima calculating the area of an ellipse, and then summing the total of all patch areas to determine total cover. Additionally, I measured tree canopy openness with a spherical densiometer (Lemmon 1956) held above each grass patch. Although U. maxima is a bunch grass with short rhizomes, it is difficult to unambi guously distinguish between ramets and genets in dense stands (particularly after the first year of establishment); therefore, I measured the area covered by each patch as opposed to attempting to determine the number of individuals. Patches were neither r eplicates nor representative of intrinsic ecological units. They were rather, a convenient way to measure U. maxima cover by identifying areas of continuous grass cover that adhered to an eliptical shape. In 2005, before the experimental burn, the U. maxim a population in the logged+burned treatment covered a total of 400 m 2 with 113 patches measured ranging 0.06 190 m 2 (median = 0.8); the control treatment (logged only) contained 398 m 2 of U. maxima with 150 patches ranging 0.05 24 m 2 (median = 1.3). After the initial censuses, I returned annually to measure both the burned and unburned (control) populations of U. maxima during the first three years after the fire (i.e., up to 4 yr post logging).
52 The 12 ha controlled burn was conducted in the late dry season (September 2005) during the height of the fire season. The plot was burned using a backing fire ignited with drip torches along 350 m of a logging road and a head fire ignited along a bulldozed fire break on the other edge of the plot. Fire behavio r varied as a result of the heterogeneous conditions created by selective logging microhabitats. The fire was extremely intense in felling gaps (~10% of the burned area) but was otherwise a low intensity surface fire. Urochloa maxima present before the fi re represented only 0.3% of the entire burn area and presumably contributed little to the overall fire behavior or intensity at the 12 ha scale. As such, this experiment tested the effect of fire on the early stages of U. maxima invasion well below the a bundance threshold at which grass fuel loads determine fire regimes. Data A nalysis I compared microhabitats on the basis of canopy openness, grass cover, and grass species richness with ANOVA followed by Tukey post hoc tests. All values are based on a si ngle equal sized plot for each microhabitat ( n = 16). Ground cover percentages were square root transformed before analysis to normalize distributions. I used linear regression to relate grass cover and species richness to canopy openness across all sites ( N = 80). Regression analysis was also used to relate U. maxima cover to native plant cover on log landings where the alien grass had non zero values for cover ( n = 8); I defined native plant cover as the sum of native grass cover plus cover by non grass v egetation. Frequencies of occurrence of each species were calculated from presence/absence data for each microhabitat type and reported as percentages. I used analysis of covariance (ANCOVA) to test the main effects of time since logging (co variate) and m icrohabitat (fixed factor) on total grass cover. I used canopy openness measurements and species presence/absence data to calculate logistic regression curves for each grass species. The calculated slopes from these regression were used as a disturbance a ffinity
53 metric (DA) for each species. The steeper the slope the more a species prefers open canopy sites; species that occur almost exclusively in undisturbed, closed canopy forest have negative slopes. Because a combination of water limitation and variati on in light availability among microhabitats is likely to influence the distributions of grass with different pho to synthetic pathways, I compared mean DA values of C 3 ( n = 8) and C 4 ( n test. All analyses were performed usi ng SPSS 12.0 statistics software (SPSS Inc. Chicago, Illinois USA). Due to the low logging intensity (4 tree ha 1 ) and the use of reduced impact logging techniques disturbed microhabitats cover only a small portion of logged stands. Felling gaps (9%), skid trails (9%), roads (1%) and log landings (0.2%) combine to cover about 19% of harvested areas. In addition to the microhabitat specific results, I calculated grass cover percentages for harvested forest as a whole weighted by the proportional area of each microhabitat. I did not perform statistical analysis on the fire effects data because the treatments were not replicated in this BACI designed experiment. Instead, I report the total area covered by U. maxima in each yearly census. Canopy openness estimat es for each population represent the weighted mean canopy openness values for all grass patches (weighted by patch size). As such, canopy measures are indicative of canopy openness where U. maxima was growing, not the forest as a whole. Results Grass S pec ies D istributions I encountered 20 grass species 1 5 yr after logging in this tropical dry forest (Table 2 1). Species included representatives from three sub families (Bambusoidea, Chloridoideae, and Panicoideae), a wide range of growth forms, and specie s with C 3 and C 4 photosynthetic pathways. Sixteen of these species are native to eastern lowland Bolivia and four are exotic. Of
54 the alien species, U. maxima and U. brizantha were introduced from Africa and are widely planted as cattle forage in the Neotro pics. The other exotics, Sorghum halapense and Cynodon nlemfuensis are pantropical weeds present at low frequencies and never abundant in my savanna woodlands (K illeen & Hinz 1992a) but instead are either common roadside weeds (i.e., native ruderals) or forest grasses (Killeen 1990; Table 2 1). Grass cover and species richness increased with increasing severity of logging disturbance. Canopy openness, one indicat or of disturbance severity, was greater in all logging microhabitats compared to unlogged forest ( Figure 2 1a). Across microhabitats the greatest mean grass cover and mean number of grass species were on log landings (37% cover, 3.3 species) and roads (17 % cover, 4.2 species; Figure 2 1b,c). Both grass cover and species richness positively correlated with tree canopy openness across all sites ( Figure 2 2a,b). Grass cover across microhabitats was independent of time since logging ( ANCOVA, for time since log ging, F = 0.06, P = 0.81). Although grass cover was typically greater in more disturbed sites, the distributions of individual species differed substantially ( Table 2 1). Disturbance affinities ranged from 6.3 for Pharus lappulaceus a forest grass that did not occur in highly disturbed sites, to 9,9 for the exotic Cynodon which, like the other alien species, occurred only on roads and log landings. Grasses with C 4 photosynthesis had greater mean disturbance affinities than C 3 grasses (5.5 and 0.15, P < test). Urochloa maxima, the most abundant alien grass encountered, formed monodominant stands (up to 91% cover) and was present in 69% of log landings and 38% of roads (Table 2 1). Regression analysis of log landings that had non zero v alues for U. maxima cover yielded a non linear negative relationship between alien grass cover and native
55 plant cover: [native plant cover] = 18.15 ln[ U. maxima cover] + 86.84 ( R 2 = 0.79, P = 0.003, n = 8). In addition to U. maxima, the native ruderals Le ptochloa virgata and Digitaria insularis and the native bamboo Guadua panicultata (a giant rhizomatous grass with woody culms) were abundant on log landings (Table 2 1). Unlike Leptochloa and Digitaria Guadua occurred at low frequency in all microhabitat s including unlogged forest. Guadua was never abundant with the exception of a single log landing where it grew to 4 m tall and covered 75% of the plot. Roads had the highest grass species richness (mean 4.2, maximum 9 species per 30 m 2 plot, Figure 2 1c ); 18 of the 20 grass species recorded were present on at least one road ( Table 2 1). In addition to the exotic U. maxima common grasses on roads were the native ruderals: Leptochloa, Panicum millegrana and Axonopus compressus I encountered only four grass species and sparse grass cover (2.4%) in unlogged forest. Of the forest grasses, three are Bambusoideae: Pharus lappulaceus is a small (50 cm tall) understory species present in 69% of plots; Guadua and Chusquea ramosissima a liana like climber wit h woody culms, were rare in unlogged plots. Also rare in unlogged forest was Lasiacis sorghoidea a caespitose Panicoideae with C 3 photosynthesis, which typically grows 1 2 m tall Lasiacis occurred in all microhabitats but reached high abundance in areas of intermediate disturbance severity with maximum cover in gaps and on skid trails of 21 and 49%, respectively. Alien G rass R esponses to F ire During three years of monitoring after burning 12 ha of partially grass invaded forest U. maxima cover increased from 400 m 2 to 1660 m 2 (a 315% increase; Figure 2 3a). Over this same time period, the control population in unburned forest decreased from 398 to 267 m 2 (a 31% reduction). As such, there was six times more alien grass in the burned forest than in the unb urned forest three years after the fire (4 yr post logging). Canopy openness measured in the
56 burned and unburned forest corresponded with changes in U. maxima cover ( Figure 2 3). Canopy openness in the burn treatment remained at its pre fire post logging l evel of roughly 60% during all three years of monitoring. In contrast, mean canopy openness in the unburned forest decreased steadily each year dropping from 47 to 20% during 4 yr post logging. Inter annual changes in grass cover were not consistent for ei ther the burned or the unburned forest ( Figure 2 3a). After increasing three fold in the first year, grass cover in the burned forest decreased slightly during the second year only to be followed by another increase during year three. Grass cover in the co ntrol area decreased over the course of the study but this decrease occurred after it doubled during the first year of monitoring ( i.e., the second year post logging). Discussion Grass S pecies D istributions The absence of alien grasses in unlogged forest and their abundance in the most disturbed sites in selectively logged stands suggests a strong link between logging and grass invasion in this forest. Positive effects of the severity of logging induced forest changes on grass cover are not surprising giv en that disturbance is an important factor influencing ecosystem invasibility (e.g., Hierro et al. 2006), and that graminoids are widely recognized for their adaptations to disturbed environments (Chapman 1996). The most abundant grass encountered was U. maxima an alien species with C 4 photosynthesis. While a host of morphological (e.g., seed size) and physiological (e.g., drought tolerance) adaptations might contribute to the invasiveness of U. maxima I expect that photosynthetic pathway (i.e., C 3 vs. C 4 ) is especially important: two consequences of logging induced canopy openness are increased light availability and increased water stress (e.g., Ashton 1992) conditions in which C 4 photosynthesis is advantageous. Indeed, a comparison of the disturba nce affinities of C 3 versus C 4 grasses showed that C 4 grasses favored open microhabitats
57 created by logging disturbances whereas C 3 species tended to be more common in closed canopy sites. In tropical dry forests where water stress typically limits plant e stablishment and survival (e.g., Gerhardt 1996; Khurana & Singh 2001), C 4 grasses appear poised to take advantage of resources liberated by low intensity reduced impact logging but only if seeds arrive at suitable recruitment sites. In the seasonally dr y tropical forest I studied, U. maxima dominated sites heavily disturbed by logging (log landings and roads) but did not invade undisturbed forest. Roads and log landings are not solely areas of severe soil and canopy disturbance, but are also sites freque nted by logging trucks that make repeated trips into the forest. Given the documented potential for motor vehicles to disperse seeds long distances ( v on der Lippe & Kowarik 2007), and the fact that alien invasive and native ruderal grasses commonly occurr ed far (>500 m) from parent populations in this forest, logging traffic probably plays a critical role in dispersing grass seeds during the timber harvest. Forest invasion by alien grasses is problematic largely because of their effect on forest flammabi lity. Urochloa maxima can grow 2 3 m tall and produces fine fuel loads of up to 17 Mg ha 1 (standing biomass plus litter, J. Veldman, unpublished data). With the exception of Guadua (a native bamboo), U. maxima is the only species in this forest that form ed monodominant stands that appeared to exclude native plant species. Primarily because of its high productivity, U. maxima was introduced to South America from the Guinea Coast of Africa during the 18 th century and is a preferred forage grass throughout t he Neotropics (Parsons 1972). The species is still cultivated for forage in the Chiquitania region, but U. brizantha is more commonly used. Interestingly, U. brizantha is not nearly as invasive as U.
58 maxima in INPA, even though both species are abundant in pastures and along roads leading to the property. The only native grass that produces fuel loads comparable to U. maxima is Guadua Although rare in this study, Guadua flourished on one log landing (75% cover), and amongst all species was second only to U. maxima in maximum percent cover. Large (100 1000 ha) monodominant Guadua stands are common throughout the region (Killeen 1990), and are characterized by bamboo canopies 2 7 m tall, fuel loads of up to 16 Mg ha 1 (standing bamboo biomass plus litter ), a nearly annual fire regime, low tree density and tree basal area, and charred standing dead trees (Veldman 2008). Vigorous resprouting from rhizomes coupled with production of large quantities of fine fuel makes native bamboo domination a plausible out come for forests subjected to canopy disturbances and fire (e.g., Gagnon & Platt 2008). Surprisingly, most forestry research in Bolivia has been conducted in areas with little or no Guadua despite the fact that G. sarcocarpa and G. weberbaueri form monodo minant stands that can limit forest succession in other parts of the Amazon basin (Griscom & Ashton 2003 2006). In contrast to U. maxima which can produce 1300 seeds m 2 annually (Veldman & Putz 2010 ), rare seed production by Guadua probably explains why this mast fruiting bamboo was not a more frequent colonizer of disturbed sites. Increased grass cover in logging affected microhabitats (which account for 19% of harvested stands) translates to a small (but two fold) increase in total grass cover between unlogged (2%) and logged forest (4%). This modest increase in grass cover belies the extent of grass invasion and the spatial distributions of grasses following logging. Grass colonization of roads and log landings results in a network of grassy clearin gs (log landings) connected by grass invaded corridors (roads) throughout the forest. Alien grasses in these sites are of consequence
59 because they extend into all parts of the forest; in my study site 100% of selectively logged areas fall within 500 m of a logging road or log landing. In the absence of further disturbance, the presence of grasses should be temporary because they will be out competed by the regenerating forest. Nonetheless, re entry by logging machines or forest fires can stimulate furthe r expansion of grass populations. Even grass populations that are initially small can serve as propagule sources for larger invasions if roads or log landing are re used. For example, U. maxima invaded skid trails in a portion of INPA apparently because sk idders re used grass invaded log landings and entered the forest along grass invaded roads thereby dispersing seeds into the forest. This invasion of skid trails not only provides anecdotal evidence that grass populations can expand due to additional distu rbance, but also afforded a site to study the effect of fire on grass invaded forest. Alien G rass R esponses to F ire An experimental burn in grass invaded forest demonstrated that forest fires can increase the extent and duration of alien grass invasion f ollowing selective logging; three years after the fire, the combination of logging and fire resulted in an invasion six times larger than logging alone (1660 m 2 versus 276 m 2 ). Increased light availability due to fire induced canopy tree mortality (e.g., C ochrane & Shultz 1998) may explain much of the positive effect of fire on alien grass population growth. Both burned and unburned forests had open canopies before the fire as a result of selective logging, but whereas the burned forest canopy showed no sig ns of recovery, the unburned forest canopy closed substantially. In addition to light availability, the creation of suitable establishment sites following the fire appears to have contributed to U. maxima population growth. The high intensity fire that c onsumed woody debris in felling gaps created large ash covered clearings throughout the burn plot. Despite no alien grass recruitment in these gaps during the first 2 yr post fire, in the final census (i.e., 3 yr post logging) I found twelve new
60 grass patc hes in burned gaps that totaled 240 m 2 and accounted for 65% of the increase in U. maxima compared to the previous year. As well as increasing the spatial extent of the invasion, the fire increased the duration of grass cover. Under any conceivable scenar io of grass population decline and forest canopy recovery beyond three years post fire (4 yr post logging), grasses in the unburned area will disappear long before they do in the burned forest. It is unclear exactly how long alien grasses can persist in th e absence of additional disturbance. My analysis of grass cover across microhabitats 1 5 yr post logging showed no effect of time. Consequently, a longer period of forest recovery would be needed to detect substantial declines in grass cover. I can, nonet heless, use the unburned forest data (4 yr post logging, Figure 2 3a) as well as data from the grass distribution study (up to 5 yr post logging) to deduce that grasses can persist for at least five years and quite possibly 6 10 years after logging even wi thout fire or re entry by loggers. Conclusion s Selectively logged areas in a tropical dry forest in eastern lowland Bolivia are colonized by both native and alien grasses, especially in areas that are also burned. These grass species should not be confuse d with the native grass flora that composes the species rich native savannas ( cerrados ) of central South America. Rather, the grasses that colonized logged forest are mostly invasive pasture grasses together with a few common native ruderals and bamboo. I identified two grasses with high potential to respond to disturbance from logging and fire. One of them is U. maxima, an alien grass that is already ubiquitous in Neotropical agricultural landscapes; the other one is Guadua paniculata a well known native bamboo that was relatively rare in my study site. The increased cover of these two grasses following selective logging, combined with their impressive fine fuel production, underscore the need for fire suppression in forests under timber management. This study was conducted in a forest that was logged at a relatively low
61 intensity, and has been protected from wild fire in recent history. In poorly managed forests or in areas harvested at greater intensities the implications for forest degradation are far greater. Understanding the vegetation changes that result from human disturbances is important at both local and global scales if we want to develop better forest management methods and improve vegetation dynamic models to predict the fates of tropical fo rests. Land uses that promote grasses over trees, whether intentionally (i.e., pasture establishment) or unintentionally (logging), substantially affect ecosystem flammability, carbon storage, nutrient cycling, evapotranspiration, and forest regeneration. Most of these factors are incorporated in models, both conceptual (e.g., Cochrane 2003) and quantitative (e.g., Cox et al. 2004 ; Golding & Betts 2008) that predict alarming fates for Amazonian forests. Models currently under development should also incorp orate the grass promoting potential of logging and fire and pay particular attention to the different forest fates that are possible under scenarios with and without invasive native and exotic grasses.
62 Table 2 1 Grass cover in five forest microhabitats 1 5 yr after selective logging. Alien species are marked by asterisks. Species are listed in decreasing order of affinity for logging disturbance as estimated from the slope of logistic regression equations relating species presence/absence data to canopy o penness in 30 m2 plots ( N = 80). Disturbance affinities (DA) for species present in < 5% of plots are listed in parentheses. Microhabitats are listed in increasing order of logging disturbance. Percentages for mean cover, maximum cover, and frequency in ea ch microhabitat are listed, respectively ( n = 16 per microhabitat). Unlogged forests Felling gaps Skid trails Log landings Roads Species C 3 /C 4 DA mean max freq. mean max freq. mean max freq. mean max freq. mean max freq. Cynodon nlemfu ensis C 4 (9.9) 6 0.2 3 6 Digitaria insularis C 4 7.3 6 4.9 27 44 0.2 3 25 Sorghum halepense* C 4 (7.2) 0.2 3 6 0.2 3 6 Urochloa brizantha* C 4 (7.2) 0.4 6 6 0.4 6 13 Chloris dandyana C 4 6.3 0.6 9 13 19 Leptochloa virgata C 4 6.1 5.5 27 56 3.8 27 75 Urochloa maxima* C 4 5.8 16 91 69 2.5 18 38 Digitaria bicornis C 4 5.5 0.4 6 25 6 Panicum trichoides C 3 4.6 6 1.1 9 25 1.3 6 50 Paspalum paniculatum C 4 3.8 0.2 3 6 0.4 3 50 Setaria poiretiana C 4 (3.3) 6 Setaria vulpiseta C 4 (3.3) 6 Panicum millegrana C 3 3 6 25 3.0 15 44 Acroceras excavatum C 3 1.6 0.4 3 19 0.6 6 31 1.9 15 31 1.3 9 25 Axonopus compressus C 4 0.9 3.2 3 9 25 Lasiacis sorghoidea C 3 0.6 0.6 6 13 5.5 21 38 7.8 49 69 0.8 12 25 0.6 3 31 Guadua paniculata C 3 0.3 13 0.2 3 13 0.9 9 13 4.7 75 6 0.4 6 6 Olyra latifolia C 3 ( 1.1) 6 6 Chusquea ramosissima C 3 3 .9 0.9 6 19 1.5 24 6 0.4 6 13 Pharus lappulaceus C 3 6.3 0.8 6 69 0.8 6 25 0.2 3 50
63 Figure 2 1 Canopy openness, grass cover density and species richness in five microhabitats arranged in order of increasing disturb ance by selective logging, 1 5 yr after harvest A) Canopy openness; habitats labeled with different letters differ at P < 0.05 (ANOVA, post hoc Tukey, n = 16). B) Grass ground cover and C) grass species richness. Total, native, and alien grasses are denot ed by white, gray, and black bars respectively; asterisks mark habitats that differ from forests in grass cover or species richness at P < 0.05 (ANOVA, post hoc Tukey, n = 16). Columns and error bars indicate mean+1SE for sixteen 30 m 2 plots per microhabit at type; values in parentheses denote maximum values.
64 Figure 2 2 Grass cover and species richness in relation to tree canopy openness 1 5 yr after selective logging for all habitats combined ( N = 80). A) Grass cover: [total grass cover] 2 = 0.0631 [ca nopy openness] + 0.476, P < 0.001; note that the grass cover data were square root transformed before analysis. B) Grass species richness based on species presence/absence data: [number of species] = 0.043[canopy openness] + 0.644, P < 0.001).
65 Figure 2 3 Responses of alien grass populations and tree canopy cover to an experimental fire in a selectively logged Bolivian tropical dry forest. Black triangles and circles denote measurements during the three year study; in gray, pre logging (2004) values are inferred from data measured in unlogged areas of the forest. A) Total Urochloa maxima cover before and 3 yr after an experimental fire in grass invaded selectively logged areas. B) Canopy openness measured above grass patches in the burned and unburned g rass invaded areas.
66 CHAPTER 3 LONG DISTANCE DISPERSAL O F INVASIVE GRASSES B Y LOGGING VEHICLES IN A TROPICAL DRY FOREST 2 Predicting responses of vegetation to environmental factors in human altered tropical ecosystems requires an understanding of the controls on plant population expansion across landscapes ( i.e. long distance dispersal) as well as of factors affecting recruitment at local scales ( i.e. microsite conditions). I studied the roles of light availability, habitat type, soil disturba nce, and seed dispersal in a selectively logged forest in lowland Bolivia where the exotic forage grass Urochloa ( Panicum ) maxima is abundant on roads and log landings but does not invade felling gaps or unlogged forest. Shade house trials and seed additio n experiments with U. maxima revealed that this C 4 grass thrives in high light but also grows in partial shade (10% full sun; but not 1% full sun), and that felling gaps, but not undisturbed forest, are suitable for grass establishment. To determine if se ed dispersal by logging vehicles explains the discrepancy between actual and potential grass recruitment sites, I collected grass seeds that fell from trucks onto log landings located long distances (>500 m) from established grass populations. Trucks dispe rsed an estimated 1800 alien grass seeds per log landing during the early dry season; automobiles also transported seeds of grass (135 seeds /vehicle). The seeds collected (and relative abundances) were the exotics U. (Panicum) maxima (97%) and U. (Brachia ria) brizantha (2%), and the pan tropical weeds Sorghum halapense (1%) and Rottboellia cochinchinensis (0.2%). Grasses invade this forest where disturbance coincides with seed 2 Reprinted with permission from: Veldman J.W., and F. E. Putz. 2010. Long distance dispersal of invasive grasses by logging vehicles in a tropical dry forest. Biotropica ( in press ).
67 dispersal by motor vehicles, while dispersal limitation apparently prevents inv asion of many sites otherwise suitable for grass recruitment ( i.e. felling and natural gaps). Introduction There is lagging (but growing) awareness of novel tropical ecosystems (Lugo 2009) and complex vegetation transitions (Putz & Redford 2010), as wel l as increasing recognition that ecosystems resulting from human activities have not been adequately studied (Hobbs et al. 2006). Although novel disturbances can simply result in altered abundances of native species, plant invasions can cause drastic chang es in community compositions, ecosystem functions, or both ( e.g. Mack & D'Antonio 1998). As scientists consider the effects of environmental change on tropical vegetation ( e.g. Barlow & Peres 2008), it will be important to understand the biology of not o nly the pool of species present in native plant communities, but also the species (native or exotic) that are likely to expand their current ranges (Higgins & Richardson 1999). Range shifts, migrations, and biological invasions cannot occur if propagules do not reach suitable establishment sites or if suitable sites are lacking. To conceptualize the suite of factors that limit recruitment, many studies are conducted within a framework of barriers or filters, physical or biological in nature, that a specie s must pass through in order to invade ( e.g. Parendes & Jones 2000). Numerous human activities break down these barriers by creating conditions suitable for invasive plant establishment ( e.g. increased resource availability via disturbance) or by enhanc ing propagule dispersal ( e.g. Hodkinson & Thompson 1997). Ecosystem altering human activities that both create appropriate establishment sites and enhance propagule dispersal of invasive species demand particular attention. Roadways represent a familiar example of the coupling of human mediated dispersal and the creation of novel environments. Roads are essentially corridors of repeatedly disturbed habitat (Gelbard & Belnap 2003) along which plant propagules are transported by motor vehicles
68 (Clifford 19 59; but see Christen & Matlack 2006). Roadside floras are often dominated by non native species, and studies from high traffic urban areas demonstrate that motor vehicles disperse a disproportionately large number of seeds of invasive species (v on der Lipp e & Kowarik 2007). Although it is plausible that long distance seed dispersal by motor vehicles occurs in remote regions of the tropics, the effects of roads on plant migrations have received little attention in the tropical conservation literature (but see Corlett 2009). Perhaps this disregard is based on the perception that tropical forests resist plant invasions (Levine 2002), or that in comparison to promoting wild fires, logging, hunting and, ultimately, agricultural conversion (Angelsen & Kaimowitz 1999 ; Nepstad et al. 2001), any roles roads play in plant invasions seem secondary. Roads that extend into forests in advance of agricultural frontiers are usually constructed for selective logging (Uhl et al. 1991). In addition to increasing forest acce ss to people, logging opens forest canopies, disturbs surface soils, and renders forests more fire prone (Uhl & Buschbacher 1985 ; Laurance 2003 ; but see Blate 2005). Of these disturbances, exposed mineral soil and increased light availability are thought t o be important for the establishment of invasive grasses from seed (Setterfield et al. 2005), and fires can further exacerbate grass invasions (D'Antonio & Vitousek 1992). Indeed, both native and alien grasses were reported to colonize a Bolivian tropical dry forest in areas most severely disturbed by selective logging and fire (Veldman et al. 2009), but how grass propagules arrived at the invaded sites remains unclear. I investigated factors expected to influence patterns of alien grass invasion follow ing selective logging in a tropical dry forest in eastern lowland Bolivia where alien grasses regularly colonize roads and log landings but not other anthropogenic microhabitats created by selective logging ( i.e. not felling gaps and skidtrails) or unlogg ed forest (Veldman et al. 2009). The
69 aggregation of alien grasses on sites most severely disturbed by logging might reflect habitat specialization due to microsite conditions, but could also be due to high propagule pressure from seeds dispersed by logging vehicles. I conducted a seed addition experiment, along with a shade house study to determine the microsite requirements for grass establishment and to determine if microsite availability limits grass establishment in felling gaps and unlogged forest. To determine whether long distance dispersal by logging vehicles contributes to grass invasion in this forest, I collected debris falling from logging trucks and inspected automobiles to look for invasive grass seeds. The results of these studies should yiel d a more complete picture of how grasses invade selectively logged tropical forest, inform strategies to mitigate invasive seed dispersal, and, draw attention to motor vehicles as dispersers of plant propagules in remote tropical regions. Methods Study S it e This research was carried out in a seasonally dry tropical forest owned by a vertically integrated logging and parquet flooring manufacturer (INPA Parket Ltd., hereafter INPA) located in a region of eastern lowland Bolivia known as Chiquitania. The stud Stewardship Council certified forest of which 900 ha/yr has been harvested since 2002 at an intensity of ~4 trees/ha. The majority of timber is extr acted during the dry season (May to October), during which trees are felled, skidded to log landings, loaded onto flatbed trucks (18 m long with 14 18 wheels) and transported 40 km to a sawmill in the town of Concepcin (population 6,000). An indirect co nsequence of logging in INPA is forest invasion by grasses, the most abundant of which is the African forage grass Urochloa maxima Alien grass distributions are
70 generally limited to roads and log landings ( i.e. the most severely disturbed microhabitats); in a survey of selectively logged areas within INPA, alien grasses were not encountered in unlogged forest, felling gaps, or skid trails (Veldman et al. 2009). A single road provides access to INPA from Concepcin and traverses a landscape mosaic of frag mented forests, swidden agriculture plots, cattle ranches, and native savannas. These anthropogenic and semi natural areas, and the roadside itself, are presumably sources of grass propagules that invade INPA. Alien grasses occur in even the most remote p ortions of the property that were logged extending some 30 km from the forest entrance. S hade H ouse and S eed A ddition E xperiments I conducted two experiments to assess the establishment requirements of U. maxima and thereby to determine if there are suit able establishment sites in the forest where U. maxima does not typically invade ( i.e. the degree to which U. maxima is microsite as opposed to dispersal limited). To assess the viability of seeds used in the experiments (seeds were purchased 2 wks before the experiments from the Union Centro Veterinario in Santa Cruz), I scattered 50 seeds into each of 12 10 x 14.5 cm plastic trays lined with paper towels, maintained water in the trays at the saturation point for the paper, and counted/removed germinate d seeds daily for 2 wks. An average of 48 % of the seeds germinated, of which 97 % germinated within the first 7 d. To estimate the light requirements of U. maxima I assessed seed germination and 1 mo seedling survival ( i.e. emergence sensu Setterfield et al. 2005) under a range of light conditions in shade houses during the early rainy season (October November 2005); shade houses were located in Santa Cruz at the Bolivian Institute for Forestry Research (IBIF; Poorter & Markesteijn 2008). I filled 48 p lastic trays (10 x 14.5 cm) 2 cm deep with forest soil and placed them in one of four light treatments: 0, 1, 10, or 70 % full sun ( n = 12 per light treatment); light
71 intensities were measured with a LI COR model LI 250 light meter (LI COR, Lincoln, Nebrask a USA). I saturated each tray with water and after the water drained ( i.e. the soil reached field capacity), I scattered 50 seeds on the soil surface; the soil was then maintained at about field capacity during the entire experiment. I counted the number of plants in each tray after 7 days to assess initial germination, and then counted and harvested entire plants after 1 mo; seedlings were washed and oven dried to constant weight at 60 C before weighing. I compared seedling survival and plant biomass bet ween light treatments by ANOVA and Tukey post hoc tests. All analyses were performed using SPSS 12.0 statistics software (SPSS Inc., Chicago, IL USA). To investigate how soil disturbance and microhabitat conditions influence U. maxima establishment, I so wed seeds in experimental plots in an area of selectively logged forest. In the early rainy season (October 2005), I randomly selected twenty felling gaps and adjacent areas of closed canopy forest within a 100 ha area that was harvested during the precedi ng dry season. In each felling gap and forest I established plots of either mechanically scarified or untreated soil in a 2 x 2 factorial complete block design with soil treatment and microhabitat as factors and gap forest pairs as blocks ( n = 20 blocks). In each felling gap and forest, I scattered 100 seeds on each of two scarified and two untreated sub plots of 0.1 m 2 such that 200 seeds were sown per 0.2 m 2 per treatment per block. For the scarified ( i.e. disturbed surface soil) treatment I used a mache te to remove all litter and plants and to expose a 1.5 m 2 area of mineral soil prior to scattering 100 seeds in the center 0.1 m 2 of each scarified area. I measured canopy cover over each sub plot using a spherical canopy densiometer held 1.3 m above the g round (Lemmon 1956). After 1 mo I counted the grass seedlings, collected the above and below ground grass biomass, and weighed the plants after oven drying at 60 C.
72 I used a two way ANOVA to test for main effects of soil treatment, microhabitat, and thei r interaction, on the number of seedlings, total grass biomass, and biomass per individual; blocks were treated as a random effect and Tukey post hoc tests were used to compare means among treatments. To reveal which sites in the seed addition experiment were suitable for grass establishment, as opposed to those with plants so small they were unlikely to survive, I counted the plots that had a mean per plant biomass > 3 mg. This biomass cut off was based on the results of the shade house experiments in whi ch the minimum biomass measured for plants exposed to 10 % full sun was 3 mg/individual; 10 % full sun was the minimum sunlight exposure required for high grass survival and growth. I with Chi square. Seed R ain f rom L ogging V ehicles During the early dry season (May 2008) we sampled alien grass seed rain falling from logging trucks ( n = 8), and collected seeds from automobiles ( n = 3) on three different log landings in the forest. Log l andings are large (~700 m 2 ) clearings where logs are organized and loaded onto trucks; abandoned log landings are often colonized by grasses. New log landings are appropriate sites for studying long distance seed dispersal by motor vehicles because they ar e the primary destination of trucks and automobiles entering the forest. I verified that each log landing used in this study was accessed by a newly created road and was at least 500 m from previously logged forest, pre existing roads, or existing alien gr ass populations with a geographic positioning system (GPS) and a map of the property. In addition I searched for grasses within 3 m of the road (200 m x 6 m sampled) leading away from each log landing to confirm that alien grasses were not present. To est imate the number of seeds dispersed by trucks to log landings I randomly positioned 12 plastic tarps (115 x 74 cm each) totaling 10.2 m 2 beneath each truck immediately
73 after it arrived at the landing. As timber was loaded onto the trucks, dirt and seeds we re dislodged and fell onto the tarps. After each truck was loaded, I removed the 12 tarps and combined the material collected into a single sample. In addition to the seed rain from log loading, I collected debris that workers swept off truck beds. Flatbed trailers can accumulate bark and soil that is removed before log loading; I collected this debris onto a 5 x 5 m tarp as it was swept off the trailer. I did not ask forest managers to clean the truck beds for the sake of this study and did not anticipate this source of seed rain. I estimated the number of seeds carried by automobiles by collecting all soil and other debris that I were able to dislodge by hand using brushes and wooden spatulas from one randomly selected front and one rear wheel well from each car sampled. I also collected seeds present on the removable interior floor mats from the passenger compartments. In comparison with the truck data, which represent a true estimate of seed rain ( v on der Lippe and Kowarik 2007), I do not know what prop ortion of the automobile seed is deposited on log landings a limitation of all studies that collect seeds attached to motor vehicles (e.g., Clifford 1959 ; Zwaenepoel et al. 2006). To determine the number of alien grass seeds in the truck and automobile debris I manually sorted through the samples to separate seeds from other material. For samples too large to process in their entirety, I analyzed a series of sub samples. For samples up to 2.5 kg, the sub samples comprised at least 10 % of the total sampl e fresh weight; for larger samples a minimum of 0.25 kg was sub sampled. I identified the seeds to species and determined the number of full seeds (as opposed to empty caryopses) in the samples by cutting each seed with a razor blade and inspecting it unde r a dissecting microscope. I report only the number of full seeds collected. I chose visual inspection of the samples rather than germination trials to assess the numbers and
74 identities of grass seeds carried to log landings by vehicles because the seeds o f the most common alien grasses in this forest are relatively large (3 6 mm) and are easily separated from soil and other debris. Furthermore, grass seedlings are extremely difficult to identify to species whereas seed characteristics are often diagnostic. Finally, required seed dormancy periods can be long and vary even within species (Usberti & Martins 2007) making reliable germination trials lengthy. Based on the number of seeds that fell onto tarps (10.2 m 2 ) during log loading I calculated total seed r ain per truck by extrapolating to the area of each truck (~48 m 2 ). To calculate total seed rain densities in seeds/m 2 for log landings I assumed 10 truck visits per landing and 700 m 2 as the area of each landing. To calculate the number of seeds per vehicl e for passenger cars I extrapolated from the two tires/wheel wells sampled to four and used the total number of seeds from the interior floor mats. By sampling only tires and wheel wells I underestimate the total number of seeds carried on the exterior of vehicles. Likewise, more exhaustive cleaning of vehicle interiors would certainly yield higher seed numbers than the floor mats alone. On the other hand, because I do not know what proportion of the seeds attached to automobiles is deposited in log landing s, my estimates of alien grass seed rain on log landings are conservative insofar as they only include seeds that fell from logging trucks. To estimate the production of seeds readily available to passing vehicles, I counted the seeds of U. maxima th e most invasive grass in INPA, along 200 m of road abandoned 1 yr prior to my study. The compact soil of the road, lack of traffic, and scarce rain during the period of seed production (early dry season) made it possible to collect fallen seeds from the su rface of the roadbed as an estimate of total availability. I collected seeds from 12 50 x 50 cm plots in June
75 2008 (soon after seed shattering) and counted the number of full seeds, as described above. I used these data as an estimate of potential U. max ima seed rain onto grass invaded roads. Results Conditions R equired for G rass E stablishment In shade houses, mean initial (7 d) germination rates were equal among treatments, but were lower than in seed viability trials (33 38% and 48 % respectively). U rochloa maxima emergence ( i.e ., germination and survival to 1 mo) was equal under high light (70% full sun; 38% survival) and moderate shade (10% full sun; 38% survival); few seedlings survived for 1 mo under deep shade (1% full sun; 3% survival) and none survived without light ( Figure 3 1a). Although emergence did not differ between 70 and 10 % full sun, the seedlings in high light were nearly four times larger than those grown in moderate shade (19 and 5 mg/plant, respectively); the few seedlings that surv ived 1 mo in 1 % full sun appeared moribund and weighed only 0.5 mg/plant ( Figure 3 1b). Although not as large as those grown in 70 % full sun, plants in 10 % light were clearly maintaining positive carbon balances as evidenced by their growth and apparent vi gor. The minimum biomass measured for any of the 12 flats under moderate shade was 3 mg/individual. In the in situ studies that follow, I used this value (3 mg/plant) as a cut off to separate suitable establishment sites from plots with plants that are so small that they are unlikely to survive. In the seed addition experiment, canopy openness values for plots located in closed canopy forest ( n = 20) and felling gaps ( n = 20) were 8 2% and 61 13% (mean 1SD), respectively. Overall, Urochloa maxima emergence was low; only 1.8 % of all seeds sown survived for 1 mo. The individuals that did survive were overwhelmingly those sown in closed canopy forest plots on scarified soil (10.5 seedlings/0.2 m 2 of 200 sown; Figure 3 2a); but despite the between trea tment differences in numbers of seedlings, total grass biomass did not differ
76 ( Figure 3 2b). This result was partly because of high variability in total grass biomass values for felling gaps ( i.e. many plots with no emergence), but was primarily due to th e huge discrepancy in plant sizes between gaps and forests; seedlings in gaps were eight times larger than plants growing in the shaded understory ( Figure 3 2c). In felling gaps, 35 % of untreated soil plots and 20 % of scarified plots had plants > 3 mg/indi vidual; in contrast only 5 % of scarified plots and none of untreated soil in the forest had a per plant biomass that indicated the site was suitable for grass establishment ( Figure 3 2d). Seed D ispersal by Logging V ehicles Log trucks were estimated to hav e dispersed 1800 alien grass seeds per log landing in the early dry season ( Table 3 1). The majority of seeds came from flatbed cleaning (1750 seeds/log landing). The seed rain during log loading was more modest: 5 seeds per truck visit or 50 seeds/landing Automobiles carried alien grass seeds attached to tires and wheel wells (50 seeds/vehicle) and on interior floor ma ts (90 seeds/vehicle; Table 3 1). Alien grass seeds collected from vehicles on log landings included three species known to colonize selec tively logged forest ( U. maxima U. brizantha and Sorghum halapenses; Veldman et al. 2009), and another exotic grass, Rottboellia cochinchinensis, not previously documented in INPA. The majority of seeds were U. maxima (97%) followed by U. brizantha (2%), S. halepense (1%), and R. cochinchinensis (0.2%). Both U. maxima and U. brizantha are highly productive African grasses planted as forage throughout the Neotropics (Parsons 1972). Sorghum halepense and R. cochinchinensis are large (2 3 m tall) pantropical weeds that are common along roadsides in Chiquitania, particularly on wet sites. The number of U. maxima seeds collected from the surface of a grass invaded road was estimated as 1270 seeds/m 2 equivalent to ~8000 seeds/m of road.
77 Discussion Although lo gging opens forest canopies and disturbs soil factors generally thought to influence the success of grass invasions ( e.g. Setterfield et al. 2005) it appears that canopy openness is more important to establishment of U. maxima in this forest. Soil dis turbance increased grass emergence in closed canopy forest, but in felling gaps ( i.e. open canopy, high light conditions) there was no effect of soil scarification on survival of sown grass seeds. Urochloa maxima seeds probably contain sufficient energy r eserves to grow through leaf litter but can only then survive where there is sufficient light. The growth and survival of plants grown in shade houses under 10 % full sun nonetheless demonstrates that this C 4 pasture grass is more shade tolerant than might be expected, and suggests that areas of logged forest with even intermediate levels of canopy and soil disturbance ( i.e. skid trails) are suitable recruitment sites. In the months following disturbance by selective logging, the forest contains more pot ential than realized recruitment sites for alien grasses. The seed addition experiment showed that felling gaps are suitable sites for U. maxima establishment. Given that alien grasses were not found in felling gaps (~10% of the area of selectively logged forest) but were common in the 1 % of logged forest occupied by roads and log landings (Veldman et al. 2009), I conclude that few seeds ever arrive in felling gaps. It is also unlikely that seeds are regularly dispersed into unlogged stands, but even if the y were, low light availability and lack of disturbed soil would preclude their establishment. This contention is consistent with the assertion by Fine (2002) that undisturbed tropical forests are perceived as resistant to invasion because many introduced s pecies, like the grass in this study, are light demanding. Seed transport by trucks appears to be the primary method of grass seed dispersal into this selectively logged forest. In the early dry season, during the peak of grass seed production and the com mencement of timber extraction activities, log trucks dispersed large quantities of
78 alien grass seeds long distances into what was old growth tropical forest. Sixty nine percent of log landings in this forest contain alien grasses (Veldman et al. 2009); I provide at least circumstantial evidence that these populations were largely established from seeds dispersed by trucks. I also found grass seeds on the tires and in the wheel wells of passengers vehicles, a well documented phenomenon (Clifford 1959), but did not estimate the number of seeds they actually deposited on log landings. The seeds collected from logging vehicles almost certainly originated from roadsides between the sawmill and the log landings; given ~8000 seeds/ m produced along an estimated 4 km of exotic grass invaded roadside, some 30 million alien grass seeds were available to vehicles traveling into the forest. In general, motor vehicle dispersal helps explain why alien grasses are limited to roads and log landings when many other appropr iate establishment sites are opened by selective logging. The mechanism of seed attachment to motor vehicles for the grasses encountered in this study appears to be different than is commonly proposed for other vehicle dispersed roadside floras. Other st udies have emphasized the importance of precipitation and mud for attaching seeds to passing motor vehicles (Zwaenepoel et al. 2006), but in this forest seeds were transported and dispersed in the absence of precipitation and were primarily carried on or i n vehicles as opposed to being stuck in attached mud. Fertile culms of U. maxima U. brizantha S. halepense and R. cochinchinensis grow 1 3 m tall with inflorescences that hang over roadways. When vehicles pass with the windows open, seeds fall into th e passenger compartments; when people exit automobiles they presumably carry some of these seeds with them (Wichmann et al. 2009). The flatbed trailers of trucks are ~1.5 m tall, an ideal height to collect grass seeds that get lodged amongst bark fragments from logs from previous trips. Indeed, the vast majority of seeds delivered to log landings in this study were swept from the beds of log trucks.
79 A lien grass invasions in forests are worrisome for various reasons including their effects on forest flamma the rainy season (when logging operations cease) makes transit of primary logging roads difficult. Clearing the roads using bulldozers has become both a nuisance and cost for fores t managers (D. Velasquez, pers. comm. ). Fortunately, simple measures could be taken during periods of grass seed production (late rainy and early dry season) to substantially reduce seed dispersal by logging activities in this forest. Truck beds should be swept off at the saw mill after delivery of timber from the forest and once again before entering the property to remove accumulated seeds. Seed transport by automobiles could be minimized if both the interior and exterior of vehicles were regularly cle aned to remove accumulated seeds. Although not included in this study, I expect that heavy equipment ( i.e. bulldozers and graders) also move invasive plant propagules. Forest managers should time road maintenance activities so that grass invaded areas are avoided during periods of seed production. When this is not feasible, clearing of roadways using machetes or mowers such that grass seeds fall to the ground at the site of seed production would limit the number of seeds that could fall onto truck beds or accumulate in the passenger compartments of automobiles. This study links vehicle dispersal of invasive species to an active invasion process in a remote area of tropical forest. The sampling design used guaranteed that collected seeds traveled at least 500 m, but given that new roads that extend up to 5 km into previously unlogged forest support populations of alien grasses and that alien grasses occur at the farthest points of the property (30 km from the entrance), it is clear that grass seeds are comm only transported much farther from parent populations. In light of my results, and the previously demonstrated potential of motor vehicles to drive biological invasions ( v on der Lippe and Kowarik 2007), biologists and
80 ecosystem managers should be aware tha t human movements can unintentionally disperse seeds even in areas intended for native ecosystem conservation through natural forest management. Table 3 1 Alien grass seeds collected from motor vehicles arriving at log landings during selective loggin g of a seasonally dry Bolivian forest (mean 1 SD). Alien grass seed rain Vehicle class Source n Seeds/Sample Seeds/Log deck Seeds/m 2 log trucks log loading 8 5 6 50 0.1 flatbed cleaning 5 350 140 1750 2.4 automobiles tires/wheel wells 3 50 80 interior floor mats 3 90 50 Total 1800 2.5 Figure 3 1 Survival and total biomass at 1 mo of Urochloa maxima seedlings emerging from 50 seeds sown and grown under 0, 1, 10 or 70 % sunlight for 1 mo. (mean+1SE; letters indicate differences between light treatments at p < 0.05, ANOVA, post hoc Tukey, n = 12). (A) Number of seedlings surviving 1 mo as a proportion of the number of seeds planted. (B) Average biomass per seedling for live plants at the end of 1 mo.
81 Figure 3 2 Em ergence of Urochloa maxima seedlings (mean+1SE) 1 mo after planting 200 seeds in 0.2 m 2 plots in a 2 x 2 factorial design combining two habitat treatments (unlogged forests and felling gaps) and two soil treatments (scarified soil and untreated soil; bar c olors: black = scarified soil; white = untreated soil). (A) Number of seedlings per 0.2 m 2 plot; there were significant effects of soil treatment, habitat type and an interaction between factors; letters indicate differences between treatments at p < 0.05 (two way ANOVA, post hoc Tukey, n = 20 treatment blocks). (B) Total grass biomass per 0.2 m 2 plot; no differences. (C) Biomass per individual for plots with live plants; effect of habitat type only ( n = 17, 15, 6, 9; plots with no plants were excluded). ( D) Proportion of plots suitable for U. maxima establishment X 2 = 10.0, df = 3, p < 0.02).
82 CHAPTER 4 GUADUA PANICULATA (BAMBUSOIDEAE) IN TH E BOLIVIAN CHIQUITAN IA: FIRE ECOLOGY AND A POTENTIAL NATIVE FORAGE GRASS 3 Guadua paniculata Munro is a fire adapted bamboo that dominates large areas of the Chiquitano dry forest in eastern lowland Bolivia. In this study I present data from eight sites (bamboo stands with 35 100 % Guadua cover) located across the Chiquitania region to 1) investigate the interact ion between Guadua fire, and trees; 2) describe the vegetation associate d with Guadua ; and, 3) estimate Guadua forage production and analyze its nutrient content s Guadua stands supported both low densities of trees and low tree basal area (240 trees/ha, 5 cm DBH; 5.8 m 2 /ha) presumably due to frequent intense fires that result from fuel loads produced by the bamboo Guadua is able to produce up to 16 Mg/ha of fine and small biomass (< 25 mm; 8 4 Mg/ha, mean1SD) that increases flammability and fire in tensity. Guadua appears able to convert forests to bamboo stands by contributing to fire induced tree mortality. As such, it will be important to include Guadua in studies of the conservation and management in this fire susceptible region. On the positive side, Guadua leaves serve as cattle forage and bamboo stands produce up to 4.7 Mg/ha of high quality forage (10.5% protein content) Although more research is needed to determine how to make the forage produced a vailable for animal production, Guadua coul d be utilized as a native forage species in a region that is experiencing high deforestation rates due to the cultivation of exotic pasture grasses. 3 Reprinted with permission from: Veldman, J. W. 2008. Guadua paniculata (Bambusoideae) en la Chiquitania Boliviana: Ecolog a de fuego y la oportunidad para un forraje nativo Revista Boliviana de Ecologa y Conservacin Ambiental 24 :65 74.
83 Introduction Disturbance regimes in the Chiquitania of Bolivia are changing due to human activities that in crease the frequency of forest fires, disturb forest s during timber extraction, and replace forest s with crops or planted forage grasses (Steininger et al. 2001). Forest fires kill trees, cause nutrient losses, and increase the risk of future fires (Lauran ce 2003). Timber extraction increases forest susceptibility to fires in large part due to the construction of logging roads that open the canopy and provide ignition sources (Nepstad et al. 1999). The expansion of cattle ranching does not simply represent deforestation but the substitution of a diversity of forest species by one or two exotic forage grasses; additionally planted pastures are frequent sources of fires that spread into adjacent forests (Uhl & Buschbacher 1985 ; Martinez M. et al. 2003) Given the current increases in fire frequency, large areas of forest affected by logging, and high rates of deforestation for cattle production it is important to know more about the plant species favored by these conditions and thus likely to increase in abun dance in Chiquitania. A native lowland Bolivian species that is well adapted to disturbances, including fire, is the bamboo Guadua paniculata Guadua has a widespread distribution and is estimated to domina (Killeen 1990). Guadua can fo rm large (up to 1000 ha) bamboo dominated patches locally known as small colonies (10 100 m 2 ) in forests. Although Guadua is a dominant species in many parts of the region (Jardim et al. 20 03; Killeen et al. 1990) it has received little attention in previous studies of Chiquitano dry forest. Normally Guadua is described as being characteristic of woodlands or scrub (i.e., cerrad o or mattoral; Killeen et al. 1990 ; Killeen & Hinz 1992a, re spectively) The Bolivian vegetation map by Na varro and Ferrreira ( 2007) includes Guadua chaparral secondary forest The fact that Guadua is simply considered a component of these vegetation types, however they are lab eled, ignores the
84 dynamic potential of bamboo to promote ecosystem flammability and fire intensity (e .g., Keeley & Bond 1999 ) that can contr ibute to forest tree mortality. Killeen et al. ( 1990 ) and Jardim et al. (2003) recognize d that Guadua est and Killeen et al. ( 1990 ) note d that Guadua stands frequently burn. Guadua like many bamboo species, resprout s readily after fires and can produce large quantities of highly flammable fine fuel ( e.g., Gagnon & Platt 2008 ). The grass fire cycle proposes that the interaction of grasses and fire represents a positive feedback that can tr ansform an ecosystem originally dominated by trees to one dominated by grasses and characterized by frequent fires G iven its c haracteristics, Guadua seems likely to fit within this grass fire framework. The broad objective of this study was to determine if the interaction of Guadua and fire d eserves greater attention in our understanding of the ecology and conservation of Chiquit ano dry forest. Because of its traditional uses in house construction, Guadua is recognized as a non timber forest product ( Sandoval 2001 ), but is less well know n as a forage for livestock. The palatability of Guadua to cattle was recognized by Killeen (1 990) but cattle ranchers generally view Guadua as a weed (B. Colombara, pers. com.). Although some ranches and communities graze cattle in bamboo stands, clearing of Guadua to sow cultivars of Urochloa spp. (formerly Brachiaria spp. and Panicum maximum ) is very common (J. Veldman, pers. obs.) Given that deforestation and the expansion of planted pasture grasses threaten Chiquitano dry forest, the second objective of this study was to provide basic informa t ion about the potential of Guadua as a native forag e Understanding the potential uses of this species is relevant given that its abundance in the region could increase due to anthropogenic forest disturbance and increased incidences of forest fires.
85 Methods The ecosystems of Chiquitania Boliva located on the southern border of the Amazon Basin are as dynamic as they are diverse. For example, as recently as 3000 years ago areas that are currently forests were savannas (Mayle et al. 2000). The native savannas of the region cerrados nia and typically burn annually (Killeen and Hinz 1992a). Although the present mosaic of forests and savannas largely correspond s with soil characteristics, topography, and hydrology (Killeen et al. 1990), c limate change in the long term (Hutyra et al. 200 5) and changing disturbance regimes in the short term ( i.e., fire, logging) are expected to alter v egetation distributions regionally (e.g., Nepstad et al. 2001; Nepstad et al. 2004) T o better understand the vegetation associated with a species that is we ll adapted to fire and understudied in Chiquitania, I analyzed data from eight Guadua stands ( minimum 25 % bamboo cover) in the P rovinces of N flo de Chaves and Velasco in the Department of Santa Cruz Bolivia ( Table 4 1). The Guadua stands are part of a mor e extensive project studying the conversion of forest to savanna and that includes 29 sites across the Chiquitania (Chapter 1) These sites were selected at random from a satellite image analysis that identified areas that changed from forest to savanna (i ncluding Guadua stands) between 1986 and 2005. Six of the eight bamboo stands included in this study are recently derived (i.e., areas that were forest in 1986) and the other two were already established in 1986. Field sampling took place during the late r ainy season and early dry season (February May 2008). S tudies on the potential Guadua forage production and quality were conducted on the Estancia Cacarachi a privately owned cattle ranch (694285 8185262; UTM WGS 84 Zone 20 S) located 22 km west of the town of San Ignacio de Velasco. Guadua stands at Estancia
86 Cacarachi were sampled during the first week of June 2008 following 9 months of leaf production after a fire in 2007. Biophysical C haracteristics of B amboo S tands To describe the vegeta t ion associat ed with Guadua and investigate the relationship between Guadua fire, and trees, I randomly positioned two to five 20 m transects in each bamboo stand ( n = 8). Along each transect I measured: 1) percent cover by Guadua by noting its presence or absence at 41 points spaced every 0.5 m; 2) the height of Guadua w h ere present; 3) Guadua biomass < 25 mm diameter i n three clip plots of 1 m 2 (3 m 2 /transect ; all culms rooted in the plot were harvested ); 4) biomass of Guadua litter in three subplots of 50 x 50 cm (0. 75 m 2 /transect), separating Guadua litter from litter from other sources. Samples were weighed in the field and subsamples were collected and dried for 48 hr at 65 C to correct for moisture content. Within 4 m of the same transects (20 x 8 m), I identified all trees >5 cm, and measured their diameter at breast height (DBH) and height. At the center and endpoints of each transect I measured tree canopy density using a spherical densitometer held at 1.3 m (Lemmon 1956). Canopy density measurements do not incl ude Guadua As an indication of the intensity of past fires, I measured the maximum char height on the trunk of each tree (Fry 2008) To estimate grazing pressure, I counted the number of cattle trails that crossed extensions of each transect (number of trai ls per 40 m ; Lange 1969 ; Pringle & Landsberg 2004) Forage P roduction and N utrient C ontent Total aboveground biomass of Guadua cannot be converted directly into an estimate of forage production because cattle only consume the leaf blades leaf sheaths, an d the smallest branches enclosed by the sheaths In a Guadua stand in Estancia Cacarachi, I randomly located four 20 m transects following the methods described above to estimate the total standing biomass of Guadua I then separated and weighed all of the material suitable as forage I dried
87 s ubsamples to correct for moisture content and then related total standing biomass (< 25 mm diameter) to potential forage production using linear regression Forage samples were analyzed for protein, phosphorous, and calcium content at the laboratory of CIAT (Centro de Investigacin Agrcola Tropical) in Santa Cruz. P rotein and phosphorous content of laminas collected in eight other Guadua stands ( Table 4 1) were also analyzed. Data and A nalysis M easurements from each transect were averaged to obtain a single value for each of the eight si tes; the values I report are the mean 1 SD for n = 8 bamboo stands. Regression analyses to relate tree, Guadua fire data and forage production were performed in SPSS 12.0 (SPSS I nc. Chicago, Illinois). Data not normally distributed w ere log transformed prior to analysis Results and D iscussion Guadua and F ire Guadua stand s in Chiquitania are characterized by frequent fires, low tree density, and open tree canopies ( Table 4 2). The sites in this study had 35 100% cover of Guadua which formed a canopy 2.6 m tall (mean) with maximum bamboo canopy heights of 3 7 m ( Table 4 2). All bamboo stands had burned within the previous two years and supported large bamboo fuel loads (84 Mg/ha; Table 4 2). The low tree densities associated with Guadua are probably due to a combination of high tree mortality and limited tree regeneration caused by frequent fires. Guadua is light demanding and its abundance (i.e., % cover) appears to depend on tree canopy openness. Guadua abundance decreased as tree canopy cover increased ( Figure 4 1). Based solely on th e s e observational data Guadua may simply grow in sites that cannot support tall forest trees and lack closed tree canopies (e.g., Killeen et. al 1 990). Likewise, one could conclude that Guadua is a species typical of secondary forests, growing where humans have
88 disturbed the forest canopy (e.g., Navarro and Ferriera 2007). Although soil fertility and canopy openess certainly influence the abundance and distributions of Guadua a notion of Guadua as merely growing were there are few trees, ignores th e possibility that Guadua is an active modifier of the ecosystems were it occurs. In the Chiquitania, Guadua can increase its own access to light by incre asing forest flammability and the intensity of fires that kill competing trees. Bamboo fuel loads measured in Guadua stands were negative ly correlated to both tree density and tree basal area ( Figure 4 2ab). In these same sites char heights on tree trunks were also negatively correlated with tree density ( Figure 4 2c), suggesting t hat the intensity of past fires influenced the number of trees present. Guadua stands had lower tree density (240180 trees/ha) and lower basal area (5.84.0 m 2 /ha) compared to f orests in the region where Guadua is absent [e.g., 914 trees/ha with basal area of 27.6 m 2 /ha in Lomerio ( Killeen et al. 1998); and 427 trees/ha with basal area of 21 m 2 /ha in INPA near Concepcin ( Villegas et al. 2009) ] The trees that co occurred with G uadua included both species typical of forest s as well and species typical of savanna s. In total I recorded 43 species of which seven accounted for 50% of the stems measured These common species include trees typical of savanna environments: Magonia pubes cens A. St. Hil. (Sapindaceae; 15%), Luehea paniculata Mart. (Malvaceae, Tilioidae; 10%), Phyllostylon rhamnoides (J. Pois.) Taub. (Ulmaceae ; 6%) ; as well as typical forest species: Co m bretum leprosum Mart. (Combretaceae; 6%), Anadenanthera culubrina (Vell .) Brenan (Fabaceae; 5%), Aspidosperma rigidum Rusby (Apocynaceae; 4%), and Eriotheca roseorum (Cuatrec.) A. Robyns (Malvaceae, Bombacoidae; 4%). Of these common trees, Anadenanthera and Phyllostylon are the only timber species. Other timber species (and t heir relative abundances) were the savanna species Tabebuia roseo alba (Ridl.) Sandwith
89 (Bignoniaceae; 2.1%), and forest species Caesalpinia pluviosa DC (Fabaceae; 1.9%), Astronium urundeuva (Allemo) Engl. (Anacardiaceae; 1 6%), Aspidosperma tomentosum Ma rt. (Apocynaceae; 1. 4%), Aspidosperma cylindrocarpon Mll. Arg. (Apocynaceae; 0 9%) y Tabebuia serratifolia (Vahl) G. Nicholson (Bignoniaceae; 0 5%). Disproportionate to their relative abundance forest species Anadenanthera and Astronium accounted for 1 5% and 13% of dead standing trunks respectively. A possible explanation is that frequent fire s are favoring savanna species better adapted to tolerate fires (e.g., Magonia ) while contributing to the low densities of forest species ( Hoffmann et al. 2009; P inard & Huffman 1997). In the two old Guadua stands only 20% of trees were forest species (80% were savanna species), whereas trees in the six newly formed bamboo stands 53% of trees were forest species (47% savanna species). The coincidence of Guadua and fire argues that more attention be given to this species by efforts to conserve and manage the Chiquitano dry forest. Underscoring the potential implications of Guadua proliferation for forest s of the region, I found a clear relationship between signs of high fire intensity and large bamboo fuel loads in areas with low tree densities. Additionally, six of the eight Guadua stands were forests 22 years prior to this study, indicating that the interaction between Guadua and fire has almost certainly favored t he expansion of bamboo at the expense of forest tree species. To advance a dynamic notion of the interactions between Guadua fires, and trees I propose two diagrams ( Figure 4 3a b) to conceptualize the interaction between Guadua and fire, and to elucidat e factors that determine the abundance of Guadua and trees in the Chiquitania. Guadua is not the only bamboo that benefits from a self catalyzed disturbance cycle to form bamboo dominate stand s within tropical forests. In Peru for example, Guadua weberau eri and
90 Guadua sarc o carpa climb and pull down trees and proliferate in the high light environment s of the tree fall gaps it helps create (Griscom & Ashton 2006). Studies of the Chiquitano dry forest have primarily focused on areas containing high densiti es of timber species and very little Guadua ( Fredericksen et al. 2000; Killeen et al. 2005; Killeen et al. 1998) As a consequence, the implications of Guadua for sustainable forest management has yet to be explored (but see Rockwell et al. 2007 for studie s on G weberaueri and G sarc o carpa ) Guadua is often abundant along roadsides and in forest gaps created by logging activities (Veldman et al. 2008), but without studies that include pre and post logging data, or that are conducted in forest s with large r Guadua populations, it is difficult to know the degree to which logging increase s bamboo abundance. In a scenario in which Guadua proliferates following logging and then further expands due to subsequent bamboo fueled fires ( Figure 4 3a), the consequence s for sustainable timber management could be catastrophic. I argue that more areas that contain Guadua should be studied in Chiquitania and that Guadua be included in forest monitoring projects. Guadua as a N ative F orage Although Guadua is associated w ith fire and forest degradation, it has a high potential to serve as cattle forage in Chiquitania given its productivity and nutrient content s ( Table 4 3; Table 4 4). The production of bamboo forage measured in this study is comparable to exotic forage gra sses in the genus Urochloa (formerly Brachiaria brizantha B. decumbens and Panicum maximum ), and can even exceed production by U. brizantha the most commonly planted forage grass in Chiquitania ( Table 4 3). Impressively, the Guadua harvested in the Esta ncia Cacarachi yielded 4.73.3 Mg/ha of forage ( Table 4 3). Based on the equation that relates total Guadua biomass to available forage ( Figure 4 4), I calculated that the other eight
91 Guadua stands in this study ( Table 4 1) produced less than Cacarachi but were nonetheless very productive (1.51.3 Mg/ha, range 0.3 4.1 Mg/ha ; Table 4 2 ; Table 4 3). In addition to productivity nutrient content (i.e., protein and minerals ) is an important aspec t of forage value for livestock The material harvested at Estanc ia Cacarachi had protein content s sufficient to meet cattle nutritional requirements ( Table 4 4). The 10.5 % protein measured in the Guadua forage is much higher tha n the values of 6 .0 to 7.5 % reported for U. brizantha and U. decumbens in other parts of the Department of Santa Cruz (Joaquin & Herrero 2003). Considering leaf laminas alone, p rotein content was even higher (13.2 % ; Table 4 4 ). The Guad u a forage collected also contained high concentrations of phosphorous and calcium, minerals critical for animal nutrition ( Table 4 4). The objective of p resenting data on the production and nutritive quality of Guadua is to suggest the possibility that this native grass be integrated with other forages for cattle production. That said there are some limitations of Guadua that need to be overcome. T o promote Guadua without any sort of management would result in bamboo stands with canopies exceeding 2 m tall and the majority of leaves out of reach of animals. Additionally, high culm densit ies can limit access to grazi ng cattle and make it difficult for humans to enter Guadua stands Another important consideration is that some populations produce spines on the culms If Guadua were to grow in a man n er that made all of its leaves accessible to cattle, this grass would l ikely be the preferred forage species in Chiquitania. However promising, the potential of Guadua as a native forage will remain unrealized until we determine how to manage its growth. Some possible management methods include: 1) planting and harvest of Gua dua as a cut (not grazed) forage; 2) opening trails into bamboo stands to facilitate cattle access; 3) pruning of Guadua culms in order to maintain them at a height accessible to cattle.
92 Unfortunately, base on this study alone, we cannot know if the high forage production estimated for Guadua can be maintained under management. More research is clearly needed to determine how to manage and utilize this species that will likely become more and more abundant in Chiquitania i f the frequency of forest fires r emains high. Given its nutritive qualit ies and high productivity it seems an obvious focus for further study. As evidence that Guadua really is palatable to livestock, one night during field sampling I found a horse with its head in the bed of my pick up t ruck ; it had e aten 2 kg of Guadua collected the previous day. Table 4 1. Locations and number of transects for eight sites used to study Guadua in Chiquitania Bolivia ( organized west to east ). The coordinate s and elevations are average values obtained w ith a GPS at the center of each transect Based on a satellite image change analysis (1986 2005, Chapter 1), New indicates n ewly formed Guadua stands in areas that were forest in 1986, and Old are those bamboo stands that were already established in 1986 Coordinates are UTM WGS 84 Zone 20 S Age Zone or nearest town Province Coordinates Elevation (m) Number of Transects New San Ramn 553792 8183011 375 3 New Lomerio 622004 8158285 417 3 New Concepcin aves 631775 8244874 412 4 New San Ignacio Velasco 687416 8187409 452 5 Old San Ignacio Velasco 696168 8184592 449 5 New Carm en de Ruiz Velasco 715546 8249468 281 2 New San Ignacio Velasco 716102 8209649 410 2 Old Paragua Velasco 725938 8264433 222 2
93 Table 4 2 Biophysical variables measured in Guadua stand s in Chiquitania ( n = 8). Mean 1SD Range (min max ) Guadua cover (%) 67 21 35 100 mean height (m) 2 6 0 7 1 3 3 5 mximum height (m) 5 0 1 5 3 0 7 0 standing biomass (Mg/ha) 6 5 3 7 1 8 14 bamboo litter (Mg/ha) 1 5 1 0 0 2 2 6 potencial forrage (Mg/ha) 1 5 1 3 0 3 4 1 f i ne fuel (Mg/ha) 8 0 4 1 2 0 16 Trees canopy cover (%) 30 22 9 63 mean canopy height (m) 4 2 3 6 0 11 mximum canopy height (m) 14 8 0 25 density(trees/ha) 240 180 25 530 basal area ( m2 /ha) 5 8 4 0 0 12 Fire char height (m) 2 2 1 8 0 6 5 2 time since fire (aos) 1 2 0 4 1 2 Grazing cattle crossing s (trails/40 m) 0 8 1 0 0 2 6
94 Table 4 3 Compariso n of forage (leaf) production and total biomass of Guadua paniculata and cultivated forage grasses in the g enus Urochloa ( including the former Brachiaria brizantha, B. decumbens, and Panicum maxima ) Species Forage (Mg/ha) Total biomass (Mg/ha) M easurement Management Site(s) Source Guadua 4,7 17,1 available no grazing, 9 mo since fire Estancia Cacarachi Thi s study Guadua 1,5 a 6,5 available various Chiquitania Table 4 1 This study U. brizantha 1,3 b 5,4 availa ble no grazing, 3 yr since fire San Ignacio de Velasco J. Veldman, unpublished data U. decumbens 1,1 6,6 Production 12 mo cattle grazing Area Integ rada de Santa Cruz Joaquin y Herrero (2003) U. maxima 5,2 b 15,0 Production 12 mo cattle grazing + fertili z er Ribeiro Preto, Bra z il Silveira et al. (2005) a calculated with equation in Figure 4 4 b calculated from data in Rivero (1999; cited by Joaqui n & Herrero 2003) to relate total dry weight to the weight of leaves.
95 Table 4 4. Concentration of protein, phosporous (P), and calcium (Ca) in forage of Guadua paniculata harvested at Estancia Cacarachi, 23 km east of San Ignacio de Velasco, Santa Cruz Bolivia (column 1). As a reference, minimum values (critical levels) and recomended levels for cattle are also included. Forage (lamina + sheath) Lamina only a Critical level b Recommended level c Guadua quality Prot ein (%) 10,5 13,2 < 7 % 11,2 good P (%) 0,15 0,15 < 17 ppm 0,22 good Ca (%) 0,2 < 72 ppm 0,22 0,35 good a mean for leaves collected in eight Guadua stands ( Table 4 1) b McDowell & Conrad (1977) cited by Velsquez Pereira et al. (1997) c Minson et al (1976) Figure 4 1. R elationship between Guadua abundance and tree canopy cover ; Guadua cover (%) = 0.88 [tree canopy cover] + 98.4; p = 0.011. The site marked by was excluded from the analysis because it was more heavily grazed than the other sites (2.6 cattle trails/40 m). If thi s outlier is included, the analysis still has a negative relationship (R 2 = 0,44; p = 0.071).
96 Figure 4 2 Relationships between trees, bamboo fuel loads, and fire intensit ies that suggest the combination of Guadua and fire influences tree abundance in the Chiquitano dry forest. A) R elationship between Guadua fuel loads (dry mass < 25 mm diameter) and tree density (>5 cm DBH; trees = 32[biomass Guada ] +502; p = 0.035). B) R elationship between Guadua fuel loads and tree basal area (basal area = 0.727 [ biomass Guadua ] +11.6; p = 0.037). C) R elationship between average char heights on tree trunks and tree density ( tree s /ha = 528 log10[1+char height] +484; p = 0.046).
97 Figure 4 3. Diagrams that incorporate fire to explain abundance s of Guadua and trees i n the Chiquitano dry forest. A) Model of the grass fire cycle in which a grass such as Guadua is part of a self catalyzing cycle that increase s flammability and replace s forest with grass dominated ecosystem s characterized by frequent fires (based o n D'Ant onio & Vitousek 1992, Keeley & Bond 1999 ). Logging and other disturbance s (e. g., wind, Gagnon & Platt 2008) can interact with the cycle and contribute to the proliferation of Guadua B) Diagram of trees and Guadua along a gradient of fire and tree canopy openness.
98 Figure 4 4 Relationship between the total standing biomass of Guadua and the mass of leaves suitable as cattle forage. Each po int represents a transect where harvested and measured the total biomass of Guadua in three 1 m 2 clip plots as wel l as the weight of the leaves and leaf sheaths. Dry forage = 0.369 [total Guadua biomass] 0.157.
99 APPENDIX A GRASS SPECIES IN NAT IVE AND DERIVED SAVA NNAS
100 Table A 1 Frequency, mean cover and maximum cover for grass species in native ( n = 12) and deriv ed ( n = 17) savannas. Species are listed in increasing value on DCA 1. Exotic species are indicated with an asterisk. Native Derived Species a frequency (%) mean cover (%) maximum cover (%) frequency (%) mean cover (%) maximum cover (%) DCA 1 DCA 2 Aristida riparia 17 0.19 1.9 4.11 0.07 Leptocoryphium lanatum 17 1.26 9.8 3.93 0.11 Sorghastrum minarum 25 0.24 2.0 3.81 0.12 Thrasya petrosa 67 2.21 13.7 3.76 0.14 Schizachyrium sanguineum 42 1.33 6.1 3.74 0.02 Aristida sp. 17 0.06 0.5 3.65 0.11 Axonopus fissifolius 17 0.07 0.7 3.47 0.07 Axonopus marginatus 17 2.20 13.7 3.46 0.39 Schizachyrium scabriflorum 25 0.20 1.0 3.33 0.17 Loudetia flammida 25 0 .75 7.8 3.29 0.22 Eragrostis articulata 6 0.03 0.5 3.27 0.14 Trachypogon 67 4.67 20.7 3.24 0.13 Gymnopogon spicatus 33 0.29 2.2 3.18 0.08 Panicum quadriglume 25 0.13 0.9 2.99 0.40 Axonopus chrysoblepharis 42 0.42 2.6 6 0.01 0.2 2.82 0.13 Elionurus muticus 75 11.48 26.3 2.74 0.25 Axonopus barbigerus 67 1.24 5.9 6 0.06 1.0 2.55 0.25 Schizachyrium microstachyum 50 0.62 2.4 6 0.01 0.2 2.35 0.27 Paspalum erianthum 58 0.96 3.9 2.28 0.26 Paspalum stellatum 58 1.82 9.8 6 0.09 1.5 2.04 0.06 Andropogon leucostachyus 25 0.24 2.0 1.90 0.02 Ichnathus procurrens 50 0.53 1.7 1.68 0.05 Hyparrhenia rufa 67 2.04 11.7 18 0.60 8.5 1.47 0.23 Andropogon selloanus 42 0.29 1.2 6 0.06 1.0 1.29 0.45 Tripsacum australe 8 0.04 0.5 6 0.17 2.9 1.19 0.65 Paspalum plicatulum 75 0.68 3.4 18 0.27 3.4 1.18 0.17
101 Table A 1 C ontinued Native Derived Species a frequency (%) mean cover (%) maximum cover (%) frequency (%) mean cover (%) maximum cover (%) DCA 1 DCA 2 Digitaria dioica 42 0.52 3.4 1.17 0.63 Setaria parviflora 33 0.16 1.0 12 0.06 0.5 0.80 0.39 Olyra spp. 18 0.33 3.8 0.62 0.20 Paspalum malacophyllum 25 0.07 0.5 24 0. 21 2.0 0.34 0.04 Guadua paniculata 42 8.58 79.5 53 18.13 72.2 0.07 0.81 Setaria scandens 17 0.06 0.5 35 0.54 5.9 0.10 0.87 Setaria vulpisetia 18 0.65 10.2 0.30 1.27 Lasiacis sorghoidea 8 0.02 0.2 18 0.17 1.5 0.38 1.48 Imperata bras iliense 50 0.41 2.4 35 3.14 25.9 0.40 3.29 Axonopus compressus 8 0.06 0.7 18 0.28 4.1 0.65 1.82 *Rottboellia cochinchinensis 12 0.14 1.5 1.09 2.54 Panicum trichoides 35 0.38 4.1 1.15 1.78 Leptochloa virgata 35 0.37 2.0 1.3 1 2.53 *Urochloa decumbens 12 1.89 28.8 1.39 1.97 Digitaria insularis 24 4.69 50.2 1.42 1.90 Digitaria ciliaris 24 0.22 2.9 1.68 0.28 Panicum laxum 8 0.02 0.2 6 0.18 3.1 2.21 0.28 Cynodon spp. 12 0.23 2.9 2.50 0.2 0 Paspalum conjugatum 12 0.52 8.3 2.87 0.93 Urochloa brizantha 8 0.02 0.2 53 13.86 75.1 3.31 0.09 Oplismenus hirtellus 6 0.03 0.5 3.39 0.30 Panicum millegrana 6 0.03 0.5 3.48 0.91 Panicum trichanthum 12 0.32 3.9 4. 19 0.03 Urochloa maxima 18 1.65 22.9 4.54 0.21 a Grasses present in just one site and therefore not included were: Acroceras zizanioides, Andropogon lateralis, Aristida capillacea, Axonopus canescens, Chloris elata, Eragrostis maypurensis, E. a mabilis, E. ciliaris, Eriochloa distachya, Guadua chacoensis, Hackelochloa granularis, Ichnathus inconstans, I. pallens, Mesosetum cayennensis, Oplismenus burmannii, Panicum peladoense, P. stenodes, P. stolonifer um, Paspalum gardnerianum, P. pictum, Schiza chyrium maclaudii, Sporobolus cubensis, S. indicus, S. pyramidalis, Zea mays
102 APPENDIX B TREE SPECIES IN N ATIVE AND D ERIVED S AVANNAS
103 Table B 1 Frequency, mean stem density, and mean basal area for tree species in native ( n = 12) and derived ( n =1 7) savannas. Species are listed in increasing value on DCA 1. Species typical of savanna are indicated by an asterisk. Native Derived Species a frequency (%) density (ind./ha) basal area (m 2 /ha) frequency (%) density (ind./ha) basal area (m 2 /ha) DCA 1 DCA 2 Caryocar brasiliense 33 94 1.87 2.37 0.17 Myrica pubes cens 50 57 0.58 6 4 0.03 2.16 0.36 Qualea grandifolia 100 339 3.93 6 4 0.03 2.08 0.06 Callisthene fasciculate 42 57 0.59 6 4 0.07 2.04 0.62 Pse udobombax heteromorphum 33 42 0.50 6 4 0.02 2.00 0.85 Brosimum gaudichaudii 42 36 0.18 1.81 0.07 Simira rubescens 0.67 6 4 0.07 1.80 0.41 Vochysia rufa 33 36 0.18 1.66 0.08 Curatella americana 92 297 4.59 1.47 0.19 Diptychandra aurantiaca 42 104 2.58 6 22 0.52 1.33 0.46 Guettarda viburnoides 33 47 0.23 6 15 0.08 1.28 0.46 Tabebuia aurea 50 42 0.61 1.11 0.06 Dilodendron bipinnatum 33 36 0.78 12 33 0.57 0.99 0.81 Phyllostylon rhamno ides 67 94 0.55 35 26 0.63 0.97 1.93 Eriotheca roseorum 8 57 0.34 12 22 0.54 0.97 1.03 Dipteryx alata 33 21 0.30 6 15 0.31 0.88 0.61 Qualea spp. 33 57 1.59 0.82 0.46 Hexachlamys boliviana 25 16 0.17 12 18 0.10 0.62 0.80 Term inalia argentea 33 68 0.91 6 11 0.15 0.60 0.18 Astronium fraxinifolium 17 10 0.10 6 15 0.22 0.55 0.73 Myrsine umbellate 6 15 0.08 0.51 1.35 Piper sp. 6 4 0.02 0.51 1.35 Protium heptaphyllum 6 4 0.01 0.51 1.35 X imenia Americana 8 5 0.01 6 7 0.06 0.46 0.74
104 Table B 1 C ontinued Native Derived Species a frequency (%) density (ind./ha) basal area (m 2 /ha) frequen cy (%) density (ind./h a) basal area (m 2 /ha) DCA 1 DCA 2 Tabebuia serratifolia 8 5 0.02 6 4 0.13 0.37 0.77 Simarouba amara 42 63 12 7 0.02 0.36 0.82 Erythroxylumsp. 8 5 0.02 12 29 0.24 0.33 0.25 Alibertia verrucosa 8 5 0.04 6 7 0.04 0.15 0.38 Cybistax antisyphilitica 18 59 0.48 0.13 1.95 Astronium ur undeuva 25 21 0.74 41 51 2.32 0.10 0.80 Physocalymma scaberrimun 8 10 0.20 24 26 0.62 0.13 0.45 Attalea speciosa 8 5 0.57 12 11 1.05 0.21 0.81 Combretum leprosum 12 7 0.02 0.24 1.37 Rhamnidium elaeocarpum 17 10 0.05 29 63 0.54 0.34 2.68 Apeiba tibourbou 8 5 0.15 29 37 0.43 0.35 0.49 Luehea paniculata 33 31 0.30 41 37 0.34 0.54 1.46 Plathymenia reticulate 8 5 0.06 6 4 0.21 0.62 3.22 Tabebuia impetiginosa 6 4 0.27 0.83 3.01 Tabebuia roseo alba 8 5 0.03 29 22 0 .35 0.85 2.14 Acrocomia aculeate 42 36 0.77 18 15 0.31 0.86 0.07 Spondias mombin 12 7 0.61 0.86 2.92 Aspidosperma tomentosum 8 5 0.03 18 15 0.35 0.93 2.62 Neea steimbachii 8 5 0.04 6 4 0.01 1.01 3.39 Casearia gossypiosperma 6 15 0.06 1.09 0.69 Rollinia herzogii 6 4 0.02 1.09 0.69 Neea sp. 6 4 0.03 1.10 0.58 Anadenanthera culubrina 8 10 0.03 35 114 2.23 1.17 2.34 Caesalpinia pluviosa 8 16 0.30 12 18 0.62 1.29 0.45 Aspidosperma cylindrocarpon 12 11 0.55 1.74 3.29 Sapium glandulosum 12 18 0.09 2.17 0.06 Guazuma ulmifolia 47 59 0.68 2.56 0.54 Cedrela fissilis 12 11 0.74 2.65 0.74
105 Table B 1 C ontinued Native Derived Species a frequency (%) density (ind./ha) basal area (m 2 /ha) frequen cy (%) density (ind./h a) basal area (m 2 /ha) DCA 1 DCA 2 Piptadenia viridiflora 12 7 0.04 2.71 1.02 Cordia alliodora 8 5 0.10 18 11 0.12 2.81 0.06 Cecropia concolor 24 44 0.48 2.88 0.06 Maclura tinctoria 18 11 0.06 2.91 0.06 Machaerium villosum 8 10 0.05 18 11 0.12 2.93 1.34 Pouteria macrophylla 12 7 0.08 2.96 1.96 Albizia niopoides 12 15 0.39 3.00 0.76 Ficus sp. 8 5 0.16 6 4 0.53 3.80 0.02 Attalea phalerata 18 11 1.79 4.21 0.08 Hymenaea courbaril 12 7 0.37 5.98 0.19 a Species present in just one site and therefore not included were: Aspidosperma rigidum Bauhinia rufa, Calycophyllum spruceanum, Cariniana ianeirensis, Ceiba samauma, Gallesia i ntegrifolia, Heliocarpus americanus, *Luehea candicans, Syagrus sancona, Urera baccifera, Vataireopsis speciosa, and Zanthoxylum coco.
106 LIST OF REFERENCES Aliscioni, S.S., L.M. Giussani, F. O. Zuloaga, E. A. Kellogg. 2003. A molecular phylogeny of Panicum (Poaceae : Paniceae): Tests of monophyly and phy logenetic placement within the Panicoideae. American Journal of Botany 90 :796 821. Angelsen, A., and D. Kaimowitz. 1999. Rethinking the causes of deforestation: Lessons from economic models. World Bank Resear ch Observer 14 :73 98. Ashton, P. M. S. 1992. Some measurements of the microclimate within a Sri Lankan tropical rain forest. Agricultural and Forest Meteorology 59 :217 235. Balch, J. K., D. C. Nepstad, P. M. Brando, L. M. Curran, O. Portela, O. de Carvalho and P. Lefebvre. 2008. Negative fire feedback in a transitional forest of southeastern Amazonia. Global Change Biology 14 :2276 2287. Barlow J., and C. A. Peres. 2008. Fire mediated dieback and compositional cascade in an Amazonian forest. Philosophical T ransactions of the Royal Society B Biological Sciences 363 :1787 1794. Betts, R. A., P. M. Cox, M. Collins, P. P. Harris, C. Huntingford, and C. D. Jones. 2004. The role of ecosystem atmosphere interactions in simulated Amazonian precipitation decrease and forest dieback under global climate warming. Theoretical and Applied Climatology 78 :157 175. Blate, G. M. 2005. Modest trade offs between timber management and fire susceptibility of a Bolivian semi deciduous forest. Ecological Applications 15 : 1649 1663. Bond, W. J. 2008. What limits trees in C4 grasslands and savannas? Annual Review of Ecology and Systematics S.39 : 641 659. Borhidi, A. 1988. Vegetation dynamics of the savannization process on Cuba. Vegetatio 77 : 177 183. Cavelier J., T.M. Aide, C. Santos A.M. Eusse, J.M. Dupuy. 1998. The savannization of moist forests in the Sierra Nevada de Santa Marta, Colombia. Journal of Biogeography 25 :901 912. Chapman, G. P. 1996. The Biology of Grasses. CAB International, Wallingford, UK. Christen, D., and G. Matl ack. 2006. The role of roadsides in plant invasions: a demographic approach. Conservation Biology 20 : 385 391. Clifford, H. T. 1959. Seed dispersal by motor vehicles. Journal of Ecology 47 : 311 315. Cochrane, M. A. 2003. Fire science for rainforests. Natur e 421 : 913 918.
107 Cochrane, M. A., A. Alencar, M. D. Schulze, C. M. Souza, D. C. Nepstad, P. Lefebvre, and E. A. Davidson. 1999. Positive feedbacks in the fire dynamic of closed canopy tropical forests. Science 284 :1832 1835. Cochrane, M. A., and M. D. Schul ze. 1998. Forest fires in the Brazilian Amazon. Conservation Biology 12 :948 950. Cochrane, M. A., and M. D. Schulze. 1999. Fire as a recurrent event in tropical forests of the eastern Amazon: effects on forest structure, biomass and species composition. Bi otropica 31 : 2 16. Corlett, R. T. 2009. Seed dispersal distances and plant migration potential in tropical East Asia. Biotropica 41 : 592 598. Cox, P. M., R. A. Betts, M. Collins, P. P. Harris, C. Huntingford, and C. D. Jones. 2004. Amazonian forest dieback under climate carbon cycle projections for the 21st century. Theoretical and Applied Climatology 78 :137 156. D'Antonio, C. M., and P. M. Vitousek. 1992. Biological invasions by exotic grasses, the grass fire cycle, and global change. Annual Review of Ecol ogy and Systematics 23 : 63 87. Daubenmire, R. 1972. Ecology of Hyparrehnia rufa (Nees) in derived savanna in north western Costa Rica. Journal of Applied Ecology 9 : 11 23. Fine, P.V.A. 2002. The invasibility of tropical forests by exotic plants. Journal of Tropical Ecology18: 687 705. Foster, J.R., P.A. Townsend, C.E. Zganjar. 2008. Spatial and temporal patterns of gap dominance by low canopy lianas detected using EO 1 Hyperion and Landsat Thematic Mapper. Remote Sensing of Environment 112 :2104 2117. Freder icksen, T. S., M. J. Justiniano, B. Mostacedo, D. Kennard, and L. McDonald. 2000. Comparative regeneration ecology of three leguminous timber species in a Bolivian tropical dry forest. New Forests 20 :45 64. Gagnon, P. R. and W. J. Platt. 2008. Multiple dis turbances accelerate clonal growth in a potentially monodominant bamboo. Ecology 89 : 612 618. Gelbard, J. L., and J. Belnap. 2003. Roads as conduits for exotic plant invasions in a semiarid landscape. Conservation Biology 17 : 420 432. Gerhardt, K. 1996. Ef fects of root competition and canopy openness on survival and growth of tree seedlings in a tropical seasonal dry forest. Forest Ecology and Management 82 : 33 48. Golding, N. and R. Betts. 2008. Fire risk in Amazonia due to climate change in the HadCM3 cl imate model: Potential interactions with deforestation. Global Biogeochemical Cycles 22 :GB4007.
108 Grime, J. P., K. Thompson, R. Hunt, J. G. Hodgson, J. H. C. Cornelissen, I. H. Rorison, G. A. F. Hendry, T. W. Ashenden, A. P. Askew, S. R. Band, R. E. Booth, C. C. Bossard, B. D. Campbell, J. E. L. Cooper, A. W. Davison, P. L. Gupta, W. Hall, D. W. Hand, M. A. Hannah, S. H. Hillier, D. J. Hodkinson, A. Jalili, Z. Liu, J. M. L. Mackey, N. Matthews, M. A. Mowforth, A. M. Neal, R. J. Reader, K. Reiling, W. Ros sFraser, R. E. Spencer, F. Sutton, D. E. Tasker, P. C. Thorpe, and J. Whitehouse. 1997. Integrated screening validates primary axes of specialisation in plants. Oikos 79 :259 281. Griscom, B.W. and P.M.S. Ashton. 2006. A self perpetuating bamboo disturban ce cycle in a neotropical forest. Journal of Tropical Ecology 22 :587 597. Griscom, B.W. and P.M.S. Ashton. 2003. Bamboo control of forest succession: Guadua sarcocarpa in Southeastern Peru. Forest Ecology and Management 175 : 445 454. Guissani L. M., J. H. Cota Snchez, F. O. Zuloaga, E. A. Kellogg 2001. A molecular phylogeny of the grass subfamily Panicoideae (Poaceae) shows multiple origins of C4 photosynthesis. American Journal of Botany 88 :1993 2012. Hierro, J. L., D. Villarreal, O. Eren, J. M. Graham, a nd R. M. Callaway. 2006. Disturbance facilitates invasion: the effects are stronger abroad than at home. American Naturalist 168 :144 156. Higgins, S. I., and D. M. Richardson. 1999. Predicting plant migration rates in a changing world: The role of long dis tance dispersal. American Naturalist 153 : 464 475. Hobbs, R. J., S. Arico, J. Aronson, J. S. Baron, P. Bridgewater, V. A. Cramer, P. R. Epstein, J. J. Ewel, C. A. Klink, A. E. Lugo, D. Norton, D. Ojima, D. M. Richardson, E. W. Sanderson, F. Valladares, M. Vila, R. Zamora, and M. Zobel. 2006. Novel ecosystems: theoretical and management aspects of the new ecological world order. Global Ecology and Biogeography 15 :1 7. Hodkinson, D. J., and K. Thompson. 1997. Plant dispersal: the role of man. Journal of Appl ied Ecology 34 : 1484 1496. Hoffmann, W. A., R. Adasme, M. Haridasan, M. T. d. Carvalho, E. L. Geiger, M. A. B. Pereira, S. G. Gotsch, and A. C. Franco. 2009. Tree topkill, not mortality, governs the dynamics of savanna forest boundaries under frequent fire in central Brazil. Ecology 90 :1326 1337. Hoffmann, W.A., and R.B. Jackson. 2000. Vegetation climate feedbacks in the conversion of tropical savanna to grassland, Journal of Climate 13 :1593 1602. Hoffmann, W. A., V. Lucatelli, F. J. Silva, I. N. C. Azeuedo M. D. Marinho, A. M. S. Albuquerque, A. D. Lopes, and S. P. Moreira. 2004. Impact of the invasive alien grass Melinis minutiflora at the savanna forest ecotone in the Brazilian Cerrado. Diversity and Distributions 10 :99 103. Holling, C.S. 1973. Resilienc e and stability of ecological systems. Annual Review of Ecology and Systematics 4 :1 23.
109 Hutyra, L. R., J. W. Munger, C. A. Nobre, S. R. Saleska, and S. A. Vieira. 2005. Climatic variability and vegetation vulnerability in Amazonia. Geophysical Research Le tters 32 :L24712. Jardim, A., T. J. Killeen, and A. Fuentes 2003. Gua de los arboles y arbustos del bosque seco Chiquitano, Bolivia. FAN, Santa Cruz, Bolivia, 324 pp. Joaquin, N., and M. Herrero. 2003. Referencias para el manejo de pasturas de gnero Brach iaria. Documento tcnico. Centro de Investigacin Tropical. Santa Cruz, Bolvia. Keeley J. E., and W. J. Bond. 1999. Mast flowering and semelparity in bamboos: the bamboo fire cycle hypothesis. The American Naturalist 153: 383 391. Khurana, E. and J. S. Sing h. 2001. Ecology of seed and seedling growth for conservation and restoration of tropical dry forest: a review. Environmental Conservation 28 :39 52. Killeen, T. J. 1990. The grasses of Chiquitania, Santa Cruz, Bolivia. Annals o f the Missouri Botanical Gard en 77: 125 201. Killeen T. J., E. E. Garcia, S. G. Beck. 1993. Gua de rboles de Bolivia. Herbario Nacional de Bolivia and Missouri Botanical Garden, 958 pp. Killeen, T.J., B. T. Louman, T. Grimwood. 1990. La ecologa paisajstica de la regin de Concepci n y Lomerio en la provincia de uflo de Chavez, Santa Cruz, Bolivia. Ecologa en Bolivia 16 :1 45. Killeen, T. J., and P. N. Hinz. 1992a. Grasses of the Precambrian shield region in eastern lowland Bolivia I: habitat preferences. Journal of Tropical Ecology 8 :389 407. Killeen, T. J., and P. N. Hinz. 1992b. Grasses of the Precambrian shield region in eastern lowland Bolivia II: life form and C3 C4 photosynthetic types. Journal of Tropical Ecology 8 :409 433. Killeen, T. J., E. Chavez, M. Pea Claros, M. Toledo L. Arroyo, J. Caballero, L. Correa, R. Guilln, R. Quevedo, M. Saldias, L. Soria, Y. Uslar, I. Vargas, M. Steininger. 2005. The Chiquitano dry forest, the transition between humid and dry forest in eastern lowland Bolivia. In T. Pennington (Ed.). Neotrop ical savannas and seasonally dry forests: plant diversity, biogeography and conservation. CRC Press, England. pp. 206 224 Lange, R. T. 1969. Piosphere Sheep Track and Dung Patterns. Journal of Range Management 22 :396 400 Laurance, W. F. 2003. Slow burn: the insidious effects of surface fires on tropical forests. Trends in Ecology & Evolution. 18 : 209 212. Lemmon, P. E. 1956. A spherical densiometer for estimating overstory density. Forest Science 2 : 314 320.
110 Lippincott, C.L. 2000. Effects of Imperata cyli ndrica (L.) Beauv. (cogongrass) invasion on fire regime in Florida sandhill (USA). Natural Areas Journal 20 :140 149. Lugo, A. E. 2009. The emerging era of novel tropical forests. Biotropica 41 : 589 591. MacDonald, G.E. 2004. Cogongrass ( Imperata cylindrica ) Biology, ecology, and management. Critical Reviews in Plant Sciences 23 :367 380. Mack, M. C., and C. M. D'Antonio. 1998. Impacts of biological invasions on disturbance regimes. Trends in Ecology & Evolution 13 : 195 198. Mack, M. C., and C. M. D'Antonio 2003. The effects of exotic grasses on litter decomposition in a Hawaiian woodland: the importance of indirect effects. Ecosystems 6 :723 738. Malhi, Y., L. Aragao, D. Galbraith, C. Huntingford, R. Fisher, P. Zelazowski, S. Sitch, C. McSweeney, and P. Mei r. 2009. Exploring the likelihood and mechanism of a climate change induced dieback of the Amazon rainforest. Proceedings of the National Academy of Sciences of the United States of America 106 :20610 20615. Martinez M., J.A., V. G. Morales, G. Z. Villegas, A. M. Malla. 2003. Fuego en el pantanal: Incendios forestales y prdida de recursos de biodiversidad en San Matas Santa Cruz. Fundacin PIEB, La Paz Bolivia, 188 pp. Mayle, F.E., D. J. Beerling, W. D. Gosling, M. B. Bush. 2004. Responses of Amazonian ecosystems to climatic and atmospheric carbon dioxide changes since the last glacial maximum. Philosophical Transactions of the Royal Society B Biological Sciences 359 :499 514. Mayle, F. E., R. Burbridge, and T. J. Killeen. 2000. Millennial scale dynamics of southern Amazonian rain forests. Science 290 :2291 2294. McDowell, L. R., and J. H. Conrad. 1977. Trace mineral nutrition in Latin America. World Animal Review. 24 :24 33. (cited by Velsquez Pereira et al. 1997). Minson, D. J., T. H. Stobbs, M. P. Hegar ty, and M. J. Playne. 1976. Measuring the nutritive value of pasture plants. In N. H. Shaw and W. W. Bryan (Eds.). Tropical pasture research principles and methods. Commonwealth Agricultural Bureaux, Farnham Royal, Bucks, England. pp. 308 337. Mokat, D.N. V.W. Bendale, S.S. Deokule, B.B. Jadhav, J. L. Mehta. 2005. Allelopathic effect of Hyptis suaveolens on rice. Journal of Ecobiology 17 :479 483. Murthy, E.N., V.S. Raju, C.S. Reddy. 2007. Occurrence of exotic Hyptis suaveolens Current Science 93 :1203 1 203. Navarro, G ., and W Ferreira. 2007. Mapa de Vegetacin de Bolivia a escala 1:250.000 The Nature Conservancy (TNC). Ed. digital. Santa Cruz de la Sierra. ISBN 978 99954 0 168 9
111 Nepstad, D., G. Carvalho, A. C. Barros, A. Alencar, J. P. Capobianco, J. Bishop, P. Moutinho, P. Lefebvre, U. L. Silva, and E. Prins. 2001. Road paving, fire regime feedbacks, and the future of Amazon forests. Forest Ecology and Management 154 :3 95 407. Nepstad, D., P. Lefebvre, U. L. Da Silva, J. Tomasella, P. Schlesinger, L. Solorzano, P. Moutinho, D. Ray, and J. G. Benito. 2004. Amazon drought and its implications for forest flammability and tree growth: a basin wide analysis. Global Change Bio logy 10 :704 717. Nepstad, D. C., A. G. Moriera, and A. A. Alencar. 1999. Flames in the rain forest: origins, impacts and alternatives to Amazonian fire. Pilot Program to Conserve the Brazilian Rain Forest. Brazilia, Brazil, 190 pp. Nepstad, D.C., C. M. Sti ckler, B. Soares, F. Merry. 2008. Interactions among Amazon land use, forests and climate: prospects for a near term forest tipping point. Philosophical Transactions of the Royal Society B Biological Sciences 363 :1737 1746. Nepstad, D. C., A. Verissimo, A. Alencar, C. Nobre, E. Lima, 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. Oliviera, P. S., and R. J. Marquis 2002 The Cerrados of Brazil: Ecology and Natural History of a Neotropical Savanna. Columbia University Press, New York, 398 pp. Parendes, L. A., and J. A. Jones. 2000. Role of light availability and dispersal in exotic plant invasion along roads and streams in the H. J. Andrews Experimental Forest, Oregon. Conservation Biology 14 : 64 75. Parsons, J. J. 1972. Spread of African pasture grasses to the American tropics. Journal of Range Management 25 :12 17. Pinard, M. A., and J. Huffman. 1997. Fire resistance and bark properties of trees in a seasonally dry forest in eastern Bolivia, Journal of Tropical Ecology 13 :727 740. Pinard, M.A., F. E. Putz, J. C. Licona. 1999. Tree mortality and vine proliferation following a wildfire in a subhumid tropical forest in easte rn Bolivia. Forest Ecology and Management 116 :247 252. Platt, W. J., and R. M. Gottschalk. 2001. Effects of exotic grasses on potential fine fuel loads in the groundcover of south Florida slash pine savannas. International Journal of Wildland Fire 10 :155 1 59. Poorter, L., and L. Markesteijn. 2008. Seedling traits determine drought tolerance of tropical tree species. Biotropica 40 : 321 331. Pringle, H. J. R., and J. Landsberg. 2004. Predicting the distribution of livestock grazing pressure in rangelands. Aus tral Ecology 29 :31 39.
112 degradation, deforestation, long term phase shifts, and further transitions. Biotropica 42 : 10 20. Quesada C. A., J. Lloyd, M. Schwarz, S. Pat io, T. R. Baker, C. Czimczik N. M. Fyllas, L. Martinelli, G. B. Nardoto, J. Schmerler,A. J. B. Santos M. G. Hodnett, R. Herrera, F. J. Luiz o A. Arneth G. Lloyd, N. Dezzeo, I. Hilke, I. Kuhlmann, M. Raessler, W. A. Brand, H. Geilmann, J. O. Moraes Filh o, F. P. Carvalho, R. N. Araujo Filho, J. E. Chaves, O. F. Cruz Junior, T. P. Pimentel, and R. Paiva 2009 Chemical and physical properties of Amazon forest soils in relation to their genesis Biogeosciences Discussion 6 : 3923 3992. Renvoize, S. A. 1998 Gramneas de Bolivia. The Royal Botanical Gardens, Kew, 644 pp. Rockwell, C. A., K. A. Kainer, C. L. Staudhammer, and C. Baraloto. 2007. Future crop tree damage in a certified community forest in southwestern Amazonia. Forest Ecology and Management 242 :1 08 118. Sandoval M, L. M. 2001. Uso y manejo del bamb en dos zonas agroecolgicas de Santa Cruz. Graduate Thesis. Universidad Mayor de San Simon, Cochabamba Setterfield, S. A., M. M. Douglas, L. B. Hutley, and M. A. Welch. 2005. Effects of canopy cover an d ground disturbance on establishment of an invasive grass in an Australia savanna. Biotropica 37 : 25 31. Shukla, J., C. Nobre, and P. Sellers. 1990. Amazon Deforestation and Climate Change. Science 247 :1322 1325. Simon, B.K. and Jacobs, S.W. 2003. Megathy rsus a new generic name for Panicum subgenus Megathyrsus Austrobaileya 6 :571 574. Smith, E. P. 2002. BACI design. Pages 141 148 in A. H. El Shaarawi, and W. W. Piegorsch, editors. Encyclopedia of Environmetrics. John Wiley & Sons, Ltd., Chichester, UK. S teininger, M. K., C. J. Tucker, P. Ersts, T. J. Killeen, Z. Villegas, and S. B. Hecht. 2001. Clearance and fragmentation of tropical deciduous forest in the Tierras Bajas, Santa Cruz, Bolivia. Conservation Biology 15 :856 866. Suding, K. N., K. L. Gross, G. R. Houseman. 2004. Alternative states and positive feedbacks in restoration ecology. Trends in Ecology and Evolution 19 :46 53. Totey N.G., D. K. Shadangi, P. K. Khatri. 1994. Allelopathic effects of van tulsi ( Hyptis suaveolens ) on germination and growth of teak ( Tectona grandis ) seedlings. Indian Journal of Forestry 17 :137 141. Uhl, C., and R. Buschbacher. 1985. A disturbing synergism between cattle ranch burning practices and selective tree harvesting in the Eastern Amazon. Biotropica 17 :265 268.
113 Uhl, C. R. Buschbacher, and E. A. S. Serro. 1988. Abandoned pastures in eastern Amazonia I. Patterns of plant succession. Journal of Ecology 76 :663 681. Uhl, C., A. Verissimo, M. M. Mattos, Z. Brandino, and I. C. G. Vieira. 1991. Social, economic, and ecologica l consequences of selective logging in an Amazon frontier the case of Tailandia. Forest Ecology and Management 46 :243 273. Usberti, R., and L. Martins. 2007. Sulphuric acid scarification effects on Brachiaria brizantha, B. humidicola and Panicum maximum seed dormancy release. Revista Brasileira de Sementes 29 :143 147. Van Deusen, P.C. 1996. Unbiased estimates of class proportions from thematic maps. Photogrammetric Engineering and Remote Sensing 62 :409 412. Velsquez Pereira J., L. McDowell, J. Conrad, N. Wilkinson and F. Martin. 1997. Nivel mineral existent e en suelos, forrajes y ganado bovino en Nicaragua. II. Macrominerales y composicin orgnica de forrajes. Revista de la Facultad de Agronoma de La Universidad del Zulia. 14 : 91 110. Veldman, J. W. 20 08. Guadua paniculata (Bambusoideae) en la Chiquitania Boliviana: Ecolog a de fuego y la oportunidad para un forraje nativo Revista Boliviana de Ecologa y Conservacin Ambiental 24 :65 74. Veldman, J. W., B. Mostacedo, M. Pea Claros, and F. E. Putz. 200 9. Selective logging and fire as drivers of alien grass invasion in a Bolivian tropical dry forest. Forest Ecology and Management 258 : 1643 1649. Veldman J.W., and F. E. Putz. 2010. Long distance dispersal of invasive grasses by logging vehicles in a tropi cal dry forest. Biotropica ( in press ). Villegas, Z., M. Pena Claros, B. Mostacedo, A. Alarcon, J. C. Licona, C. Leano, W. Pariona, and U. Choque. 2009. Silvicultural treatments enhance growth rates of future crop trees in a tropical dry forest. Forest Ecol ogy and Management 258 :971 977. v on der Lippe, M. and I. Kowarik 2007. Long distance dispersal of plants by vehicles as a driver of plant invasions. Conservation Biology 21 :986 996. Wichmann, M. C., M. J. Alexander, M. B. Soons, S. Galsworthy, L. Dunne, R. Gould, C. Fairfax, M. Niggemann, R. S. Hails, and J. M. Bullock. 2009. Human mediated dispersal of seeds over long distances. Philosophical Transactions of the Royal Society B Biological Sciences 276 :523 532. Zwaenepoel, A., P. Roovers, and M. Hermy. 2006 Motor vehicles as vectors of plant species from road verges in a suburban environment. Basic and Applied Ecology 7 :83 93.
114 BIOGRAPHICAL SKETCH Joseph W. Veldman spent his formative years in the dunes, forests, and beaches on the shores of Lake Michigan He attended Hope College, in Holland Michigan, were he studied b iology and c hemistry and ran cross country and track. It was during his time in college that he developed an interest in ecolog ical research. Encouraged by Professor K. Greg Murray he studi ed seed chemical defense and attended a tropical ecology field course in Costa Rica. After college, a brief stint in the private sector and an attempt to run semi professionally convinced Joseph to apply to graduate school. In the fall of 2003 he arrived a t the University of Florida to what was then the Department of Botany and to begin a career of conservation research and teaching at the college level. officemate Bonifacio Mostacedo encouraged him to develop a project in Bolivia In 2005 he began working with the Bolivian Institute of Forestry Research (IBIF) on grass invasions and fire in selectively logged tropical dry forests. Later, with funding from the Garden Club of America and a NSF Doctoral Disserata tion Improvement Grant, he expanded his studies to cover a large region of eastern lowland Bolivia. Much of his time in Bolivia was spent living in tent s cooking on fires, bathing in rivers, and benefiting from the hospitality of ranchers and rural villag ers. After two years of fieldwork, Joseph left his life and friends in Bolivia to return to the United States to complete his rich society le ft him longing for a life that feels more real. He struggles with the guilt of his privileged position in the world but is nonetheless grateful for the opportunities he has had. Joseph hopes that even if his dissertation research does not change much in the world, that the personal relationships a nd experiences o ver the last six years have had positive, if intangible, impacts