Coweeta LTER Research Program

Current Research 2002-2008

Current Initiative II: Ecosystem Responses to Variation in the Socio-Natural Template

C-IIA. Baseline Conditions of Stream Fish.
Summary: To assess and predict the effects of land-use change on biota first requires quantifying natural levels of variation in organisms, and determining the factors that affect population size in the absence of land-use change.

We propose determining the relative importance of density-dependent and density-independent processes to long-term population regulation and demography of fishes by continuing our long-term quantitative fish population sampling in the three permanent Coweeta gradient sites (3rd - 5th order streams). We will match the results to removal experiments and, resources permitting, construct a model of fish population responses to environmental change. This study parallels research in II.C.4 addressing similar questions about forest tree assemblages. Fish populations and habitat availability will be sampled through 2005 using the methodologies of Freeman et al. (1988) and Grossman and Freeman (1987). This round of sampling will yield up to 4 generations of data for the dominant species at the three sites, and 14 will thus represent one of the longest quantitative data sets on fish population assemblages. We will use Akaike’s Information Criterion (Anderson et al. 2000, Burnham and Anderson 2001) to quantify the individual and combined effects of density-dependent (e.g. resource limitation) and density-independent (e.g. natural disturbance) processes on data patterns. This technique avoids many of the statistical pitfalls present in descriptive studies of population regulation (Anderson et al. 2000, Burnham and Anderson 2001), but if possible we will also examine our data using hierarchical Bayesian models as proposed in II.C.4 below. Manipulative experiments (e.g., removals) will be designed to test for the relative importance of density-dependent vs. density-independent forces after the long-term descriptive data patterns are analyzed. With descriptive and manipulative experiment results we will be able to construct environmentally-based population models (Individual-Based Models similar to Jaeger et al. 1997) to forecast the response behavior of fish populations to changes in environmental regimes and disturbance frequency.

C-IIB. Understanding The Role of Sedimentation.
Summary: Our research will evaluate responses at the assemblage, population, and organismal levels, as well as the concomitant effects of temporal and spatial heterogeneity.

A unifying theme of both present and historic research in the Coweeta LTER Program is the quantification of ecological responses to natural and anthropogenic disturbance on levels ranging from the organism to the ecosystem. The stream flora and fauna of southern Appalachia evolved in predominantly forested systems with shaded streams that were clear, cool, and unproductive. Sediments were well sorted and primarily derived from instream sources and hillslope failures (i.e., debris flows and landslides). This changed as humans altered their land-use patterns on both local and regional levels, so that increased sedimentation from surface runoff is now one of the most deleterious impacts on aquatic ecosystems (Waters 1995). Current species composition furthermore depends on sedimentation legacies, not only on present sedimentation patterns (Harding et al. 1998, Scott and Helfman 2001). In a series of integrated studies we propose quantifying land-use influences on historic and present patterns of erosion and sedimentation, as well as the spatial extent and mechanisms driving responses of organisms and ecosystems to this anthropogenic disturbance. Our research will evaluate responses at the assemblage, population, and organismal levels, as well as the concomitant effects of temporal and spatial heterogeneity. Results from this integrated set of sedimentation studies will provide a key input to our Initiative 3 forecasting efforts. Our general hypothesis is that changes in land use produce predictable changes in sedimentation that then drive biotic processes in watersheds where the change is occurring.

We will test the following specific hypotheses:
1. Process-based, spatially-explicit soil erosion and transport models can accurately predict changes in suspended sediment yield caused by changes in land use.

2. Incorporating historic land use will improve our ability to explain current suspended sediment yields in southern Appalachian streams.

3. Changing land use patterns cause altered in-stream habitat (particularly sediment amount and distribution), which leads to altered fish assemblages resulting in part from upstream invasion by widespread generalists.

4. Increases in suspended sediments produced by changes in land use will reduce foraging success, energy accumulation, and population size of native water-column fishes.

C-IIB1. Sediment Erosion and Yield Through Time.
Summary: We will quantify the impacts of human land use on sediment generation and input in decadal time steps from 1910 to the present through a series of field-measurement and fine grained, process-based sediment models.

Although sedimentation is the primary source of anthropogenic changes in stream ecosystem structure and function (Harding et al. 1998), our ability to quantitatively predict sediment erosion, transport, and deposition is poor. Swift [, 1988] found that soil erosion from roads introduced significant amounts of sediment to 15 streams, and Swank et al. (2001) attributed large increases in soil erosion and stream sedimentation to human land uses, particularly agriculture. Bolstad and Swank (1997) found that as streams progressed through increasingly developed watersheds, stream turbidity also increased leading to important reductions in water quality during storm flow (Figure 5). However, we know of no attempts at process-based predictions of the cumulative temporal and spatial effects of multiple land uses and land-use change on sedimentation for the southern Appalachian region. A century of theoretical and empirical studies have identified the mechanisms by which sediments are eroded and transported to streams: Soil particles are dislodged by raindrop impacts and fluvial shear, then incorporated to surface runoff and transported down slope. Sedimentation processes are well understood in three-dimensional (two spatial + time) and four dimensional (three spatial + time) systems, but these processes have rarely been incorporated into sediment generation and transport models due to complexity, computational limitations, and ill conditioning (Flanagan and Nearing 1995). Standard sediment models use explicit estimation of shear stress and resistance to generate erosion and transport rates. However, this requires spatially-detailed information for variables that are typically sparsely sampled at inappropriate spatial resolutions. Land use, soil characteristics, climate, and slope drive sedimentation, but are currently only known at coarse resolutions over broad scales, e.g., 30-meter horizontal sampling interval. Such constraints lead to the use of lumped parameter models (e.g., the Universal Soil Loss Equation, USLE) and its derivatives (the revised and the modified USLE). However, the numerous limitations of lumped parameter approaches restrict our ability to predict past and current sediment inputs at the accuracy and spatial detail needed for robust characterizations of past, current, and future sediment impacts on stream ecosystems. We will quantify the impacts of human land use on sediment generation and input in decadal time steps from 1910 to the present through a series of field-measurement and fine grained, process-based sediment models. We hypothesize that process-based erosion, and transport models will provide accurate estimates of sediment delivery to streams, and correlate with observed sedimentation rates. By quantifying the effects of historic and contemporary land use on sediment inputs to southern Appalachian watersheds we can investigate the consequences of these changes for stream communities as well as the bioenergetics and behavior of individual species. Figure 5. Sediment inputs are higher with increased building densities and result in more turbid waters, especially during storm flow conditions (Bolstad and Swank 1997, Figure 3). 16 We will generate data for eight watersheds spanning the range of current and past land use conditions (e.g., from minimal to substantial agricultural and urban land uses). Sites will be co-located in the Little Tennessee and French Broad Rivers with other proposed studies; the selection of study sites will be guided by the historic land use coverages developed in II.B.1 and additional factors (e.g., current land use, stream size, gradient, elevation, and geology). We will measure surface transported sediment using sediment traps and other standard methods, and instream turbidity using stage, flow, and frequency samplers. We will also work with cross-sectional and other stream morphometry measurements as outlined in II.B.3.c. The cartographic base for the study watersheds produced in II.B.1 will be augmented with fine-resolution terrain data (1-2 m) so we can represent near-stream depositional zones. Process-based models will be adapted to distributed network and grid approaches (Cochrane and Flanagan 1999), and applied to our study watersheds. Spatially-explicit land use data from each time period will be used to condition a landscape sediment model, and the resulting predictions will be compared to the sediment generation chronologies developed in II.B.3.c. Sedimentation chronosequences will be collected in each modeled watershed and use as independent checks on predicted sediment inputs.

C-IIB2. Homogenization of Stream Fish Assemblages.
Summary: We will examine the effects of land-use change on the characteristics of stream fish assemblages in the southern Appalachians.

The continuum of land conversion in the upland areas of the region corresponds to fish assemblages dominated alternatively by native specialists or introduced generalists. From an initial assemblage characterized by species specialized for life in relatively cool, clear, shallow, low nutrient, rocky and diverse habitats (e.g., darters, benthic minnows, sculpin, and brook trout), land conversion eventually leads to a mixed assemblage of introduced species tolerant of warmer, slower, nutrient-rich habitats with increased sediment loads (e.g., sunfishes, some suckers, some minnows, and perhaps some catfishes). The end-point of the shift in fish assemblage composition is regional and national homogenization (e.g. Rahel 2000). In 1997 and 1998 we sampled fish assemblage structure at sites in the Little Figure 6. Relative abundance of (A) highland endemic versus (B) cosmopolitan fish species as a function of land use intensity at 36 sites in the Little Tennessee and French Broad river basins (Scott and Helfman 2001, Figure 2). 17 Tennessee and French Broad drainages, and found no significant correlation between overall fish diversity and land-use intensity. However, a significant relationship was revealed by dividing the fish fauna into regional, highland endemics and widespread, generalist species (Figure 6 Scott and Helfman 2001, Scott et al. in review). The results indicate that upland areas of high endemism are being invaded, displaced, and homogenized by native species capitalizing on habitat degradation. We hypothesize that intermediate habitat conditions result from conversion of forestland to agricultural and suburban uses, and lead to progressive homogenization of the fish fauna (Figure 7). Large-scale homogenization is therefore dependent on summed events at smaller geographic scales so that our objective is to identify the intermediate steps facilitating the process of regional homogenization.

We will test our hypothesis of homogenization-via invasion in streams in the Little Tennessee and French Broad drainages across a gradient of landuse types that will include sites used in developing and validating the sediment generation model described in II.C.2.a. Assemblage characteristics, stream geomorphology, water quality and sediment distribution will be measured and our findings combined with results from the sediment generation model to determine how anthropogenic factors influence biotic responses that diminish the integrity of southern Appalachian stream communities.

C-IIB3. Effects of Suspended Sediments on Fish Foraging Success and Habitat Use.
Summary: We will examine whether increases in suspended sediments deleteriously affect the prey-capture success and ultimately habitat selection and population size of several common water-column species in the Coweeta, Little Tennessee, and French Broad Drainages.

Perhaps the most common impacts on streams of deleterious land-use change are increases in suspended and settled fine sediments (Waters 1995). Descriptive studies of stream fishes have detected an inverse correlation between fish diversity and siltation (Waters 1995). However, the mechanisms for these declines are unknown. The necessity for a mechanistic explanation for the relationship derives from the fact that increased siltation is certain to be a by-product of future land-use change in western North Carolina (e.g., increased urban and suburban development, and Figure 7. Postulated time course over which homogenization occurs in highland streams with increasing watershed disturbance (Scott and Helfman 2001, Figure 3). 18 agriculture).

Siltation can negatively affect fishes in a variety of ways including reductions in spawning habitat, decreased prey abundance, and decreased foraging success (Figure 8). We have previously shown that turbidity levels associated with frequent storm events in the Coweeta drainage can significantly decrease the foraging success of rainbow trout (Oncorhynchus mykiss) (Barrett et al. 1992). The longterm consequences of these decreases are unknown, but they are certain to result in decreases in individual fitness, which ultimately may result in decreased population size. We will use the methodology of Barrett et al. (1992) to determine whether increased turbidity significantly decreases the foraging success of several species of abundant drift-feeding fishes. We will probably select members of the Cyprinidae for which there is little extant data even though most southern Appalachian fishes belong to the order. We will choose 2-5 species that occur in sites to be used in validating the sediment generation model, and relate our results to a new optimalforaging habitat selection model developed for Coweeta fishes (Grossman et al. 2002). Time permitting, we will construct a bioenergetically-driven Individual-Based Model (Jaeger et al. 1997) to link decreases in foraging success to shifts in habitat selection and population persistence.

C-IIC. Understanding The Role of Organic Matter.
Summary: Understanding short- and long-term impacts of land use change on organic matter dynamics are critical to forecasting future impacts on ecosystem processes. We propose continuing a series of measurements in Big Hurricane Branch and Hugh White Creek (the reference stream) using the same methods we have used at monthly, annual, 5-year, and 10-year intervals since 1974 (Stone and Wallace 1998, Webster et al. 1999, Benfield et al. 2001, Webster et al. 2001).

Organic matter inputs to terrestrial and aquatic systems (senesced leaves, roots, and stems) represent significant additions of energy and nutrients into systems as diverse as primary and disturbed forests, suburban parklands, agricultural fields, and the managed plantings around shopping malls. Energy transfer and material recycling through litter inputs to soils and streams are among the most fundamental of ecological processes (Webster et al. 1995, Wallace et al. 1997). The two studies in this section examine the recovery of streams from significant changes in organic matter inputs from logging, and the interaction of litter quality, microclimate, and biota in soils and streams. II.C.3.a. Recovery of Headwater Streams from Logging. Watershed 7 (WS 7) was clear cut in 1975-76 and we have extensively studied the effect of logging on Big Hurricane Figure 8. Increased turbidity significantly reduces the distance at which foraging rainbow trout are able to detect prey (Barrett et al. 1992, Figure 1). 19 Branch, which drains WS 7. The recovery of Big Hurricane Branch subsequently became a core research activity of the Coweeta LTER Program, and after 15 y of study we made long-term (100+ y) forecasts of the trends in stream processes following forest disturbance (Webster et al. 1992). These forecasts were based on the strong connectivity of forest streams and their riparian vegetation, largely related to the importance of large wood. The measurements will be used to validate and refine our previous forecasts.

The specific re-measurement studies we proposed for Big Hurricane Branch and Hugh White Creek are:
1. Leaf breakdown in 2004-5. Leaf breakdown rates in Big Hurricane Branch are faster than pre-clear cut rates and reference site rates after 23 years post-clear cut (Figure 9).

2. Benthic macroinvertebrate production in 2003-4.

3. Particulate transport in 2004-5. Particulate transport was measured intensely in both streams in the 1970s (Gurtz et al. 1980) and the 1980s (Webster et al. 1990), but has not been measured since.

4. Allochthonous inputs and benthic organic matter standing crop, including wood, in 2005-6. 5. Continuous DOC export in 2002- 2008. We have measured DOC concentrations monthly since 1977. 6. Stream cross sections in 2006. Shortly after WS 7 was logged, we established permanent cross-sections on both streams to measure long-term changes in streambed morphology at approximately 10- year intervals.

We also propose continuing a study of Cunningham Creek where we initiated in 1986 a long-term experiment of logs in streams (Wallace et al. 1995). We propose to make during the current funding cycle annual measurements of benthic communities at three sites upstream and downstream of where logs were added in 1988, and one re-measurement of the decay status of these logs.

C-IIC1. Links Among Land-Use Change, Litter Inputs and Litter Processing.
Changes in land-use result in changes in litter inputs and litter processing. The most obvious are gross Figure 9. Predicted and measured standing crops of leaves in Big Hurricane Branch and Hugh White Creek in 1994-95, showing that both measured and simulated values are significantly lower in Big Hurricane Branch (Upper panel from Webster et al. 2001). 20 changes in the quantities of litter entering soils and streams as natural plant communities are replaced by such structures as homes, businesses, and parking lots. There are also subtle changes resulting from shifts in the quality of litter entering soils and streams, the microclimate in which litter is processed, and the communities of fauna and flora participating in decomposition processes. Models of nutrient and energy flux in developed landscapes based solely on quantities of litter input are likely to be misleading since they do not incorporate the interactive effects of litter quality, microclimate, and biota. We propose to combine monitoring and experimental procedures to explore the interactive effects of litter quality, microclimate, and biota on the decomposition of plant litter along gradients of land-use in southern Appalachia.

Our primary questions are:
1. Do the relative roles of litter quality, microclimate, and biota on rates of litter processing vary along land-use gradients?

2. How do interactions among these key variables change along land-use gradients?
With previous funding from an LTER Supplement and other sources we explored the effects of macroinvertebrates on decomposition processes in upland, riparian, and stream habitats at the Coweeta and the Luquillo LTER sites. We tested the general hypothesis that the exclusion of macroinvertebrates would have increasing effects on the decomposition process as the quality of leaf-litter increased. We developed the hypothesis from studies of insect folivores on green leaf tissue, and it appears to hold true for senesced leaves. For example, the reduction in decomposition rates of leaf material in streams at Coweeta and Luquillo following the exclusion of macroinvertebrates is most pronounced on high quality litter (Powell 2001). We have also shown that the effects of fauna on decomposition are habitat-dependent; for example, macroinvertebrates have a greater impact on the decomposition of oak litter in riparian zones than in upland forest (Hunter et al., unpublished data, Figure 10). Simply put, litter quality, fauna, and habitat-type matter to the decomposition process and interact in complex ways. We will select sites within and beyond the Coweeta basin that reflect current variation in land-use across the southern Appalachian region and include the following habitat types: (a) oak hardwood, (b) production agriculture, (c) suburban parkland, and (d) urban housing development. We anticipate selecting three replicates of each habitat-type for a total of 12 sites to examine decomposition patterns in soils and streams using established procedures. Natural variation in the quality of litter, C inputs, N, phenolics and lignin will be measured and key abiotic variables including temperature, moisture, pH, stream flow and sedimentation will be estimated. Macroconsumer access to litter will be manipulated by mesh size (terrestrial) and electric exclosures (streams). For our control, we will transfer a standard Figure 10. Reduction in decomposition rates of leaf material in streams at Coweeta and Luquillo following the exclusion of macroinvertebrates is most pronounced on high quality litter. 21 reference litter to each field site, and a natural litter from each site to both a “common garden” and a “common stream” site.

C-IID. Climatic and Site Controls of Forest Form and Function.
Predicting forest response to global and regional change requires a mechanistic understanding of environmental controls on plant productivity, demography, diversity, and ecosystem function. Environmental factors such as available moisture, light, and soil nutrients influence plant growth and productivity. These factors vary geographically with regional climate and are modified locally by elevation and topographic position. Natural and anthropogenic disturbances may further alter local conditions, and the effects of past land use on soils and vegetation may persist for decades. Enhanced understanding of population- and ecosystem-level responses to environmental variability is needed to anticipate responses of southern Appalachian ecosystems to climate and land-use change. We propose an integrated set of experimental and observational studies to understand how variation in climatic and site characteristics controls productivity, overstory tree demography, understory herbaceous diversity, microarthropod diversity, and ultimately forest form and function. Studies will be conducted in three landscapes that span a regional climate/elevation gradient (Table II.3): Nancytown, GA (~350 m, 34o30' N), and the Coweeta Basin (~1000 m, N35o03' N), and Mars Hill, NC (~730 m, 35o45' N). The landscapes are located along north/south climatic and bio-geographic gradients in the Southern Blue Ridge. Variation in temperature and precipitation among these three landscapes will permit us to conduct research under different climatic regimes. However, climatic effects may be mitigated by site-level edaphic factors such as light availability, soil moisture, and/or nutrient levels. Moreover, these edaphic factors are correlated with topographic position and land use history. By selecting study plots within landscapes to vary edaphic factors and by experimental manipulations, we will measure the interaction between broad-scale climatic gradients and fine-scale, local edaphic variation. Within each landscape, replicated measurement plots will be selected and used in the research activities described below to analyze above- and belowground interactions. Each plot will vary in topographic position (i.e., ridge or cove). To address land use history, paired study plots differing in land-use history will be located in cove sites representing previously-farmed forested sites and non-farmed forested sites. Since ridge sites have typically never been farmed they represent a non-farmed forested site.

The questions that organize our research are:
1) What underlying mechanisms cause primary production to change with moisture and temperature?

2) Are differences in net productivity driven by changes in gross production, respiration, phenology, allocation, or water use?

3) How do changes described above affect the primary limits on tree growth and demography including gross production, net production, allocation, defense, and dispersal?

4) How do tree and understory herbaceous species respond to variation in climate and land-use history, and how do these factors influence species composition?

5) How do soil microarthropod communities respond to climate gradients, and how are population size, density and/or diversity affected by land-use history?

Table II.3.

C-IID1. Controls on Productivity.
Summary: Our objective in this funding cycle is to parameterize the full fecundity schedules as a function of tree size.

Productivity influences individual plant demography by determining the energetic resources allocated to growth, reproduction, storage, and defense. Potential differences among species in energy allocation are the basis for the hypothesis that tradeoffs involving trophic interactions (e.g., competition for limiting resources and selective predation) and life history (e.g., longevity vs. colonization ability) are responsible for coexistence of similar species (e.g., Rees et al. 2001). Total production in southern Appalachian forests is similar to other eastern deciduous forests, but relatively low given the exceptionally high precipitation and temperature relative to other temperate forests (Day and Monk 1977). Fine-scale analyses have shown strong trends in aboveground productivity related to both elevation and topographic position (Bolstad et al. 2001), and although the specific mechanisms have not been conclusively established, productivity decreases from low to high elevation (Figure 11). While a portion of productivity decrease is due to the length of the growing season, it also changes as a function of species composition and their physiological acclimation across the gradient (Bolstad et al. 1999, Mitchell et al. 1999).

We propose quantifying the effects of present climate variations on productivity and allocation in the dominant forest tree species of the southern Appalachian Mountains, and hypothesize that:

1. Higher net productivity at low elevation is primarily due to lower respiration, longer growing seasons, and higher rates of net photosynthesis.

2. Total production will be low on ridge sites (irrespective of elevation) in direct relation to where summer water deficits prevail. Belowground allocation will increase on ridges in response to low moisture availability, and relative reproductive allocation will be low.

Two 30 x 30 m plots will be established in both ridge and cove positions in each of the three study landscapes (n=12 plots). Stems of all trees >1.5 m tall will be mapped, and productivity and demographic parameters measured using standard mensuration procedures. Forest structural measurements will include: height, diameter, leaf area and mass, leaf nitrogen, root mass by depth, understory composition and structure, soil N and C, N mineralization rates, and litterfall by species and tissue type. The physical environment will similarly be measured at each site by way of soil and air temperatures, relative humidity, soil moisture, and below-canopy radiation. One site on each landscape (n=3) will be designated an intensive measurement plot, on which above-canopy radiation, windspeed, and stem temperatures will be recorded. We will link individual tree responses to stand dynamics by comparing productivity differences among species and crown classes with growth, mortality, and reproductive effort. Reproductive effort will by estimated from seed trap and census data (Clark et al. 1999b, LaDeau and Clark 2001). Mortality and its relation to growth rate will build on the initial census and modeling efforts of Wyckoff and Clark (2001) and incorporate the diameter censuses to be carried out in 2002 and 2004. For fecundity and growth, we will relate demographic and physiological responses from annual rates. Variation in net primary productivity will be evaluated with simulation modeling. Our objective in this funding cycle is to parameterize a spatially-explicit and species-specific productivity model using detailed physiology data collected over the past 10 years (Sullivan et al. 1996, Bolstad et al. 1999, Mitchell et al. 1999, Vose et al. 1999). Sensitivity analyses will be used to evaluate the relative importance of tissue-specific respiration, growing season length, and net photosynthesis in regulating productivity across the landscape. Model results will be validated with data collected from the three study landscapes. Figure 11. ANPP is significantly related to elevation in southern Appalachian forests, and productivity decreases from cove to ridge and from low to high elevations. 24 Belowground allocation will be measured intensively at the cove and ridge sites at Coweeta and Nancytown using a combination of minirhizotron images collected monthly and soil cores (Hendrick and Pregitzer 1993, Shan et al. 2001). Eight soil cores from each plot will be collected at the beginning of the study to establish a quantitative relationship between minirhizotron data and root mass. Additional cores will be collected throughout the study to check our ability at predicting changes in standing crop from minirhizotron production and mortality rates. We will establish a relationship between root length and numbers at the beginning of the study by digitizing images and estimating total root length. Thereafter, the dynamics of root numbers will be used to calculate length production and mortality. We have tested this technique using existing datasets and found numbers to be accurate (r2 >0.90) predictors of length; the technique also yields a 75% savings in time (Crocker et al. in review). Less intensive measures of belowground allocation will be made across the gradient using the total belowground carbon allocation (TCBA) approach of Raich and Nadelhoffer (1989). We will measure soil respiration monthly, and when possible, bi-weekly during the growing season using a Li-Cor 6400 unit. Annual soil CO2 efflux will be estimated using soil temperature data. Gravimetric soil moisture will be measured on three cores for each sampling period to correct for the effect of summer droughts (common along the gradient) on respiration rates (McDowell et al. 2001). Litterfall will be measured annually from the standard litter traps long-used on the Coweeta gradient sites. We will estimate fine root production for sites where it was not measured with cores and minirhizotrons using fine root production/TCBA relationships from the Coweeta and Nancytown sites.

C-IID2. Controls on Overstory Demography.
Summary: Ecological theory predicts and empirical evidence supports the idea that the maintenance of diversity depends on tradeoffs. Trophic tradeoffs result from patterns of consumption while life history tradeoffs are generally expressed as variation in the timing of reproductive effort.

In plants, life history and trophic tradeoffs are linked through patterns of allocation that affect growth, seed size, fecundity, dispersal, and survivorship (Connell and Slatyer 1977, Loehle 1988, Tilman 1988, Clark 1991, Rees et al. 2001). Tradeoffs are most important in forests at recruitment stages when they likely reflect resistance and tolerance to herbivores (Janzen 1970, Connell 1978, Harms et al. 2000, HilleRisLambers et al. 2002), and canopy gaps (Clark et al. 1998, Hubbell 2001). In forests, the trophic tradeoff hypothesis predicts that species coexist as a consequence of tradeoffs in competitive ability for resources and/or susceptibility to herbivores. In effect, species in optimal sites are abundant because they realize maximum growth rates, experience low mortality, and/or allocate more resources to defense and consequently have low herbivore losses. The life history tradeoff hypothesis predicts that some species escape competitive exclusion by episodically reaching sites missed by their more dispersal-limited competitors. In effect, they may be poorer competitors yet persist by virtue of obtaining sites that poor dispersers fail to reach or by release from frequency and/or density dependent regulation where populations are dense.

We therefore hypothesize that:
1. Tradeoffs in life history and resource acquisition mediate patterns of coexistence and diversity of trees in southern Appalachian forests;

2. Variation in conditions between gap and non-gap sites, and along gradients in elevation, provide the environmental heterogeneity on which life history and trophic tradeoffs are expressed;

3. Tradeoffs change the relative ranking of species in their recruitment success at seed and seedling stages along gradients in climate and light availability. We propose experimental recruitment studies on seed and seedling plantings that include selective exposure to vertebrate and invertebrate seed and seedling predators across moisture and elevation gradients, in gap and non-gap settings. We will include the dominant canopy species (n=10) in the region, and carry out experiments at (a) the non-farmed cove sites in each study landscape, (b) the natural gaps in the three study landscapes, and (c) the established experimental gaps in the Coweeta Basin (independent funding to Clark). At each site we will establish four 2- m x 1-m exclosures against vertebrate and invertebrate herbivores consisting of sunken barriers (for small mammals) topped with agricultural cloth (for deer and invertebrates) (Beckage et al. 2000). Previous studies show that fine grade agricultural cloth does not significantly reduce light levels reaching plants. Each exclosure will be associated with a control plot to which herbivores have free access. Our overall design will consist of 3 landscapes x 8 sites (4 gap, 4 non-gap) x 2 exclosure levels x 4 replicates, or a total of 192 sampling units. In each replicate we will plant 10 seeds and 6 seedlings of each species, and record the following response variables: germination, growth, mortality, allocation to defensive compounds, and herbivore losses. Allocation to defense will be estimated monthly from phenolic microanalyses based on 2mm disks removed from foliage and analyzed for hydrolysable tannins, condensed tannins, total phenolics, and astringency (modified from Hunter and Forkner 1999, Klaper et al. 2001). Losses will be monitored with biweekly censuses of seed and seedling damage, beginning with damage to newly fallen seed (pre-dispersal loss rate). We will quantify for each species the effects of recruitment limitations in each setting and determine the extent to which differences may support the notion that tradeoffs contribute to potential persistence.

C-IID3. Controls on Understory Diversity.
Summary: Forest understory plants are small, have relatively short life spans, mature sooner, and are therefore easier to manipulate experimentally than trees. As such, they provide an excellent model experimental system for how plants may respond to climatic variation and land use change (Figure 12).

We hypothesize that:
1. The demographic performance and geographical limits of many forest understory herbs are determined by soil moisture, in particular the length and severity of the summer drought and available light levels (PAR) during the growing season;

2. Past land-use history has altered the size and spatial variability of soil carbon and nutrient pools, and therefore influences present-day habitat quality for these species;

3. Species with specialized habitat needs and limited dispersal ability will be more vulnerable to both climatic variation and landscape change.

Herbaceous species diversity shifts to weedy species when patches are smaller, or when the past disturbance regime in the forest has been more intense. Our goals in this study are to discover how abiotic factors (i.e., nutrients, moisture, or light) limit the local and regional distributions of a set of representative species, and how land use history interacts with these factors to affect the suitability of specific sites. Species under consideration for detailed studies include Anemone quinquefolia, Arasaema triphyllum, Botrychium virginianum, Hepatica acutiloba, H. nobilis, Viola canadensis, Polygonatum biflorum, Smilacena racemosa, and Goodyera pubescens. We already have extensive distributional data on all of these species, and we have preliminary demographic data (independent funding to Pulliam) on several of them. The hypotheses concerning the impacts of local environment and land use history on plant demography and distribution will be tested both by measuring demography across the full range of conditions where the species normally occur and introducing them into microhabitats where they do not normally occur. Six study sites will be (or have already been) established on each of the three study landscapes in closed-canopy, mixed deciduous forest. Two study sites on each cove position will be located on previously-farmed mesic locations and four sites will be placed on non-farmed locations that include two mesic and two xeric positions (i.e., coves or ridges). A 24m x 20m demography plot will be located on each site and divided into 120, 2m x 2m cells. All individuals (ramets) of selected forb species and tree seedlings will be marked and monitored for survival, growth, and reproduction. During the growing season, light levels, soil moisture, soil temperature and nitrogen, and phosphorous mineralization will be monitored on 16 or more points within each plot. Rates and spatial dependency of potential nitrogen mineralization will be assessed as a function of land-use history. We will also establish small (5m x 5m) experimental plots on each site in the immediate vicinity (50-100m) of each demography plot (a total of 48 small plots). Preliminary identification of light gaps and light level measurements (PAR) will be used to stratify the small experimental plots so that on each site half the plots are in low-light and half in high-light patches. On each landscape there will be 2 to 4 replicates of 6 treatment conditions (mesic, high light, previously-farmed; mesic, low light previously-farmed; mesic, high light, non-farmed; mesic, low light, non-farmed; xeric, high light, non-farmed; and xeric, low light, non-farmed). Each treatment condition is further divided into four subplots. Twelve of the small experimental plots on the non-farmed locations will be used as “common garden” experiments to focus on 4 forb species (Polygonatum biflorum, Goodyera pubescens and two others to be determined) selected to represent a range of life history and dispersal strategies. Seeds, seedlings, and adults of each species will be collected at each of the four study landscapes and grown under the same conditions in all twelve gardens. This will test the degree of local adaptation as determined by growth and survivorship. We will also include plants from other locations even if they do not grow naturally at the experiment location as part of our test of factors limiting the geographic range of species. The remaining 36 small plots (2 replicates x 6 treatment conditions x 3 landscapes) will be used for local (within landscape) transplants and ecophysiological measurements. The transplant experiments will be used to test the range of local conditions that plants from one locale can tolerate. Within the constraints of plant material available for transplanting, seeds, seedlings and adults of all four species will be grown across the full range of moisture, light, and land-use history, however, we will vary one factor at a time focusing on moisture first, followed by light, and concluding with land-use history. We will measure physiological response of 27 individual plants to light, nutrients, and moisture on each small plot with a LiCor 6400 photosynthesis machine equipped with a built in light source.

C-IID4. Controls on Microarthropod Diversity.
Much of forest diversity is underground and consists of bacteria, protists, fungi, and many animal groups. We will compare soil microarthropod populations between sites with different land-use histories to help us predict the effects of future land-use change on soil biota critical to the decomposition process (Swift et al. 1979, Reynolds et al. 2000, Reynolds and Hunter 2001).

We hypothesize that:
1. Increases in the size of canopy openings, and the consequent decrease in canopy inputs such as frass, litter, and throughfall, will lead to changes in soil microarthropod communities, and particularly decreases in collembola and oribatid mites.

2. Rates of decomposition will decrease as canopy inputs are reduced. We have already established 5 transects at Coweeta containing 10, 1m-square plots that encompass gap and adjacent control areas. We will establish additional transects in the cover locations at the Nancytown and Mars Hill landscapes in matched non-farmed and previously farmed sites. Microarthropods will be sampled seasonally from soil cores on each plot with additional samples collected immediately prior and subsequent to the gap creation at Coweeta (n=50). Microarthropods extracted from soil cores will be sorted into the following categories: collembola, three suborders of mites, and “other.” Collembola and oribatid mites will be identified to species (oribatids) or morpho-species (collembola) for testing our diversity hypothesis. Measures of decomposition will only be carried out at the Coweeta gap plots. We will use 15cm x 15cm litter bags containing 2.5 g of red oak/red maple leaves (Reynolds et al. 2002). Twelve bags will be set out adjacent to 6 of the soil microarthropod plots after gap establishment at Coweeta (n=360). The study will be conducted for 2 years, with bags brought in for weighing and microarthropod extraction every other month.