Project Summary

The Luquillo Experimental Forest Long-Term Ecological Research program (LUQ) focuses on the long-term dynamics of tropical forest ecosystems characterized by large-scale, infrequent disturbance, rapid processing of organic material, and high habitat and species diversity. Research by LUQ has stimulated a new appreciation of the significance of large-scale disturbances in tropical forested ecosystems and the key role of the biota in shaping the response to these events. Hurricanes occurring one and 10 years after the LTER program began permitted us to capitalize upon landscape-scale natural experiments which we continue to follow closely. Among our most important findings from these natural experiments is that detrital dynamics plays a central role in forest recovery by influencing carbon and nutrient storage and flow.
The central theme we propose is that disturbance, through its effects on detrital dynamics, is a dominant ecosystem driver in the Luquillo Experimental Forest. A common feature of all disturbances is the generation of dead organic matter that must be processed by the biota as part of the ecosystem's response to disturbance. Pulses of detritus shift the flow of energy within the food web, modify the availability and distribution of nutrients, and feed back on the composition and productivity of plant and animal communities. Rapid processing of detritus distinguishes the Luquillo Experimental Forest (LEF) from other forested LTER sites, where decomposition takes two to 20 times as long. In this proposal, we combine long-term measurements, field experiments, simulation modeling, and cross-site comparisons to address five questions: (1) How do climatic factors, litter quality, and detritivore diversity regulate decomposition of detrital pulses? (2) How do terrestrial and aquatic food webs differ in response to detrital pulses? (3) What is the effect of disturbance frequency on nutrient cycling, plant community composition, and the accumulation of soil organic matter? (4) To what degree is the export of carbon and nutrients from watersheds a result of soil characteristics that are affected by detrital dynamics? (5) How do elevationally related changes in climate impact plant and detritivore communities, and how do these feed back on the quantity and quality of litter inputs and decomposition? Through our focus on disturbance and detrital dynamics, we build on existing strengths in integrating community and ecosystem ecology and the research opportunities provided by infrequent, large-scale disturbance, high diversity, and pronounced elevational gradients in the LEF.
Our research will be conducted in two spatial contexts. In mid-elevation tabonuco forest, we will continue long-term measurements of ecosystem response to hurricanes, landslides, and anthropogenic disturbance. We will initiate an experiment mimicking increased frequency of hurricanes to investigate the effect of increased detrital inputs on nutrient cycling, community composition, and organic matter accumulation. We also will manipulate key functional groups of invertebrates to gain a better understanding of similarities in detrital processing between aquatic and terrestrial food webs. In a landscape context, we will establish new plots to examine the effect of elevationally related changes in climate, plant communities, and decomposers on detrital processing. Manipulative experiments will compare the relative importance of abiotic and biotic controls on decomposition in terrestrial and aquatic habitats. Long-term measurements of hydrological and nutrient fluxes in watersheds will relate soil characteristics to stream nutrient and organic matter losses and provide information to gauge the effects of future disturbances. Simulation models of key population, community, biogeochemical, and landscape processes provide null-model predictions to inform these new observations and experiments.

Section 1 Results From Prior NSF Support

Long-Term Research in the Luquillo Experimental Forest 2 & 3; NSF Grants DEB-9411973 and DEB-9705814, Oct. 1994 - Nov. 2000 [$3,600,000], and DEB-00805238, Nov. 2000 - Oct. 2002 [$1,400,000])

The Luquillo Experimental Forest LTER program (LUQ) has shown how a diverse biota interacts with a varied disturbance regime and environmental gradients to determine habitat structure, nutrient cycling, community organization, and food web relations in terrestrial and aquatic ecosystems (Walker et al. 1991, 1996a, 1996b, Lugo & Scatena 1995, Covich et al. 1996, Reagan & Waide 1996). Building on a history of research in forestry (Wadsworth 1995) and ecosystem ecology (Odum & Pigeon 1970), the Luquillo LTER program focused initially on the importance of hurricanes and landslides within four forest types in the Luquillo Experimental Forest (LEF). We emphasized the significance of the biota in restoring ecosystem productivity after disturbance events at two study sites (El Verde and Bisley; see Fig. 2.1.1) in one of these forest types, tabonuco. In preparation for experimental studies of disturbance planned for LTER 1, we established long-term measurements of ecosystem characteristics to understand the spatial and temporal dynamics of tabonuco forest. These measurements provided background to assess the impacts of Hurricane Hugo and associated landslides in 1989 and to evaluate biotic responses to these disturbances. In LTER 2, we continued our intensive studies of the effects of and response to hurricanes and landslides, while expanding our focus to encompass other forms of disturbance including historical and present-day agriculture. Hurricane Georges in 1998 provided an opportunity to assess the interaction between disturbance events. To test our understanding of forest processes against a broader range of biotic and abiotic conditions, LUQ expanded its research from mid-elevation tabonuco forest (200 to 600 m) to the summits of the LEF (1000 m) during the first two years of LTER 3. We are now poised to 1) investigate how the effects of repeated disturbance (canopy opening and pulses of detritus) propagate through the system to modify other forest characteristics, and 2) examine how a key ecosystem process, detrital dynamics, changes along elevationally related gradients of climate and biodiversity.

LUQ was initiated in 1988 and renewed in 1994 and 2000. During the last 13 years, we have provided a platform for research within Puerto Rico and for comparisons across the LTER Network by establishing a point of reference in which climatic conditions are continuously warm and wet and the biota diverse. LUQ involves researchers from the University of Puerto Rico, the International Institute of Tropical Forestry and Forest Products Lab (USDA Forest Service), and 11 mainland institutions. Since its inception, LUQ has produced 348 peer-reviewed journal articles, 89 book chapters, 7 books, and 32 scholarly reports as well as 56 theses and dissertations. Please see our web page (http://luq.lternet.edu) for a complete list.

1.1 Major Findings
Research conducted by the Luquillo LTER program has changed how ecologists view the importance of large-scale disturbances in maintaining rain forest biodiversity and ecosystem functions (Sugden 1992, Lodge et al. 1994, Whittaker 1995, Miller & Lodge 1997, Willig & Walker 1999). Pioneering work on the effects of hurricanes on tropical forests and the interaction between natural and anthropogenic disturbances has stimulated reconsideration of the resilience of rain forest ecosystems and contributed to the development of a new vision of rain forests as dynamic systems (Walker et al. 1996a, Walker 1999, Foster 2000). This new vision is widely applicable to forested ecosystems with other characteristic large-scale disturbances such as fires or droughts. Comparison of our results with those from other forests (Boose et al. 1994, Turner et al. 1997, Dale et al. 1998, Walker 1999, Foster 2000) is leading us towards a conceptual model of how disturbance structures LEF ecosystems. Although disturbance takes many forms, the conversion of live biomass into detritus is a factor common to all disturbances (Sousa 1984, Dale et al. 2001). The mechanisms by which ecosystems process pulses of detritus thus determine their response to disturbance by regulating critical ecosystem fluxes and storages. Hence, our results strongly support the emerging concept that disturbance, through its effects on detrital dynamics, is a dominant ecosystem driver in the Luquillo Experimental Forest.

1.1.1 Disturbance regime: Although we anticipated in LTER 1 that disturbance was a major force structuring LEF ecosystems, the degree to which this prediction has been substantiated is nevertheless surprising. In 13 years, two major and four lesser hurricanes have affected our study sites (Fig. 2.1.1), widespread landslides have occurred associated with specific landscape features (Walker et al. 1996b), and floods and droughts have modified stream environments and populations (Covich et al. 1996, 2000). In addition to documenting the effects of these natural disturbances, we have found that legacies from various human disturbances (mainly agriculture and logging) in some areas of tabonuco forest have long-lasting effects (Thompson et al., in press b). This cumulative evidence reaffirms our initial view of disturbance as a dominant driver of structure and function in LEF ecosystems (Waide & Lugo 1992).
The effects of hurricanes are widespread and severe and have thus provided the major focus for our research on disturbance impacts. Hurricanes strike the LEF on average every 55-60 years and pass within 60 km every 20-25 years (Scatena & Larsen 1991). Simulation of historical hurricanes has demonstrated a strong, cumulative spatial pattern in impact (Boose et al. 1994; see Fig. 2.1.1). Tabonuco forest models have predicted strong effects of repeated hurricanes on long-term forest composition (Doyle 1981) and ecosystem characteristics (Sanford et al. 1991), but these predictions have yet to be verified experimentally.

1.1.2 Hurricane disturbance and recovery: During Hurricane Hugo defoliation and branch loss were widespread and, depending on site exposure, 18 to 76% of trees were snapped off or uprooted (Walker 1991, Basnet et al. 1992). Mortality of large trees varied spatially from 0 to 37% (Lugo & Scatena 1996). However, tree mortality was low enough in many areas that species composition changed little (Zimmerman et al. 1994, Walker 1995). Massive mortality of fine roots occurred after Hurricane Hugo (Silver & Vogt 1993). Animals were greatly affected by changes in habitat structure and food resources (Waide 1991a,b). Frogs and snails that rely on litter for shelter or food increased in abundance following Hugo (Woolbright 1996, Willig et al. 1998). Nectar- and fruit-feeding birds and bats declined following both hurricanes (Gannon & Willig 1994). In streams, high flows moved storm-generated detritus into debris dams. Increased detritus and microbial biomass led to increased detritivore abundance, particularly Atyid shrimps (Covich et al. 1991, 1996, Johnson et al. 1998, Johnson & Covich 2000).
Hurricanes Hugo and Georges opened the forest canopy and produced large pulses of litterfall and woody debris (Lodge et al. 1991, Scatena et al. 1996). Hugo deposited approximately 1 kg/m2 of litter on the forest floor and the subcanopy vegetation intercepted and suspended a like amount. Total litterfall from Hugo was 1.2 times the annual litterfall, but delivered in just one day. Nutrient delivery to the forest floor was proportionally greater, because hurricane litterfall and fine root inputs had high concentrations of N and P (Lodge et al. 1991, Lodge & McDowell 1991). Nutrient losses to streams were smaller than expected, given the amount of damage, because soil retention and nutrient uptake by pioneer plants quickly restored aboveground nutrient pools and ecosystem function (Scatena et al. 1996, Silver et al. 1996, McDowell et al. 1996). While highly dynamic in their chemistry compared to streams in other ecosystems (e.g., HBR, Bormann & Likens 1981), streams draining the LEF went from elevated levels of nitrate and potassium to normal within 1 to 2 years of the hurricane (Schaefer et al. 2000).
Sprouting of damaged trees and recruitment of pioneer trees quickly restored forest structure (Walker 1991, Brokaw et al., in press), producing a temporary increase in net primary productivity (NPP; Scatena et al. 1996). Within five years, many ecosystem state variables and functions (e.g., leaf litterfall, aboveground forest biomass, NPP, some soil nutrient pools) and the abundances of many plants and animals had returned to or were approaching pre-Hugo levels (Zimmerman et al. 1996). Exceptions include pioneer trees and shrubs, particularly Cecropia schreberiana, whose abundances remained high long after the hurricane (Brokaw 1998).
In the short term (1 to 2 yr), the combination of a detrital pulse and canopy opening increased light levels (Fernández & Fetcher 1991) and soil moisture (Steudler et al. 1991), increased nutrient inputs to forest floor and streams, and altered N and P cycling (McDowell et al. 1996, Silver et al. 1996), increased nitrate concentrations in streams and soil (Silver & Vogt 1993), elevated rates of denitrification (Steudler et al. 1991), increased the functional diversity of soil microbial communities (Willig et al. 1996), differentially affected survival of litter fungi (Lodge & Cantrell 1995), differentially affected germination and seedling survival in pioneer and late successional species (Guzmán-Grajales & Walker 1991), and accelerated growth (Scatena et al. 1996). In particular, the post-hurricane pulse of detritus and the decrease in canopy biomass redirected energy away from herbivores to detritivores, with significant effects on abundance and population structure. The relative importance of canopy opening and the pulse of detritus in regulating these effects has yet to be determined.
We are beginning to understand the longer-term dynamics of post-hurricane detrital pulses. Decomposition of both leaf litter and wood is extremely rapid in the LEF compared to other sites within the LTER Network (Gholz et al. 2000). The combination of warm, moist conditions in the litter layer after Hugo led to complete decomposition of tabonuco leaves in less than a year and of 3 to 6 cm wood in 7 to 16 years (Vogt et al. 1996). Ongoing studies of bole decomposition suggest a residence time of a few decades (versus centuries in temperate forests). The decomposition of coarse woody debris (> 5 cm diameter) can regulate productivity for long periods of time (up to 13 years) by immobilizing nutrients as it decomposes (Zimmerman et al. 1995b). Simulations of long-term dynamics of tabonuco forest using the CENTURY model predicted that repeated disturbance by hurricanes would result in higher soil organic matter (SOM), P availability, N-mineralization, and productivity and lower forest biomass (Sanford et al. 1991). Simulation results suggested that higher productivity from repeated disturbance was driven by increased SOM through its effect on P availability. Recent data show significantly more SOM and N under decomposing logs from Hurricane Hugo (Lodge et al. 2001). Silver et al. (1996) however, found no change in SOM after Hurricane Hugo and suggested that increased decomposition and soil respiration rates may have balanced the SOM added. A thorough examination of this suggestion and predictions from Sanford et al. (1991) requires long-term experiments.

1.1.3 Human disturbance and recovery: Human disturbance was widespread in tabonuco forest during the last century (Foster et al. 1998), and these disturbances exhibit strong legacies. For example, the Luquillo Forest Dynamics Plot (LFDP), a 16-ha grid at El Verde, has been free of human disturbance since the 1930s, but its tree species composition reflects past land use more strongly than damage from hurricanes in 1932, 1989 and 1998 (Thompson et al., in press b). In general, second growth forests in the LEF demonstrate the legacy of human disturbance in their altered species composition rather than in their biomass, tree density, or species richness (Zimmerman et al. 1995a, Thompson et al., in press b). In streams, we have shown strong impacts of water diversion on organisms and ecosystem processes, and we developed instream flow models to help manage water resources (Pringle 1997, March et al. 1998, Benstead et al. 1999, Pringle et al. 1999, Scatena & Johnson 2001).

1.1.4 Interactions between disturbances: We have shown how disturbance effects are conditioned by previous disturbances (Willig & Walker 1999). High rainfall associated with hurricanes initiates many landslides (Scatena & Larsen 1991), but these occur mainly where roads have changed local topography (Guariguata 1990). Hurricanes Hugo and Georges were comparable in strength, but Hugo felled and delimbed so many susceptible trees that Georges, coming nine years later, had much less effect (Brokaw et al., in press). The tabonuco forest is less resistant to hurricanes in areas that have been disturbed by humans. In the LFDP, Hurricane Hugo caused most damage to trees in areas cleared 70 years ago, because secondary tree species were more susceptible to damage than were old-growth species (Zimmerman et al. 1994, Everham 1996, Thompson et al., in press b). These anthropogenically altered areas are likely to remain in secondary forest indefinitely because the high return frequency of hurricanes perpetuates destruction of and recolonization by secondary tree species.

1.1.5 Beyond tabonuco forest: Ascending the Luquillo Mountains, climate becomes cloudier, wetter, and cooler, and forests become shorter, denser, less species-rich, and less productive (Waide et al. 1998). Oxygen depletion in saturated soils limits productivity and decomposition at high elevations (Silver et al. 1999, McGroddy & Silver 2000). Thus, elfin forests at 1000 m were slow to return to pre-hurricane levels of productivity compared to tabonuco forest (Walker et al. 1996b). We extended our understanding of tabonuco forest processes to the full elevational range of the LEF by integrating CENTURY with a climate model developed by LUQ (TOPOCLIM; Wooster 1989) and a primary production model (Marley 1998) coupled with a geographic information system. The resultant landscape model (L-CENTURY) provided reasonably accurate estimates of spatial and temporal variability in gross and net primary production as well as storages and fluxes of soil organic carbon and nitrogen (Wang 2001, Wang et al. in press). However, the scarcity of appropriate data from higher elevations prevented a full validation of the model predictions.

1.1.6 Cross-site studies: As the only tropical LTER site, LUQ plays a valuable role in cross-site comparisons within the LTER network and between the LTER network and other tropical sites (Heneghan et al. 1998a,b, Thompson et al., in press a; see Section 2.6). For example, in the Long-term Intersite Decomposition Experiment (LIDET) our diverse, warm site had higher decomposition rates than less diverse, cooler sites (e.g., KNZ, CWT) but had rates similar to those in other wet tropical sites (Costa Rica; Gholz et al. 2000). Comparative studies between LUQ and NWT show how the biota and climate affect decomposition and soil processes (González & Seastedt 2001). The Lotic Intersite Nitrogen Experiment (LINX) showed that in our N-rich site, ammonium turnover in streams is very rapid, due largely to nitrification rates that were higher than those at any other site in the study (Peterson et al. 2001, Merriam et al. 2002). Results from LUQ and the USGS-sponsored cross-site Water, Energy, and Biogeochemical Budgets (WEBB) project have been instrumental in establishing the importance of temperature in controlling global variation in weathering rates (McDowell & Asbury 1994, White & Blum 1995, White et al. 1998).

1.1.7 Major publications: We summarized results of LUQ research in three books (Lugo & Lowe 1995, Reagan & Waide 1996, Walker 1999) and two special issues of Biotropica (Walker et al. 1991, 1996a). LUQ scientists led or participated in publications of cross-site results on the relationship between primary productivity and species richness (Waide et al. 1999, Gross et al. 2000, Scheiner et al. 2000, Mittelbach et al. 2001), on the effects of large infrequent disturbances (Turner et al. 1997, Dale et al. 1998), on the biology of key forest species (Lugo et al. in press), and on measuring primary productivity (Clark et al. 2001a,b). A synthesis of LUQ results to date is being prepared for the LTER-Oxford book series. The Supplementary Documentation provides a list of LUQ publications during the past six years (total = 271).

1.1.8 Human resources: Ten post-doctoral associates have worked with LUQ during the past six years. Fifty-six students have completed doctoral or masters' degrees working with LUQ (see list of publications), 26 during the past six years. Fifty students (62% female, 66% minority) have been involved in Research Experience for Undergraduates (REU) projects, supported by supplemental funds to the LTER project or by a site grant to the University of Puerto Rico. For the current census of the LFDP, we supported (A. W. Mellon Foundation grant) over 50 student volunteers and technicians, a first experience for many in the tropics. Six high schools throughout Puerto Rico have experimental plots near their schools and participate in a research network with LUQ as part of the Schoolyard LTER program (Lugo 1999).

1.2 Response to Previous Reviewers' Comments
We submitted a renewal proposal in 2000 and were asked to prepare a new proposal for the period 2002-6. Thus, the present proposal will support the final four years of LTER 3. We were asked to address the following major points in this new proposal:

1.2.1 Conceptual framework: Based on an evaluation of results to date, LUQ principal investigators jointly identified the need for a better understanding of the interaction between disturbance and detrital dynamics as our next major research priority. While recognizing the need for future work on a variety of topics, we agreed upon an integrative conceptual framework for research in LTER 3: disturbance, through its effects on detrital dynamics, is a dominant ecosystem driver in the LEF. The scientific justification for this conceptual framework arises from results presented above and is considered in detail below. In brief, we propose to continue our focus on disturbance through a closely integrated approach to linked ecosystem processes, the production of detritus by disturbances and the role of the biota in recycling of carbon and nutrients via decomposition. Emphasis on disturbance and detrital dynamics also provides a framework for our recent initiative to extend our research to the entire elevational gradient represented in the LEF. We have revised and reduced the number of proposed hypotheses and field projects, such that these all contribute directly to the investigation of disturbance and detrital dynamics. We also have revised modeling and long-term measurement programs to eliminate redundancy and to insure that these guide and support the new research.

1.2.2 Data management: LUQ data sets are accessible through a new searchable web interface on our revamped web page (http://luq.lternet.edu). All data sets and metadata have been updated and are freely available (see Supplementary Documentation).

1.2.3 Cross-site comparisons: Cross-site research has always been a strength of LUQ, but our 2000 proposal did a poor job of making that point. We describe cross-site research projects in detail in the body of the proposal (Section 2.6).

Section 2 Proposed Research

2.1 Introduction

Incorporating disturbance regime as a fundamental attribute of an ecosystem or landscape has revolutionized the way scientists view ecological systems (Sousa 1984, Pickett & White 1985, Walker 1999). Numerous studies show that disturbance imparts a distinctive signature to the structural and functional attributes of most ecosystems, including tropical forests (Garwood et al. 1979, Sanford et al. 1985, Gómez-Pompa & Kaus 1990, Walker et al. 1991, 1996a, Ashton 1993, Clark et al. 1995, Walker 1999, Willig & Walker 1999). In the Luquillo Mountains, hurricanes are the dominant disturbance type, and, with landslides and land use, constitute a disturbance regime whose impacts range from decadal to generational (Walker et al. 1991, Waide & Lugo 1992, Lugo & Scatena 1995, Walker et al. 1996a). Previous research by the Luquillo Experimental Forest LTER program focused on understanding the effects that hurricanes in 1989 and 1998 had on the structure and function of mid-elevation tabonuco forest and how the biota responded to these disturbances (Fig. 2.1.1). This focus, together with studies of landslides and human land use, provided insights into the key characteristics of disturbance that alter forest function in tropical montane ecosystems over long time scales. One primary effect of disturbance is to redistribute organic matter from live biomass compartments to the detrital pool. The dominant biotic and abiotic processes that contribute to the production and decomposition of detritus and to the subsequent fate of associated C and nutrients are shown in bold in Fig. 2.1.2. These processes, which define detrital dynamics, play a central role in the recovery of forest structure and function by regulating carbon and nutrient storage and flow. Thus, disturbance, through its effects on detrital dynamics, is a dominant ecosystem driver at the Luquillo LTER site.
The disturbance regime determines the temporal, spatial, and chemical characteristics of litter inputs to the detrital system and thereby controls critical parameters and dynamics of terrestrial and aquatic compartments of the LEF ecosystem. Disturbance produces a pulse of detrital material (fine litter and coarse woody debris), modifies the microclimate, and changes decomposer species composition. The key factor, however, is the pulsed nature of inputs of detrital material (Sousa 1984), and thus the long-term response of the biota to disturbance is constrained by the quality, quantity, and distribution of detritus. Changes in microclimate and species composition (particularly detritivores) modify the rate at which the detrital pulse propagates through the system. Our previous research in tabonuco forest has shown that post-disturbance conditions determine the response rate and trajectory of the modified ecosystem (Willig & Walker 1999) by influencing germination, growth, survival, nutrient export, soil conditions, plant community composition, and consumer trophic structure. In this proposal, we build upon our earlier studies to focus on the critical linkages between disturbance and the production and processing of detritus, and the influence of these linkages on the long-term dynamics of ecosystems. We address five questions fundamental to understanding the interaction between disturbance and ecosystem function in the Luquillo Mountains:

(1) How do climatic factors, litter quality, and detritivore diversity regulate decomposition of detrital pulses?
(2) How do terrestrial and aquatic food webs differ in their response to detrital pulses?
(3) What is the effect of disturbance frequency on nutrient cycling, plant community composition, and the accumulation of soil organic matter?

C

Figure 2.1.1. Hurricane damage to (A) forest near El Verde Field Station showing (B) detritus deposited on forest floor. Winds and litterfall deposited at Bisley Experimental Watersheds (C) by storms experienced at the LEF during 1988 - 2001. A century of maximum exposure to hurricane winds (D; as estimated by the HURRECON and EXPOSE models, Boose et al. 1994, unpublished) shows that all areas of the LEF sustained severe damage sometime during 1886 to 1996 and that topographic position modifies patchiness of wind damage. Legend shows none to high damage categories, yet most of the surface was highly or very highly damaged over 100 years.

Fig. 2.1.3. Mean annual temperature and precipitation of the Luquillo Experimental Forest relative to other US-LTER sites in the Long-term Intersite Decomposition Experiment. As a result of higher rainfall and temperatures, decomposition in the LEF takes less time than in any of the other US-LTER sites (Gholz et al. 2000).

(4) To what degree is the export of carbon and nutrients from watersheds a result of soil characteristics that are affected by detrital dynamics?
(5) How do elevationally related changes in climate impact plant and detritivore communities, and how do these feed back on the quantity and quality of litter inputs and decomposition?
We have assembled a diverse scientific team to address these questions (Table 2.1.1).
Detrital dynamics and its linkages with disturbance have broad relevance to the fields of ecology and biogeochemistry and contribute to our understanding of global change. Controls on the fluxes of organic matter through ecosystems have important feedbacks to soil C sequestration, atmospheric CO2, and the export of terrestrial C to rivers and oceans (Schlesinger 1999). Although detrital food webs are the major pathway for carbon mineralization in many ecosystems and release most of the essential nutrients for primary producers, relatively few studies have explored the functional aspects of detrital food webs compared to other food webs (Moore et al. 1993, Jefferies et al. 1994, Polis & Hurd 1996, Hooper et al. 2000).

2.2 Conceptual Approach

Disturbance alters ecosystems in many ways, and response to repeated disturbance may occur at both ecological and evolutionary time scales. A comprehensive approach to understanding the effect of disturbance on ecosystems recognizes this complexity and addresses it by focusing on critical regulating processes. In the LEF (Fig. 2.1.3), high, seasonally constant temperatures and rainfall, high biotic diversity, and repeated large-scale disturbance regulate the rapid rates of growth and decomposition that characterize Luquillo and many other tropical rain forest ecosystems. By affecting the timing, magnitude, and quality of inputs of detritus into the system, disturbance drives the temporal and spatial dynamics of decomposition. The rate at

Table 2.1.1. Participants in LTER 3. An Executive Committee comprised of the four PIs and three rotating senior personnel (current members in bold) directs the research program. The LTER Program is also supported by additional key scientific personnel: a Data Manager, the Project Director of the Luquillo Forest Dynamics Plot (LFDP), and the Director of El Verde Field Station.

	Participant	Present Affiliation	Specialty	Years of Tropical Experience
Principal Investigators	J.K. Zimmerman	University of Puerto Rico	Plant ecology	17
	A.E. Lugo	Int. Inst. Tropical Forestry, USDA FS	Ecosystem analysis, nutrient cycling	42
	D.J. Lodge	Forest Products Lab, USDA FS	Nutrient cycling, fungal systematics	20
	N. Brokaw	University of Puerto Rico	Forest ecology	28
Senior Personnel	G. Belovsky	Notre Dame University	Population and ecosystem modeling	2
	A. Covich	Colorado State University	Stream ecology	26
	T. Crowl	Utah State University	Quantitative analysis, stream ecology	14
	E. Cuevas	University of Puerto Rico	Decomposition, belowground processes	25
	G. González	Int. Inst. Tropical Forestry, USDA FS	Decomposition, nutrient cycling	8
	C. Hall	SUNY-ESF	Modeling, stream ecology	16
	P. Klawinski	William Jewell College	Insect ecology	4
	W. McDowell	University of New Hampshire	Biogeochemistry	20
	C. Pringle	University of Georgia	Stream ecology	15
	W. Silver	University of California – Berkeley	Biogeochemistry	21
	F. Scatena	Int. Inst. Tropical Forestry, USDA FS	Geomorphology, hydrology	25
	J. Thomlinson	University of Puerto Rico	Landscape ecology, GIS	9
	R. Waide	University of New Mexico	Avian ecology	31
	L. Walker	University of Nevada – Las Vegas	Succession, primary production	18
	M. Willig	Texas Tech University	Invertebrate ecology, quantitative analysis	25
	X. Zou	University of Puerto Rico	Nutrient dynamics, earthworm ecology	13
Data Manager	E. Meléndez-Colom	University of Puerto Rico	Information management	13
Project Director LFDP	J. Thompson	University of Puerto Rico	Forest ecology	15
Director El Verde Field Station	A. Ramírez	University of Puerto Rico	Stream ecology, insect systematics	8

which detrital pulses are processed has implications for other ecosystem fluxes and storages (e.g., nutrient cycling and export, soil organic matter, biomass, productivity), as well as for community and population structure and dynamics. Climate and the biota also influence ecosystem processes, but these factors themselves are mediated by disturbance. The Luquillo LTER program focuses on a comprehensive understanding of these complex interactions by integrating population, community, and ecosystem ecologists under a common conceptual approach.

Our present approach incorporates long-term measurements of forest and stream response to natural and anthropogenic disturbance, associated short- and long-term manipulative experiments to develop a process-level understanding of results from our long-term measurements, validation of this understanding through parallel experiments and measurements along gradients of climate and species richness, and comparison of results from LUQ with other LTER and non-LTER sites. By integrating investigations of disturbance, organic matter processing and accumulation, nutrient cycling and export, consumer populations, and productivity, our research addresses the five core LTER research areas.

1) Measurements of long-term changes in climatic, biotic, and biogeochemical characteristics resulting from disturbance in tabonuco forest (Section 2.3)
We use standard measurements of key organisms, processes, and ecosystem characteristics to identify and quantify changes in tabonuco forest in the LEF. This approach takes advantage of natural (e.g., hurricanes, landslides, droughts, floods) and anthropogenic (e.g., previous land use) disturbances to examine the effect of and response to disturbance. While natural disturbances are not controlled experiments, the existence of pre-disturbance data and topographic variation in the intensity of disturbance permits us to quantify the severity of effects among and within disturbance types. Simulation models (HURRECON, CENTURY) provide information about the effects of hurricanes of different intensities and paths. This approach includes the collection and management of core long-term datasets. We will add two new long-term datasets in LTER 3: one on coarse woody debris, an important structural and functional feature of the forest floor, the other on herbivory, a process which under certain circumstances may be an important regulator of detrital processing. In addition, we will expand collection of selected core data sets to other forest types within the LEF, as described in (3) below.

2) Experiments to determine how disturbance alters detrital dynamics (Section 2.4)
To examine key processes identified through long-term measurements we use focused, short- and long-term experiments. In LTER 3 we will focus experiments on the processes that contribute to and control detrital dynamics in tabonuco forest. We will conduct experiments that address the importance of different elements of disturbance (e.g., detrital pulse, changes in microclimate) on decomposition (Question 1), the functional characteristics of invertebrate detritivores on the forest floor and in streams (Question 2), the potential for interactions between autotrophic and detrital food webs, and the effect of disturbance frequency on organic matter accumulation, species composition, and the export of carbon and nutrients (Question 3).

3) Experiments and measurements along gradients of climate and species richness (Section 2.5)
In the first part of LTER 3, we have initiated a gradient analysis of forest communities and ecosystem attributes with elevation. Results from this analysis will increase our understanding of the relative roles of climate, litter quality, and decomposer communities in determining decomposition rates in terrestrial and aquatic habitats (Question 5), evaluate landscape-scale patterns of SOM and nutrient stores predicted from simulation models, and determine the role of soil characteristics influenced by detrital dynamics in controlling carbon export in streams (Question 4). This gradient analysis will allow us to scale results from forest stand to landscape perspectives and will establish a research infrastructure throughout the LEF on which to base the future development of our research program.

4) Comparison of results from LUQ with other LTER and non-LTER sites (Section 2.6)
Continued participation in ongoing network comparisons (e.g., LIDET, LINX) will be augmented by involvement in new cross-site studies. These new comparative studies will focus on developing an improved understanding of processes involved in detrital dynamics across diverse ecological systems.
In the following four sections, we provide detail about each of these approaches and how they address the five scientific questions raised above. A final section indicates how research conducted under these approaches will lead to a synthetic understanding of disturbance and response in tropical forest ecosystems. Additional methodological details are provided on our website (http://luq.lternet.edu).

2.3 Measurements of Long-Term Changes in Climatic, Biotic, and Biogeochemical Characteristics Resulting from Disturbance in Tabonuco Forest

2.3.1 Background and approach: Measurements of long-term changes in climatic, biotic, and biogeochemical characteristics in the LEF were initiated to determine background levels of spatial and temporal variability, detect rare or gradual events, measure deviations from background as a result of disturbance, and quantify the responses to disturbance. Some of these measurements (e.g., climate, primary production, species abundances) were continuations from earlier research, and others were initiated to address specific goals of the LTER program. Long-term measurements originally were intended to provide a backdrop for watershed-scale manipulation experiments, but the occurrence of Hurricane Hugo and associated landslides refocused our research on these critically important types of disturbance.
Naturally occurring disturbances provide opportunities to observe responses to forces that have shaped ecosystems over evolutionary time. Measurement of the effects of disturbance against background levels constitutes a kind of natural experiment, which, although without a control, can provide important insights into the dynamic behavior of ecosystems. In Caribbean forests, hurricanes are the dominant disturbance type affecting ecosystem structure and function, but other kinds of disturbance also have important effects. Infrequent events such as hurricanes have strong, multi-year to decadal effects on forest and stream functioning (Scatena et al. 1996, Schaefer et al. 2000), while drought has more subtle but equally important long-term impacts on some organisms (Covich et al. 1996, 2000). Likewise, the effects of human disturbance, such as clear-cutting, agriculture, and gradual shifts in the landscape mosaic (Silver et al. 1996, Thomlinson et al. 1996, Silver et al. 2000) can persist for decades and longer (Zimmerman et al. 1995a, Willig et al. 1996). Understanding these effects requires long-term measurements.
Studies of these natural experiments of disturbance and response involve a series of long-term measurements of meteorology, hydrology, ecosystem parameters, and plant and animal populations. These core measurements also reveal long-term changes not associated with episodic events and provide context for short-term, manipulative experiments, such as those on detrital processing (see Section 2.4). Long-term measurements (Table 2.3.1) were defined in LTER 1 and refined as our understanding of tabonuco forest dynamics increased. Long-term measurements are conducted principally at two study sites, El Verde Research Area and the Bisley Experimental Watersheds, in the Espíritu Santo and Mameyes drainages (Fig. 2.3.1). Additional measurements are made in the Río Blanco Watershed, in landslides throughout the LEF, and in sites extending to elfin forest at the summits of the Luquillo Mountains (Fig. 2.3.1).

2.3.2 Disturbance experiments: Hurricanes in 1989 and 1998 and landslides associated with these hurricanes and other storms provide the setting for long-term measurements of forest response. These measurements include:

(1) Environmental properties (e.g., light, nutrients, moisture, temperature) that vary with disturbance size, age, and origin;
(2) Biological properties that are expected to vary with environmental properties (e.g., population density, species composition, growth, nutrient uptake, reproductive success, and carbon fixation);
(3) System-level properties that emerge from the effects of the disturbance regime on the mutual interaction of abiotic environment and biota (e.g., decomposition, nutrient cycling, SOM, resilience, and food web structure).

Measurements of these properties address the following hypothesis, as articulated in LTER 2:

Hypothesis 1: The response to disturbance and the subsequent trajectories of recovery are determined by 1) location along abiotic gradients, 2) the abiotic and biotic conditions resulting from disturbance, and 3) biotic processes subsequent to disturbance. The relative importance of these three factors will vary with the severity of the disturbance. (Brokaw, Covich, Klawinski, Lugo, McDowell, Ramírez, Scatena, Silver, Thompson, Waide, Walker, Willig, Zimmerman, Zou)

Rationale - Results from LTER 1 and 2 demonstrate that understanding of ecosystem recovery after disturbance requires knowledge of 1) the spatial distribution of resources needed for growth (e.g., water, light, and nutrients distributed along gradients of slope, aspect, elevation, and geology), 2) legacies of natural and anthropogenic disturbances, and 3) the properties of nutrient storage and cycling, decomposition, reproduction, dispersal, establishment, growth, and survival inherent in the biota. In LTER 3, we continue measurements of these factors in tabonuco forest to compare responses to specific disturbances and to understand the complex interactions that result from serial disturbances.
Workplan: meteorology - Meteorological measurements answer questions of how average, unusual, and changing climatic patterns affect LEF ecosystems. In LTER 1 we initiated meteorological measurements (Table 2.3.1) at multiple sites conforming to LTER Level 3 weather stations. Those data and historical records were extrapolated to the whole LEF using mechanistic models (Wooster 1989, Hall et al. 1992, García-Martinó et al. 1996a,b). We augmented these permanent stations with short-term measurements of environmental variables associated with particular experiments. Long-term measurements revealed the surprising incidence of droughts, a natural phenomenon that has important impacts on some organisms (Covich et al. 1996, 2000). In the first part of LTER 3, we initiated intensive micrometeorological measurements in tabonuco forest. In addition, we expanded our long-term meteorological measurements along the elevational gradient to help understand how climate

Table 2.3.1. Long-term measurements associated with the Luquillo LTER research program. Legend: B = Bisley, E = El Verde, L = Landslides, P = Pico del Este, LEF = other areas in the LEF.

Measure	Initiation	Funding	Location	Frequency
METEOROLOGY
Rainfall	1975, 1987 1994	NSF, DOE USFS, Mellon	E, B, P, LEF	Hourly and daily totals Daily NOAA records to turn of century
Temperature (air and soil)	1975, 1987 1994, 1997	NOAA, NSF, USFS, USGS, Mellon	E, B, P, LEF 12 stations	Hourly and daily totals with Level 3 Stations
Humidity, wind speed and direction	1988, 1994	NSF, USFS, USGS, Mellon	B, E, P, LEF 6 stations	Continuous measurement with Level 3 Stations and vertical profiles at El Verde
Light (PAR, total radiation, albedo)	1988, 1994,	NSF, USFS, USGS	E, B, P, LEF	Continuous measurements at 4 stations
HYDROLOGY
Stream discharge	1945	USGS, USFS	LEF	Daily discharge, 18 historical streams, 12 currently active
Throughfall	1987	USFS, NSF	B, LEF	Daily and weekly in Bisley
CHEMISTRY (major cations and anions)
Rain	1983, 1988	NSF, NADP NSF	E, B, P	Bulk and wet-only; wkly samples major ions
Throughfall	1988, 1994	UPR, NSF	B	Weekly
Streamwater	1983, 1988 1997 (DON)	UPR,NSF, NASA	B, E, LEF	Weekly samples from 8 long-term streams; Periodic sampling of 13
Litterfall	1989, 1994	NSF, USFS	B, E, P	2 week sampling
Groundwater, soil water, soil oxygen	1988, 1994	NSF, USFS, Mellon, USGS	LEF	Weekly to periodic
VEGETATION
Forest structure and composition (i.e., density, composition, biomass)	'>1988	'>NSF, USFS, Mellon	B, E, LEF	1-5 year intervals
Belowground biomass	1988	NSF, Mellon	B, E	Yearly
Canopy structure and leaf area index	1989	NSF	E, B	Every 3 years
CWD distribution	2002	NSF, USFS	E, B	Every 3 years
Seedling dynamics	1989	NSF, USFS	B, E	Monthly to yearly
Flowering phenology	1989	NSF, USFS	B, E,	Weekly to monthly
Herbivory	2002	NSF	E	Yearly
Litterfall, litter decomposition	1987, 1994, 1996	NSF, USFS	B, E, L, P, LEF	Weekly to periodic
Arial;'>Landslide revegetation	'>1988	'>NSF	LEF	Every 6 months to yearly
Abandoned pasture revegetation	1996	NSF	LEF	Yearly
FAUNA
Key species inventory	1988	NSF	E, B	Yearly
DISTURBANCE
Inventory of gaps, landslides, and stream  channel change	1988	NSF,USFS, USGS	B, E, LEF	Yearly to periodic

Figure 2.3.1. Vegetation of the Luquillo Experimental Forest. Currently, the vegetation is classified into four forest types, described by dominant species or physiognomy. Lower elevations of the LEF (200-600 m) include tabonuco (Dacryodes excelsa) forest, the subject of LTER 1 and 2, which is studied intensively at El Verde Field Station and the Bisley Experimental Watersheds. Elfin forest occurs at the summits of the Luquillo Mountains over 900 m where conditions are extremely wet, cloudy, and windy; Pico del Este is a commonly used study site. At intermediate elevations, colorado (Cyrilla racemiflora) and palm (Prestoea montana) forest occur, with the latter occupying the steepest slopes and riparian areas. Our current and proposed research is aimed at extending our detailed understanding of disturbance and ecological response in tabonuco forest to all vegetation types. Efforts to document community variation in biotic communities are utilizing elevational transects in the Espíritu Santo (ES), Mameyes (MA) and Río Blanco (RB) watersheds.

affects detrital processing (Section 2.5). Both of these measurement programs use state-of-the-art wireless technology being implemented under a separate grant from NSF.
Workplan: hydrology and nutrient cycling - Our measurements of hydrology and nutrient cycling (Table 2.3.1) are used to show how fundamental ecosystem processes are affected by disturbance, and how they change over the long term and along the elevational gradient. Our primary measurements include whole-watershed mass balances for nutrients and other elements (measuring precipitation inputs, biomass accumulation of trees, and stream exports; e.g., McDowell & Asbury 1994); measurement of internal nutrient fluxes (throughfall, litterfall, and litter decomposition; Fig. 2.3.2); and changes in nutrient stocks in soils and vegetation in forested plots as well as during primary succession on landslides. During LTER 1 and 2, these measurements demonstrated the impacts of hurricanes, landslides, and clearcutting on nutrient cycling (Zimmerman et al. 1995b, Zarin & Johnson 1995a,b, Scatena et al. 1996, Silver et al.1996, McDowell et al. 1996, Walker et al. 1996a,b, Schaefer et al. 2000, Lodge et al. 2001). Several of our projects have shown that coarse woody debris (CWD) is particularly important in regulating productivity in tabonuco forest (Zimmerman et al. 1995b, Miller & Lodge 1997, Waide et al. 1999, Fig. 2.3.3), and thus we have begun long-term studies of the distribution of CWD at several sites (Harmon & Sexton 1986). During LTER 3, we will continue all of these basic measurements of hydrology and nutrient cycling at our current sites and expand to include more measurements at higher elevations (Section 2.5).

Fig. 2.3.2. Rainfall, throughfall, and litterfall collected in the Bisley Experimental Watersheds from 1988 - 2002. Note that litterfall is plotted on a log scale to illustrate large inputs of litter during hurricanes and to show the response of fruitfall.

Fig. 2.3.3. Decomposing logs have a significant effect on soil carbon (Lodge et al., in prep.). Data compare logs felled in Hurricane Georges (G1 - G4; 0.6 yr prior to sampling) and Hurricane Hugo (H1 - H4; 9.6 yr) and show significant increases in soil carbon under logs vs. soils sampled nearby. There were parallel differences in total nitrogen and soil microbial biomass.

Workplan: vegetation - The principal goals of long-term vegetation measurements (Table 2.3.1) are to quantify the effects of and the responses to disturbance, and to provide information to interpret changes in faunal populations and ecosystem properties. Information from annual measurements of forest plots at Bisley is used to assess change at fine temporal scales. These data have been instrumental in demonstrating the rapid recovery of biomass and nutrient capital after Hurricane Hugo (Scatena et al. 1996). Shortly after Hurricane Hugo, we established the 16-ha Luquillo Forest Dynamics Plot (LFDP) at El Verde to track vegetation changes at fine spatial scales. In the LFDP, all trees and shrubs ³ 1.0 cm dbh (c. 130,000 stems) have been marked, measured, and mapped (Thompson et al., in press a). The size of the LFDP is necessary to understand the dynamics of a species-rich community in a heterogeneous landscape. Three censuses of the LFDP (the third with Mellon Foundation support) have given us an unprecedented, detailed record of stability and change in a tropical forest subjected to two severe hurricanes (Fig. 2.3.4; Brokaw 1998, Brokaw et al., in press) and various human land uses (Zimmerman et al. 1994, Willig et al. 1996, Thompson et al., in press b). Results from the LFDP underscored the long-term persistence of secondary species after land use, which has important implications for detrital processing, because primary and secondary species differ in decomposition rate. Plant responses also are measured over the long term in 20 landslides throughout the LEF (Walker et al. 1996b; Fig. 2.3.5) and in two clearcuts at Bisley (Scatena et al. 1993, Silver & Vogt 1993, Silver 1994, Silver et al. 1994).
In addition to continuing long-term vegetation measurements, we will initiate measurements to determine long-term patterns of herbivory in two ways. As in most tropical forests, herbivores mainly eat young leaves (Coley & Barone 1996), which flush during May and

Fig. 2.3.4. Results from the 16-ha Luquillo Forest Dynamics Plot showing (lefthand graph) the stability of species composition in response to Hurricane Hugo (1989). Note the exceptional increase in the abundance of the pioneer Cecropia schreberiana (Brokaw 1998). The temporal response to hurricane damage is in contrast to the legacy of human disturbance evident in species distributions (right-hand graph). The northern (upper) portion of the plot was clear-cut in the 1920s and the present-day forest is dominated by Casearia arborea, a successional species. The lowermost portion of the plot was subject to selective logging and maintains a near native composition of species, including tabonuco (Dacryodes excelsa; Thompson et al., in press b).

Fig. 2.3.5. A model of landslide succession for the Caribbean developed by Walker et al. (1996) from long-term and retrospective studies showing the importance of slide stability (dashed vs. solid lines) and soil organic matter for the rapidity with which landslides become revegetated.

June in the tabonuco forest or following disturbance (Angulo-Sandoval & Aide 2000). We will measure percent herbivory on new leaves of focal species marked each year in May and June. The second measurement will be of inputs of green leaf litter and insect frass to the forest floor. These measures will be used to gauge the changes in rates of herbivory during recovery from disturbance, with the ultimate goal of evaluating the role of herbivory in succession.
Workplan: fauna - Long-term measurements continue to address questions of how animal and microbial communities respond to disturbance (Covich et al. 1996, Secrest et al. 1996, Woolbright 1996, Gannon & Willig 1997, Huhndorf & Lodge 1997, Reed 1998, Willig et al. 1998, Pyron et al. 1999, Lodge et al. 2001). Surveys of stream decapods, mollusks, birds, frogs, lizards, and arthropods are coordinated with plant surveys in the LFDP and Bisley to determine the importance of habitat structure and resource availability in faunal response to disturbance (Table 2.3.1). Key terrestrial detritivores are monitored in both the LFDP and Bisley watersheds (Willig & Camilo 1991, Secrest et al. 1996, Willig et al. 1998) and, since 1998, in a series of plots surrounding the LFDP (Klawinski, unpub.). Detritivores in aquatic habitats (shrimp) have been monitored since 1988 (Covich et al. 1991, 1996). These measurements will be continued under this proposal.

2.4 Experiments to Test how Disturbance Alters Detrital Dynamics in Tabonuco Forest

2.4.1 Background and approach: Results from observations of disturbance and response in LTER 1 and 2 define the gaps in our knowledge of tabonuco forest and establish a priority for experiments to be conducted during the next four years. Strong hurricanes reposition living plant materials (e.g., leaves, flowers, fruits, and branches) from the forest canopy to the forest floor (Zimmerman et al. 1994, 1995b) increasing the levels of detritus on the ground. Light level, soil moisture, and temperature simultaneously increase after disturbance (see Prior Results for more details) and may interact with the pulse of detritus to affect subsequent community and ecosystem changes. For example, increased light promotes the establishment of pioneer and shrub species, but the litter deposited by the hurricane inhibits the germination of these same species (Guzmán-Grajales & Walker 1991). Populations of herbivores, detritivores, and decomposers also respond to both altered microclimate and shifts in resource availability (Waide 1991b, Covich et al. 1994, 1996, Lodge 1996, Zimmerman et al. 1996, Willig et al. 1998). Nutrient pools change dramatically over a period of 6 to 18 months as a function of the rapid decomposition of high-quality leaf litter, decreased uptake because of fine root mortality, and microbial immobilization (McDowell et al. 1996, Silver et al. 1996, Schaefer et al. 2000). Over the long-term, decomposition of low quality CWD and regenerating vegetation regulate nutrient losses from the forest system (Zimmerman et al. 1995b, Scatena et al. 1996). Hence, the two primary effects of hurricane disturbance, changes in microclimate and redistribution of biomass, propagate through the system in complex ways.
A fundamental priority for future research becomes clear from examining results from LTER 1 and 2: the need to assess the independent effects of detrital inputs, microclimate, and different functional groups of invertebrates in detrital processing after hurricanes. This assessment will depend on experimental manipulations and modeling that focus on predicting how changes resulting from disturbance interact to determine the subsequent behavior of the system. Our present understanding of hurricane impacts comes from measurements of the effects of naturally occurring hurricanes on tabonuco forest and comparisons with similar disturbances in other forests (Walker et al. 1991, 1996a). These measurements are informative but cannot tease apart the effects of various aspects of hurricane disturbance and suffer from the lack of a control or reference condition. Experimental and modeling approaches described below complement the long-term measures described in Section 2.3.
We will establish an experimental arena where the primary hurricane effects - changes in microclimate and redistribution of the forest canopy to the forest floor - can be studied independently. Long-term manipulations will increase the frequency of hurricane effects against the background levels of natural disturbance to test predictions about the impacts of an increased disturbance frequency on species composition, productivity and storage and export of C and nutrients (Sanford et al. 1991). We also will establish short-term experiments to determine the role of different groups of invertebrate detritivores in processing detritus on the forest floor and in streams, and we will measure interactions between detrital and autotrophic food webs.
Results from biotic manipulations will be used to develop a Trophic Interaction Model (TIM) for tabonuco forest that will incorporate the role of biota in the processing of detritus on the forest floor and in streams. TIM emphasizes the roles of functional groups in decomposition and other ecosystem processes. There is growing awareness that the autotrophic food web (primary producers, herbivores and their associated carnivores) and detrital food web (detritus produced by the autotrophic food web, decomposers and their associated carnivores) mutually influence each other and affect ecosystem processes (DeAngelis 1992, Adams & Wall 2000, Hooper et al. 2000, Palmer et al. 2000). Thus, combining autotrophic and detrital food webs into integrated models (such as TIM) incorporates the ecological interactions created by biodiversity into ecosystem functioning (nutrient release, primary production, C storage). This type of model has provided exciting new insights into how biota accelerate or decelerate ecosystem processes (DeAngelis 1992, Pastor & Naiman 1992, Jefferies et al. 1994, Polis & Strong 1996, Polis et al. 1996, Pastor & Cohen 1997, Belovsky & Slade 2000, Ponsard et al. 2000, Crowl et al. 2001).

2.4.2 Experiment 1 - Canopy Trimming Experiment: This long-term experiment will increase the frequency of simulated hurricane effects above background levels to once every six years. The experiment will determine effects of repeated disturbance of the forest canopy and increased detrital inputs to the forest floor on germination, growth, survival, nutrient cycling, soil conditions, and trophic structure. Climate change models predict increased frequency and intensity of Caribbean hurricanes (Emmanuel 1987, Goldenberg et al. 2001), and our goal is to evaluate predictions regarding the effects of an increased rate of hurricane disturbance on tabonuco forest (Sanford et al. 1991). The experiment also is designed to decouple the effects of canopy disturbance (e.g., light levels, temperature, moisture, etc.) from those of increased detrital inputs on rates of detrital processing and resultant community and ecosystem processes. Manipulations and measurements of detrital processing, SOM, and soil properties associated with SOM will continue for at least three more funding periods (until 2024). In the short-term, we will use faunal manipulations nested within the canopy trimming experiment to measure the strength of interactions between autotrophic and detrital food webs in the context of hurricane-like disturbance. These results will be directed specifically at parameterizing the Trophic Interaction Model (see below). This experiment also will provide a physical and intellectual focal point for the project participants.
The Canopy Trimming Experiment has two parts: 1) a forest canopy manipulation with measurements of coupled changes in microclimate, structure, and biota and their associated impacts on ecosystem processes, and 2) specific manipulations of the biota that assess the importance of components of the food web. Measurements of ecosystem parameters within the experimental treatments will address the following hypotheses:

Hypothesis 2: Short-term dynamics of key response variables after disturbance will be a function of the interaction between microclimate and detrital inputs, whereas long-term dynamics (particularly of SOM and NPP) will be a function of detrital inputs. (All)

Rationale - Zimmerman et al. (1996) summarized short-term (5 yr) patterns in response to hurricanes (Fig. 2.4.1) but were unable to establish causality for many of these patterns because of the lack of controlled experiments. For example, net primary productivity following Hurricane Hugo was augmented by regeneration of pioneer species responding to high light levels (Scatena et al. 1996), but the effect was reduced to some degree by nutrients immobilized in decomposing woody debris (Zimmerman et al. 1995b) and low capacity for uptake because of root mortality (Silver & Vogt 1993). Decomposition of high quality, hurricane-generated leaf litter, which should decompose rapidly, was slowed by drier post-hurricane conditions (Ostertag et al., in revision). Populations of snails initially declined in response to changed microclimate then rebounded quickly, apparently in response to increased ground cover and regeneration of the understory (Willig & Camilo 1991, Secrest et al. 1996, Willig et al. 1998). These examples (and others; Zimmerman et al. 1996) suggest that interactions between microclimate and detrital pulses are important for short-term forest dynamics and emphasize the necessity of controlled experiments to decouple their effects.
Long-term predictions are derived from the CENTURY model, which was parameterized for the tabonuco forest (Sanford et al. 1991) using detailed information on carbon and nutrient stocks in above- and belowground biomass (Odum & Pigeon 1970). Simulations with hurricanes rates, and organic soil P. A hurricane-disturbed forest was predicted to have higher overall productivity than a forest without disturbance (Sanford et al. 1991). The Canopy Trimming Experiment will allow us to test the contention (Sanford et al. 1991) that increased forest productivity in hurricane disturbed forests is the result of increased available levels of nutrients mineralized from high SOM rather than via effects on the canopy, i.e., via the maintenance of a young, open-canopy, fast-growing forest.

Fig. 2.4.1. Idealized 5 yr trajectories of responses of different components of wet subtropical forest in the Luquillo Experimental Forest, Puerto Rico, to disturbance caused by Hurricane Hugo (from Zimmerman et al. 1996). The shaded portion of the graph represents +/- 15 percent variation indistinguishable from pre-hurricane values. Curve A: transient (<1 yr) increase. Examples include forest floor biomass, some soil nutrient pools, and nitrate concentrations in streams. Curve B: slow increase and return to pre-hurricane levels. Examples are net primary productivity, abundance of Atya shrimp in streams, of adult coqui frogs, and of the terrestrial snail, Cepolis squamosa. Curve C: catastrophic decrease and subsequent rise above pre-hurricane levels. Examples are aboveground pools of potassium and magnesium and of several species of terrestrial snails. Curve D: catastrophic decrease and return to near pre-hurricane levels. Examples are tree biomass and tree density. Curve E: a catastrophic decline and steady increase, but not to within 15 percent of pre-hurricane levels. An example is fine total litterfall. Curve F: catastrophic decline and little recovery until 5 yr post-hurricane. Examples are fine root biomass and abundance of the walking stick, Lamponius portoricensis. Both of the latter returned to near normal values subsequently.

Workplan: forest canopy and structure manipulations - Four blocks of four plots (30 x 30 m) will be identified on the basis of similar slope, soil characteristics, and forest canopy species composition; aspect will be the blocking variable. This plot size was chosen to provide sufficient space for long-term monitoring of plot responses as well as the biotic manipulations discussed below. This level of replication has proven sufficient to measure ecosystem responses to manipulations of detritus in previous experiments in tabonuco forest (Zimmerman et al. 1995b, Walker et al. 1996b). Two plots within each block will have the branches of the canopy trees trimmed by a professional arborist to open the canopy, which is roughly equivalent to the action of a hurricane with sustained winds of more than 150 kph. The recurrence interval for hurricanes of this strength is 55-60 yrs for the LEF (Scatena & Larsen 1991). The other two plots will not experience canopy manipulation and will be subject to normal hurricane frequency. This experiment will be initiated in year 2 of funding because of the need to coordinate the logistics of an experiment of this scale and conduct one year of pre-manipulation measurements.
The experimental manipulations will create four treatments in a 2 x 2 factorial design in each block:

(1) Canopy trimmed and removed biomass distributed on forest floor. This simulates the changes in microclimate (openness) and structure (redistribution of biomass) created by a hurricane;
(2) Canopy trimmed and removed biomass eliminated. This simulates the changes in microclimate (openness) created by the hurricane without the associated change in forest structure (redistribution of biomass);
(3) Canopy untrimmed with canopy biomass from a trimmed plot distributed on the forest floor. This simulates the changes in forest structure (redistribution of biomass) created by the hurricane without the associated change in microclimate;
(4) Canopy untrimmed and no canopy biomass added to forest floor. This maintains the forest unmodified by hurricane disturbance.

Within each plot measurements will be made within a core (20 x 20 m) area to minimize edge effects. The measurements and their frequency are listed in Table 2.4.1. Sampling for some variables will be quarterly for one year prior to and after the canopy manipulations and annually thereafter until the next scheduled manipulations. The inner plots will be divided into 1 x 1 m quadrats and randomly allocated to be measured for either plant or invertebrate densities, or for soil characteristics (10 quadrats each; Guzmán-Grajales & Walker 1991, Walker et al. 1996b). Vertebrates will be recorded using the entire inner plot. Litterfall will be measured using baskets placed 1 m above the ground. Litter samples will be collected biweekly. Root production and turnover will be measured using sequential ingrowth cores. Soil nutrient pools and fluxes will be measured using standard methods (Robertson et al. 1999). Export of nutrients will be measured using soil solution collected from tension lysimeters. Statistical analyses will be made using repeated measure MANOVA by block. Measured ecosystem variables (Table 2.4.1) will be compared to CENTURY predictions (see below).
In addition to distinguishing the relative importance of microclimatic changes versus detrital inputs in determining responses of tabonuco forest to severe hurricane damage (Zimmerman et al. 1996), this experiment will also determine the contribution of decreased nutrient uptake to the flush of nutrients into groundwater and streams (McDowell et al. 1996, Schaefer et al. 2000) by isolating two potential causes of fine root mortality, lack of carbohydrate from leaves versus soil drying (a third being the physical disruption of root systems; Silver & Vogt 1993).
Canopy manipulations will be repeated every six years. This interval was chosen because it is approximately ten times the average frequency of storms of this magnitude (Scatena & Larsen 1991) and many relevant ecosystem parameters return to near pre-hurricane values after five years (Zimmerman et al. 1996).

Table 2.4.1. A list of measurements to be made in the Canopy Trimming Experiment and associated food web manipulations.

Two sets of biotic manipulations will be conducted within the main treatments of the Canopy Trimming Experiment: one will manipulate detritivory (detrital-based food web) and the other will manipulate herbivory (autotroph-based food web). These experiments will employ small-scale mesocosms and removal of organisms to isolate the potential effects of food web components thought to influence ecosystem processes. Because of the intense maintenance requirements of these manipulations, they will be conducted for only two years.

Hypothesis 3: The absence of invertebrate detritivores will have strong effects on detrital dynamics, retarding decomposition rates and related processes. Microclimatic changes associated with canopy opening will reinforce these effects, but the addition of detritus will buffer the effects of canopy opening. (Belovsky, Crowl, González, Klawinski, Lodge, Willig, Zou)

Rationale - Cross-site studies involving LUQ show that removal of microarthropods from the litter layer (Heneghan et al. 1998a) and litter microarthropods and macrofauna from the soil (González & Seastedt 2001, Liu & Zou, in press) significantly slows litter decay rates in terrestrial ecosystems. González and Seastedt (2001) reported that faunal effects on litter breakdown can reach 66% in tabonuco forest. We believe that these effects are important under typical conditions, but that they will be modified by changes due to disturbance. Specifically, we predict a three-way interaction between the main effects of the Canopy Trimming Experiment and the presence/absence of invertebrate detritivores, as stated in the hypothesis.
Workplan - Detritivores will be manipulated with selected measurements from Table 2.4.1 made in each subplot: [Note: protocol was changed in December 2003; naphthalene and eletroshocking are not being used in this experiment.]

(1) Invertebrate detritivores excluded - microbial decomposition only. Subplots (2 x 2 m) will be trenched and barriers placed to prevent earthworm immigration. Earthworms will be eliminated from subplots by electroshocking (Liu & Zou, in press). Naphthalene will be placed at ground level in each subplot to exclude arthropod detritivores (González & Seastedt 2001). These subplots will be isolated in corners of the inner plots at a distance of 5 m from monitoring areas and other experiments. The experiment will eliminate essentially all invertebrate detritivores (and understory herbivores) to examine microbial decomposition of litter in the context of the Canopy Trimming Experiment.
(2) Invertebrate and microbial decomposition. Subplots will be trenched and barriers placed but earthworms will not be removed nor will naphthalene be applied. These subplots will be placed in the remaining inner plot corners.
Data from the experiments on the detrital-based food web will be analyzed using repeated measures MANOVA by block. Estimates for particular effects will be used to develop modeling parameters (see 2.4.4).

Hypothesis 4: Presence of herbivores will significantly alter patterns of detrital processing by differentially reducing the abundances of early successional plant species. This effect will be most pronounced under open canopy conditions. (Belovsky, Crowl, Waide, Willig)

Rationale - Considerable recent evidence suggests that interactions between autotrophic and detrital food webs can strongly influence fluxes of C and nutrients (DeAngelis 1992, Pastor & Naiman 1992, Adams & Wall 2000, Hooper et al. 2000, Palmer et al. 2000). This experiment specifically addresses the impact of species with low- vs. high-quality litter on detrital dynamics, and how this impact varies with the presence of herbivores. Following disturbance, and in the absence of herbivores, fast-growing plant species will dominate and their high quality litter should prime decomposition and rates of nutrient cycling, producing a positive feedback to plant growth. Where present, herbivores should preferentially consume high-quality leaves and hence reduce the average quality of litter inputs. Under open canopy conditions, where the difference in growth rates between fast- and slow-growing species is greatest, the effect of herbivores on detrital processing should be disproportionately greater.
Workplan - Herbivores and the plant community will be manipulated in screened cages (2 x 2 x 3 m). Each plot will contain four cages, randomly located but at least 5 m from the detrital-based food web experiments to prevent interference. The litter and soil down to 2 cm will first be cleared and homogenized between plots within the block and then redistributed in each cage to minimize variability. The caged area also will have understory vegetation removed and will not contain tree trunks. Crossed wooden walkways will be placed within each cage, for access with minimal disturbance to the soil. Cage roofs will be constructed to deflect litter input from the canopy. Thus, subsequent litter inputs will come only from the artificial community and any consumers that might be present.
Within each cage a simplified understory community of two species will be transplanted from nearby forest. Piper glabrescens is a common, fast-growing shrub in closed canopy forest and openings. Manilkara bidentata is a common, slow-growing canopy tree. Individuals of each species approximately 1 m tall will be planted within each subplot and measured for size (total branch length and leaf number) as an estimate of aboveground biomass. We will manipulate numbers of an herbivore, the walking stick Lamponius portoricensis. These are common herbivores in the understory of tabonuco forest and are generalists feeders that nonetheless exhibit a preference for P. glabrescens and are present throughout the year (Willig et al. 1993). Cages (two in each plot) will be stocked with adult individuals of an average size to achieve the average forest density and sex ratio (Willig et al. 1993). Initial stocking densities are at the forest average, because this would be the number at the time that a hurricane strikes. Walking sticks will be surveyed in the cages every month and the numbers will be reduced or augmented to maintain experimental densities. Two cages will contain no herbivores, to measure ecosystem processes in the absence of any consumers. The autotroph-based food web experiment will include the measurements listed in Table 2.4.1.
The experiments with the autotroph-based food web will be run for two years and will be analyzed using repeated measures MANOVA by block. Estimates from post-hoc comparisons for particular effects will be used to develop modeling parameters for TIM (see below).

2.4.3 Experiment 2 - Detritivore Functional Group Experiment: The contributions of different groups of invertebrates to decomposition and nutrient mineralization after hurricanes are not fully understood. In stream ecosystems, these two groups of decapod invertebrates are necessary for the complete processing of hurricane-generated detritus (Covich et al. 1999, Crowl et al. 2001, March et al., in press). Terrestrial ecosystems have groups of detritivores that are analogous in function to stream invertebrates (Schowalter 2000). Experimental manipulations in which we tease apart the independent effect of particular groups of the biota in aquatic and terrestrial ecosystems are proposed to address two hypotheses:

Hypothesis 5: Within ecosystems: decomposition rates will be most rapid in the presence of all detritivore functional types. The effect of excluding functional groups will vary depending on the group excluded. (Crowl, González, Klawinski, Pringle, Ramírez)

Hypothesis 6: Between ecosystems: exclusion of analogous functional groups will have parallel effects on decomposition rates in the two ecosystems. (Crowl, González, Klawinski, Pringle, Ramírez)

Rationale - Diverse assemblages of invertebrates consume disturbance-generated detritus on the forest floor and in streams in tabonuco forest. In tropical headwater streams, shrimps are often the main group of detritivores, followed by crabs and insects (Covich & McDowell 1996, Covich et al. 1999). Variation in shrimp densities among streams results in large differences in particle export to downstream reaches at landscape levels (Pringle et al. 1999). Two shrimp species, Xiphocarus elongata and Atya lanipes, are important processors of detritus in headwater streams (Crowl et al. 2001, March et al., in press). In terrestrial ecosystems, millipedes and isopods are mainly responsible for the fragmentation of litter, while mites and collembolans are important microbial grazers (Schowalter 2000). As in the stream, invertebrates play a significant role in litter decomposition in terrestrial ecosystems (Heneghan et al.1998a, González & Seastedt 2001, González et al. 2001), but we lack information on the importance of particular functional groups of detritivores (e.g., fragmenters, grazers) in terrestrial ecosystems. Furthermore, our past studies of aquatic ecosystems have only considered species of decapods and not other detritivore groups (e.g., insects). This set of experiments will investigate the relative importance of different functional groups of invertebrates in detrital processing in terrestrial and aquatic environments of the tabonuco forest. The experiments are based on a single pulse of detritus as occurs after a hurricane and provide a base for comparing detrital food webs between terrestrial and stream habitats. The results from this experiment will allow us to investigate the utility of TIM at finer scales of biotic resolution (i.e., among detritivores).
Workplan - In terrestrial habitats we will create a series of enclosures (2 x 2 m; Lawrence & Wise 2000) that will limit the passage of litter invertebrates. We will extract litter invertebrates from each of these plots and then recolonize them with normal densities of combinations of fragmenters and microbial grazers (see Table 2.4.2 for details). Mites and collembolans are grazers, and millipedes and isopods are fragmenters. These are the four most numerically dominant groups of detritivores in tabonuco forest litter (Pfeiffer 1996). All enclosures will be covered with window screening to prevent passage of invertebrate detritivores. Five replicates of each treatment will be constructed, and 12 litterbags will be placed in each enclosure. The bottom layer of these bags will be made from fiberglass window screening (1 mm mesh), and the top layer of these bags will be 0.5 cm plastic mesh. This will
allow the entry of larger detritivores. These bags will contain either 5 g of air-dried, fresh leaves of Cecropia schreberiana or Dacryodes excelsa (tabonuco), common fast- and slow-growing species, respectively, that have been used previously (Crowl et al. 2000).
One bag per month will be removed at random from each enclosure. We will extract invertebrates from the bags and then oven-dry the litter and weigh, grind, and determine its ash-free dry weight. Subsamples will be analyzed for nutrient content and secondary chemistry. The results will allow us to determine the degree to which proposed functional groups affect rates of decomposition individually and in combination. Because of the large number of potentially important decomposer groups and their interactions, logistic considerations do not allow us to consider more than a single disturbance and two litter types in this initial experiment.
To determine the relative roles of the different functional groups of stream detritivores (shrimp, insects, crabs), we will use multiple exclusion methods at multiple spatial scales. This study builds on our collective experiments in which we have used a combination of whole-pool manipulations (Crowl & Covich 1994, Crowl et al. 2001), electrical exclusion patches (March et al. 2001, in press, Pringle 1996, Pringle et al. 1999, Ramírez 2001), and litter bags (Ramírez 2001) to exclude insects, shrimps, and crabs in a nested design (Table 2.4.3). To exclude all invertebrate taxa, we will use fine mesh litter bags that prevent insect colonization. The insects-only treatment will consist of coarse mesh litter bags placed within an electrified fence.

Table 2.4.2.Summary of the Decomposer Functional Group Experiment. The design for the experiment parallels the aquatic study of Crowl et al. (2001) by excluding grazers, fragmenters and both in a 2 x 2 factorial design.

Decapods will be excluded by a combination of electricity and manual removal from fenced pools. We will employ a split-plot design to minimize the total number of pools necessary. Bags will contain 5 grams of fresh C. schreberiana or D. excelsa litter (Crowl et al. 2001). Leaves will be prepared by the methods described in Heneghan et al. (1998a) and removed from each replicate on weeks 2, 4, 8, 16, and 32 or until litter has been completely decomposed within any treatment plot. Upon removal, invertebrates will be extracted from the litter, oven-dried, weighed, and ground, and ash-free dry weight will be determined. Subsamples also will be analyzed for nutrient content.

2.4.4 Synthesis and integration: We are working with two models to interpret the results of the experiments described above, CENTURY and TIM. These models take complementary approaches to understanding ecosystem function. CENTURY is a linked production-decomposition model operating at relatively large scales that aggregates much biotic detail. TIM is a food chain model operating at relatively small scales that explicitly considers effects of functional groups thought important to ecosystem function. The first tropical version of CENTURY was developed for tabonuco forest (Sanford et al. 1991) because it was the only tropical site where detailed data on C and nutrient stocks in above- and belowground biomass had been gathered (Odum & Pigeon 1970). This model greatly aided previous experimental studies of detrital effects on ecosystem productivity in the LEF (Zimmerman et al. 1995b) and has been extended to the entire LEF landscape using GIS and other models we developed (Wang 2001, Wang et al. in press). LUQ researchers are currently working with the developers of CENTURY to improve the P submodel (Silver et al. 2000) and to incorporate variable soil moisture. The refinements of P-cycling in the model are important to this proposal because Sanford et al. (1991) emphasized mineralization of elevated organic P to explain increased NPP after repeated hurricane disturbance. Comparisons of predictions of measured variables in the Canopy Trimming Experiment using the revised version of CENTURY should greatly enhance our understanding of the treatment effects and the utility of the model for understanding disturbance effects on production and decomposition in wet tropical forests.

Table 2.4.3. Summary of the Aquatic Decomposition Experiment. The design builds on previously reported experiments (Crowl et al. 2001, March et al. 2001) and will use a nested design with insect exclusions (fine/coarse litter bags) and macro-fauna exclusions (electric exclosures) nested within whole pool manipulations. At the whole-pool level, shredders and filter feeders will be excluded in 2 x 2 factorial design.

_____________________________________________