Final Report: Assessment of forest disturbance in the mid-Atlantic region: a multi-scale linkage between terrestrial and aquatic ecosystems

EPA Grant Number: R826110
Title: Assessment of forest disturbance in the mid-Atlantic region: a multi-scale linkage between terrestrial and aquatic ecosystems
Investigators: Eshleman, Keith N. , Galloway, James N. , Gardner, Robert H. , Pitelka, Louis F. , Seagle, Steven W. , Webb, James R.
Institution: University of Maryland , Oregon State University , University of Virginia
EPA Project Officer: Packard, Benjamin H
Project Period: October 1, 1997 through September 30, 2002
Project Amount: $697,834
RFA: Approaches to Multi-scale Ecological Assessment in the Middle Atlantic Region (1997) RFA Text |  Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Ecosystems


The overall objective of this research project was to develop, test, validate, and demonstrate an analytical framework for assessing regional-scale forest disturbance in the mid-Atlantic region by establishing a multi-scale linkage between forest disturbance and forest nitrogen (N) export to surface waters. The specific objectives of this research project are to: (1) characterize forest composition, recent disturbance (specifically gypsy moth defoliation) history, and annual N export for intensively studied watersheds; (2) model N export from intensively studied watersheds because of disturbance; (3) verify N export as an indicator of disturbance at subregional scales; (4) scale forest point data to landscape scales; and (5) correlate spatial and temporal patterns of N export and forest species composition changes with forest disturbance at the regional scale.

Forests are the dominant land cover type in the mid-Atlantic region of the United States, despite a long history of intensive logging and clearing of the land for settlement and agriculture before the 20th century, and a more recent decline in forest cover resulting from urban and suburban development. Presently, the U.S. Environmental Protection Agency (EPA) estimates that the Chesapeake Bay Watershed is composed of 60 percent forest, 29 percent agriculture, and 10 percent urban land. The forested lands of the Chesapeake Watershed and of the entire mid-Atlantic region are an important natural and economic resource, providing fiber for building materials, paper, and fuel. Forested lands also provide a habitat for wildlife and endangered floral and faunal species, and recreational opportunities for human inhabitants of the area. Due to their close coupling with the regional hydrological cycle, forested lands provide important economic benefits to the region in the form of water supplies for domestic, industrial, and consumptive uses. Relative to other land uses, surface and ground water supplies produced by forested lands are typically found to be of higher quality, although various negative effects of forest management practices on water quality have been documented after the experimental manipulations of watersheds were first conducted during the early 1960's. Several recent studies also have demonstrated surface water quality impacts associated with forest defoliations by the gypsy moth larva (Lymantria dispar). These studies have confirmed that nonharvesting forest disturbances can significantly affect rates of loss of solutes to streamwater, which is consistent with stream responses following both experimental and natural forest disturbances.

It was hypothesized that excessive N leakage (export) from forested watersheds is a potentially useful, integrative "indicator" of negative changes in forest function that occur in synchrony with changes in forest structure and species composition. Our research focused on forest disturbance associated with recent defoliations by the gypsy moth larva at spatial scales ranging from small watersheds to the entire region. The project attempted to establish a multi-resource linkage between forests and surface waters in the mid-Atlantic region using data and models developed from: (1) watershed-scale studies of intensively studied systems; (2) synoptic-scale surveys of resource conditions (including soils, geologic class, forest class, forest disturbance, and surface water quality); and (3) remotely sensed information.

Summary/Accomplishments (Outputs/Outcomes):

Supplemental funding was provided in 2000 and 2001 to perform several additional research tasks that involved more extensive use of remote-sensing imagery for quantifying forest defoliation at multiple spatial and temporal scales such as: use of high temporal satellite data (including MODIS and AVHRR) and high spatial resolution satellite data (i.e., Landsat 7) to characterize the specific timing, extent, and intensity of gypsy moth defoliation; and use of the temporally explicit representations of disturbance and a nitrate-N export model to assess rates of nitrate-N leakage from disturbed forests to downstream surface waters in the mid-Atlantic region.


Fluxes of dissolved N, primarily in the form of nitrate, from forested watersheds can contribute both to the acidification of acid-sensitive surface waters, and to the eutrophication of downstream receiving waters, particularly coastal and estuarine ecosystems. The scientific literature is replete with studies of dissolved N leakage from forested watersheds in the Northeastern United States, Canada, and Western Europe. Many of these studies have interpreted long-term increases or short-term patterns in dissolved N concentrations in surface waters as evidence of a situation known as "nitrogen saturation," in which the supply of N from the atmosphere to a forest exceeds the demand for N by watershed floral and microbial organisms. Because of a legacy of over-fertilization, the leakage of dissolved N from forested watersheds to surface waters is a particularly important issue for watersheds of the Chesapeake Bay, which is presently the subject of a major ecosystem restoration.

Our analyses suggested that patterns of N leakage from several intensively monitored mid-Appalachian forested watersheds during the late 1980's and early 1990's, displayed considerable temporal and spatial synchrony with outbreaks of defoliation by the gypsy moth larva (a non-native forest insect pest). This evidence suggested very strongly that forest disturbance may be an important contributor to N leakage in the mid-Appalachian region, and that poorly documented insect defoliation-rather than premature N saturation of intact forest ecosystems-needs to be considered as a possible explanation of N leakage from forested watersheds in the mid-Appalachian region and elsewhere (Eshleman, et al., 1998).

A logical first step in quantitatively testing the hypothesis that insect defoliation produces an increase in nitrate-N export was taken by evaluating the ability of a simple, empirical model to explain temporal changes in annual N export from gaged forested watersheds in the years following disturbance. A simple linear systems response model was proposed to describe the N flux from a forested watershed subjected to a large-scale disturbance of vegetation, such as a defoliation by gypsy moth larvae (Eshleman, 2000):

where: Nw(t) is the nitrogen export from watershed w; U(t - tau) is a unit N export response function (UNERF); Dw(tau) is the proportion of forested watershed disturbed at time t (0 less than or equal to Dw(tau) less than or equal to 1); and Bw is the baseline N export from watershed w in the absence of disturbance. The term unit in this case refers to the N export response to a complete disturbance of the watershed (i.e., when 100 percent of the watershed area is disturbed). The model includes two terms: (1) convolution of a UNERF with the proportion of the watershed disturbed at time tau, representing the integral response of the watershed to disturbance (Nw,D), and (2) baseline N export response from the watershed without disturbance (Bw). The model represents the well-known response of a linear system to an "impulse" (in this case a disturbance) and is governed by the principles of proportionality and superposition, which allow the responses of the system to partial or multiple impulses be predicted by convolution once the UNERF is known. The model makes the inherent assumption that the biogeochemical response of a forested watershed to insect defoliation is linear both in space and in time. This means that: (1) partial defoliation produces a nitrate-N response that is proportional to the area of the watershed defoliated; and (2) multiple defoliations in time produce nitrate-N responses that are additive in time (i.e., they can be superimposed). Annual N export data from five defoliated watersheds were analyzed during this part of the study. Several forms of linear UNERF models-parameterized by deconvolution of annual time series of N export using linear programming or by a least squares method-were generally found to be minimally biased and to explain relatively high percentages (38-98 percent) of the total variation in annual N export (Eshleman, 2000). Despite their neglect of spatial and temporal ecosystem non-linearities, these linear systems models appear reasonably robust, making them at least as useful as their more complex non-linear brethren for purposes of biogeochemical regionalization.

A second major step in the research involved the development, verification, and application of a regional version of the UNERF model. Given the large amount of data available for Shenandoah National Park (SNP) in Virginia, where the biogeochemical effects of gypsy moth defoliation have been relatively well documented, we were able to develop lithology-based UNERF models and test predictions from these models against field data from a set of 10 ungaged watersheds in SNP (Eshleman, et al., 2001). The results from the model testing largely verified the utility of the simplistic, lithology-based UNERF model (see Figure 1), although we observed some over-prediction of nitrate-N losses for the siliciclastic watersheds. Although nitrate-N leakage from disturbed forested watersheds has been almost universally attributed to a combination of biological and hydrological factors, a predictive model based on three lithologic classes was unexpected. However, lithology might indirectly predict N losses resulting from disturbance in an area like SNP, where there is a relatively tight coupling between bedrock, soil moisture, and forest species composition. It is well known that groundwater residence time and the occurrence of viable springs in SNP are predicted largely by lithology. For example, it has been shown that there are very large storages of groundwater that are common in the deep (3-20 m) regolith underlain by the basaltic rocks of the Catoctin Formation. In contrast, very few springs are observed in areas underlain by the metamorphosed sandstones and siltstones of the Antietam, Harpers, and Weverton Formations because of the extreme resistance to weathering of these siliciclastic rocks; these geological substrates promote rapid runoff and provide little storage of groundwater. Groundwater residence times and springs in the granitic Pedlar Formation are intermediate between the basaltic and siliciclastic bedrock, as predicted by its intermediate resistance to weathering. Therefore, we speculate that both forest vegetation and hydrological response of SNP watersheds are significantly-but indirectly-determined by substrate; it is these factors that largely control the rate and magnitude of the nitrate leakage response to forest defoliation. Forests on the siliciclastic rocks respond most quickly to the short hydrologic residence time of these systems and also to the dominance on these sites of drought-tolerant chestnut oak (Quercus prinus), whose leaves are a preferred food of the gypsy moth larva. Forests on the basaltic rocks produce the greatest rate of nitrate-N leakage per "unit" defoliation, likely because of the much greater density of forest vegetation on these richer sites; the nitrate-N response of the basaltic watersheds lags the response of the siliciclastic watersheds, presumably because of longer hydrologic residence time. Finally, the nitrate-N responses of the granitic watersheds are the lowest per unit defoliation and the most attenuated, which can perhaps be explained by the high water-holding capacity of a deep saprolite that is known to exist in these deeply weathered systems.

Following model testing for specific watersheds, we made park-wide estimates of annual nitrate-N export for the 21-year period from 1980-2000, using the same linked geographic information system and modeling approach; we set Bw equal to 0.10 kg/ha-yr, a value close to the mean value obtained for the 10 ungaged watersheds. It was assumed that no gypsy moth defoliation occurred in SNP during the years prior to the first mapped defoliation in 1986, and that defoliation since 1993 has been negligible. Predicted park-wide estimates of annual nitrate-N export suggest a dramatic transient response to gypsy moth defoliation during the late 1980's and early 1990's (see Figure 2). The model predicts that park-wide annual average nitrate-N export from SNP forests increased exponentially from a rate of 0.1 kg/ha in 1986 to a rate of 1.84 kg/ha in 1993; as defoliation declined in the early 1990s, the model suggests that average annual nitrate-N export exponentially declined from its 1993 peak to an estimated flux of about 0.3 kg/ha in 2000.

Figure 1. UNERFs for Four Gaged Watersheds in Shenandoah National Park, Virginia: Paine Run (PAIN), Piney River (PINE), Staunton River (STAN), and White Oak Run (WOR). Composite derived lithology-based UNERFs also shown (inset). Reprinted from Eshleman, et al. (2001).

Figure 2. Gypsy Moth Defoliation and Predicted Annual Nitrate-N Export in SNP (waters years 1980 through 1999). Revised and reprinted from Eshleman, et al., 2001.

The simple animations provide a general but integrated picture of the watershed disturbance model. Two aspects of the animations are particularly noteworthy: (1) geology-driven spatial heterogeneity in the N export process within a watershed (and within SNP); and (2) the combined effect of this spatial heterogeneity, with additional spatio-temporal variation resulting from the interacting patterns of defoliation and the time-lag of the disturbance response. Given such spatial and temporal complexity, the ability of the UNERF to produce robust estimates is even more impressive.

SNP forests appear to be characteristic of other N-limited second-growth forests in the Eastern United States that leak little N under undisturbed conditions, despite receiving relatively large inputs of N from atmospheric deposition sources. Although these regional-scale predictions cannot be validated using field data, we conclude that vegetation disturbances can apparently cause major changes in N input-output balances, with potentially important ramifications for low-order forest streams and downstream receiving waters. Interestingly, the annual rates of nitrate-N leakage from SNP-even following severe disturbances-are much lower than the rates that have been usually attributed to forests in the Chesapeake Bay Watershed for purposes of quantifying sources and sinks of nitrogen among different land use or land cover classes. Therefore, our research results have important ramifications for watershed modeling, and clearly suggest that models that treat forest nitrogen cycling as static in time and uniform in space will have significant difficulties simulating nitrogen export during periods when forest vegetation is disturbed.

Compared to our success in parameterizing a simple empirical disturbance model that could predict annual nitrate-N export from defoliated forests as a function of lithology and disturbance history, we had much lesser success in mapping forest composition and identifying and mapping changes in forest composition that were expected responses to disturbance. In particular, both at the watershed and regional scales, interpretation of our highly detailed database of more than 3,500 plot-based measurements of forest composition and associated variables remains incomplete. Hopefully, this interpretation will be completed in the near future. Similarly, our ability to estimate forest compositional changes from sequential Forest Inventory and Analysis (FIA) surveys was severely hampered by problems of data comparability and availability, yet there may be some opportunities to complete these analyses synergistically with some newer projects that have recently been initiated. On the other hand, our research showed that remote sensing of forest disturbances and forest changes using satellite imagery appears to offer some significant advantages-both in terms of cost and accuracy-over conventional mapping techniques at multiple scales ranging from local to landscape. However, additional research is needed to further test, refine, and validate these methods and make them available to forest ecologists, hydrologists, and modelers so that they can be used to better quantify, understand, and ultimately manage ecosystem changes that are difficult to otherwise discern.

Journal Articles on this Report : 8 Displayed | Download in RIS Format

Other project views: All 24 publications 9 publications in selected types All 8 journal articles
Type Citation Project Document Sources
Journal Article Cole CA, Brooks RP. A comparison of the hydrologic characteristics of natural and created mainstem floodplain wetlands in Pennsylvania. Ecological Engineering 2000;14(3):221-231. R826110 (Final)
R824803 (1998)
R824803 (Final)
R824905 (1999)
R824905 (Final)
  • Full-text: ScienceDirect - Full Text HTML
  • Abstract: ScienceDirect - Abstract
  • Other: ScienceDirect - Full Text PDF
  • Journal Article Debinski DM, Ray C, Saveraid EH. Species diversity and the scale of the landscape mosaic: do scales of movement and patch size affect diversity? Biological Conservation 2001;98(2):179-190. R826110 (Final)
    R825155 (Final)
  • Full-text: ScienceDirect-Full Text HTML
  • Abstract: ScienceDirect-Abstract
  • Other: ScienceDirect-Full Text PDF
  • Journal Article Eshleman KN, Morgan II RP, Webb JR, Deviney FA, Galloway JN. Temporal patterns of nitrogen leakage from mid-Appalachian forested watersheds: role of insect defoliation. Water Resources Research 1998;34(8):2005-2016. R826110 (1999)
    R826110 (Final)
  • Full-text: Wiley-Full Text PDF
  • Abstract: Wiley-Abstract
  • Journal Article Eshleman KN, Gardner RH, Seagle SW, Castro NM, Fiscus DA, Webb JR, Galloway JN, Deviney FA, Herlihy AT. Effects of disturbance on nitrogen export from forested lands of the Chesapeake Bay watershed. Environmental Monitoring and Assessment 2000;63(1):187-197. R826110 (1999)
    R826110 (2000)
    R826110 (Final)
  • Abstract: Springer-Abstract
  • Other: Ingenta-Abstract
  • Journal Article Eshleman KN. A linear model of the effects of disturbance on dissolved nitrogen leakage from forested watersheds. Water Resources Research 2000;36(11):3325-3335. R826110 (1999)
    R826110 (2000)
    R826110 (Final)
  • Abstract: Wiley-Abstract
  • Journal Article Eshleman KN, Fiscus DA, Castro NM, Webb JR, Deviney Jr. JF. Computation and visualization of regional-scale forest disturbance and associated dissolved nitrogen export from Shenandoah National Park, Virginia. The Scientific World Journal 2001;1(Suppl 2):539-547. R826110 (2001)
    R826110 (Final)
  • Abstract from PubMed
  • Full-text: TheScientificWorldJournal-Full Text PDF
  • Abstract: TheScientificWorldJournal-Abstract
  • Journal Article Eshleman KN, Fiscus DA, Castro NM, Webb JR, Herlihy AT. Regionalization of disturbance-induced nitrogen leakage from mid-Appalachian forests using a linear systems model. Hydrological Processes 2004;18(14):2713-2725. R826110 (1999)
    R826110 (2000)
    R826110 (2001)
    R826110 (Final)
  • Abstract: Wiley-Abstract
  • Other: Harvard-Abstract
  • Journal Article Jakubauskas M, Kindscher K, Debinski D. Spectral and biophysical relationships of montane sagebrush communities in multi-temporal SPOT XS data. International Journal of Remote Sensing 2001;22(9):1767-1778. R826110 (Final)
    R825155 (1997)
    R825155 (1999)
    R825155 (Final)
  • Full-text: Iowa State University-Full Text PDF
  • Abstract: Taylor and Francis-Abstract
  • Supplemental Keywords:

    forests, forest ecosystem, forestry, disturbance, nitrate-nitrogen, nitrogen saturation, watershed, defoliation, gypsy moth, regionalization, Shenandoah National Park, Chesapeake Bay, water quality modeling, linear system, linear model, verification, lithology, remote sensing, mid-Atlantic, forest composition, Landsat, MODIS, AVHRR, NDVI, Environmental Monitoring and Assessment Program, EMAP, EMAP-MAHA, MRLC, Appalachian Plateau, Valley and Ridge, Blue Ridge, Virginia, VA, EPA Region 3., RFA, Ecosystem Protection, Environmental Exposure & Risk, Scientific Discipline, Geographic Area, Water, Ecosystem Protection/Environmental Exposure & Risk, Hydrology, Nutrients, Water & Watershed, exploratory research environmental biology, Environmental Chemistry, Ecosystem/Assessment/Indicators, Ecological Effects - Environmental Exposure & Risk, Aquatic, Mid-Atlantic, Ecological Risk Assessment, Ecology and Ecosystems, Scaling, Watersheds, Ecological Indicators, nutrient transport, environmental monitoring, aquatic ecosystem, EMAP, nutrient supply, remote sensing, ecological exposure, ecological effects, watershed management, ecosystem assessment, N deposition, temperate forest ecosystems, nutrient flux, forest ecosystem, spatial scale, forest ecosystems, modeling, Chesapeake Bay watershed, conservation, environmental consequences, regional scale impacts, ecological assessment, ecological impacts, ecosystem management, gypsy moth, regional scale, terrestrial, aquatic ecosystems, water quality, nitrogen compounds, assessment methods, stress responses, Environmental Monitoring and Assessment Program, integrated ecological assessment, remotely sensed data, defoliation, interactions, nutrient fluxes, land use, nitrogen, Environmental Monitoring & Assessment Program, indicators, Chesapeake Bay

    Relevant Websites: Exit Exit Exit

    Progress and Final Reports:

    Original Abstract
  • 1998
  • 1999 Progress Report
  • 2000 Progress Report
  • 2001 Progress Report