|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 NEOTROPICAL DRY FORESTS OF TH E CARIBBEAN: SECONDARY FOREST DYNAMICS AND RESTORATION IN ST. CROIX, US VIRGIN ISLANDS By BRIAN F. DALEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Brian F. Daley
3 To my wife, Jennifer, for supporti ng me in this and in all things.
4 ACKNOWLEDGMENTS Thank you to my parents, William and Tricia Daley, for giving me the fundamental tools and building bloc ks I needed to accomplish this. Thank you to my lovely wife, Jennifer, for helping me use those tools proper ly while also being my best friend. I would also like to thank Ariel Lugo for cha llenging me to try and prove my beliefs and for providing generous inst itutional support to make it possible.
5 TABLE OF CONTENTS page ACKNOWLEDG MENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 LIST OF ABBR EVIATIONS ............................................................................................. 9 ABSTRACT ................................................................................................................... 10 CHA PTER 1 INTRODUC TION .................................................................................................... 12 Secondary Dr y Fore sts ........................................................................................... 12 Successi on ............................................................................................................. 13 Invasive Exot ic Spec ies .......................................................................................... 14 Influence of Exotic Pl ants on Succession ............................................................... 16 Summary ................................................................................................................ 17 2 LAND COVER CHANGE ANALYSIS US ING LANDSAT SATELLITE IMAGERY FOR ST. CROIX, US VIRGIN ISLANDS ................................................................. 20 Introduc tion ............................................................................................................. 20 Methods .................................................................................................................. 23 Study Area ........................................................................................................ 23 Image Proce ssing ............................................................................................. 24 Image Classification and Gr ound-truthed Poin t Data. ...................................... 25 Land Cover Change Analysis ........................................................................... 27 Fragmentation .................................................................................................. 28 Forest Cover Change and R oads ..................................................................... 28 Results and Discussion ........................................................................................... 29 Classified Images ............................................................................................. 29 Land Cover Change ......................................................................................... 31 Forest Change .................................................................................................. 33 Fragmentation .................................................................................................. 35 Forest Cover Change and R oads ..................................................................... 36 Conclusi on .............................................................................................................. 39 3 DIVERSITY AND STRUCTURE OF SEC ONDA RY FORESTS OF VARYING AGES ...................................................................................................................... 51 Introduc tion ............................................................................................................. 51 Methods .................................................................................................................. 56
6 Study Area ........................................................................................................ 56 Field Data Co llection ........................................................................................ 56 Data Anal ysis ................................................................................................... 57 Results .................................................................................................................... 58 Discuss ion .............................................................................................................. 62 Conclusi ons ............................................................................................................ 66 4 ENRICHING DEGRADED CARIBBEAN DRY FOREST WITH RARE NATIVE SPECIES ................................................................................................................ 74 Introduc tion ............................................................................................................. 74 Methods .................................................................................................................. 78 The Site ............................................................................................................ 78 Species Se lection ............................................................................................. 81 Seeds ............................................................................................................... 81 Planting ............................................................................................................ 82 Results .................................................................................................................... 83 Seeds ............................................................................................................... 83 Seedling Su rvival. ............................................................................................. 84 Seedling Gr owth ............................................................................................... 84 Discuss ion .............................................................................................................. 85 Conclusi on .............................................................................................................. 89 5 CONCLUS ION ........................................................................................................ 94 LIST OF RE FERENCES ............................................................................................... 98 BIOGRAPHICAL SKETCH .......................................................................................... 107
7 LIST OF TABLES Table page 2-1 Names and descriptions of the four land cover classes used in the land cover classification of 1992 and 2002 Landsat im ages of St. Croix, U.S. Virgin Islands ................................................................................................................ 41 2-2 Overall accuracy assessment of t he four-class images for 1992 and 2002. These images are the base layers utilized in fragmenta tion analysis. ................ 42 2-3 Accuracy assessment for the three cove r classification used to detect forest cover change. These images serve as the base layers for land cover change analysis ............................................................................................................. 43 2-4 Cross-tabulation matrix compar ing classified images for 1992 and 2002 expressed as a percentage of the total st udy ar ea. ............................................ 43 2-5 Seven land scape metr ics ................................................................................... 44 3-1 Tree species encountered during field sampling in St. Croix .............................. 68 3-2 Description of the num ber of observations and the mean species diversity ....... 69 3-3 Age ranges (yr) of five forest cove r classes ........................................................ 69 4-1 Botanical and common nam es of the 4 native tree species grown for this study and a brief descripti on. .............................................................................. 90 4-2 Fruit and seed data summary for four nativ e species. ........................................ 91 4-3 Results from PROCMIXED analysis of fixed effects on plant height over time indicate significant diffe rences between species. ............................................... 91
8 LIST OF FIGURES Figure page 2-1 Classified land cover maps for St Croix ............................................................ 45 2-2 Total area of four land cover classe s in 1992 and 2002 in St. Croix, U.S. Virgin Is lands. ..................................................................................................... 46 2-3 Transition within each of the four land cover classes during the 10 year study period. ................................................................................................................ 47 2-4 Spatial distribution of areas experiencing change (shown in red) between 1992 and 2002 relative to the thr ee stable land co ver types. ............................. 48 2-5 An idealized graph of the percentage of forest cover on the island of St. Croix through time. ....................................................................................................... 48 2-6 Distribution of forest gain and forest loss from 1992 to 2002 shown in relation to topography. ..................................................................................................... 49 2-7 Change area relative to the distance from main roads in St. Croix. .................... 50 3-1 A) St. Croix, US Virgin Islands and the distribution of 50 vegetation sampling transects in secondary forests from 5 age classes ranging from 3 to <50 years and B) an idealized pair of para llel vegetation sampling transects, divided into 10m segm ents ................................................................................. 70 3-2 Recording canopy structure and compos ition data along 50m long transects .... 71 3-3 Frequency of groups of t he most common plants in secondary forests of St Croix. .................................................................................................................. 72 3-4 he mean number of species recorded per observation for each of 5 age classes of secondary forest in S t. Croix, U.S. Vi rgin Islands. ............................. 73 4-1 Field data collection at the plant ing site. ............................................................. 92 4-2 east Square Means of percent survival after 18 months for four tree species and for mulched and un mulched pl ants .............................................................. 92 4-3 Growth patterns of four trees species plotted along with monthly rainfall (cm). ........................................................................................................ 93 4-4 Final heights by species after 18 months, with and without mulch treatment, expressed as a percentage of original height. Error terms are SE. .................... 93
9 LIST OF ABBREVIATIONS ETM+ Enhanced Thematic Mapper. An 8-band multi-spectral scanning radiometer associated with Landsat 7. The primary new feature added to the ETM+ scanner is the multispectral band with a 15 m spatial resolution. GIS Geographic Information Systems. A computer software program that stores and processes spatial and other data. TM Thematic Mapper. An Earth obs ervation sensor associated with the Landsat 4 & 5 platforms. The sensor has a total of 7 bands, 6 spectral reflectance bands with a 30 m spatial resolution and a single temperature band
10 Abstract of Dissertation Pr esented to the Graduate School of the University of Florida in Partial Fulf illment of the Requirements for t he Degree of Doctor of Philosophy NEOTROPICAL DRY FORESTS OF TH E CARIBBEAN: SECONDARY FOREST DYNAMICS AND RESTORATION IN ST. CROIX, US VIRGIN ISLANDS By Brian F. Daley May 2010 Chair: Daniel J. Zarin Major: Forest Resources and Conservation Neotropical dry forests exist today mainly as secondary forests heavily influenced by exotic plants. This project analyzes landcover change and secondary dry forest dynamics in three distinct phases (land cover change, secondary forest succession and forest rehabilitation), using St. Croix, US Virgin Islands as an example. Using Landsat satellite images and other data layers, I created classified land cover maps of St. Croix for 1992 and 2002. Forest was the dominant (56%) cover type on both dates, followed by development, grassland/pastures and water. A land cover change analysis comparing the two images revealed that 15% of the study area experienced a change either to (8%) or from (7 %) forest. Grassland was the cover most likely to change and decreased from 16% to 11%, conv erted primarily to developmen t. The overall result is a landscape trending toward younger forest s, and increased forest fragmentation and development. In a second study, vegetati on data from a chronos equence of secondary forests was analyzed for changes to forest st ructure, species com position and presence of exotic species. T he leguminous exotic tree Leucaena leucocephala was by far the most frequently observed tree and dominated all stands, except those over 50 years old. Species diversity was significantly ( p <0.001) higher for forests in the two oldest
11 age classes and there was a strong trend towa rd increasing canopy complexity with increased age. However, age cl ass accounted for only a sma ll portion of variability in species diversity, indicating other influenci ng factors. Slope, elevation, aspect and soil were not significant and sites with long hi stories of intensive agricultural land-use remained low in species diversity and dominated by exotics >50 years after abandonment. In a third experimen t, a gap planting method fo r establishing four rare native tree species was tested on a site ex periencing arrested succession. All four species successfully established at >69% su rvival in 3m diameter gaps artificially created in exotic tree stands. A mulch treatment significantly ( p <0.01) increased survival, but not growth. This study demonstrat es on St. Croix forest is the primary land cover type and secondary forests are pr edominant forest type. The species composition of these forests is dynamic, but they tend to be domin ated by exotic tree species and over 50 years of natural succession is insufficient time for secondary forest stands to advance beyond the earliest succe ssional stages. These results can be applied to jump-start native forest succession and rehabilitate degraded secondary forests.
12 CHAPTER 1 INTRODUCTION Secondary Dry Forests The dry forests of the Cari bbean are one of the worlds bi odiversity hot spots due to their species richness and their high level of endemism. Despite their high species diversity and being spr ead over a large part of the Earth, dry forests remain far less studied and less protected than tropical rainfo rests (Quesada et al., 2009). Dry forests account for approximately 42% of the worlds tr opical forests, but tend to occur in areas where there are high density human populations, which makes them susceptible to clearing for agriculture and human devel opment and makes conservation more challenging (Murphy & Lugo, 1986) In their map of the gl obal distribution of tropical dry forests, Miles et al. (2006) es timate that just over 1 million km2 of closed canopy tropical dry forest remain. Habitat loss and invasive species remain as significant threats to biodiversity (Vitousek et al., 1997). The basic mechanisms delivering the threats are considered to be; climate change, forest fragmentation, fire, conversion to agriculture and other pressures from high human population (Miles et al., 2006). Roughly 66% of dry forests have already bee n converted to other uses throughout their global range (Portillo-Quintero & Sanchez-A zofeifa, 2010) and mature dry forests are predicted to effectively disappear in the next few decades (Quesada et al., 2009). Only 23,000 km2 (4.5%) of dry forest in the Americas in protected, and much of that is considered to be secondary fore st. However, Caribbean Island nations protect a higher percentage of their dry fore st than do countries in Central and North America (3.9%) (Portillo-Quintero & Sanchez-Azofeifa, 2010).
13 Secondary forests form as a result of human activity and account for over 40% of the forests in the tropics and continue to exp and in global range (Brown & Lugo, 1990). They are highly dynamic and variable in thei r structure and function and likely represent the forests of the future (Quesada et al., 2009). Secondary forests extend throughout the Caribbean region and are the dominant forest cover type in many of the Caribbean islands. The three main US Virgin Islands (St. Croix, St. Th omas and St. John) are approximately 60% forested with subtropical moist forest in the higher elevations and subtropical dry forest in the lowlands. The US Forest Service describes the forests as generally young and undeveloped st ands that reflect historic and current land uses (Brandeis & Oswalt, 2007). St. John has the highest percentage (20%) of mature secondary forest due to the protection by National Park Service while 8% of stands in St. Thomas are mature. The younger fo rests have low amount s of downed woody debris, standing snags and trees of the large tr unk diameters (>45 cm dbh) (Brandeis & Oswalt, 2007). St. Croix, the largest of the th ree US Virgin Islands and the site of this research project, has a centuries long agric ultural tradition and the current landscape is a matrix of agriculture, residential and ur ban development. These dynamics and the relatively recent introduction of exotic agricultural weed species make the secondary forests of St. Croix an ideal laboratory fo r studying secondary dry forest succession. Succession Forest succession is defined as the di rectional change through time of the species composition and struct ure of a forest stand where other outside environmental factors remain effectively constant (Fi negan, 1984). It is perhaps the most studied process in forest ecology. Primary succession refers to colonization of newly created land, such as lava flows or land ex posed by receding glaciers and secondary
14 succession occurs after a natural or human disturbance. The study of Neotropical secondary forest succession focuses on the revegetation of land after anthropogenic stand clearing disturbances, such as land cleared for agriculture and then abandoned. It is difficult to make generalizations about a protean process like succession. However, the process is generally thought of as a se ries of stages and the pattern of species replacement frequently moves in a somewhat predictable way from the simple to more complex and diverse (Guariguata & Oste rtag, 2001). Severely disturbed forests typically go through four stages of regeneration; 1) stand initiation, 2) thinning or exclusion, 3) transition or understory regeneration, and 4) steady-state or old-growth (Kozlowski, 2002). The actual species composition of a stand may not be predictable and will remain dynamic, but the change comes from predi ctable, describable sources both allogenic and autogenic. So me of the most widely accepted aspects of succession are that plants compete for a finite amount of resources and that differential response among species leads to changes in stand composition. The most basic functional groups in this competition have traditiona lly been considered sun-lo ving pioneer species and shade tolerant forest trees, including forest emergents (Finegan, 1996), although more recent work has demonstrated success in classifying plants into more detailed functional groups based on growth c haracteristics (Chazdon et al., 2010). Invasive Exotic Species Exotic species are those that are intr oduced by humans to a new area outside of their historical native range. It would be im possible to determine the actual number, but the vast majority of species introductions result in the individual dying without reproducing and going unnoticed outside of cultivat ion. In more rare cases a species is able to successfully produce off-spring unassisted and becomes naturalized to a new
15 area. In the most extreme cases the introduced species thrives, reproduces in large numbers, spreads and causes ecological damage to the new ecosystem. The damage can include reduced ecological function, alte red disturbance regimes, or large-scale exclusion/replacement of native species. In cases where the new species quickly spreads its range to the detriment of nat ive species or the whole ecosystem, the organism is considered invasive. Invasive exotic species are becoming increasingly common as humans move plants and animals, (intentionally and accidentally) around the Earth and now represent the second greatest threat to global biodi versity. Invasive plants are often characterized as aggressive, fast growing species whose establishment and spread is favored by ecol ogically open and disturbed si tes (Barnes et al., 1998). The presence of invasive species in an area however, may reflect poor health of that environment rather than characteristics of t he invading plant (Lugo, 1992). In the Neotropics, forest conversion for agric ulture and subsequent abandonment creates a condition where agricultural weeds are frequently present at the star t of the secondary succession. In the US Virgin Islands the most co mmon woody plant in regenerating pastures is locally referred to as tan-tan (Leucaena leucocephala ). The fast-growing, nitrogenfixing, leguminous tree is believed to have been introduced to the US Virgin Islands in the 1940s for pasture improvemen t. It thrives in full sun and low-nutrient soils, tolerates heavy defoliation and grazing, resprouts from stumps and r oots and plants begin copiously producing long-lived seeds in as little as one year. L. leucocephala can grow in all altitudes and soil types on St. Croix except for beach sand. The variety of L. leucocephala most pervasive in the Virgin Islands is the Hawaiian-variety native to
16 Mexico described by Pound and Cairo (1983) as producing many seeds and having the ability to become an aggressive weed. Today it is the single most common plant in St. Croix and the other US Virgin Islands (B randeis & Oswalt, 2007). Its distribution however is patchy indicating some limiting fa ctor. The plants intolerance to shade and initially weak seedlings (Pound and Cairo, 1983) are characteristics which do not aid in colonizing the shaded understory of a healthy forest and likely limits its spatial distribution. Influence of Exotic Plants on Succession A heavy influence of invasive species in an area can sometimes result in slowed or even arrested succession (Sarmiento, 1997; Chapman et al., 1999; Roth, 1999; Hau & Corlett, 2003). In these case s the plants that initially co lonize the site dominate and somehow prevent forest species from establis hing. There are many studies in Puerto Rico that focus on secondary forest devel opment as a chr onosequence on abandoned pastures or other agricultural sites of vary ing ages (Aide et al., 1995; Rivera & Aide, 1998; Aide et al., 2000; Pascarella et al., 2000; Zimmerman et al., 2000; Lomascolo & Aide, 2001; Rudel et al., 2001; Bernard, 2002; Weaver & Chinea, 2003; Lugo, 2004; Francis & Parrotta, 2006). Many of thes e studies concluded that introduced species can function to stabilize the local microclimat e, facilitate the establishment of native forest species in their understory and eventua lly give way to species rich secondary forests. Though the Virgin Islands have many environmental similarities to Puerto Rico, the land-use history and conditions are unique and studies of secondary forest succession are far fewer. For example, Ray and Brown (1995a) found that 50 year old sites in St. John were dominated by L. leucocephala while 125 year old forest surrounded by protected park land no longer had the species as a dominant member of
17 the forest canopy or the understory. The Na ture Conservancy proper ty in St. Croix was used for intensive agriculture for dec ades and was abandoned and unmanaged since 1960. A forest inventory of the 10 ha. site found that L. leucocephala was the most abundant canopy species accounting 39% of the stems/ha and 36% of the basal area (m2/ha). Forty-five years after abandonment invasive species accounted for 46% of regeneration in the understory and the population of native species was comprised almost entirely of only three primary succe ssional species (Adam & Ryan, 2003). The authors concluded that forest regenerati on through succession did not seem to be occurring and that this site appeared similar in composition, structure and function to many other sites on the island and around the Caribbean. Summary St. Croix may be on the leading edge of a region-wide phenomenon as other islands als o make the shift toward tourism and industrialized economies. This study is the analysis of the secondary forests in St. Cr oix and was conducted in three distinct parts. First I measured the extent of se condary forests and other land covers on an island-wide scale. Geographic Information Systems (GIS) are ideally suited to measure landcover change at this spatia l scale. To show change, land cover maps were created using a combination of Lands at satellite data from 1992 (TM) and 2002 (ETM+) and other spatial data layers, and t hen compared. The 10 year time period is sufficient to record actual changes in land cover across t he landscape (Lambin, 19 99; Lambin et al., 2001) that are not simply temporary fluctuations, such as seasonal changes or changes in land-use intensity within a land cover class. I evaluated if St. Croix is gaining or losing total area for forest, grassland/agricultu re and development as well as identify the spatial distribution of that change. Analysi s of the switches between one land cover
18 class to another, combined with additional ancillary data sets, helped to identify drivers and trends in the landscape. A comparative analysis of the structure and species co mposition of secondary forests of varying ages evaluated the overall state of secondary forests in St. Croix. Field data was collected from a set of both random and purposefully selected sites and interviews conducted with prop erty owners were used to c onfirm land-use history and actual stand age. The main objective was to quantify the rate that secondary forests increase their species diversity and canopy st ructure. Likewise, an increased frequency of exotic species in stands of varying ages is an indicator of a slowed rate of secondary forest recovery through succession. Stand age is unlikely the only factor determining the rate and extent of recovery of secondary forests. Additional abiotic factors were analyzed by combining s patial layers (includi ng slope, elevation, aspect and soil) with field observations of vegetation in a GIS. For highly impacted sites dominated by invasive species succession can become stalled or arrested and some form of interv ention is required to restore secondary forests. I assessed experimental forest rehabilitation methods on such a site on St. Croix, owned and managed by The Nature Cons ervancy. If managed properly, these secondary forest sites have the potential to both regain their ecological function and to serve as refugia for rare or endemic species. Four species of native tree seedlings were grown from seed and planted in artifici ally created gaps in canopies of exotic trees. One half of the plants were randomly assigned a mulch treatment. Plant survival, plant height, canopy closure and rainfall were measured to determine the performance of each species in varying light conditions and the ef fect of the mulch
19 treatment on growth and survival. The method was successful overall and could be used to jump-start succession in undeveloped secondary forests in St. Croix and other parts of the Caribbean.
20 CHAPTER 2 LAND COVER CHANGE ANALYSIS USIN G LANDSAT SATELLITE IMAGERY FOR ST. CROIX, US VIRGIN ISLANDS Introduction Caribbean islands offer a unique setting for studying land cover change bec ause they have high population densit ies in topographically complex landscapes within fixed boundaries. In the Caribbean, where land area is limited and human population density is high, secondary forests ar e the most important and most rapidly expanding forest cover type (Lugo & Helmer, 2004). The rate and causes of tropical deforestation have traditionally received considerably more attention than forest cover increases and declines in the rate of forest loss (Mather & Needle, 1998), despite increases in forest area from places as diverse as Costa Rica, China, and India (Mather & Needle, 1998; Rozelle et al., 2000; Foster & Rosenz weig, 2003; FAO., 2006). In the western hemisphere forest increase results from hi storic economic shifts away from land intensive agriculture to manufacturing and is often accompanied by out-migration of agricultural workers from frontiers to urban centers (Rudel et al., 2005). Caribbean islands of Puerto Rico and the Dominican Republic have also experienced landscape wide reforestation in conjunction with a shift agric ulture to industrializa tion (Rudel et al., 2001; Grau et al., 2003). The va st majority of secondary fore sts arising from agricultural abandonment are not pl anned or managed, yet occur at the landscape level. Forest Transition Theory (FTT) describes a process where a trend of forest loss in a region slows and eventually results in a net gain in forest area (Mather, 1992). FTT assumes that land owners act in a rational way that brings the greatest financial reward (Angelsen, 2007). The single fund amental factor driving the transition is adjustments to the agriculture sector over time, but other factors can st rengthen or weaken the process
21 (Mather & Needle, 1998). This usually me ans a decrease in the agricultural sector results in abandoned farmland becoming forest ed and is also frequently associated with an outmigration of agricultural workers from rural areas to urban areas. St. Croix has transitioned from an almost completely forested landscape, to almost completely deforested during the height of sugar cane production, to modern times where these data indicate there is a steady-state of over 50% forested landscape. Forest Transition Theory relies largely on macro-economic factors to explain how large scale changes in forest cover can occur at the landscape level. Specifically, how industrialization and population migration can lead to large scale increases in forest cover as a result of abandonment of agriculture (Mather & Needle, 1998). Rudel (2005) and others have thoughtfully applied the Forest Transition Theory to several regions of the world to reveal consistent pat terns in several countries and distinct trends that are unique to the eas t and west hemisphere. Changes in St. Croixs forest cover hav e not been studied until relatively recently and there are few quantified detai ls available on forest cove r change over time in the Virgin Islands. The islands economy was once agricultural, focused primarily on growing sugar cane, cotton and tobacco (Weaver, 2006), and today in more industrialized. Neighboring Puerto Rico is 40 times greater in land area than St Croix (only 215 km2) but has similar climate, land-use histor y, forest types and is the site of extensive forest cover research studies. Fore st cover in Puerto Rico increased from 8% to 50% over 50 years that coincided wit h industrialization of the economy and landscape level abandonment of agricultural sites (Grau et al., 2003; Helmer, 2004; Marin-Spiotta et al., 2007).
22 Many Caribbean islands in the Lesser Ant illes chain are similar in size to St. Croix and the Virgin Islands and share simila r colonial histories, economics, land-use patterns and forest types. Unfortunately t here has been little to no land cover mapping in most of these islands. Results from an experimental method of mapping land cover and forest structure using Landsat ETM+ and course-resolution lidar are now available for the U.S and British Virgin Island (Kenneway et al., 2008). Also, preliminary estimate of land cover change have been conducted for some of the Windward Lesser Antilles (St. Kitts, Nevis, S t. Eustatius, Grenada and Barbados) indicating the same trend of increasing forest cover as land under cu ltivation has declined and development and construction increased in the lo wlands (Helmer et al., 2008). I examined landcover change in St. Croi x to address three objectives (1) To create accurate classified images of St. Croix for two dates (1992 and 2002) using a combination of remotely sensed and groundtruthed data. These snapshots describe land cover on the island at two specific poi nts in time and are the starting point for change detection. (2) To quantify and describe the changes that occur during the 10 year time period, using Forest Transition th eory to put those observations into context (Mather & Needle, 1998; Rudel et al., 2005). Ten years is considered an appropriate time period for understanding the processes that lead to forest loss and degradation (Lambin, 1999). (3) To identify the drivers effecting this change. Determining which land covers interact will identify which human land uses lead to the changes observed in the landscape. I hypothesis that land cove r change increases as proximity to roads increases. When roads provide accessibi lity to a previously isolated area, anthropogenic disturbance and la nd cover increases in Amazonian frontier (Laurence et
23 al., 2002; Asner et al., 2005) and tropical mo untain landscapes (Southworth & Tucker, 2001). However, St Croix and other Caribbean islands hav e high population densities within a finite space and inelastic borders. Methods Study Area The island of St. Croix (17 N 64 W) is among the Leeward Caribbean Islands, is the largest of the three main US Virgin Islands and is the easternmost possession of the United States. St. Croix is 10 km from nor th to south and 35 km from east to west, encompassing roughly 215 km2. The terrain in the north has rugged hills and steep slopes while the south is characterized by rolling lowlands and the coast is fringed with coral reefs. Virgin Islands soils are generally molisols derived from volcanic and marine parent material. O ne of the most common soil types is locally referred to as caliche, which is white in color, has slow drainage and a pH. >8.0. Soil in the north tends to be shallower and eroded while deeper soil and the mo st productive agricultural land are concentrated in the central va lley and the south (Davis, 2000). Trade winds blow from the east all year and are most prominent in the winter months. The warm Caribbean Sea stabilizes air temperatures so that diurnal temperature changes approximate annual fluctuations. There is a pronounced eastwest rainfall gradient similar to many of the less mountainous is lands of the Lesser Antilles, where the east end receives less t han 75 cm annually and the west end over 150 cm. Rainfall tends to be distributed i rregularly throughout the year and between years. The variable rainy season is from September to December but rainfall can be concentrated during tropical storms and hurric anes. Prolonged periods of drought occur in most years, typically from March to July.
24 The majority of the island is considered to be of the subtropical dry forest ecological life zone while the hilly, northwest portion of the island is subtropical moist forest (Ewel & Whitmore, 1973). The first European to arrive in St Croix was Christopher Columbus in 1493, who described a completely forested island landscape with a small population of Carib Indians. T here were sporadic small scale agriculture and timber extraction under Fr ench and other European rule for the next 250 years. Real change began in 1733 under Danish rule when the island was divided into a grid system, and cleared for agriculture; sugar cane became the primary crop. By the end of the 1700s St. Croixs remaining forests were in more isolated hilly areas and wood was being imported to fire mills t hat processed cane and made rum. Image Processing To detect changes in St. Croixs landcov er over a 10 year period, classified landcover maps were created using two sa tellite images; Augus t, 1992 Landsat TM and January 12, 2002 Landsat ETM+ image. The classifications for each date were conducted separately. Precipitation differences between the two dates were addressed by considering rainfall data at three m onths and three weeks before each image date and determining the potential effect this could have on classifying vegetation types (Jensen, 2005). The 10 year time period was selected to minimize detection of shortterm changes such seasonal fluctuations in forest or agricultural fields being freshly cleared or temporarily left to fallow and to capture meaningful change such as forest clearing and secondary forest regeneration. All image processing was conducted using Erdas Imagine software (ERDAS, 2003). Both images were separately georectified to 1:48,000 scale USGS topographic maps using a polynomial model with 150 point pairs resulting in a root mean squared error (RMS) of 0.7 pixels (<23 m) for each. This level
25 of error is acceptable for detecting changes (Wang et al., 2004), especially in a semiarid environment (Townshend et al., 1992). Careful visual inspection of the overlaid images confirmed there was minimal misr egistration. Appropriate atmospheric corrections were made (Song et al., 2001) incl uding those for atmospheric variation, solar angle (Jensen, 2005) and radiometric ca libration using dark object subtraction (Chavez, 1996). Using the layer stack func tion, I added seven additional spectral and spatial layers. NDVI (Normalized Differenc e Vegetation Index) is derived in Landsat from bands 1 and 2 (visible red and near infr ared) and is used to measure plant health or detect differences in vegetation types. The final spectral la yers were the three principle components layers resulting from a texture analysis of the Landsat image for both dates. The three non-spectral layers el evation, slope (United States Geological Survey, 2007) and a soils layer (Natural Re sources Conservation Service, 2000) were also added so the classification was not bas ed solely on reflectance values, thus improving class seperability. Pixels in cloud cover were treated as having no data and any pixels with no data were not used in the change analysis. Clouds account for 6% of the 1992 image and 8% of the 2002 (and are not ed in the classified images for each date) and account for a combined loss of 12% of the total study area in the two date change images. Image Classification and Ground-truthed Point Data. A ground-truthed data set were created in 2004 and used to train the classification model and to assess accura cy of the resulting im ages. A total of 246 points were randomly distributed across St. Croix, uploaded into GPS units, and trainin g sample data were collected at each point. Land use and land cover characteristics were collected at each point and additional veget ation characteristics were collected for
26 the forested sites (Lu et al. 2004) such as canopy structure and species composition. All of the points were later assigned to one of four land cover cla sses; forest, grass, developed and water. Other minor cover ty pes are included in these general headings and a complete description is in Table 2-1. The 246 points were overlaid on the stacked images and data extracted from pixels contacting the points in all layers. Data from one half of the points were used to create a uni que signatures for the class and the other half used in accuracy assessment (Jensen, 1996) All pixels in the images were compared to the signatures and assigned to a class using Erdas Imagines supervised classification technique. The same 246 poi nts were overlaid on a 1992 black and white digital orthophotograph to classify the 1992 Landsat image because field verification for that date was impossible. Basic aerial phot o interpretation techniques were used to assign the points to a single cover class. The same signature creation technique and classification process was used for both dates. Accuracy of the classified images was asse ssed by overlaying the other half of the field verified points on the classified im ages for each date and tallying the correct and incorrect classifications. Results are expressed in a confusion matrix that includes an overall accuracy, producer erro r (errors of inclusion) user error (error of exclusion) and for each class on each date, as well as an overall kappa coefficient that measures accuracy relative to random chance (Foody 2002). There is no universally accepted measure for assessing classification accuracy that is appropriate for all situations, but calculating multiple accuracy term s, as done here, is preferable. These land cover classification results were further verified by comparing them with three other independent estima tes of forest cover on St. Croix. One is a land cover
27 classification mapping project (University of the Virgin Is lands, 2002) completed through traditional photo-interpreting techniques of a 1999 aerial photograph. Secondly, the US Forest Service first Forest Inventory A nalysis (FIA) was conducted in 2004 (Brandeis & Oswalt, 2007) via intensive field data samp ling and resulted in a ground-truthed map of forest cover types. Finally, results from multiple years of t he US Department of Agricultures Census of Agriculture were used for comparison of changes in area to the agricultural sector (National Agricultural St atistics Service, 2005; National Agricultural Statistics Service, 2009). These data are interesting and uniquely independent because they are derived through interviews and surv eys with farmers and are not spatially explicit. Land Cover Change Analysis Change detection was conducted to deter mine the nature, spatial extent and pattern of l and cover change (MacLeod and Conga lton, 1998). I show the results in a cross-tabulation change matrix to effectiv ely quantify persistence, change and the type of change observed (Pontius et al., 2004). The matrix describes (1) net change in each category, (2) gain, loss and swap by class, and (3) stability or persistence in each class. Water was retained as a separate class although there is very little fresh water on St. Croix. Pixels changing to or from water were few and re present either tidal change or ephemeral water fluctuation and not anthropogenic cover change. The results of the changes in forest and grass area are descr ibed within the context of the Forest Transition Theory (FTT), the change from forest loss to fo rest expansion for a given geographic area that occurs as a cons equence of broader demographic, economic, and/or political changes (Mather, 1992).
28 Fragmentation Landscape metrics were calculated with the software Fragstats 2.0 for the classified images of the two dates. Fragstats provides a suite of spatial statistics and descriptive metrics at the patch, class and landscape lev el. I used several measures of area, edge and shape determine the fragmentation of each class. The variables Area, Largest Patch Index (LPI), and Average Patc h Size (APS) measure the entire area occupied by a class, the area of the lar gest patch within a class, and the average area of the patches within a single class. Number of Patches (NP) is a simple count and a direct measure of fragment ation where more patches within a fixed area means increased fragmentation. Land scape Shape Index (LSI) is a unitless measure of the overall complexity of the class shape. Clumpiness and Interspersion-Juxtaposition Index (IJI) are measures of whether patches from a single class are clustered together or intermixed with other class types. Tak en together these measur es give a detailed picture of the spatial characteristics of land cover types in a landscape and are an effective measure of fragmentation (Nagendra et al., 2003; Southworth et al., 2004). The landscape size is equal for the two dates, so comparisons are simple to interpret. Fragstats was also used to calculate thes e same metrics for the change image that compares the differences between the two dates. In this way the spatial characteristics of stable patches can be compared with patches that experienced change. Forest Cover Change and Roads The amount of forest cover change in an area is frequently associated with roads. Because roads provide increased human access, the change is presumed to be anthropogenic. I tested this relationship in St. Croix by overlaying a layer of primary roads on 1992-2002 forest cover change layer. The roads were buffered in a GIS to
29 create eight distance classes (0-100, 100-200, 200-300, 300-400, 400-500, 500-700, 700-1000, and 1000+ m). These distances were c hosen so that each contained at least 1,000 terrestrial ha2. Water was excluded from the analysi s because it is not associated with anthropogenic change in St. Croix. The percentage of stable land cover and forest change were analyzed to determine if chang e area decreased with increased distance from roads. Results and Discussion Classified Images The four-class images for 1992 and 2002 (Fi gure 2-1) show that forest is the dominant cover type and that land cover class distribution patterns are relatively stable between dates. Forest covers the rugged slopes of the northwest and higher elevations along the north coast and eastern hills. Lar ge grassy pastures are concentrated along the south eastern coast with smaller pasture s occurring in a mix ed matrix throughout the lowlands. The south central coast is a designated industrial zone where the airport, an oil refinery and other commercial enterpris es are located. Other development is concentrated in patches throughout the centra l valley. Water encircles the coast and the only inland bodies of water are the ephemeral salt ponds in the east and west ends of the south shore. Producer and user accuracy for each cla ss, overall accuracy and kappa statistics are given in Table 2-2. There were two ma in sources of classifi cation error. Some pastures on forest edges or within mature patches of woody regrowth were misclassified as forest, while patchy c oastal scrub forest in the dry east end was occasionally misclassified as either pasture or even developed (bare ground). The second type of error was between low densit y residential and forest. Accuracies for
30 1992 and 2002 of 72% and 75% were achieved, with kappa statistics of 61% and 66%. These levels may be acceptable for indivi dual image dates, but are lower than the preferred 85% accuracy for images in a change analysis (Thomlinson et al., 1999). However, calculating user and producer accu racy levels and kappa statistics allows us to use these data with knowl edge of their strengths and weak nesses and is the current best method for measuring accuracy of classified images (Foody, 2002). Because forest cover change is the focus of this study, the images were further simplified to focus on forest change and to im prove accuracy. A raster recode process combined the developed class with grass co ver. The new Forest/Water/Other image was assessed for accuracy, which improved to 83% and 91% for 1992 and 2002, respectively. These data are used in forest cover change analysis throughout this study (Table 2-3). Water is retained as a separat e class because it is easily distinguished from other cover types. These classified images describe St. Croix as being 56% forested in both 1992 and 2002. The forest cover estimate by the Rapid Ecological Assessment (University of the Virgin Islands, 2000) used different methods the calculations, but concluded that St. Croix had 53% forest cover in 1999 and a sim ilar distribution of cover types, with the majority of forests concentrated in the hilly northwest portion of the island. The results of the FIA conducted in 2004 found St. Croix to be 53% forested, though the definition of forest was more liberal to include areas of potential forest (Brandeis & Oswalt, 2007). The authors characterized the forests as la rgely secondary stands in early successional stages and often dominated by a si ngle exotic plant species, Leucaena leucocephala. The most mature secondary stands were found to be concentrated in the hills of the
31 northwest part of the island. These comparisons corroborate the finding in this study that St. Croix is slightly mo re than 1/2 forested and there was little net change in forest area between 1992 and 2002 Land Cover Change Class by class comparison was conducted here using a cross tabulation matrix (Table 2-4). Results quantify two types of landscape change between categories; gross change for each class and net change and s wap (Pont ius et al., 2004). The net change among classes between 1992 and 2002 reveal s a largely steady-state forested landscape with a distinct trend of increas ing human development and a proportional decrease in cultivated grassland for pasture (F igure 2-2). Forest cover remained stable at 56%, (10,540 in 1992 to 10,590 ha in 2002 ) as did water (1,290 ha and 1,250). Grass cover, however decli ned substantially (over 1,000 ha or 34%) and there was a simultaneous increase in developed area by t he same amount. There are no naturally occurring grasslands in St. Croix and the vast majority of grass co ver is pasture for grazing cattle sheep and goats. An independent, non spatial data set was used to corroborate this finding of rapid change in grass cover during the 10 year study period. The Virgin Island Department of Agriculture conducts an Agricultural Census every five years in conjunction with the US Department of Agriculture. These data ar e collected through surveys and interviews with farmers and the dates closely approximate the time period covered in this study. USDA data shows that between 1998 and 2002 (Nat ional Agricultural Statistics Service, 2005) the area in pasture in St. Croix decreased from 4,050 to 2,880, a 29% decrease in pasture area. Although not confirmed thr ough site visits nor assessed for accuracy, this independent data demonstrates the same trend in pasture loss gives additional
32 confidence to these findings. It is interesting to note that St. Croixs largest dairy farm has declared bankruptcy and closed since the time this data was collected likely perpetuating the trend of pasture land coming out of production and converting to either secondary forest or being developed. A more detailed look at land cover change is achieved analyzing gross change within categories and swap (Figure 2-3). Water, for example, remained stable at 7% for both dates, but spatial change within the clas s is visible, primarily with the two ephemeral salts ponds on the south coast of t he island (Figure 2-1). These sites are a combination of salt flats, sand, water and mangrove fores t, and the water level varies with tides and rainfall. The salt pond in the south west corner of the island changed from developed to water while the brackish Gr eat Salt Pond in the south east drained, exposing mangroves and sand. The develop ed class also includes bare soil and sand because they have similar spectral signatures (Table 2-1). These changes in forest, developed and grass are either to or from water and are considered natural annual fluctuations of the coastal environment, therefore, they are not included in the analysis of anthropogenic change. Forest cover experienced no subst antial net change, but there was approximately 1,500 ha gain and loss of forest area. Forest gain primarily came at the expense of grass cover, as 1,040 ha of gr assland reverted to secondary forest during the 10 year study period. The other 500 ha came from the developed class and careful visual inspection confirmed this findings. The spatial distribution of all the stable classes and all types of change is depicted in Figure 2-4. A percentage of this change can be attributed to error as development is generally considered a permanent cover
33 type. However, the majority of changes from developed to forest occurred in the mixed coastal scrub forest of the dry east end. The steep slopes have patches of bare rock and were classified as developed in the 1992 im age. The change likely represents real forest development. However, a possible contri buting error factor is that 40% more rain fell in the three months prior to the 2002 image t han the 1992 image. Such a rainfall increase can cause denser, fuller tree canopies and result in pixels that are a mixture of grass and scrub being classed as grass during drought and as forest during high rain. The other noteworthy location is the highly patchy cluster of small patches around a large residential ne ighborhood. These patches are lots cleared to bare ground for development and allowed to regenerate into a low patch of woody cover during an extended building process. Forest Change A primary objective of this study was to determine the extent and distribution of forest cover change. To focus on forest change only, the classified image was recoded to a Forest/Non-forest classification by combining the pasture and development classes into a single group. Water was detected highly accurately and changes within the water class represent natural changes, such as tidal fluctuation and variation in ephemeral water bodies. Mangrove forests, for exam ple, frequently appear to fluctuate between water and forest cover during change det ection. The 1992 and 2002 images were recoded into three classes (forest, water, other) and assessed for accuracy resulting 82.88% and 91.43% respectively and no singl e class less than 70% for users or producers error (Table 2-3). Comparing t he two images resulted in a change matrix with nine possible categories.
34 Over the 10 year period gross forest area remained stable, at 56% of the study area. Of the 19,000 ha in t he study area, 9,200 ha (48%) were stable forest and 5,500 ha (29%) were stable non-forest areas. N on-forest areas include both developed and grass. Stable water and pixels swapping to or from water are combined into a single category because they account for a relati vely small total area (7%) and represent natural change rather than anthropogenic change. The spatial distribution of forest gain and forest loss is shown in relation to stable land covers in Figure 2-4. Changes in population, worker wages and government programs can have profound effects on the agricultura l sector that drives Forest Transition. A generalized graph of the historical changes in forest cove r on St Croix appears in Figure 2-5 and is similar to that described for many other countries in the western hemisphere. In St. Croix, land prices and workers wages hav e increased sharply in recent decades, making agriculture a much less attractive option for land owners. St. Croix is small enough that when workers discontinue farming on their family land and take wagepaying jobs, it does not result in the out -migration frequently discusses in FTT because workers can simply commute by car. However, Virgin Islands Department of Agriculture programs give 95% property ta x exemptions for properties registered in agricultural production. The least labor intensive type of agr iculture is pasture/livestock. The result is St. Croixs agricultural sector is compri sed mainly of infrequently maintained pastures where property tax exemption is more impor tant than earnings from actual livestock production. Larger commercial farms close due to changing economic realities, while the smaller land holders maintain their agriculture exemptions in order to avoid property taxes. Evidence of this is found in the US DA Census of Agriculture which finds the
35 average pasture farm in St. Croix decreas ed in size from 54 ha in 1992 to 33 ha in 2002. Many of the smallest pastures are infrequently main tained and fluctuate between grassy pasture and advanced woody regrowth. Data from this study captures this phenomenon, finding that during t he 10 year time frame, almo st 15% of the island was converted either from forest to pasture or from pasture to fore st. This indicates that total forest area in St. Croix may have achieved a post-FT steady state, but that the spatial distribution of forest remains a dynamic, shifting mosaic. The current reality is strongly influenced by the territorial governments ta x incentive and forest area would almost certainly experience a sharp increase if this incentive were to discontinue. Fragmentation Fragstats metrics for the three terrestrial classes in 1992 and 2002 are given in Table 2-5, as well as data for forest gain and forest loss. Grass cover experienced a large decrease in area, Number of Patches (NP), and a 13% reduction in the Average Patch Size (APS) as well as a decrease in average complexity of the shape of the class (LSI). While forest area remains stable between the two dates, the NP increased by 156, a strong indication of fo rest fragmentation. Forest LPI decreased slightly and the APS decreased by 29%. These numbers descri be a stable net area of forest becoming divided into a greater number of smaller patches. Interestingly, as forest becomes fragmented, developed areas are growing and merging t ogether. The total area and average patch size of the developed cla ss increased substantially while the NP decreased by over 100. The spat ial distribution of all types of forest change is shown in Figure 2-7 with colors distinguishing between the cover type the area changed from/to. Both reforestation and deforestation are conc entrated in the lower elevations, away from slopes and patches of both types of change are frequently found in association
36 with each other. Visual inspection at finer s patial resolution reveals that deforestation is usually the result of specific human activiti es such as construction or clearing fallow fields and reforestation results from cessation of human activity (the maintenance of pastures) and fields reverting to fallo w. Although one can differentiate between reforested and deforested areas during this time period, it is more meaningful to consider both types of change as parts of a single dynamic fluctuation process. This change is brought about by human activity and pasture management and residential development are the core drivers. Spatial analysis using Fragstats indica tes that reforestation and deforestation patches are highly similar in terms of av erage patch size, largest patch and edge density. Change patches have a smaller maximum size and average size than stable patches as well as a higher edge density than other patch types. An Interspersion and Juxtaposition Index (IJI) was ca lculated for each of the 6 classes. The index provides a measure (0 100) of intermixing of patch types by looking at each patch and what patch type is adjacent to it. A score of 1 i ndicates a class is adjacent to pixels of a single cover type and a 100 score means a class is equally adjacent to all other types. Reforestation and Deforestation both had IJI=37 while stable forest (45.6), stable water (53.5) and stable development ( 63.1) all scored higher. Forest Cover Change and Roads Roads frequently serve as a proxy for t he presence of human activity in land cover change studies and land at greater dist ances from roads is considered less accessible. I hypothesized that a small island like St. Croix with a high population density would not show the traditional pa ttern of increasing cover change with decreasing distance to roads because the entire island can be considered part of the
37 urban/wildland interface. In other words, humans have access the whole island, so forest cover change will not vary with road distance. I tested this by using eight road buffer classes to determine if any observed changes occur gradually. The results indicate that forest gain and forest loss occurred in conjunction with each other spatially and that gain and loss occurred in equal proportions within each distance category. Contrary to predictions, the proporti on of the land experiencing forest change decrease as distance to roads increased (Figure 2-7). However, the trend only becomes observable at the 500-700 m distance class and over 73% of the island falls within 500 m of a primary road and 94% is within 1,000 m. The variation on St. Croix occurs suddenly over a relatively short distance as previous studies have reported the findings over distances of >130 km (Laurence et al., 2002; Southworth et al., 2004). At a similar spatial scale, Helmer (2004) found that newly developed areas in Puerto Rico occur closer to roads although even Puerto Rico is over 40 times greater in area than St. Croix. I interpret this to m ean that only a very small percentage of St. Croix is inaccessible enough to humans that land cover change patterns differ. Together these data describe a shifting mosaic of land covers undergoing change in a post-agricultural context. St. Croixs landscape is roughly 50% forested, consisting of primarily young, secondary fore st stands (Brandeis et al., 2009) by forest cover and livestock production in pastures is the most dominant agri cultural practice. The Department of Agriculture Census data shows a sharp decline in revenues and land area for almost all types of agricultu re, especially livestock and pastures. Grass cover area also declined sharply between the 1992 and 2002 images and there was a corresponding increase in development. Howe ver, there was no net increase in forest
38 area indicating St. Croix may have achieved the new steady state in forest area predicted by FT theory (Mather & Needl e, 1998). There may be some unique characteristics due to the size of the island. With only 215 km2, there are few inaccessible areas and migration to urban ar eas does not occur due to declines in agriculture. Spatial analysis of the forest change areas indicates that patches of forest gain and forest loss have similar sized patches and both occur in lowland areas and in transition zones where all three terrestrial co ver types tend to be mixed (as indicated by low measures for Clumpiness and IJI in T able 2-5). The net effect is increased fragmentation of forest cover, trending away from large patches of stable forest and toward smaller patches of young secondary forest. If the trends identified in this study continue, St. Croix will see the continued spread of development on pr ime agricultural lands and forest land becoming increasingly young, fragmented and relegated to hillsides. This is surely a daunting prospect for an island that already imports the overwhelming ma jority of its food and is currently advertising itself in the tourism indust ry as The Nature Island. If St. Croix is to maintain a percentage of green spac e or forested areas around residential developments, efforts must be made to prot ect land as it comes out of agricultural production and before it becomes developed. St. Croix does not yet have effective zoning and development regulations and the te rritorial government is ineffective in maintaining protected areas or creating new ones. A recent example is the territorial senate acting to lease protected coastal land to a casino developer, despite passage of a voter referendum to block the move. The federally funded Forest Legacy Program and Forest Stewardship Program offer realistic ways for land-owners to take their
39 properties out of agricultural production while maintaining both their tax exempt status and forest cover. It is st ill uncertain if these programs will be effective in the Virgin Islands. Conclusion These classified images provide accurate, spatially explicit descriptions of land cover on St. Croix on two dates. The chang e analysis describes a forest dominated landscape (56%) experiencing a high degree of change. Total forest area remained stable, but there was subst antial change within the class and a strong trend toward younger secondary forest and increased forest fragmentation. Approximately 1,500 ha of forest were lost, with the majority conv erted to development and 1,500 ha of forest were gained, the majo rity coming from regener ating grassland. Over 5% of the island changed from grassland to developed during the 10 year time span. The loss of grassland was corroborated by independent non-spatial data collected by the Virgin Islands Department of Agriculture. These findings indicate St. Croix is still experie ncing Forest Transition and ec onomic realities favor increased urbanization at the expense of grassland while net area of fore st remained stable. I tested the hypothesis that changes in forest cover decrease with increased distance to roads and conclude that forest change area only varies at the greatest distance values. The pattern is not as pronounced as in other studies and occurs over a smaller spatial scale. Also, patches of both forest gain and forest loss are spatially (patch size, location and distribution) sim ilar and are two parts of the same process where small patches of land are irregularly cleared and then left fallow. The process is driven by macroeconomic changes (increasi ng worker wages, population density and
40 real estate prices) and re-enforced by local policies offering property tax exemption to land that is registered as agricul turally productive in name only.
41 Table 2-1 Names and description s of the four land cover classes used in the land cover classification of 1992 and 2002 Landsat im ages of St. Croix, U.S. Virgin Islands Cover Class Description Forest All forest types with >25% canopy coverage including coastal scrub, early successional stands, secondary dry fo rests and mature multi-layered moist forest. Grass Pastures, regenerating pastures wit h patches of woody regrowth covering <25% woody regrowth, recr eational fields, golf courses and large lawns. Developed Urban, mixed residential, paved roads, parking lots, bare ground, beach sand and mud flats. Water Coastal sea water, salt ponds and estuaries.
42 Table 2-2. Overall accuracy assessment of the four-class images for 1992 and 2002. These images are the base layers utilized in fragmentation analysis. Class name Reference total Classified total Number correct Producer accuracy User accuracy 1992 Forest 52 45 29 55.77% 64.44% Grass/pasture 14 20 11 78.57% 55.00% Developed 24 26 20 83.33% 76.92% Water 21 20 20 95.24% 100.00% TOTAL 111 111 80 Overall Accuracy 72.07% Overall Kappa 61.35% 2002 Class name Reference total Classified total Number correct Producer accuracy User accuracy Forest 52 53 36 69.23% 75.00% Grass/pasture 14 16 10 71.43% 62.50% Developed 20 18 15 75.00% 83.33% Water 19 18 18 94.74% 100.00% TOTAL 105 105 79 Overall Accuracy 75.24% Overall Kappa 66.41% [Producer accuracy (# correct/reference total) me asures errors of exclusion and User accuracy (# correct/classified total) meas ures errors of inclusion.]
43 Table 2-3. Accuracy assessment for the three cover classifica tion used to detect forest cover change. These images serve as the base layers for land cover change analysis. Class name Reference total Classified total Number correct Producer accuracy User accuracy 1992 Forest 52 45 39 75.00% 88.64% Non-forest 38 47 33 86.84% 70.21% Water 21 20 20 95.24% 100.00% TOTAL 111 111 92 Overall Accuracy 82.88% 2002 Class name Reference total Classified total Number correct Producer accuracy User accuracy Forest 52 52 48 92.31% 92.31% Non-forest 34 35 30 88.24% 85.71% Water 19 18 18 94.74% 100.00% TOTAL 105 105 96 Overall Accuracy 91.43% [Producer accuracy (# correct/reference total) me asures errors of exclusion and User accuracy (# correct/classified total) meas ures errors of inclusion.] Table 2-4. Cross-tabulation matrix co mparing classified images for 1992 and 2002 expressed as a percentage of the total study area. 2002 Total 1992 Loss Forest Grass Developed Water 1992 Forest 47.78% 3.17% 4. 86% 0.00% 55.80% 8.03% Grass 5.49% 6.36% 4.12% 0.00% 15.97% 9.61% Developed 2.75% 0.95% 17. 07% 0.62% 21.40% 4.33% Water 0.03% 0.00% 0.79% 6.01% 6.83% 0.82% Total 2002 56.05% 10.48% 26.83% 6.63% Gain 8.28% 4.12% 9.77% 0.62%
44 Table 2-5. Seven landscape metrics [Area (% of total area), LPI (Largest Patch Index), NP (Number of Patches), APS (Average Patch Size), LSI (Landscape Shape Index), Clumpiness and IJ I (Interspersion and Juxt aposition Index)] that describe fragmentation for the three terrestrial land cover classes on both image dates. DATE CLASS AREA % LPI (%)NP (#)APS (ha) LSICLUMPINESSIJI 1992 DEVELOPED 21.29 4.3910634.1639.99 0.8977.09 GRASS 16.07 2.9113762.4139.42 0.8860.50 FOREST 56.00 38.277835.8428.03 0.9170.71 2002 DEVELOPED 26.83 4.349465.6039.56 0.8885.20 GRASS 10.52 2.4510102.0931.53 0.8760.72 FOREST 56.19 37.559394.1628.87 0.9175.22 19922002 FOREST TO NONFOREST 8.18 <0.121830.71NA 0.7837.47 NONFOREST TO FOREST 7.32 <0.120800.67NA 0.7737.51
45 A 1992 B 2002 Figure 2-1. Classified land cover maps for St. Croix, A) 1992 and B) 2002. Green=forest, yellow=grassland/pasture, grey=development, blue=water and white= no data (normally as a result of clouds).
46 Figure 2-2. Total area of four land cover classes in 1992 and 2002 in St. Croix, U.S. Virgin Islands. Values for 1992 appear on the left hand side for each of the cover types and 2002 values appear on the right. 0 2000 4000 6000 8000 10000 12000 ForestPasture/GrassDeveloped Waterha 1992 2002
47 A 1992. B 2002. Figure 2-3. Transition within each of the f our land cover classes during the 10 year study period. 4A) Horizontal bars represent the total area in each of four land cover classes in 1992. The colors within the bars indicate how that same area became classified in 2002. 4B) hor izontal bars represent the total area in each of four land cover classes. The colors within the bars indicate how that same area was classified in 1992. Note that the total area in forest remains stable between the two dates while developed increases and grass decreases. 0%10%20%30%40%50%60% Forest Grass Developed Water Forest Grass Developed Water 0%10%20%30%40%50%60% Forest Grass Developed Water Forest Grass Developed Water
48 Figure 2-4. Spatial distribut ion of areas experiencing ch ange (shown in red) between 1992 and 2002 relative to the three stable land cover types. Figure 2-5. An idealized graph of the percentage of forest cover on the island of St. Croix through time. 0% 50% 100% Pre European Colonization (before 1500) Pre European Colonization (before 1500) Sugar Cane Era (1700) Purchase by United States (1916) TodayLand Area in ForestTime
49 Figure 2-6. Distribution of forest gain and forest loss from 1992 to 2002 shown in relation to topography. Forest loss to fo rest appears in red and loss to grass in indicated by orange. Forest gain fr om pasture is indicated with green and forest gain from developed class is blue. Much of the forest gain from the developed class was previously bare gr ound, sand or mud flats and does not represent urban development converted to forest.
50 A B Figure 2-7. Change area relative to the distan ce from main roads in St. Croix. A) The spatial distribution of t he six road buffer distance classes and B) the percent land area in each distance class occupi ed by forest gain and forest loss. 0 500 1000 1500 2000 2500 3000 3500 4000 0% 2% 4% 6% 8% 10% 12% ha Percent Area within buffer classDistance from primary road (m) Area in buffer class Deforested Reforested
51 CHAPTER 3 DIVERSITY AND STRUCTURE OF SECONDA RY F ORESTS OF VARYING AGES Introduction Native tropical and sub-tropical dry fore sts are high in species diversity and endemism and are becoming increasingly rare across the world. Forests are lost primarily due to conversion of forestl and to agriculture. Subsequent abandonment leads to creation of secondary forests, wh ich have been the dominant forest cover type in the tropics for decades (Brown & Lugo, 1990) accounting for over 60% of all tropical forests and increasing (FAO, 2006). T hese secondary forests form under unique conditions that often include altered disturba nce regimes, and harsh localized climatic and edaphic conditions which results in st ands with novel species compositions. Secondary forests will likely play an increasing role in conservation of forest biodiversity (Chazdon et al., 2009) especially in fragmented landscapes. There is a decline in global net rate of forest loss, attributed primarily to an increase in secondary forests in certain regi ons of the world (FAO., 2006). The species composition of secondary forests often contains a mix of pioneer and mature forest trees as well as native and exotic plants which reflects the unique conditions under which they were formed (Lugo & Helmer, 2004) Native mature forest species are frequently not present if there are few remn ant trees standing after disturbance and the modified environmental conditions inhibit seed dispersal of large-seeded species. Introduced exotic species play an increased role and agricultural weeds are especially common in the earliest stages of regeneration. Fo rests combining aspects of mature forests and young secondary stands can be c onsidered hybrid systems, including secondary forests with remnant mature patc hes, stands with severe a reduction in
52 mature forest trees such as heavily thinned stands, and with elevated presence of exotic species (Hobbs et al., 2008). Tropi cal secondary forests can provide a wide range of services in addition to conservation of biodiversity includ ing fuel wood, small game (Roth, 2001) and sites for agroforestry practices (Vieir a et al., 2009). As these spontaneously regenerating, frequently unmanaged se condary forests become increasingly common, so has the desire fore st rehabilitation strategies that return ecological function to the systems A land manager could facilitat e establishment of desirable species to produce a relatively stab le system that provi des ecological goods and services (Seastedt et al., 2008; Jackson & Hobbs, 2009). Secondary forests can be dynamic and re silient, quickly replacing species over time and developing structure and diversit y (Brown & Lugo, 1990). Human land use history is among the most impor tant factors determining the rate and extent of forest recovery. The duration and intens ity of the previous land us e greatly influence the rate of recovery as well as the structure and s pecies composition of the secondary forests that follow abandonment. Unfortunately, intensit y of land use is a difficult variable to measure and quantify. Recovery can be rapid if the previous land use was not intense and a propagule source is nearby (Guariguata & Ostertag, 2001; W eaver & Chinea, 2003). The secondary forests that regenerate c an recover similar levels of ecological function and species diversity over decades (Lugo & Helmer, 2004), but tend to have unique species composition with fewer native species and greater do minance by exotic species (Marin-Spiotta et al., 2007; Brandei s et al., 2009). When native forests are cleared and new species introduced, a novel ecosystem can arise, comprised of a completely new combination of species (Hobbs et al., 2008) Forests cleared for
53 agriculture have recovered species diversit y, stem density and basal area after 50 years of abandonment in Puerto Rico, New England and el sewhere without any human intervention (Grau et al., 1997; Martin et al., 2004), th ough other estimates predict nearly 200 years may be required to reach mature forest basal area and biomass values, even on lightly used land (Saldarriaga et al., 1988). Succession is influenced by in situ factors as well as the conditions of t he surrounding landscape. Intact soil seed banks and stump sprouting lead to diverse fo rest that can resemble the species diversity of original stands. Human al tered disturbance regimes, such as the introduction of fire can eliminate soil s eed sources and stump resprouts. Repeated burns reduce soil nutrient levels such a ni trogen and phosphorous and significantly slow rates of forest regrowth as well as cause the system to be more susceptible to future fires (Zarin et al., 2005). In Puerto Rico, studies of sec ondary forests regener ating on abandoned agricultural land suggest that the least impacted sites (coffee and fruit production) recover species diversity more quickly and c ontain more rare and endemic species than sites with high impact forms of pasture and agr iculture (Rivera & Aide, 1998; Pascarella et al., 2000; Thompson et al., 2002). The recovery rates and process in these studies are important for St. Croix and t he rest of the Lesser Antilles because of their proximity, similarity in ecotypes, forest species compos ition and relatively simi lar land-use history. However, the differences between St. Croix and these studies in also noteworthy. St. Croix is only 1/40 the size of Puerto Rico and has simpler topography and fewer ecotypes. Many secondary forest succession studies there are in the higher mountain regions where land use was low intensity (such as shade coffee production and fruit
54 orchards) for short time periods. St. Croix has few high mountains, making most of the island accessible for human activity. Burnin g and tilling were part of intense agricultural production of sugar cane and cotton practiced throughout the lowlands of St. Croix for centuries (Weaver, 2006). Intens ity of land-use history is known to have long lasting on forest recovery rates and species composit ion (Thomson et al., 2002). In addition, studies in mountain areas with high rainfa ll, surrounded by protected forest area (Pascarella et al., 2000; Thompson et al., 2002) are likely to experien ce faster recovery than the mosaic of former pasture and agricultural fields in the lowlands of St. Croix. Most studies of succession in secondary fo rests tend to focus of specific study sites and studies examining the effects on land use across the entire range of environmental gradient are uncommon (Brandei s et al., 2009). The Forest Inventory Analysis (FIA) in Puerto Rico and the Virgin Islands are utilized in this study because they are island wide, although the FIA does not address land use history or intensity. Studies of secondary forest on abandoned agricul tural sites in nearby island such as Jamaica indicate early site conditions sele ct against native tree establishment (McLaren & McDonald, 2003). Prolonged intense agriculture followed by persistent low-level disturbance to regenerating forests in the Do minican Republic resulted in severely hampered or arrested succession, even (Roth, 1999). Simple st ands dominated by invasive species persist after the disturbance is eliminated. A similar type of arrested succession has also been described on former pastures in fragmented mountain landscape of South America (Sarmiento, 1997) and Africa (Chapman et al., 1999). Other studies similar to the abandoned pasture in the dry sub-tropical forest in St. Croix have been in Puerto Rico (Aide et al., 2000: Zimmerman et al., 2000: Lugo & Helmer,
55 2004). The sites in these studies contain plant communities with a highly altered species composition where exotic invasive species represent a large percentage of the stems and aboveground biomass. These provide high level of detail on a small area over fairly uniform environmental condi tions and none have a comprehensive sampling to incorporate island wide conditions. The goal of this study is to determine the extent to which native forests are regenerating in St. Croix, U.S. Virgin Is lands and how exotic invasive plants may influence the process. I assess the compos ition and structure of secondary forests in St. Croix, across a temporal gradient (s tand age) and the full spectrum of land-use gradients on the island. Secondary forest stands ranging in age from two to fifty years are analyzed to determine how species rich ness changes over time and are also compare to some of the most mature stands. If secondar y forests can recover their structure and composition over time than non-interference can be a low-cost method of achieving forest restoration in targeted areas In the Virgin Islands, however, forests occur in a highly fragmented mosaic of agr iculture and development where invasive plants play a prominent role in succession. Do these newly introduced plants function in the ecological role of pioneer species just as the native pioneers do, or do they have a longer lasting influence on succession? I assess this by determining: 1) what suite of plant species are present in recovering fore sts of varying ages? 2) How does species diversity of a site vary with age? 3) How do other factors, such as land-use history, slope and soil type influence the pattern of forest recovery?
56 Methods Study Area St. Croix in the U.S. Virgin Islands prov ides an excellent opportunity to study post agricultural secondary forests. The small size (215 Km2) and relative isolation of an oceanic island limit some of t he confounding factors (such as the influence of pollutants or native/exotic organisms from neighboring systems) in herent in studying forest ecology in an uncontrolled environment. St. Croix experienced widespread forest clearing for agriculture, primarily sugarcane, cotton and pastures, during the colonial period starting in the 1500s and eventually resu lted in native forests remaining in only isolated fragments (Weaver, 2006). Like ne ighboring Puerto Rico, economic changes starting around the 1950s resulted in wides pread agricultural abandonment and forest regeneration (Helmer, 2004). Se condary dry forests are by far the most dominant forest cover type in St. Croix and represent the futu re forests of the territory (Brandeis & Oswalt, 2007). This study analyses if St. Cr oixs secondary forests are able to recover their native species diversity over time without human intervention. Field Data Collection I collected field data during the summers of 2004 and 2005. Random points from a previous land cover classification study were used to select a portion of the vegetation sampling sites. Howe ver, many of the points both were difficult to physically access and a challenge to determine the actual age of the forest stand. Therefore, additional sites were purposef ully selected in accessib le locations where land use history and stand age could be accurately determined through personal knowledge of the property, property owner interviews and field data. Land owner interviews and 1992 black and white aerial photogr aphs helped determine stand age at each site. The
57 secondary forests sampled were assigned to one of five age classes (Table 3-3). Vegetation data was collected along a total of 50 transects measuring 3 x 50 m (as described by Aide et al., 2000) on 30 sites ( n =30). The transects were broken into 10 m segments and presence/absence was recor ded for all tree species, which were identified to the species leve l (Figure 3-1). Canopy height and structure, percent ground cover, presence of vines, lianas orchids and epiphytes were also recorded (Figure 3-2). Waypoints were recorded at the start and end of each transect using a Garmin, e -trex GPS unit. Data Analysis The vegetation data from the transec ts were analyzed in SAS 9.1 using a generalized linear mixed model (proc GLIMMIX) that treated sites as a random effect and transects as spatially auto correlated. Data was summarized by age class and compared for differences. Simple linear regression was performed to determine the effect stand age had on species dive rsity of the secondary forest sites. In addition to age, spatial data were also analyzed (proc GLIM MIX) relative to species diversity and the location of age classes. These field data and their geographic attributes were combined in a geodatabase using a Geographic Information S ystem (GIS) where additional layers were added, including soils, slope, and el evation using the GIS software package ArcGIS 9.1. Results from the 2002 land cover classification of St. Croix (see Chapter 2) and results from the 10 year land cover c hange analysis were also included in the geodatabase and used to help confirm details on site land-use history. Field data from this project were cross referenced with the re sults of the Forest Inventory Analysis (FIA) conducted in 2004 by the U.S. Forest Service (Brandeis & Oswalt, 2007). Sampling
58 density for the Virgin Islands FIA was increa sed because of the sm all land area in the territory and contains detailed information on forest species composition, stem density, basal area, above ground bioma ss, downed woody debris a nd other data on the sites sampled. Exact locations of FIA data points are kept confidential, so transect locations and forest cover polygons were sent to the Forest Service FIA la b where they produced summary tables use in this analysis. Results Across all sites, a total of 101 woody spec ies were identified. Table 3-1 lists the most common species and their frequency (num ber of observations where the species occurred divided by the total number of observations). The vast majority are tree species while some of the most common li anas and woody shrubs are also included. The most frequently encountered tree species by far was the exotic invasive Leucaena leucocephala found in 67.5% of the samp le segments (Table 3-1). Capparis indica is a medium sized tree, ubiquitous to the Virgin Islands dry forest, and was the most frequently (27%) encountered nat ive among all age classes. The thorny, exotic Acacia macracantha is a common pasture weed found in 24 % of the plots sampled. Although native, Cordia alba is considered a weedy tree and Lantana camera is an early successional shrub associat ed with highly disturbed sites and each were found in roughly 20% of all sites. Borrarea succulanta, Guapira fragrens and Eugenia monticola are among the most common nativ e trees of St. Croixs dry forest and were also found in approximately 20% of samples among all age classes. Like C. indica they are medium sized trees that produce berry-like fruit containing animal dispersed seeds. Figure 3-3 presents species diversity tr ends across age classes for three plant groups. Sun-demanding exotic tree species like L. leucocephala A. macracantha and
59 the larger Albizia lebbeck generally had high frequencie s peaking in younger age classes and decreasing sharply in older forest stands. The shade-tolerant Triphasia tripholia invades mature forest and establishes in the understory. It has a patchy distribution but with extremel y high density where it occurs. Native trees generally tended to increase in frequency with stand age, like the most common species listed in Figure 3-3B. Tabebeuia heterophylla is a notable exception to the trend. It is an early colonizer, fairly long-lived, medium-sized tree with small, paperlike, and wind-dispersed seeds. These results indicate it may be intolerant of dense shade as it was rarely found in older stands. Several other woody species also demonstrated distinct trends, such as Capparis flexuosa a liana that can also grow as a low shrub and is so common in St. Croix that it is liste d separately. Other liana species are grouped into a single category. These and the shade tolerant native shrub Erythroxylum brevipes increased in frequency with increasing st and age (Figure 3-3C). Native, shade-intolerant plants such as Lantana camara and B. simaruba were found is small numbers in all age classes, but especially in young forest. Species rank order and relative frequency varied when sites were analyzed by age class. The most frequently encounter ed species in age classes 1-4 was the invasive exotic L. leucocephala which was found in 90% of segments in Age Class 1 (Figure 3-3). The second and third ranked spec ies in Age Class 1 (0-5 years) were the fast-growing, sun-demanding shrubs, L. camera and the euphorbia, Croton flavens. Two of the first native trees to colonize secondary forests are C. alba which ranked 4th and was found in 22% of Age Class 1 se gments and C. indica, ranked 9th (8%). In Age Class 2 (5-10 years) L. leucocephala still ranked 1, but its frequency decreased to 69%
60 (Figure 3-3). The 2nd and 3rd ranks were held by the lar ger and longer lived pioneer trees A. lebbeck (53%) and C. alba (24%). Age Class 3 (10-20 years) sites up to 20 years old still had L. leucocephala as the most frequent (65%) species, followed by the A. macracantha (34%). Longer lived, s hade-tolerant trees like B. succulanta (24%) and Swietenia mahogoni (18%) factored prominently in th ese stands. The three youngest forest classes shared other characteristics; canopies tended to have a single layer and were often patchy, the liana C. flexuosa was found in roughly 16% of the segments and four of the ten most frequently encountered species were exotic. Age Class 4 (20-50 y ears) stands still had L. leucocephala ranked 1 (62%), however, indicator species C. indica (48%), B. succulanta (43%) and E. monticola (34%) all factored prominently in the stands. The stands we re still somewhat patchy, but tended to have two distinct layers with shrubs L. camera (41%), Tecoma stans (31%) and Croton discolor (24%) dominating brightly lit por tions of the understory. Age Class 5 (>50 years) stands were the only sites where native C. indica (79%) ranked 1 instead of L. leucocephala Exotic species accounted for only one of the top ten ranked species, but the aggressive and shade-tolerant T. tripholia was found in 60% of Age Class 5 segments. The plant thrives in the forest understory and produces copious amount of sweet, juicy berries with bird dispersed seeds throughout the year. The percentages in this class are likely skewed somewhat due to a smaller sample size within the class due to the lim ited number of sites with 50 year old forest and known land-use histories. In addition to variation in species co mposition, age classes also differed significantly in species diversity (average nu mber of species enco untered per segment).
61 Results from the GLIMMIX procedure i ndicated a significant difference ( p < 0.001) in species diversity between age classes (Table 32). The trend was for species diversity to increase as stand age increased. Post hoc mean separ ation using the Bonferroni procedure for multiple tests of inequality indicated t hat age classes fall into two groups. Classes 1, 2 and 3 did not differ from each other and class 4 and 5 do not differ from each other (Figure 3-4). As expected, species richness generally increased with stand age but there are other factors influencing species diversity fo r age classes. Species diversity within Age Class 4 and 5 ranged fr om 5 to 25 species per segment. One 50 year-old transect at The Nature Conservancy was dominated by L. luecocephala and T. trifolia, with one or two individual native trees per segment. For some sites it was difficult to determine exact stand age because information from l and-owners seemed to co nflict with either field conditions, the aerial photographs or both. Stand age did not account for the high degree of variability in t he older age classes. In fact, there are several transects in age classes 4 and 5 with the same di versity as the youngest site s in the study. There are other mechanisms influencing successional pathways to cause the unexplained diversity results. Segments in age class 4 and 5 with less than six woody species in a 30 m2 and with all of those species considered invasive exotic or primary successionary species is a strong indication of seve rely hampered or arrested succession. Results from the GLIMMIX proc edure indicated no significant ( p < 0.05) difference between stand age for variables elevat ion and aspect (Table 3-3). For slope, Age Class 5 stands differed significantly from other sites while there was no detectable trend among the younger forest classes. Prev ious studies have found that forests,
62 especially older forests, tend to occur at higher elevations and on steeper slopes. The explanation for the trend is that steep slopes are low quality sites for farming or development and higher elevation translates to more difficult access. Finding that this trend does not apply in St. Croix is surprising, at first and one possible explanation is that site selection was not completely random and therefore not re presentative of all conditions. A more likely explanation relates to spatial scale. It is likely that St. Croixs small size makes all parts of the island acce ssible. In the land cover change analysis in the first chapter of this document, I tested whether or not land cover change tended to occur more frequently in areas located closer to roads, a phenomenon commonly demonstrated in continental systems over large areas. My resu lts gave evidence for the trend, but only at the greatest distances m easured, which accounted for only 5% of the island area. I concluded that increased dist ance from roads in such small increments did not translate to decreased human accessibi lity; therefore change remained relatively constant at almost all distances. Resu lts from the land cover change study also indicated that forest areas had a higher average elevation and slope than developed or grass areas. However, with the methods used in this study I was unable to find significant differences ( p < 0.05) between age classes wit hin the forest land cover. Discussion The goal of this study was to determine th e extent to which native forests are regenerating in St. Croi x, the influence exotic invas ive plants have in forest succession and for how long. Almost all of St. Croi xs forests are considered to be secondary forests (Little & Wadsworth, 1964) and results from Chapter 2 indicate that >8% of the island may be covered in forests less than 10 years old. Results from this study indicate that species diversity within young secondary forests of St. Croix is relatively
63 low and there is a high degree of dom inance by a single invasive plant, L. leucocephala. The canopies tend to be patchy, less than 5 m tall, and have no emergent trees. Older secondary stands have higher species diversity and an increased presence and diversity of native trees, however even after 50 years L. leucocephala is still the most common species. Four common native trees should be considered indicator species for successfully recovering native secondary forests; Capparis indica Bourreria succulanta Guapira fragrens and Eugenia monticola Abundance of C. flexuosa and other lianas was lowest in young stands and increased wit h stand age, and may also be an indicator of improving forest health. Studies in ra inforest tree gaps have found lianas to be positively correlated to pioneer tree abundance and frequently resulted in lianadominated gaps with stalled forest succession (Schni tzer et al., 2000). This is not the case in St. Croix. These indicator species also tolerate a wide range of conditions and are potential candidates for reintroduction when rehabilitating the most degraded secondary forest sites. The FIA found L. leucocephala to be the most abundant tree in St. Croix and assigned it the highest importance value of any tree species (Brandeis & Oswalt, 2007). (The Forest Service classifies the spec ies as native, although locally the species is considered to have been in troduced by the US Department of Agriculture during the 1950s.) Brandeis and Oswalt (2007) concluded L. leucocephala stands were distinct in that they were undeveloped, even-aged, largely monospecif ic with high stem density (4,235 stems/ha) and classified t hem as reversions, a cover type that was not found in the other U.S. Virgin Islands or in any of the United States. Thes e FIA findings support
64 my finding that the plant shows an unusual degree of dominance in St. Croix that alters the succession process. Relative to continental forest systems or large islands with diverse topography and life zones, St. Croix woul d naturally have lower spec ies diversity and comparisons between sites should be made with this in mind. In nearby Puerto Rico, a study of The Cartajena Lagoon Wildlife Refuge recorded species diversity in a secondary dry forest two years and 10 years after fi re and grazing were excluded from the area (Weaver & Chinea, 2003). The Cartejena Refuge (CR) has similar soil, rainfall, slope and elevation gradients to those found across St. Croix. The survey found a total of 103 tree species with L. leucocephala being the most common and showed a high degree of dominance by a small suite of exotic species on both dates. Of the 103 tree species identified, 43 combined to account for only 2.2% of the st ems. The 10 most common species in St. Croix (Table 3-1) occurred in a similar number of plots in CR, the notable differences being the common vine ( C. flexuosa) and the shrub ( L. camera ) were nearly absent in the Puerto Rico site. Despite these similarities, species diversity was far greater in CR, which had 23 common tree species (defined as being found in >10% of plots) compared to 14 common tree species in St. Croix. Dominance by a single species was also greater in St. Croix as L. leucocephala was found in 67% of plots and only 33% of plots in CR. It was not possible to compare L. leucocephala dominated stands with regenerating forest comprised of primarily native tree spec ies because none could be found. In a matter of decades St. Croix s eems to have transformed from having no L. leucocephala in forests to having little to no forest regeneration without it. The
65 ramifications of this on succession are difficult to determine. The young stands in this study are compared to >50 year old forests that began regenerating when L. leucocephala was not as prevalent on St Croix or may have been absent. By comparison, St. John of the US Virgin Islands has roughly 66% of it land area protected by US National Parks Service. An inventory of five secondary forest stands within the park boundaries, ranging in age from 35 to 125 years, found that the youngest stand was similar to those in St. Cr oix, with a long agric ultural history and a short, simple secondary forest structure dominated L. leucocephala (Ray & Brown, 1995a). The three sites over 100 years old also had long histories of intense agricultural use, but t heir abandonment may have predat ed the introduction of L. leucocephala. The shade intolerant legume was not found in any of >100 yr St. John sites, which could be the resu lt of either successional exclusion or that the plant was never present. Studies of forest succession on abandoned agricultural sites are many and have indicated that tropical fore sts are potentially resilient and can recover structure and species diversity quickly (see reviews by (B rown & Lugo, 1990; Guariguata & Ostertag, 2001; Chazdon et al., 2009). When the intensity and longevity of human land use is mild, forest recovery can be quite rapid (Ai de et al., 2000). Forest recover is further enhanced by close proximity to healthy fore st patches and other seed sources. Regardless of the speed of recovery, sec ondary forests bear the legacy of previous human land use in their uni que species compositions St. Croix likely represents the most extreme setting for testing the lessons learned from the previous studies. Historical land uses were intense as in sugar cane
66 production that frequently ut ilized fire for pest control. A fter productivity decreased land was maintained as pasture for centuries, in many cases. However, the greatest difference between the previo us studies and St. Croix (and the rest of the Lesser Antilles) is an issue of scale. St. Croix s small size made almost the entire island accessible to humans, resulting in a high percentage of forest removal (Little & Wadsworth, 1964). The island is also distant from co ntinental sources of genetic material and has a relatively simple topo graphy, causing a relatively low species diversity and higher degree of endemism. Finally, the introduction of invasive organisms seems to have further reduced the forests ability to regenerate. The cumulative result of these facts is sec ondary forests with a relatively low rate of recovery A state of arrested succe ssion has been suggested in other secondary forests dominated by exotic species, including tropi cal dry forests (Sarmiento, 1997; Chapman et al., 1999; Roth, 1999; Sandor et al., 2003; Kalacska et al., 2004). The results from this study indicate a higher degree of dominance from exotic species for a longer period of time than in other studies. In fact, study sites such as the >50 year old secondary forest stand sampled at The Nature Conserv ancy in Estate Little Princess posses forest structure and diversity indicative of a five to ten year old stand. Sites like this are common and require human inte rvention as natural succession does not occur under the present conditions. Conclusions During the entire field data collection proces s for this study not a single site was found where St. Croixs Caribbean dry fore st was regenerating without the heavy influence of exotic species. The current reality is that, young secondary forests are
67 completely dominated by exotic species, regar dless of location, previous land use or any biotic or abiotic factors. After 50 years without distur bance other aggressive exotic plant species are still present, but no longer dominate forest canopies in some sites. These results indicate that secondary fore sts of St. Croix tend to recover species diversity and canopy structure over time. Howeve r, multiple sites in this study exhibited strong signs of arrested succession, in cluding undeveloped canopies, low species diversity and dominance by exotic species. An additional confounding factor is that 6070 yr old secondary forests began regenerati on under conditions that did not include island-wide domination of Leucaena leucocephala and the this has had on successional pathways cannot be easily determined. Stand age accounted for only a portion of the variability in species diversity in this study. Land use history can play an important role in the rate of forest recovery, but quantifying severity of previ ous land-use is challenging, even with detailed information from land owner interviews. Attempts to correlate slope, aspect and elevation to stand age were inconclusive. The secondary forests arising from abandoned agricultural sites on St. Croix have novel species compositions where exotic species play a major role even half a century after abandonment. The most likely explanation is the small spatial scale of the island has caused the forests to reach some type cro ss some threshold of resilience from where they are no longer able to recover unassisted. St Croix is likely on the leading edge of this phenomenon as doz ens of small islands in the Lesser Antilles are transitioning out of agricultura l and into manufacturing and tourism-based economies that will likely lead to forest transitions under similar circumstance to St Croix.
68 Table 3-1. Tree species encountered during fi eld sampling in St. Croix with, botanical names, species code, species origin (N= native to St. Croix before European settlement, I= introduced) life form (T =tree, S=shrub, L=liana and P=palm) and frequency (expressed as a percentage of the number of segments where the species was observed over the total number of segments). Genus Species Code Origin Life Form Frequency Leucaena leucocephala leu leu I T 67.5% Capparis indica cap ind N T 27.0% Acacia macracantha aca mac I T 24.2% Capparis flexuosa cap fle N L 21.0% Lantana camera lan cam N S 20.0% Bourreria succulanta bou suc N T 19.0% Cordia alba cor alb N T 18.7% Albizia lebbeck alb leb I T 15.9% Guapira fragrens gua fra N T 15.5% Eugenia monticola eug mon N T 15.1% Swietenia mahagoni swe mah I T 14.7% Croton flavens cro fla N S 13.9% Tecoma stans tec sta I S 13.5% Triphasia tripholia tri tri I S 12.7% Erythroxylum brevipes ery bre N T 12.3% Eugenia rhombia eug rho N T 11.1% Randia aculeata ran acu N T 9.9% Eugenia biflora eug bif N T 9.1% Tabebeuia heterophylla tab het N T 8.7% Eugenia foetida eug foe N T 7.9% Genipa americana gen ame I T 7.9% Casearia sylvestris cas syl N T 7.5% Eugenia aeruginea eug aer N T 6.7% Croton discolor crot dis N T 5.6% Psidium guayava psi gua I T 5.6% Swietenia macrophylla swi mac I T 5.6% Faramae occidentalis far occ N T 5.2% Krugiodendron ferrum kru fer N T 5.2% Coccoloba uvifera coc uvi N T 4.7% Crossapetalum rhacoma cros rha N S 4.3% Andira inermis and ine N T 4.3% Capparis baduca cap bad N T 3.9% Coccothrinax alta coc alt N P 3.9%
69 Table 3-2. Description of the number of observations an d the mean species diversity (number of species obse rved per segment) with standa rd error for each of five age classes of secondary forests in St. Croix, U.S. Virgin Islands. Totals are given for the experimental desi gn and plant species. Significant differences (p<0.001) for species ric hness are indicated with unique letters. Age classes without differences ar e marked with the same letter. Age Class Unique sites (n) Transects SegmentsMean Species SE 1 9 12 60 3.15 b 1.63 2 3 9 45 3.82 b 1.89 3 7 13 65 4.17 b 2.43 4 8 11 55 8.17 a 2.56 5 3 5 25 10.79 a 4.62 Total 30 50 250 5.42 3.58 Table 3-3. Age ranges (yr) of five forest cover classes and the average elevation (m), aspect ( ), and slope ( ) and corresponding standard error of the means. Significant difference ( p <0.032) between age classes for the variable are indicated with *. Results are derived from 250 observations in 30 independent samples ( n =30). Age Class Age yrs Elevation (m) SE Aspect ( ) SE Slope ( ) SE 1 0-5 59.5 118 201.8 32.6 7.7 3.2 2 5-10 145 204.3 204 39.6 10.7 5.1 3 10-20 48.2 204.3 275.1 48.7 9.5 5.4 4 20-50 42 125.1 162.6 33.1 5 3.4 5 >50 113.4 104.5 210 36.1 21.1 3.8
70 A B Figure 3-1. A) St. Croix, US Virgin Islands and the distribution of 50 vegetation sampling transects in secondary forests from 5 age classes ranging from 3 to <50 years and B) an idealized pair of parallel veget ation sampling transects, divided into 10m segments 3 m 10 m 50 m St. Croix, US Virgin Islands
71 AB Figure 3-2. Recording canopy structure and co mposition data along 50m long transects in A) a three year old regenerating pastu re comprised primarily of the exotic legume Leucaena leucocephala in the thin overstory and the exotic grass, Panicum maximum below and B) sampling vegetation in a patchy, 25 year-old secondary forest with dow ned woody debris from Cordia alba growing on a site previously used for agriculture.
72 A Exotic Trees B Native Trees Figure 3-3. Frequency of groups of the most common plants in secondary forests of St Croix. A) Exotic trees and shrubs. Sun loving species such as L. Leucocephala demonstrate decreasing freq uency with increasing stand age, while share tolerant exotics such as T. trifolia tend to increase in frequency. B) Native trees show a trend of increasing frequency as stand age and species diversity increase. C) Frequen cy of lianas and shade tolerant shrubs ( E. brevipes ) increases with stand age while the pioneer shrub species L. camara decreases. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 12345Percent FrequencyAge Class aca mac alb leb cor alb leu leu tri tri 0% 10% 20% 30% 40% 50% 60% 70% 80% 12345Percent FrequencyAge Class bor suc cap ind eug mon gua fra tab het
73 C Other Plants Figure 3-3. Continued Figure 3-4. he mean number of species recorded per observation for each of 5 age classes of secondary forest in St. Croix, U.S. Virgin Islands. 0% 10% 20% 30% 40% 50% 12345Percent FrequencyAge Class bur sim cap fle lan cam lianas ery bre 0 5 10 15 20 25 012345Species per observationAge Class
74 CHAPTER 4 ENRICHING DEGRADED CARIBBEAN DRY FOREST WITH RARE NATIVE SPECIES Introduction Secondary forests are the most extensiv e forest cover type in the tropics, accounting for over 40% of total forest ar ea (Brown & Lugo, 1990). In the Caribbean, secondary dry forest is the fastest expandin g forest vegetation type. They have a simpler structure than intact native forests, with lower species diversity and greater degree of dominance by a few, of ten alien species. Secondary forests in the Caribbean tend to emerge on old fields that are characterized by soil degradation, more frequent fires, depleted soil seed banks, harsh climatic conditions and high competition with grasses and other pioneer species. The legacy of previous land-use influences the ecosystem function and struct ure for decades, centuries and longer (Foster et al., 2003). In cases where the duration and inte nsity of the disturbance is not severe, secondary forests in the Caribbean have prov en resilient and can regenerate naturally and recover much of their structure and ecologi cal function (Pascarella et al., 2000). It has been suggested that alien tr ee species function in the recovery process to stabilize local microclimate and facilitate natural r egeneration though recovery may take over 80 years (Vieira et al., 2003). Novel ecosyst ems are increasing recognized as the likely result when forest regenerates after la rge-scale anthropogenic disturbance (Hobbs et al., 2009). Lugo & Helmer (2004) have used the term new forests to label the unique structure and species com position. Novel ecosystem s tend to have native and introduced species living toget her under new environmental conditions and disturbance regimes and introduced tree species tend to dominate, especially during the first few decades of regeneration. These novel ecosystems offer unique conservation
75 opportunities but may also requite novel management and restoration approaches (Seastedt et al., 2008; Hobbs et al., 2009). These systems are well documented in the Caribbean, including Puerto Rico and the Virgin Isl ands (Lugo & Helmer, 2004; Brandeis et al., 2009). Abandoned pastures and agricultural sites present unique environmental conditions that favor the est ablishment of fast growing ex otic species. Lack of seed dispersal is a primary barrier to native forest regeneration (Holl et al., 2000), and is especially important on sites where frequent burning ha s eliminated the soil seed bank Even where nearby seed sources provide adequat e seed rain other barriers to seedling establishment may be present (Zimmerm an et al. 2000). Seed predation, low germination, and early seedling mo rtality are common in old fi elds and can significantly influence successional processes. Light and soil moisture also influence seedling survival and the response varies significantly among species (Augspurger, 1984; Loik & Holl, 1999; McLaren & McDonald, 2003). Seedling competition with grasses, agricultural weeds and invasive woody species can also inhibit successful estab lishment (Nepstad et al., 1991; Aide et al., 1995; Vieira et al., 2003). For example, in t he U.S. Virgin Islands, thirty five year old secondary forests adjacent to native stands were found to have sparse canopies dominated by thorny vines and a single exotic species, Leucaena leucocephala (tantan), accounted for 58% of the species im portance and 95% of t he understory seedlings (Ray & Brown, 1995a). The term disc limax forest has been used to describe the common condition of Antillean dry forest s with a canopy and understory dominated by two or three aggressive exotic species and no recruitment of native species in the under
76 story (Roth, 1999). Such fo rests do not develop increasingly complex assemblages and structures and species compos itions and are said to demons trate arrested succession. The causes can range from any combination of; depleted in situ seed banks, seeds arriving from outside sources being unable to establish and compete, and increased disturbance frequency that ultimately results in a steady state. Disclimax forests are distingu ishable from novel ecosystems by the rate at which they develop structure and species diversity. The novel systems or new forests are initially dominated by a few exotic species when human activity ceases on the site. The exotic species stabilize the site microclim ate and facilitate native species establishment in the understory. Mature forest species eventually replace pioneer species. After 40 years the forest structure and biomass are sim ilar to native forests, but exotic species play a more prominent role and reflect the l egacy of human use of the site. Neotropical forests can quickly recover their structure, ec ological function and di versity if there are nearby propagule sources and the extent and t he intensity of the land use was not too severe (Guariguata & Ostertag, 2001). When this is not the case and barriers impede succession, the result is a disclimax fore st. Arrested succession, when successional sequences are interrupted or even halt ed, has been documented in a number of systems (Lugo, 1997; Chapman et al., 1999) The cause is frequently the establishment of recalcitrant vegetation (including grasses and lianas) that prevents mature forest species from establishing (Sarmiento, 1997; Schnitzer et al., 2000). Arrested forests in the Caribbean are dominat ed by a few species of woody vegetation forming a single low, uniform canopy and have an understory comprised of the same canopy species (Ray & Brown, 1995a; Roth, 1999). Unlike the new forests, they are
77 not a refugia for native species, have low species diversity and pr ovide poor wildlife habitat. Disclimax forests with arre sted succession require some form of intervention to restart succession and restore native diversity. Species selection for restoration projects is an important c onsideration and there are several approaches. Exotic species can be justified in the most ex treme circumstances such as in mining restoration (Parrotta & Knowles, 1999), and plantations have been suggested as a way of stabilizing site conditions and promoting establishment of native species (Lugo, 1997). If the project goal is ecological re storation of native forest habitat, exotics species should only be considered a last resor t. Some foresters have tended to rely on the few commercial species for which data and seeds are readily available instead of looking toward native trees that can fill t he role and generating baseline data for these species. Rehabilitation in harsh condition s such as recently abandoned pastures with strong competition from exot ic grasses and full-sun condit ions, rapid growing pioneer species may be appropriate to initiate the succession process (Zimmerman et al., 2000). Dry forest restoration studies have found planting seedlings is more successful than direct seeding (Ray & Brown, 1995b; McLaren & McDonald, 2003; BonillaMoheno & Holl, 2009), that pioneer species general ly perform better than mature forest species (Holl et al., 2000; Hau & Corlett, 2003) and using many species is preferable to using few (Van der Putten et al., 2000). An additional intervention at a later successional stage can also be made to introduce a larger number and broader range of species to enhance diversity (Lamb et al., 2005). Establishing fast growing shrubs
78 has also been shown to increase tree seedling su rvival (Holl, 2002). Using fruit bearing trees and bird perches in restoration is considered promising, but a more complex management practice and difficult to eval uate (Wunderle, 1997; Galindo-Gonzalez et al., 1998; Toh et al., 1999; Shiels & Walk er, 2003). Unfortunately, there are no universally applicable rules and understanding the unique site c onditions is essential to selecting the appropriate restoration method and species for a given site. In this study I examined the efficacy of planting native tree seedlings in artificial gaps created in a disclimax forest canopy cu rrently dominated by invasive plants, primarily L. leucocephala The objective of the project was to determine the efficacy of the treatments and performance of the species selected. The goal of the research was to develop an effective method for establishing native plants in abandoned old fields to reinitiate forest succession and restore ecol ogical function and native diversity to this forest type. Methods Artificial gaps were created in the secondar y forest to give s eedlings a temporary competitive advantage over the surrounding canopy dur ing t he establishment phase. Mortality, seedling height and gap closure were measured over the course of two dry seasons. The relative performance of four native tree species and t he effect of a mulch treatment on growth and survival were analyzed. The covariates rainfall and gap canopy closure were also analyz ed for their effect on growth. The Site This study was conducted in St. Croix, U.S Virgin Islands on Estate Little Princess (ELP), a property owned and maintained by T he Nature Conservancy (lon. 64.0689 N, lat. 17.4050 W). ELP is a coastal site wit h gently sloping (0-10%) topography under 50
79 m elevation. The soil types on site are Si on clay and very stony Arawak gravely loam, which derive from volcanic and marine parent material and are molis ols typical to St. Croix (Davis, 2000). The soils are white in color with a pH. of >8.0 and are locally referred to as caliche and c onsidered poor for farming. EL P receives an average of 75 cm/yr rainfall, but is highly variable between years. Trade winds blow from the east all year, with increased intensity in the winter months. Daily temperature fluctuations approximate annual change. The Estate was purchased by the first gov ernor of St. Croix from the Danish West India Company in 1743 and had alread y been cleared for agriculture by the French. In 1750 a great house, windmill, sugarcane processi ng plant and as many as 36 row houses for enslaved African workers were built. Records indicate that most of the tropical forest cover was hand cleared for crops and remaining pockets of vegetation were periodically burned. The largest area of ELP was reserved for sugar cane production and was cultivated for over 200 years (Burgel & Scopinich, 2002). Cotton and tobacco were planted in the 1800s, but with shortlived success. Low productivity led to conversion to ca ttle pasture during most of the 20th century, until 1961 when formal agriculture ceased in 1955. Apar t from minor landsc ape plantings, the land has not been cultivated since that time. The 24 acre property is mostly fenced, but remains accessible along the eastern Long Reef Beach side. Prevailing trade winds from the east travel over open water for many kilometers before reaching ELP and th ere are no substantia l forest fragments within three kilometers, both limiting fact ors to secondary fore st succession. The closest weather station to the site is the Un iversity of the Virgin Islands Agriculture
80 Experiment Station, which m easured approximately 150 cm of rainfall per year during the time of this study. The agricultural land-use history of ELP is typical for St. Croix. After 54 years of abandonment ELP possesses the type of alien-dominated secondary forest that has become ubiquitous on old fields in St. Croi x a low, semi-open canopy with low species diversity, dominated by a few fast growing ex otic plant species. The dominant species is L. leucocephala and the forest type is locall y referred to as tan-tan bush. Before planting, data were collected wit hin ELP along three randomly distributed vegetation transects, measuring 50 m long by 2 m wide. Transects were sub-divided into 10 m segments to determine species frequen cy, composition and fo rest structure. The forest canopy was a patch y, semi-open single layer, 3-4 m in height, with low species diversity and dominated by a few mostly exotic species. L. leucocephala was found in high density in 13 of 15 segments. The weedy Cordia dentata (white manjack) and Acacia macracantha (stink casha) appeared as single plants with spreading crowns and were found in 13 and eight segments, respec tively. Dense thickets of the exotic, thorny bush Triphasia tripholia (sweet lime) grew in 8 segments, wherever the canopy was most open. Single indi viduals of the native trees, Bourreria succulanta (pigeonberry), Capparis indica (white caper) and Guapira fragrens (black mampoo) were found in several segments. These seven species accounted for all of the tree diversity in 14 of the 15 segments. These data are consistent with findings of other studies in the region (Ray and Brown, 1995) and are consistent with the disclimax condition described by Roth (1999).
81 Species Selection Previous work evaluated growth and surv ival of native tree species planted in full-sun in grassy field instead of using artifici al gaps. Research indicated that 30 cm tall seedlings of 12 native dry forest tree spec ies can be successfully established (90% survival) in regularly maintained plots without supplemental water (Daley and Zimmerman, 2002). In this ex periment, I selected species that could either quickly exploit the gap more rapidly than competitors, or subsist in the shaded canopy understory. I gave preference to rare or unde r-utilized tree species with these attributes (Table 4-1). Seeds Fruits of the four tree s pecies were collected at ma turity and weighed. Their seeds were removed, cleaned, counted, treated and sown in soil-free Promix in germination trays within 48 hours of colle ction. The four pregerminative seed treatments were; water (24 hour soak in distilled water), boil, (30 second dip in 100 C water), GA3 (one hour soak in 2,000 ppm Gibberellic acid solution), and control (Vozzo, 2002). There were 20 seeds per treatment with a minimum of four replications for each species. Trays were maintained in a screened shade-house, the potting media was kept moist and germination was recorded over a 90 day period. Time to germination of the first seed was recorded and germination was considered complete when there was no new germination in a replication for tw o weeks. Successful seedlings were transplanted to promix filled polybags when they produced their third true leaf and a slow-release, pelletized 20-20-20 fertilizer wa s mixed into the media. Seedling were grown in the shade house for between 3-6 months and then hardened off for one month in bright sunlight prior to planting. Germination results were inverse sine transformed
82 and a GLM was used to run an analysis of va riance (ANOVA) test for differences between treatment means. Planting A backhoe was used to clear a 120 m long by 3 m wide transect to facilitate creation of, and access to, 26 arti ficial gaps wit hin the dense secondar y forest (Figure 41). A total of 26, gaps were hand-clear ed with machetes and evenly distributed on either side of the transect. The gaps were 3 m in diameter, with gap centers spaced at least 5 m from each other and 5 m from the edge of the cleared transect. Gap centers were marked with stakes and native tree seedli ngs were planted in the four cardinal directions at a distance of 1.5 meters from center. Four plants we re randomly assigned to each of the gaps for a total of 104 (26x4) plants. Gaps were randomly assigned a treatment of mulch or no-mu lch. A 20 cm layer of st raw hay was applied around the tree seedlings in the mulched gaps. No supplem ental water or fertilizer was given to the seedlings in the field. Seedlings were planted in December 2006, coinciding with the end of the hurricane season and the peak of the seasonal ra ins. The data were collected 13 times over an 18 month period that included two complete dry seasons. Plant height was measured from the ground to the base of the highest le af on the planting day and approximately monthly therea fter. When dead plants were found the likely cause of death was noted. Canopy gaps were hand-cleared periodica lly as the surrounding vegetation encroached into the gaps. Canopy closure was measured on the same 13 dates using a spherical densiometer by facing each of the four cardinal directions. Canopy closure was calculated for each gap on each date from t hese four readings. Plant survival data
83 were analyzed using the SAS statistical package procedure GENMOD which handles binary data. Post hoc comparisons were made using Wald Chi-squares. Plant height data were analyzed using the PROC MIXED (SAS) which is a type of analysis of repeated measures that accounts for both ra ndom and fixed effects(Littell et al., 1996). Plant heights were logtransformed to ens ure normality of the data. Subject trees varied in their original size and the mix ed procedure accounts for this difference. Models accounting for original height diffe rence and use total height over time or absolute growth rate are found to provide the best prediction and clearest interpretation of results (MacFarlane & Kobe, 2006). When analyzing the factors affecting pl ant growth, the independent variables species and mulch treatment were considered discreet and rain, rain-lag and canopy closure were continuous covariates. Rain is the total rainfall in the 30 days prior to data collection and rain-lag 31-60 days prior. The gap effect was accounted for with a random statement in the m odel. Results are report ed as significant when p <0.05. Standard deviation and standard error of the means are provided as appropriate and noted. Results Seeds Fruit and seed data, most effective pregerminative treatment and germination rate are listed in Table 4-2. All of the data provided in t he table are based on averages of multiple seed collec tions and at least four seed germination replications. At least one pregerminative treatment initiate d germination in three weeks or less for all four species. At least one treatment significantly ( p <0.05) increased the germi nation rate relative to control for C. rickseckeri and G. officinale but not for S. polyphylla and S. conocarpum
84 Fruit of C. rickseckeri contains a single, hard, woody stone with one to th ree embryos. Removing the embryos is destructive, so t he germination rate provided is per stone and is above 100%. The thin pods of S. polyphylla contain highly variable quantities of paper-like seeds that germinate at low rates, but large numbers are easily collected. The 30-second hot water treatment significantly ( p <0.05) decreased germination rates for all species except for C. rickseckeri Seedling Survival. Of the 104 seedlings planted in the 26 gaps a total of 94 (90%) survived at least 18 months. Four of the plants desiccated soon after planting or during the dry season and six were severely browsed by animals, most likely by an introduced deer species ( Odocoileus virginianus) Four of the six browsed plants were S. conocarpum Survivorship data are illustrated in Figure 42. Significant differences were found between species ( p <0.05). A post-hoc analysis using Chi squares found S. conocarpum had lower average survivorship than other species, but did not include C. rickseckeri because there was 100% survival for that species. There were significant differences ( p =0.015) between plants in mulched and un-mulched gaps because eight of the ten dead plants were un-mulched. Percent canopy closure did not have a significant effect on survivorship. Intera ctions between treatm ents effects (mulch and percent canopy closure) were also non-significant. Seedling Growth Figure 4-3 illustrates the overall growth patte rn of all four species relative to the mulch treatment, canopy closure and rainfall. For six of the 13 months when measurements were taken, average height change by species was under 1 cm or an actual decrease in height due to drought-induced wilting or die-back. The canopy
85 closure measure follows the rainfall distribut ion pattern closely. During drought the canopy experienced leaf loss and during periods of higher rainfall there was new growth which closed the canopy gaps somewhat. As expected, plant growth patterns differed between species, but the gener al trend across species was for increases during high rainfall and flattened curves during drought. Results of the mixed analysis procedure fi xed effects on plant height over time (Table 4-3) indicated highly significant differences (p <0.0001) between species mean heights over time (month*species). The m onth*mulch treatment was also significant, but less so ( p <0.034). Mulch and the interacti on between mulch and species were nonsignificant. Mean separat ion between species using an adjusted Tukeys test places the species in three groups. Figure 43 indicates significant differences between species height over the length of the experiment. All three continuous covariates (rain, ra in-lag, canopy closure) had a significant ( p <0.0001) effect on seedling he ight (Table 4-3). Predicted values of LSmeans of logheight were calculated for each species and plotted against the range of rainfall values observed during the experiment. As expec ted, rain and rain-lag both had positive effects on height. The canopy closure effe ct was negative and also highly significant ( p <0.0001) on height when adjusted for rainfall. Canopy effect varied by species, but was negative for three species and had no effect on S. conocarpum Discussion As other island in the Lesser Antilles follow s imilar economic patterns that lead to Forest Transition the patterns of secondary fore st succession in St. Croix are likely to be repeated. Puerto Rico, St. Croixs larger and more populac e neighbor, has undergone a dramatic transformation from an agricultural-based economy wit h as little at 6% forest
86 cover in 1940, to an urban industrial economy wit h over 40% forest cover today (Birdsey & Weaver, 1987; Helmer et al., 2002; Grau et al., 2003). This land cover change resulted from a combination of Puerto Ric os unique relationship wit h the United States and internal policies that lead to rapid reforestation (Koop & Tole, 1997), although reforestation was not a stated policy goal and the reforestation was unmanaged (Martinuzzi et al., 2007). St. Croix shares a similar economic relationship with the United States that has contributed to r apid post-agricultural changes including industrialization and subsequent increases in tourism, land values, GDP, and secondary forest cover. How the process has differ ed on the smaller island has not yet been studied, but preliminary work indicates that secondary forests are the most important cover type in St. Croix, but that the island has a larger percentage of land in fluctuation than Puerto Rico. That is, land is still being cleared in some places while secondary forests regenerate in others. The fruit seed and shade house germination data here is similar to findings in previous work with G. officinale (Daley & Zimmerman, 2008) Seeds for this same species germinated at a rate of 5-15% w hen planted directly in the field in an experiment in St. John, U.S. Virgin Islands (Ray & Brown, 1995b). This may be the first published seed data on C. rickseckeri and S. conocarpum Greenhouse germination rates were high for all species except for S. polyphylla Average S. polyphylla seed weight is similar to that reported for Puerto Rico t hough the germination rate is approximately half (Francis, 1993). No s eeds from any species showed any sign of dormancy.
87 Availability of soil moisture is well known to affect the survival rate of seedlings. While I did not measure soil moisture directly, mulched plants had significantly higher survival than non-mulched plants, and I recomm end using any type of mulch possible. The disadvantage, of course, is the incr eased labor, but there are several benefits including; reduced competition with weeds, improved soil moisture and ease in identifying planted areas in the field. The dry season began earlier than expect ed the winter the seedlings were planted. There were two small rain events after planting followed by four months of drought (Figure 4-2). I antici pated a higher mortality rate than observed due to drought, but attribute the success to selecting highly drought-tolerant native plants. I recommend establishing field plants in October in the Virg in Islands to minimize the risk. Shade was not a significant factor on seedl ing survival in this study. Rainfall is often the single greatest factor in determini ng survival and growth in dry forest seedlings. Although growth rate s varied significantly between species, the importance of rainfall did not. Predicted values for all four species had the same slope across rainfall gradients for both rain and rain-lag variables. Mulch effect was significant on height, but only temporarily as the mulch broke down over the course of the experiment. Droughts (Dec-April and May-Aug) were th e periods of lowest average height increase and highest mortality. The mulch treatment likely alleviated some of this stress through reduced competition from weeds and from increased soil moisture. Supplemental watering would be impractical in establishing field plants and survivorship in this project was high enough to conclude it unnecessary.
88 Shade can also significantly affect t he survival and growth range of planted seedlings, but shade does not affect species equally (Augspurger, 1984; Hooper et al., 2002). Field studies in the Caribbean have concluded that planting seedlings is the most effective method of establishing plan ts and that varying amounts of shade have increased growth and survival relative to full sun (Ray & Brown, 1995b; McLaren & McDonald, 2003). In this study the artifici ally created gaps created varying degrees of heavy shade with canopy closure values for gaps ranging from 55 98%. Since canopy closure had a negative effect of growth it is possible that larger gaps may create conditions suitable for more sun-demandi ng species. A disadvantage to the gap method is that the canopy thins during t he dry season and exposes the seedlings to increased sun when they are the driest. During periods of high rainfall when the seedlings have moisture available, gaps t end to close due to growth of the surrounding canopy. Four rare endemic species that would not likely have established on their own were selected for planting. Two of the pl anted species were multi-stemmed shrubs or small trees ( S. polyphylla and S. conocarpum) and two were upright trees ( C. rickseckeri and G. officinale ). The two shrubby trees we re the best performers as measured by change in height. This measure is a conservative indicato r of their relative productivity compared to the single-stemmed species because these plants were multistemmed and grew in a spreading V-shaped canopy. Both S. polyphylla and S. conocarpum rapidly filled the space created by the gap. G. officinale also has a somewhat spreading canopy architecture, t hough less so. This species achieved the lowest overall height, but plants appeared in very good health and grew consistently in
89 rain, drought and shade. Individuals of this dense hardwood species are known to live for hundreds of years (Little & Wadswort h, 1964) and begin producing flowers after one year and viable seeds after two year s (Daley & Zimmerman, 2004). The most surprising result for the height variable was the performance of C. rickseckeri This species was selected primarily due to its rapid height increase in a preliminary study. In a full-sun, regularly mowed planting, 50 cm C. rickseckeri saplings grew to an average of over 400 cm in a 30 month period without any supplemental watering (Daley and Zimmerman, 2004). That sa me growth rate would have resulted in a 200 cm average height increase by the end of this study instead of the average height of 20 cm average increase recorded here. The species has not previously been described as a pioneer, but appears to require full sun for its char acteristic rapid growth. Conclusion Disclimax forests experiencing stalled or arrested succession require intervention to restart the natural regeneration process. The secondary forest used in this project was typical of old-field succession in St. Croix, U.S. Virgin Islands. Fifty four years after agricultural abandonment the canopy remained dominated by three pioneer species ( L. leucocephala A. macracantha and C. alba ). The seasonally dry secondary forest presented multiple barriers to native seedling establishment and the goal of the project was to test a method of establishing rare native plants under these conditions. Th is project successfully established four endemic species in artificial gaps in a disclim ax forest dominated by exotic species. Seedlings were established in the wet s eason and required no supplemental water. Seedling survivorship was improv ed 15% by using hay mulch.
90 Table 4-1. Botanical and co mmon names of the 4 native tree species grown for this study and a brief description. Family, Genus species (common) Description Boraginacaea Cordia rickseckeri Millsp. (orange manjack) Fast growing pioneer species with animal dispersed seeds and an upright, single-stemmed habit to 12-15 m tall. Highly drought tolerant and endemic to the dry forest of Puerto Rico and St. Croix. Zygophyllacaea G uaiacum officinale Linn. (lignum vitae) Relatively slow growing, long-lived, hard wood tree with animal dispersed seeds and an upright but spreading habit. It is high ly drought tolerant and grows well in both the understory and full sun. This plant appears on the U.S. Virgin Islands Endangered Species List Fabacaea Senna polyphylla Jacq. (desert cassia) Medium growth rate, shade-intolerant, multistemmed shrub or small tree (3 m) has small seeds reported to be disper sed by cattle. Native to lowland dry forest of Hispaniola, Puerto Rico and the Virgin Islands, it is both salt and drought tolerant. Solanacaea Solanum conocarpum Dunal (purple maron bacuba) Medium growth rate, mult i-stemmed bush or small tree (3 m), seeds are likely animal dispersed. Known only in the U.S. Virgin Islands, only 200 plants are believed to remain in the wild. The petition for U.S. endanger ed species status was rejected in 2006 due to lack of information. Little is known about the ecology of the species.
91 Table 4-2. Fruit and seed data summary for four native species. Standard error of the means is provided. Seed germination treatments tested were; 24 hour water soak, 1 hr soak in 2,000 ppm Gibarellic acid (Ga3), hot water and control. Cordia rickseckeri Guaiacum officinale Senna polyphylla Solanum conocarpum Fruit weight (g) 5.19 0.49 0.72 0.054 0.275 .028 12.02 2.4 Fruit/kg 193 18 1,389 104 3,672 418 86 7 Seed/fruit 1.00 0.01 1.68 0.49 12.93 4 81.3 10.7 Seed weight (g) 1.89 0.20 0.31 0.03 .049 .004 .0496 .007 Seed/kg 530 60 3,219 251 22,350 1,960 20,550 2,950 Germination begins (days) 15 15 18 8 Germination ends (days) 30 45 40 18 Most effective treatment Water Ga3 Control Control Germination rate 140%* 71% 27% 95% [* Multiple embryos are contained in a single stone for this species. Germination rate is given as the number of seedlings per stone.] Table 4-3. Results from PROCMIXED analysis of fixed effects on pl ant height over time indicate significant differences between species. Results also indicate the effect of the mulch treatment and the interaction had no significant effect on height. Month and the interactions with month indicate that height increases were not constant over time. The e ffects of the three covariates (rain, rain_lag and canopy closure) were highly significant. Effect Num DF Den DF F -Value Pr > F Species 3 193 5.54 0.0011 mulch 1 32.3 0.46 0.503 mulch*species 3 157 2.54 0.0583 month 1 244 262.64 <.0001 month*species 3 231 8.28 <.0001 month*mulch 1 233 4.55 0.034 rain 1 706 156.44 <.0001 rain_lag 1 919 299.07 <.0001 canopy 1 691 340.37 <.0001
92 A B Figure 4-1. Field data co llection at the planting site. A) 3m by 100m transect is cut in a dense, low secondary forest. B) student research assistant collects seedling height data in an artificial gap A B Figure 4-2. east Square Means of percent surv ival after 18 months for four tree species and for mulched and unmulched plants A) mean survivorship by species appears on the left and B) averages by mulch treatment are on the right. Error is expressed as SE. Signific ant differences are noted with different letters. A A B 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% C.ric n=28 G.off n=29 S.pol n=27 S.con n=20Percent survivalSpecies B A 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% No mulchMulchPercent survivalTreatment No mulch Mulch n =52
93 Figure 4-3. Growth patterns of four trees species plotte d along with monthly rainfall (cm). Canopy closure is plotted on a s eparate Y-axis on t he right. Species averages for the mulch treatment (+) are shown with dashed lines and averages for no-mulch treatment (-) have solid lines. Figure 4-4. Final heights by species afte r 18 months, with and witho ut mulch treatment, expressed as a percentage of orig inal height. Error terms are SE. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 20 40 60 80 100 120Dec 06 Jan 07 Feb 07 Mar 07 Apr 07 May 07 Jun 07 Jul 07 Aug 07 Sep 07 Oct 07 Nov 07 Dec 07 Jan 08 Feb 08 Mar 08 Apr 08 May 08 Jun 08 Rainfall/Plant Height (cm)Month Year rain Cric Crick + Goff Goff+ Spol Spol+ Scon Scon+ Closure Canopy Closure 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% C.rickG.offS.polS.conPercent height increase from originalSpecies No Mulch Mulch
94 CHAPTER 5 CONCLUSION The results of this research provide t he reader with detailed, spatially explicit descriptions of land-cover distribution on St. Croix for two distinct points in time (1992, 2002). A second study resulted in species-spec ific des criptions of the composition and structure of secondary forest s of St. Croix with a focu s on the influence of exotic invasive plants in succession. A final study tested rehabilitation methods for forests experiencing arrested succession through the introduction of ra re native tree species. Together the projects identify landscape level trends, describ e stand level problems with the succession process in highly impacted sites on the island and test methods or addressing the problem by stimulating succession. St. Croi x is used a model system to describe Forest Transition problems on small Caribbean island such as those in the Lesser Antilles. I project that the forest transition problems currently being experienced in St. Croix will be repeated as other small islands in the Windward and Leeward chains go through the same economic and social changes. More specifically the change analysis de scribes a forest dominated landscape (56%) experiencing a high degree of change. There was no net change in forest area, but there was substantial change within the forest cl ass and a strong trend toward younger secondary forest and increased forest fragmentation. Approximately 1,500 ha of forest were lost, with the majority c onverted to development and 1,500 ha of forest were gained, the majo rity coming from regener ating grassland. The largest transition between classes result ed in over 5% of the island changing from grassland to development during the 10 year time span. The loss of grassland was corroborated by independent non-spatial data collected by the Virgin Islands
95 Department of Agriculture. These findings indi cate St. Croix is still experiencing Forest Transition and economic realit ies favor increased urbaniza tion at the expense of grassland while net area of forest remained stabl e. Tax incentives likely exert a strong influence on the process by rewarding low-qua lity, low-production pasture. Secondary forest growth would likely increase sharply if these agricultural tax incentives were removed or alternative incentives for protection of secondary forests were supported. I tested the hypothesis that changes in forest cover decrease with increased distance to roads and concluded that forest change area only varies at the greatest distance values. The pattern is not as pronounced as in other studies and occurs over a smaller spatial scale. Also, patches of both forest gain and forest loss are spatially (patch size, location and distribution) sim ilar and are two parts of the same process where small patches of land are irregularly cleared and then left fallow. The process is driven by macroeconomic changes (increas ing worker wages, population density and real estate prices) and re-enforced by the property tax exemption for land that is registered as agriculturally productive in name only. During the entire field data collection proces s for this study not a single site was found where St. Croixs dry forest was r egenerating without the heavy influence of exotic species. The current reality is that, young secondary forests are completely dominated by exotic species, r egardless of location, previous land use or any biotic or abiotic factors. After 50 y ears without disturbance other aggressive exotic plant species are still present, but no longer dominate forest canopies in some sites. Other sites are totally dominated by only two or three in vasive exotics and show no signs or successional progression. These results indi cate that secondary forests of St. Croix
96 tend to recover species diversity and canopy st ructure over time. However, multiple sites in this study exhibited strong signs of arrested succession, including undeveloped canopies, low species diversity and dominance by exotic species. An additional confounding factor is that 60-70 yr old secondary forests began regeneration under conditions that did not incl ude island-wide domination of Leucaena leucocephala and the this has had on successional pathways cannot be easily determined. Stand age accounted for only a portion of the va riability in species diversity in this study. Land use history can play an important role in the rate of forest recovery, but quantifying severity of previ ous land-use is challenging, even with detailed information from land owner interviews. Attempts to correlate slope, aspect and elevation to stand age were inconclusive. The secondary forests arising from abandoned agricultural sites on St. Croix are novel ecosystems with uniq ue species compositio ns where exotic species play a major role even half a century after abandonment. The most likely explanation is the small spatial scale of t he island has caused the forests to reach some type cross some threshold of resilience from where they are no longer able to recover unassisted. St Croix is likely on the l eading edge of this phenom enon as dozens of small islands in the Lesser Antilles are transitioning out of agricultural and into manufacturing and tourism-based economies that will likely lead to forest transitions under similar circumstance to St Croix. Secondary forest suffereing arrested succe ssion require intervent ion to restart the natural regeneration process. The field site in the third experiment in this project was typical of old-field succession in St. Croix, U. S. Virgin Islands. Fi fty four years after
97 agricultural abandonment the canopy remained dominated by three pioneer species ( L. leucocephala A. macracantha and C. alba ). The seasonally dry secondary forest present ed multiple barriers to native seedling establishment and the goal of t he project was to test a met hod of establishing rare native plants under these conditions. This project successfully established four endemic species in artificial gaps in a disclim ax forest dominated by exotic species. Seedlings were established in the wet s eason and required no supplemental water. Seedling survivorship was improved 15% by using hay mulch. Compared to other methods such as direct seeding or large scale clearing and planting, this gap-planting method offers the opportunity to successfully establish native plants across large areas at relatively low cost. Such a method is r equired in St. Croix today as an estimated 150 ha of new secondary forests begin regenerat ing on abandoned agriculture every year.
98 LIST OF REFERENCES Adam, M., and C. Ryan 2003. Ev aluation of Estate Little Princess Forest Ecosystem: assessment of restoratioon needs The Nature Consewrvancy. Aide, T. M., J. K. Zimmerman, L. Herrera M. Rosario, and M. Serrano 1995. Forest recovery in abandoned tropical pastures in Puerto-Rico. Forest Ecology and Management 77: 77-86. Aide, T. M., J. K. Zimmerman, J. B. Pa scarella, L. Rivera, and H. Marcano-Vega 2000. Forest regeneration in a chronose quence of tropical abandoned pastures: Implications for restoration ecology. Restoration Ecology 8: 328-338. Angelsen, A. 2007. Forest cover change in space and time: Combining the Von Thunen and Forest Transition theories. Asner, G., D. Knapp, E. Broadben t, P. Oliveira, M. Keller, and J. Silve 2005. Selective Logging in the Brazilia n Amazon. Science 310: 480-482. Augspurger, C. 1984. Light r equirements of neotropical tr ee seedlings. A comparative study of growth and survival. Journal of Ecology 72: 777-795. Barnes, B., D. Zak, S. D enton, and S. Spurr 1998. Page s 774. Forest Ecology. John Wiley & Sons, Inc., New York, USA. Bernard, R. I. 2002. Early effo rts for forest restoration in Puerto Rico: Lessons from the past and management tools for the future. In ternational Instit ute of Tropical Forestry, internal report : 9 pages. Birdsey, R. A., and L. P. Weaver. 1987. Fo rest area trends in Puerto Rico. Southern Forest Experiment Station, Ashville, North Carolina. Bonilla-Moheno, M., and K. Holl 2009. Direct seeding to restor e tropical mature-forest species in areas of slash-and-burn agriculture. Restoration Ecology : 1-8. Brandeis, J. T., E. Helmer, H. Marcano-Ve ga, and E. Lugo Ariel 2009. Climate shapes the novel plant communities that form af ter deforestation in Puerto Rico and the U.S. Virgin Islands. Forest Ecology and Management 258: 1704-1718. Brandeis, T., and S. Oswalt. 2007. The Status of the U.S. Virgin Island's forests, 2004. United States Department of Agricultur e Forest Service, Southern Research Station, Ashville, NC. Brown, S., and A. Lugo 1990. Tropi cal secondary forests. Journal of Tropical Ecology 6: 1-32.
99 Burgel, J., and K. Scopinich. 2002. The Natu re Conservancy, Estate Little Princess Master Plan. The Nature Conservancy, Christiansted, St. Croix U.S. Virgin Islands. Chapman, C. A., L. J. Chapm an, L. Kaufman, and A. E. Zanne 1999. Potential causes of arrested succession in Kibale National Park, Uganda: growth and mortality of seedlings. African Journal of Ecology 37: 81-92. Chavez, J., P.S., 1996. Image based atmospheric corrections revisited and improved. Photogrammetry Engineering and Remote Sensing 62: 1025-1036. Chazdon, R. L., B. Finegan, R. Capers, B. Salgado-Negret, F. Casanoves, V. Boukili, and N. Norden 2010. Composit ion and dynamics of functional groups of trees during tropical forest succession in northeatern Costa Ric a. Biotropica 42: 31-40. Chazdon, R. L., C. Peres, D. Dent, D. Sheil, A. E. Lugo, D. Lamb, N. Stork, and S. Miller 2009. The potential for species conservation in tropical secondary forests. Conservation Biology 23: 1406-1417. Daley, B. F., and T. W. Zimmerman 2004. Growth and phenology of outplanted tree seedlings. Caribbean Urban and Community Forestry, St. John, U.S. Virgin Islands, May, 2004. 2004. University of the Virgin Islands, [insert place of publication]. Daley, B. F., and T. Z. Zimmerman 2008. Ge rminating five tree species native to the U.S. Virgin Islands. Tree Planters Notes 52: 10-15. Davis, J. R. 2000. Soil Survey of the Unit ed States Virgin Islands. United States Department of Agriculture. Erdas, I. 2003. Leica Geosyst em GIS & Mapping LLC. Ewel, J. J., and L. J. Whitmo re 1973. The Ecological Life Zones of Puerto Rico and the U.S. Virgin Islands. Forest Service Research Paper. Fao, F. A. A. O. 2006. Global Forest Re sources Assessment, 2005. United Nations Food and Agriculture Organization, Rome, Italy. Fao. 2006. Global Forest Resources Assessment, 2005. United Nations Food and Agriculture Organization, Rome. Finegan, B. 1984. Forest succession. Nature 312: 109-114. Finegan, B. 1996. Pattern and process in neo tropical secondary forests: The first 100 years of succession. Trends in Ecology & Evolution 11: 119-124. Foody, G. M. 2002. Status of land cover cl assification accuracy assessment. Remote Sensing of Environment 80: 185-201.
100 Foster, A., and M. Rosenzweig 2003. Economic growth and the rise of forests. The Quarterly Journal of Economics 118: 601-637. Foster, D., F. Swanson, J. A ber, I. Burke, N. Brokaw, D. Tilman, and A. K. Knapp 2003. The importance of land-use legacies to ecology and conservation. BioScience 53: 77-88. Francis, K. J. 1993. Seeds of Puerto Ric an trees and shrubs: a second installment. Pages 5 in U. S. F. Service, editor. Southern Forest Experiment Station, New Orleans, LA. Francis, K. J., and A. J. Parrotta 2006. Veget ation response to grazing and plating of Leucaena leucocephala in a Urochloa maximum -dominated grassland in Puerto Rico. Caribbean Journal of Science 42: 67-74. Galindo-Gonzalez, J., S. Guevara, and V. J. Sosa 1998. Bat and bird-generated seed rains at isolated trees in pastures in tr opical rainforest. Conservation Biology 14: 1693-1703. Grau, H. R., M. Aide, J. K. Zimmerman, J. R. Thomlinso n, E. Helmer, and X. Zou 2003. The ecological consequences of socioeconomic and land-use changes in postagriculture Puerto Rico. BioScience 53: 1159-1168. Guariguata, R. M., and R. Ostertag 2001. Neotropical secondary forest succession: changes in structural and functional characteristics. Forest Ecology and Management 148: 185-206. Hau, B. C. H., and R. T. Corl ett 2003. Factors affecting the early survival and growth of native tree seedlings planted on a degraded hillside grassland in Hong Kong, China. Restoration Ecology 11: 483-488. Helmer, E., T. Kennaway, D. Pedreros, M. Clark, H. Marcano-Vega, L. Tieszen, T. Ruzycki, S. R. Schill, and S. Carrington 2008. Land cover and forest formation distributions for Sr. Ketts Nevis, St. Eustatius, Grenada and Barbados from decision tree classification of cloud-clear ed satellite imagery. Caribbean Journal of Science 44: 175-198. Helmer, E., O. Ramos, T. D. M. Lopez, M. Quinones, and W. Diaz 2002. Mapping the forest type and land cover of Puerto Rico, a component of the Caribbean biodiversity hotspot. Caribbean Journal of Science 38: 165-183. Helmer, E. H. 2004. Forest conservati on and land development in Puerto Rico. Landscape Ecology 19: 29-40. Hobbs, R. J., E. Higgs, and J. A. Harris 200 8. Novel ecosystems: implications for conservation and restoration. Trends in Ecology & Evolution 24: 599-605.
101 Hobbs, R. J., E. Higgs, and J. A. Harris 200 9. Novel ecosystems: implications for conservation and restoration. Trends in Ecology & Evolution 24: 599-605. Holl, K. D. 2002. Effect of shrubs on tree seedling establishment in an abandoned tropical pasture. Journal of Ecology 90: 179-187. Holl, K. D., M. E. Loik, E. H. V. Lin, and I. A. Samuels 2000. Tropical montane forest restoration in Costa Rica: Overcoming barriers to dispersal and establishment. Restoration Ecology 8: 339-349. Hooper, E., R. Condit, and P. Legendre 2002. Re sponse of 20 native tree species to reforestation strategies for abandoned farmland in Panama. Ecological Applications 12: 1626-1641. Jackson, S. T., and R. J. Hobbs 2009. Ecologic al restoration in the light of ecological history. Science 325: 567-569. Jensen, J. R. 1996. Introcuc tory Digital Image Processing. Prentice-Hall, Englewood Cliffs, New Jersey. Jensen, J. R. 2005. Introductory digita l image processing; A remote sensing perspective. Pearson Prentice Ha ll, Upper Saddle River, NJ. Kalacska, M., G. A. Sanchez-Azofeifa, J. C. Calvo-Alvarado, M. Quesada, B. Rivard, and H. D. Janzen 2004. Species compositio n, similarity and diversity in three successional stages of a seasonally dry tropical forest. Forest Ecology and Management 200: 227-247. Kenneway, T., E. Helmer, M. Lefsky, J. T. Brandeis, and K. Sherrill 2008. Mapping land cover and estimating forest structure using satellite imagery and course resolution lidar in the Virgin Islands. Journal of Applied Remote Sensing 2: 1-27. Koop, G., and L. Tole 1997. Measuring differ ential forest outcomes: A tale of two countries. World Development 25: 2043-2056. Kozlowski, T. T. 2002. Physiol ogical ecology of natural regeneration of harvested and disturbed forest stands: implications for forest management. Forest Ecology and Management 158: 195-221. Lamb, D., D. P. Erskine, and A. J. Parrotta 2005. Rest oration of Degraded Tropical Forest Landscapes. Science 310: 1628-1632. Lambin, E. F. 1999. Monitoring forest degradation in tropical r egions by remote sensing: Some methodological issues. Global Ecology and Biogeography 8: 191-198.
102 Lambin, F. E., B. L. Turner, H. Geist, B. S. Agbola, A. A ngelson, W. J. Bruce, T. O. Coomes, R. Dirzo, G. Fischer, C. Folke, P. S. George, K. Homewood, J. Imbernon, R. Leemans, X. Li, F. E. Moran, M. Mortimore, P. S. Ramakrishnan, F. J. Richards, H. Skanes, W. Steffen, D. G. Stone, U. Svedin, A. T. Veldkamp, C. Vogel, and J. Xu 2001. The cause of land-use and land-cover change: moving beyond the myths. Global Environmental Change 11: 261-269. Laurence, W., F, A. K. Albernez, G. Schrot h, P. M. Fearnside, S. Bergen, E. Venticinque, M, and C. Da Costa 2002. Predictors of deforestation in the Brazilian Amazon. Journal of Biogeography 29: 737-748. Littell, C. R., A. G. Milliken, W. W. Str oup, and D. R. Wolfinger. 1996. SAS System for Mixed Models. SAS Institute In c, Carey, North Carolina. Little, E. W., and F. H. Wads worth. 1964. Common Trees of Puerto Rico and the Virgin Islands. U.S. Department of Agriculture, Forest Service, Washington, D.C. Loik, M. E., and K. D. Holl 1999. Photosynth etic responses to light for rainforest seedlings planted in abandoned pasture, Costa Rica. Restoration Ecology 7: 382391. Lomascolo, T., and T. M. Aide 2001. Seed and seedling bank dynamics in secondary forests following hurricane georges in Puer to Rico. Caribbean Journal of Science 37: 259-270. Lugo, A., and E. Helmer 2004. Emerging fo rests on abandoned land: Puerto Rico's new forests. Forest Ecology and Management 190: 145-161. Lugo, A. E. 1992. More on exotic species. Conservation Biology 6: 6. Lugo, A. E. 1997. The apparent paradox of r eestablishing species richness on degraded lands with tree monocultures. Forest Ecology and Management 99: 9-19. Lugo, A. E. 2004. The outcome of alien tree invasions in Puerto Rico. Frontiers in Ecology and the Environment 5: 265-273. Macfarlane, W. D., and K. R. Kobe 2006. Se lecting models for capturing tree-size effects on growth-resource relationships Canadian Journal of Forest Resources 36: 1695-1704. Marin-Spiotta, E., R. Ostertag, and W. L. Silver 2007. Long-Term patterns in tropical reforestation: Plant community composition and aboveground biomass accumulation. Ecological Applications 17: 828-839. Martinuzzi, S., A. W. Goul d, and R. M. O. Gonzalez 2007. Land development, land use, and urban sprawl in Puerto Rico integr ating remote sensing and population census data. Landscape and Urban Planning 79: 288-297.
103 Mather, A. S. 1992. The Forest Transition. Area 24: 367-379. Mather, A. S., and C. L. Needl e 1998. The Forest Transition: a theoretical basis. Area 30: 117-124. Mclaren, K. P., and M. A. Mcdonald 2003. The effects of moisture and shade on seed germination and seedling survival in a tropi cal dry forest in Jamaica. Forest Ecology and Management 183: 61-75. Miles, L., A. C. Newton, R. S. Defries, C. Ravilious, I. May, S. Blyth, V. Kapos, and J. Gordon 2006. A global overview of the conservation status of tropical dry forests. Journal of Biogeography 33: 491-505. Murphy, P. G., and A. E. Lugo 1986. Ecology of tropical dry forest. Annual Review of Ecology and Systematics 17: 67-88. Nagendra, H., J. Southworth, and C. Tucke r 2003. Accessibility as a determinant of landscape transformation in western Hondur as: links in pattern and process. Landscape Ecology 18: 141-158. National Agricultural Statistics Service 2005. 2002 Census of Agriculture, Virgin Islands of the United States. Pages 42 in U. S. D. O. Agriculture, editor. National Agricultural Statistics Service 2009. 2007 Census of Agriculture; Virgin Islands of the United States, territory and is land data. Pages 44 in U. S. D. O. Agriculture, editor. Natural Resources Conservation Service, U. S. D. O. A. 2000. Soil survey of the United States Virgin Islands, San Juan, Puerto Rico. Nepstad, D., C. Uhl, and A. Serrao 1991. Recuperation of a degr aded Amazonian land forest recovery and agricul tural restoration. Ambio 20: 248-255. Parrotta, J. A., and O. H. Knowle s 1999. Restoration of tropica l moist forests on bauxitemined lands in the Brazilian Am azon. Restoration Ecology 7: 103-116. Pascarella, J. B., T. M. Aide, M. I. Serrano, and J. K. Zimmerman 2000. Land-Use History and Forest Regeneration in t he Cayey Mountains, Puerto Rico. Ecosystems 3: 217-228. Pontius, G. R., Jr, S. Emily, and M. Menzie 2004. Detecting important categorical land changes while accounting for persistence. Agriculture, Ecosystems & Environment 101: 251-268. Portillo-Quintero, C. A., and G. A. Sanchez-Azofeifa 2010. Extent and conservation of tropical dry forest in the Am ericas. Biological Conservation 143: 144-155.
104 Quesada, M., G. A. Sanchez-Azofeifa, M. Alva rez-Anorve, K. Stoner, L. Avila-Cabadilla, J. C. Calvo-Alvarado, A. Castillo, M. Espirito-Santo, M. Fagundes, G. Fernandes, J. Gamon, M. Lopezaraiza-Mikel, D. Lawrenc e, C. P. L. Morellato, S. J. Powers, F. De S. Neves, V. Rosas-Guerrero R. Sayago, and G. Sanchez-Montoya 2009. Succession and management of tropical dry fo rests in the Americas: Review and new perspectives. Forest Ecology and Management 258: 1014-1024. Ray, G., and B. Brown 1995a. T he structure of five successi onal stands in a subtropical dry forest, St. john, U. S. Virgin Islands. Caribbean Journal of Science 31: 212222. Ray, G. J., and B. J. Brown 1995b. Restoring Caribbean dry forests Evaluation of tree propagation techniques. Restoration Ecology 3: 86-94. Rivera, L., and M. Aide 1998. Forest recovery in the karst region of Puerto Rico. Forest Ecology and Management 108: 63-75. Roth, L. C. 1999. Anthropogenic change in subt ropical dry forest during a century of settlement in Jaiqui Picado, Santiago Pr ovince, Dominican Republic. Journal of Biogeography 26: 739-759. Roth, L. C. 2001. Enemies of the trees? Subsistence farmers and perverse protection of tropical dry forest. Journal of Forestry 99: 20-27. Rozelle, S., J. Huang, S. Husain, and A. Za zueta. 2000. China: From afforestation to poverty alleviation and natural fore st management. World Bank, Washington D.C. USA. Rudel, T. K., T. O. Coomes, F. E. Moran, F. Achard, A. Angelsen, J. Xu, and E. Lambin 2005. Forest transitions: towards a globa l understanding of land use change. Global Environmental Change 15: 23-31. Rudel, T. K., M. Perez-Lugo, and H. Zichal 2001. When fields revert to forest: Development and spontaneous reforestation in post-war Puerto Rico. Professional Geographer 52: 386-397. Saldarriaga, J., D. We st, M. L. Tharp, and C. Uhl 1988. Long-term chronosequence of forest succession in the upper rio negro of Colombia and Venezuela. Journal of Ecology 76: 938-958. Sandor, B., S. J. Meiners, S. T. A. Pickett, and M. L. Cadenasso 2003. Plant colonization windows in a mesic old field succession 6: 205-212. Sarmiento, O. F. 1997. Arrested succession in pastures hinders regeneration of tropandean forests and shreds mount ain landscapes. Environmental Conservation 24: 14-23.
105 Schnitzer, A. S., W. J. Dalling, and W. Carson 2000. The impact of lianas on tree regeneration in tropical forest canopy gaps : evidence for an alternative pathway of gap-phase regeneration. Journal of Ecology 88: 655-666. Seastedt, T. R., R. J. H obbs, and K. Suding 2008. Management of novel ecosystms: are novel approaches required? Frontiers in Ecology and the Environment 6: 547553. Shiels, B. A., and R. L. Walker 2003. Bird perches increase food forest seeds on Puerto Rican landslides. Restoration Ecology 11: 457-465. Song, C., C. E. Woodcock, K. C. Seto M. P. Lenney, and S. A. Macomber 2001. Classification and change detection using landsat TM data: When and how to correct atmospheric effects? Re mote Sensing of Environment 75: 230-244. Southworth, J., D. Munroe, and H. N agendra 2004. Land cover change and landscape fragmentation--comparing the utility of continuous and discrete analyses for a western Honduras region. Agricult ure, Ecosystems & Environment 101: 185-205. Southworth, J., and C. Tucker 2001. The influence of accessibility, local institutions, and socioeconomic factors on forest cover change in the mountains of Western Honduras. Mountain Research and Development 21: 276-283. Thompson, J., N. Brokaw, J. K. Zimmerman, R. B. Waide, E. M. E. Iii, D. J. Lodge, C. M. Taylor, D. Garcia-Montiel, and M. Fluet 2002. Land use hi story, environment, and tree composition in a tropical forest. Ecological Applications 12: 1344-1363. Thomson, J., N. Brokaw, J. K. Zimmerman, R. Waide, E. Everham, J. D. Lodge, C. Taylor, D. Garcia-Montiel, and M. Fluet 2002. Land use history, environment and tree composition in a tropical fo rest. Ecological Applications 12: 1344-1363. Toh, I., M. Gillespie, and D. Lamb 1999. The ro le of isolated trees in facilitating tree seedling recruitment at a degraded sub-tropical rainforest site. Restoration Ecology 7: 288-297. Townshend, G. R. J., O. C. Justice, C. Gurney, and J. Mcmanus 1992. The Impact of misregistration on change detection. IE FE Transactions on Geoscience and remote sensing 30: 1054-1060. United States Geological Survey 2007. 1:24,000 digital line graphs, U.S. Virgin Islands, Reston, Virginia. University of the Virgin Islands, C. D. C. 2000. Rapid ecological assessment. University of the Virgin Islands, St. Thomas, VI. U.S.A. University of the Virgin Islands, C. D. C. 2002. Rapid ecological assessment. University of the Virgin Islands, St. Thomas, VI. U.S.A.
106 Van Der Putten, W. H., S. R. Mortimer, K. H edlund, C. Van Dijk, V. K. Brown, J. Leps, C. Rodriguez-Barrueco, J. Roy, T. A. Dia z Len, D. Gormsen, G. W. Korthals, S. Lavorel, I. Santa Regina, and P. Smilauer 2000. Plant species diversity as a driver of early succession in ab andoned fields: a multi-site approach 124: 91-99. Vieira, I. C. G., A. S. Silva De Almeida, E. A. Davidson, T. A. Stone, R. D. Carbalho, C. J. Guerrero, and J. Benito 2003. Classif ying successional forests using Landsat spectral properties and ecological charac teristics in eastern Amazonia. Remote Sensing of Environment : 470-481. Vieira, L. M. D., K. Holl, and F. Peneireir o 2009. Agro-successional restoration as a strategy to facilitate tropical fore st recovery. Restoration Ecology 17: 451-459. Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo 1997. Human domination of Earth's ecosystems. Science 277: 494-499. Vozzo, J. A. E. 2002. Tropical Tree Seed M anual. Pages 899 in J. A. Vozzo, editor. USDA Forest Service, Agriculture Handbook #721. U.S. Forest Service, Washington D.C. Wang, G., G. Gertner, S. Fang, and A. Anderson 2004. Mapping vegetation cover change using geostatistical methods in bitemporal landsat TM images. IEEE Transactions of Geoscience and Remote Sensing 42: 632-643. Weaver, L. P. 2006. Estate Thomas Expe rimental Forest; Research History and Potential. USDA Forest Service, Internat ional Institute of Tropical Forestry, San Juan, Puerto Rico. Weaver, L. P., and D. J. Chinea 2003. Secondary subtropical dr y forest at the La Tinaja tract of the Cartagena Lagoon Na tional Widlife Refuge, Puerto Rico. Caribbean Journal Science 39: 273-285. Wunderle, M. J. 1997. The role of animal seed dispersal in accelerating native forest regeneration on degraded tropi cal lands. Forest Ecology and Management 99: 223-235. Zarin, J. D., E. A. Davidson, E. Brondizio, I. C. G. Vieira, T. Sa, T. Feldpausch, E. Schuur, R. Mesquita, F. E. Moran, P. Delamonica, M. J. Ducey, G. Hurtt, C. Salimon, and M. Denich 2005. Legacy of fi re slows carbon accumulation in Amazonian forest regrowth. Frontiers in Ecology and the Environment 3: 365-369. Zimmerman, J. K., J. B. Pasca rella, and T. M. Aide 2000. Ba rriers to forest regeneration in an abandoned pasture in Puerto Rico. Restoration Ecology 8: 350-360.
107 BIOGRAPHICAL SKETCH Brian Daley graduated from St Johns Preparatory School in Danv ers in 1987 and obtained a Bachelor of Science degree in sociology from Sale m State College in 1992. He served as a Peace Corps volunteer in Costa Rica from1994-1996 and was awarded a Master of Science in Natural Resource Management from the School of Forestry at the University of Montana in 1999. In that same year he began work at the University of the Virgin Islands, Agriculture Experiment Station as an Agroforestry Research Specialist. Since 2010 he has been a Natural Resources Specialist and Partner at the private firm, Geographic Consulting LLC. Brian Daley was awar ded a Doctor of Philosophy from the University of Florida, School of Fore st Resources and Conservation in 2010.