Final Report: Development And Testing Of A Multi-Resource Landscape-Scale Ecological Indicator: Forest Fragmentation, Structure, and Distribution Relative to Topography

EPA Grant Number: R826598
Title: Development And Testing Of A Multi-Resource Landscape-Scale Ecological Indicator: Forest Fragmentation, Structure, and Distribution Relative to Topography
Investigators: Seagle, Steven W. , Townsend, Philip A.
Institution: University of Maryland Center for Environmental Science
EPA Project Officer: Packard, Benjamin H
Project Period: October 1, 1998 through September 30, 2003
Project Amount: $683,374
RFA: Ecological Indicators (1998) RFA Text |  Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Ecosystems

Objective:

Indicator Characteristics

The overall goal of this research project was to combine remote sensing and digital elevation models (DEMs) in developing a multiresource indicator. A landscape-scale ecological indictor should possess four distinct characteristics, that is, it must be: (1) interpretable in terms of ecological function; (2) quantifiable over a significantly large geographical area; (3) comparable over both space and time; and (4) readily repeatable over time. Ecological indicators seldom meet all of these criteria, but at the landscape scale, some combination of remotely sensed variables will help meet these criteria. In addition, multiple studies have shown that topographic variables are strong correlates of ecosystem processes, and thus measures derived from widely available DEMs are useful for predicting ecological function.

The specific objectives of this research project were as follows:

1. We evaluated the role of topography, an indirect driver of soil moisture conditions and forest productivity in central Appalachian landscapes, as a factor influencing the breeding productivity of forest interior songbirds in the mountains of western Maryland. By developing the ability to predict spatial patterns of breeding productivity from topographic variables, it may be feasible to map avian habitat quality continuously over large areas.

2. We sought a novel means of predicting forest vertical complexity from synthetic aperture radar (SAR) imagery. Forest bird diversity is correlated with the complexity of forest vertical structure. Application of this technique presents the potential to map forest structure (and consequently its correlate, avian diversity) continuously over large landscapes.

3. We developed metrics that can represent simultaneously both the quality of avian habitat and the likelihood of nitrogen export from multiple-use watersheds.

Indicators that are capable of providing status information on multiple resources simultaneously are valuable for environmental management. Forest interior songbird reproductive success is a terrestrially based indicator of forest ecosystem health within multiple-use landscapes. Similarly, nutrient export from multiple-use watersheds has been viewed as an aquatic-based indicator of landscape integrity.

Summary/Accomplishments (Outputs/Outcomes):

Avian Community Reproductive Success and Topography

The hypothesis that breeding productivity of forest interior songbirds is correlated significantly with forest productivity gradients created by topography was tested using study sites in the Appalachian Plateau and the Ridge and Valley Provinces in western Maryland. We selected mature, upland deciduous forest sites within the forest interior to focus the study on the breeding response of songbirds to topographic variation in forest productivity. Ten study sites were selected on the plateau; 12 sites were selected within the Ridge and Valley. These sites also provided empirical data for our remote sensing studies of forest structure.

Site index, functionally defined as average tree height in a 50-year-old stand, was chosen as a long-term integrative measure of forest productivity. From an examination of tree cores, forest stand structure, and topographic site conditions, we concluded that site index provides an extremely good estimate of site quality. More importantly, site index could be predicted reliably from a linear combination of topographic indices (relative slope position, percent slope, and aspect) that can be derived from DEMs. Thus, site index can be predicted over large landscapes within the central Appalachians.

We hypothesized that forest bird fledging success could be modeled as a function of forest productivity (i.e., site index). All 22 study sites were surveyed during two breeding seasons (1999, 2000) to estimate the density and fledging success of forest songbirds. We found that the density of forest interior songbirds was significantly greater in the wetter climate of the Appalachian Plateau than within the drier climate of the Ridge and Valley, and also was positively correlated with site index. A severe drought during the summer of 1999 likely influenced the pattern of fledging success during that year. The documentation of a positive correlation between forest productivity and the fledging success of forest interior songbirds is a unique contribution of our research project. Nonetheless, the fact that a positive correlation was not observed during a drought year indicated that temporal variation in weather patterns also may influence spatial patterns in songbird production. We conclude that site index confidently can be applied as an indicator of potential habitat quality.

The correlation of avian community reproductive success with forest productivity implies bottom-up control in the forest food web. We thus predicted that forest productivity would be correlated positively with invertebrate detrital food web productivity, and subsequently with avian predator productivity. Such correlations would provide a basis in ecological theory to interpret indicators of avian habitat quality. Invertebrates were sampled during the avian breeding season of 2000, and biomass was correlated with both site index and reproductive success of the most common forest floor foraging bird species (ovenbird; Seiurus aurocapillus). Litter invertebrate biomass was correlated positively and significantly with site index. Separate regressions for the Ridge and Valley and Appalachian Plateau indicate a stronger relationship between site index and invertebrate biomass for the Ridge and Valley. The proportion of ovenbird territories fledging young was correlated strongly and positively with both invertebrate biomass and site index. Our results clearly find a strong bottom-up linkage across trophic levels, propagating from primary producers to top predators. Furthermore, because forest site index is a function of topographic position, topography is a key determinant of food web productivity.

We developed statistical models to predict both site index and avian habitat quality from topographic variables. We used these models to ask whether human land use development impacts the production of interior songbirds by the systematic development of the most productive breeding grounds. A random sample of ten 10 x 10 km landscapes from each province was compared to test for differences in predicted potential habitat quality. Human land use development within our study areas clearly impacts interior songbird habitat because of development of the most potentially productive breeding grounds. However, this bias in land development pattern is strongest within the Ridge and Valley. This spatial modeling approach has several implications for conservation planning. By extrapolating habitat quality based on forest pattern and topography, one easily can detect areas of high-quality habitat that can be targeted for conservation. Also, by applying this model to the entire landscape, one can detect which areas have the highest potential to increase landscape-scale habitat quality.

Remote Sensing of Forest Structure and Complexity

A landscape-scale geometric model of forest canopy structure was developed to characterize forest vertical structure on a 30 x 30 m pixel-by-pixel basis. The model was designed to translate field measurements of canopy height, leaf area index, cover, density, and basal area by strata (canopy, subcanopy, and shrub/sapling layer) into a three-dimensional model of canopy volume. Specifically, heights to the top and bottom of each canopy layer for multiple trees per plot were used to estimate the volume of canopy filled, assuming a net elliptical-cylindrical shape to the volume of material "filling" a pixel (independent of individual trees). Canopy cover and leaf area data were used to estimate the actual amount of material filling the volume of vegetation within a pixel. Analytical geometry was used to determine the amount of foliage at 2-meter increments within the canopy, producing profiles of the forest canopy that bore substantial similarity to ground observations of the forest profile. Two measures of heterogeneity, the Shannon-Weiner index of diversity and the Gini coefficient of inequality, were used to compute synthetic indices of forest structure that characterized the fullness ("diversity") of the canopy from top to bottom (Shannon-Weiner index) and the degree of inequality in coverage among canopy layers from the top to the bottom (Gini coefficient). Together, the Shannon-Weiner index and the Gini coefficient adequately represent the structural complexity of the forests.

Because forest structure is critical to avian and insect fauna, we were interested in the potential for modeling aspects of forest structure across landscapes. To accomplish this, we linked the structural measures to remotely sensed imagery from multispectral (e.g., Landsat Thematic Mapper [TM] and enhanced TM+) and SAR sensors (Japanese Earth Resources Satellite-1 [JERS-1], Radarsat, European Remote Sensing Satellite-2 [ERS-2]). Each sensor was used because of differing sensitivities to aspects of forest structure, with Landsat data most sensitive to greenness and leaf area, JERS-1 sensitive to ground-level canopy structure, Radarsat sensitive to middle and top level structure, and ERS-2 sensitive to forest structure at the top of the canopy. Multiple linear regression of structural variables (especially Gini, Shannon-Weiner, basal area, and height of top of canopy) indicated that forest structure could be predicted from imagery with R2 between 0.55 and 0.80. This indicates great promise for mapping forest structure using widely available satellite remote sensing data sets.

Many forest bird species and ecosystem processes are dependent on or influenced by the forest understory. The evergreen understory species rosebay rhododendron (Rhododendron maximum L.) and mountain laurel (Kalmia latifolia L.) dominate the understory layer in Appalachian forests in localized areas. Using inexpensive and readily available leaf-off Landsat TM image data, the distribution of evergreen understory communities was mapped at an accuracy level of 87.1 percent in the Ridge and Valley study area and 82.9 percent in the Appalachian Plateau. Because of the patchy distributions of forest understory species and their important impact on forest structural complexity, we examined the spatial and temporal dynamics of evergreen understory communities. Although the spatial extent of the evergreen understory did not change over a 20-year time series of satellite imagery, K. latifolia and R. maximum exhibited notable fluctuations in growth vigor. Comparisons of data from the two physiographic provinces showed that the growth rate of K. latifolia was not influenced by the difference in precipitation amount between the two zones. The growth rate of R. maximum was increased by the moister conditions present in the Appalachian Plateau relative to the Ridge and Valley. In contrast, K. latifolia growth rates were identical for both provinces. Our research project suggests that at least two mechanisms potentially allow successful regeneration of canopy trees in evergreen understory areas—sporadic death among K. latifolia individuals and variability in the growth vigor (and hence leaf area) of K. latifolia and R. maximum. The results suggest that overstory regeneration bottlenecks caused by a dense evergreen understory only should exist in localized areas that are ideal for R. maximum and/or K. latifolia growth and that have not experienced large perturbations.

The nutrient cycling model NuCSS was applied to estimate carbon sequestration and nitrogen and phosphorous storage for the forests with and without an evergreen understory. Carbon storage increased by between 1,631 and 4,825 kg/ha after 50 years with the addition of evergreen understory layer vegetation in the simulated forests. Nitrogen storage increased by 41 to 224 kg/ha as a result of the evergreen understory. Nitrogen storage was greater in the forest floor as compared to the soil when a larger amount of R. maximum was present because of the large amounts of recalcitrant litter produced by this species compared to K. latifolia. Phosphorous storage in K. latifolia and R. maximum was large relative to their standing biomass.

Water Quality, Avian Habitat, and Topography

The final objectives of our research were to: (1) extrapolate avian habitat quality results to a regional scale; (2) analyze the relationships between watershed nutrient export, spatial patterns of forest land cover, and topography; and (3) integrate avian habitat quality and landscape/topographic characteristics that conserve nutrients to develop indicators that are simultaneously functional for forest interior birds and water quality. These objectives required regional databases of watershed nutrient export, land cover, and topography (DEMs). Nutrient data have been available to support these objectives for some time. Our original goal was to use Multi-Resolution Land Cover Data for the land cover database and a DEM of the Chesapeake Bay Watershed from the National Elevation Data Set. Unfortunately, both of these products had systematic errors that would propagate through spatial extrapolations and invalidate indicator values. At notable time and monetary expense, we reclassified a series of Landsat images to achieve (in mid-2003) an appropriate land cover database. It was beyond the technical scope of this project to correct the regional Chesapeake DEM; however, in December 2003 we acquired a new Chesapeake Watershed DEM from the National Elevation Data Set. This new product is much improved, and we have begun to employ it along with our reclassified land cover maps to continue our work on regional extrapolations.

Conclusions:

Our results are extremely encouraging for the development of landscape-scale ecological indicators that reflect both forest productivity and avian habitat quality. Our ability to predict these ecological parameters using only Landsat imagery to derive land cover and DEMs to derive topographic parameters is particularly exciting because it represents, to our knowledge, the first opportunity to map either of these parameters on a continuous basis over large geographic areas. Furthermore, because of their reliance only on widely available databases that, in the case of Landsat imagery, is updated annually, mapping and spatial analysis of these indicators of landscape integrity can be repeated readily over time. Our research project also clearly suggests that these results are grounded in the basic ecological principle of bottom-up control over ecosystem and food web structure. Overall, these results meet virtually all criteria of a functional and interpretable ecological indicator.

It has been recognized for decades that forest vertical complexity is an important parameter determining avian community diversity. This complexity also may be related to various ecosystem functions such as light capture and forest productivity. Our results demonstrate that by innovative application of new radar remote sensing platforms, it is now feasible to map forest vertical structure continuously and repeatedly over large landscapes. Thus, this indicator of forest complexity also conforms to virtually all of the characteristics of a functional ecological indicator.

With newly developed or recently available databases, we continue our research on the correlation of forest topographic distribution with nutrient and sediment export from watersheds. Concerns over nonpoint nutrient pollution have not abated over time, and there are continuing efforts to retain or develop forest land cover as a tool to control nutrient pollution. Our results will be pertinent to this debate by helping to define where forest cover will be most effective in nutrient control within watersheds. Moreover, by mapping forest topographic distribution over large landscapes we hope to use our results as a spatial indicator of nutrient control and forest bird habitat quality.


Journal Articles on this Report : 5 Displayed | Download in RIS Format

Other project views: All 38 publications 7 publications in selected types All 7 journal articles
Type Citation Project Document Sources
Journal Article Chastain Jr. RA, Currie WS, Townsend PA. Carbon sequestration and nutrient cycling implications of the evergreen understory layer in Appalachian forests. Forest Ecology and Management 2006;231(1-3):63-77. R826598 (Final)
  • Full-text: Science Direct
    Exit
  • Abstract: Science Direct
    Exit
  • Other: Science Direct
    Exit
  • Journal Article Chastain RA, Townsend PA. Use of Landsat ETM and topographic data to characterize evergreen understory communities in Appalachian deciduous forests. Photogrammetric Engineering and Remote Sensing 2007;73(5):563-575. R826598 (Final)
  • Full-text: ASPRS
    Exit
  • Abstract: ASPRS
    Exit
  • Journal Article Chastain RA, Townsend PA. Role of evergreen understory shrub layer in the forests of the central Appalachian Highlands. The Journal of the Torrey Botanical Society 2008;135(2):208-223. R826598 (Final)
  • Abstract: BioOne
    Exit
  • Journal Article Seagle SW, Sturtevant BR. Forest productivity predicts invertebrate biomass and ovenbird (Seiurus aurocapillus) reproduction in Appalachian landscapes. Ecology 2005;86(6):1531-1539. R826598 (2002)
    R826598 (Final)
  • Abstract: ESA
    Exit
  • Journal Article Sturtevant BR, Seagle SW. Comparing estimates of forest site quality in old second-growth oak forests. Forest Ecology and Management 2004;191(1-3):311-328. R826598 (2002)
    R826598 (Final)
  • Full-text: Science Direct
    Exit
  • Abstract: Science Direct
    Exit
  • Other: Science Direct
    Exit
  • Supplemental Keywords:

    water, watersheds, land, animal, ecosystem, indicators, scaling, modeling, terrestrial, aquatic, habitat, biology, ecology, hydrology, nutrient cycling, Environmental Monitoring and Assessment Program, EMAP, surveys, Landsat, remote sensing, Chesapeake Bay Watershed, U.S. Environmental Protection Agency, EPA, EPA Region 3, Maryland, Mid-Atlantic Highlands, forest interior, fragmentation, topography, digital elevation models, DEMs, synthetic aperture radar, SAR, forest birds, gypsy moth, invertebrate biomass, forest, forest understory, forest structure., RFA, Scientific Discipline, Geographic Area, Water, Ecosystem Protection/Environmental Exposure & Risk, Hydrology, Ecology, Water & Watershed, Ecosystem/Assessment/Indicators, Ecosystem Protection, Forestry, Ecological Effects - Environmental Exposure & Risk, Mid-Atlantic, Ecological Risk Assessment, Biology, Geology, EPA Region, Watersheds, Ecological Indicators, risk assessment, remote sensing, landscape indicator, multi-level indicators, stream ecosystems, Region 3, bird habitat, ecosystem indicators, estuarine ecosystems, gypsy moth, Mid-Atlantic Highlands, terrestrial, aquatic ecosystems, water quality, stress responses, defoliation, land use

    Progress and Final Reports:

    Original Abstract
  • 1999
  • 2000 Progress Report
  • 2001 Progress Report
  • 2002 Progress Report