Grantee Research Project Results
Final Report: Integrated Assessment of Watersheds
EPA Grant Number: R828684C003Subproject: this is subproject number 003 , established and managed by the Center Director under grant R828684
(EPA does not fund or establish subprojects; EPA awards and manages the overall grant for this center).
Center: Center for Integrated Multi‐scale Nutrient Pollution Solutions
Center Director: Shortle, James S.
Title: Integrated Assessment of Watersheds
Investigators: Brooks, Robert P. , Weller, Donald E. , Havens, Kirk , Brinson, Mark M. , Rheinhardt, Rick D. , Hite, Jeremy T. , King, Ryan , Easterling, Mary M. , Bishop, Joseph A. , Rubbo, Jennifer , Armstrong, Brian K. , Baker, Matthew , O'Brien, David
Institution: Pennsylvania State University , Smithsonian Environmental Research Center , East Carolina University , Virginia Institute of Marine Science
EPA Project Officer: Packard, Benjamin H
Project Period: March 1, 2001 through February 28, 2005 (Extended to February 28, 2006)
RFA: Environmental Indicators in the Estuarine Environment Research Program (2000) RFA Text | Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Water , Aquatic Ecosystems
Objective:
This was one of four projects under the Atlantic Slope Consortium (ASC) Center. The objectives of this research project were to: (1) develop and test indicators of the biogeochemical health and integrity of watersheds; (2) relate those indicators to environmental conditions; (3) assess the predictability of landscape characteristics to indicator responses; and (4) use those predictions to characterize the effects of watershed discharges on downstream riverine and estuarine health.
The ecological condition of streams, wetlands, and riparian areas depends upon land use and other activities upstream. Therefore, classifying regions first by social choice and physiographic setting allows better distinction of variation caused by natural sources (slope, soils, extreme natural events, etc.) and those caused by human activities (hydrologic alteration, pollution, etc.). Certain physical, chemical, and biotic indicators respond to the degree and type of human activities and thus can predict these effects on the condition of aquatic ecosystems. Indicators are desired that respond to different types of questions, are useful at various spatial and temporal scales, and are relevant in a range of physiographic settings, watershed land-use types, or social choices.
All of the indicators described here address ecological condition. The spatial scale differs among indicators, but generally applies either to characterizing a small watershed (e.g., 14-digit hydrologic unit code [HUC] or smaller) or to a particular site or reach of stream. Indicators associated with Level 1 Landscape assessments typically use remote sensing or GIS scenes and generally apply to watershed scales; however, the aggregation of data at a watershed scale also can be used to interpreted to apply the condition of a downstream point, whether it is a site or a reach.
Indicators that require field observations (Level 2 Rapid or Level 3 Intensive assessments) can provide information for interpreting ecological condition at the site level as well as upstream conditions. In fact, instream biotic indicators (e.g., Indices of Biotic Integrity [IBIs]) may reflect conditions upstream more than they do the surrounding habitat in the floodplain and riparian zone. Individual site conditions, if randomly selected within a watershed, can be aggregated to provide a statistically based estimate of a stream network at small watershed scales.
The development of indicators and their calibration requires substantial effort. The actual practice of assessing the condition of a site or a watershed, however, can take advantage of these efforts and apply them to resource management or regulatory programs. Consequently, managers do not incur the cost of development and calibration when indicators have been developed for the region of interest or for particular programmatic purposes. Indicators developed for the ASC generally use the same metrics used in indices for other regions, but they are calibrated within regional climatic, soil, biotic, and cultural conditions. If indicators are not calibrated within a physiographic region, they may be ineffective at separating variation caused by natural sources from those caused by human alterations.
There were two main thrusts in the ASC analyses of small watershed segments: (1) developing an integrated assessment protocol for simultaneous rapid assessment of the conditions of streams, adjacent wetlands, and adjacent riparian zones at an assessment point; and (2) developing improved geographic models for predicting the chemical and biological conditions of streams from watershed characteristics, particularly land cover.
Summary/Accomplishments (Outputs/Outcomes):
This is a summary for the NCER Web Site.
Scientists from Pennsylvania State University, East Carolina State University (ECU), and the Smithsonian Environmental Research Center (SERC) cooperated in the development of the integrated rapid method for simultaneously assessing the condition of streams and the adjacent wetlands and riparian zone. To develop and test the assessment protocols, field observations were collected from small watersheds throughout the Atlantic Slope. Twenty-four study watersheds (14-digit HUCs) were selected to represent the range of ecoregion and land-use types in the Atlantic Slope. Field teams sampled about 20 randomly selected stream locations within each selected watershed and measured conditions in the stream and near-stream zones. The resulting measurements were synthesized to produce a Stream, Wetland, and Riparian (SWR) index for the entire stream/near-stream complex and to develop other indicators that applied to separate components of that complex. ECU scientists supplemented this effort with an evaluation of the influence of beaver impoundments as a potential indicator.
SERC scientists led the geographic modeling effort, which focused on improving the methods and models for using watershed characteristics, particularly land cover, as landscape indicators of stream water quality and biotic condition. This effort exploited available water quality data from previous SERC studies of streams draining small watersheds and available data on physical, chemical, and biotic condition of streams from the Maryland Biological Stream Survey. We used statistical models to relate the chemical and biological data (dependent variables) to independent variables derived from analyzing digital watershed maps with a GIS. We focused on the independent variables, including physiographic province and land cover, especially cropland and developed land. We were especially interested in how the spatial arrangement of land cover moderated its influence on stream responses, so we explored new methods and metrics for accounting for two important aspects of spatial arrangement: the distance of disturbed areas to assessment points and the presence and distribution of riparian buffers along hydrologic flow paths connecting disturbed areas to streams. We used correlation, regression, multiple regression, and threshold analysis to relate responses to land cover, distance-weighted land cover, and new metrics describing riparian buffer distribution.
The development of the SWR index demonstrated that rapidly assessed field indicators can be tailored into effective tools for quantifying stream, wetland, and riparian condition within physiographic regions and land-use categories. The analysis also revealed patterns in stressor distributions among physiographic regions and social choices categories. Hydrologic modifications and measures of sediment and erosion were by far the most dominant stressors in all physiographic regions and land uses. Vegetation modification also was very prevalent, and invasive species were particularly common in Coastal Plain and mountain settings more than in the Piedmont. Coastal plain streams had fewer nearby stressors than streams in other physiographic provinces, but the numbers of stressors for urban lands were similar across provinces.
The geographic modeling effort developed enhanced methods for identifying and calibrating landscape indicators of stream responses, and those methods were applied to yield some specific recommended indicators. We identified a number of spatial challenges that arise in relating land cover to stream responses, and we presented statistical methods to surmount those challenges. Our work also produced methodological improvements in integrating community data into response indices and in automatically delineating watershed boundaries. Although land cover alone is a useful indicator of stream condition, we showed that indicator models can be improved by incorporating information on the spatial patterns of land cover through distance-weighting of source areas or through new functional riparian metrics that consider distribution of riparian buffers along hydrologic flow paths connecting source areas to streams. Our analyses show that both the amount and spatial arrangement of cropland and development in a watershed can have a significant impact on nutrient discharges. Similarly, the amount and spatial arrangement of developed land (or impervious surface) significantly affect the response thresholds of stream macroinvertebrate communities. Compared to traditional ways of quantifying riparian patterns, our functionally based riparian metrics were more interpretable and more independent of watershed land cover. Analyses incorporating those metrics show that riparian buffer configuration is correlated with reduced nutrient discharges in some but not all provinces within the Chesapeake Bay watershed.
The general conclusions summarized above are embodied in the example results for selected indicators presented in the next section.
Inverse-Distance Weighted Cropland
Croplands closer to water bodies can be stronger sources of sediment and nutrients to aquatic systems, whereas discharges from more distant croplands may be attenuated by a variety of processes along transport pathways before reaching a waterbody.
This index is based on the proportion of cropland in a watershed, modified by giving greater weight to cropland areas closer to waterbodies or sampling stations, while still including some effect of more distant croplands. The metric is calculated from digital land cover, elevation, and stream maps using a GIS. For every pixel in a watershed, we calculate the horizontal distance to a waterbody or sampling station along the steepest descent flow line determined by landscape topography. If the distance is measured to streams, flow lines derived from digital elevation surfaces must first be modified to match the stream maps. All pixels are weighted by the inverse of this distance (1/distance), and weighted cropland properties are estimated by dividing the sum of all weighted cropland pixels by the sum of all weighted pixels in the watershed.
In one application, distance weighting of cropland proportion improved predictions (higher r2) of stream nitrate concentrations measured in the Maryland Biological Stream Survey for small coastal plain watersheds (< 600 ha or 1,500 ac), but not for larger watersheds. In another test using stream chemistry data from 429 Chesapeake Bay subwatersheds in 4 physiographic provinces, distance weighting cropland (distance to streams) improved predictions of stream nitrate concentration the Coastal Plain, but not in the other 3 physiographic provinces. Distance weighted cropland proportion can range from 0 to 100 percent and will differ increasingly from simple cropland proportion as cropland distribution becomes less uniform, perhaps by preferential location of croplands on uplands or on floodplains.
The percentage of cropland in a watershed is a commonly used indicator of sediment and nutrient concentrations in streams. Distance weighting can improve predictions for small watersheds, but the benefits of distance weighting vary among physiographic settings.
Inverse-Distance Weighted Developed Land
Urban development adjacent to a stream reach and throughout the watershed can degrade stream ecosystems, but near-stream development may have a stronger effect. By weighting land cover near the reach more heavily, the metric preferentially emphasizes local, acute effects of development (e.g., riparian forest removal, dumping hazardous materials) while still incorporating the effects of more distant development (watershed-scale hydrological modification, nonpoint source runoff).
Inverse-distance-weighted (IDW) index of developed land gives greater emphasis to developed land near a feature of interest (e.g., a stream reach) than to developed land located farther away. The metric is calculated from digital land-cover and stream maps using a GIS. Within an individual watershed, every pixel is assigned a distance (meters) to an assessment point using simple Euclidean distance. Pixels then are weighted by the inverse of their distance (1/distance) to the assessment point. The sum of distance-weighted developed land (residential and commercial) is divided by the sum of distance-weighted total land in the watershed to yield a distance-weighted developed land percentage.
Relationships between developed land and macroinvertebrate assemblages were examined among 295 Coastal Plain streams in Maryland. Assemblages exhibited an ecological threshold between 21 and 32 percent when unweighted developed land in watersheds was used. Beyond 32 percent, the probability was almost 100 percent that all streams were impaired biologically. The apparent developed-land threshold dropped to as low as 18 percent, however, when weighted by its inverse distance to the sampled reach of stream, with 100 percent certainty of a threshold above 23 percent IDW developed land. IDW percent developed land can range from 0 to 100 percent, and deviates most significantly from unweighted percent developed land among watersheds with distinctly different spatial patterns of urbanization. Higher values indicate greater probabilities of stream impairment, particularly above 18-23 percent in Coastal Plain streams.
Developed land contributes to stream impairment. Distance-weighted developed land provides a more discriminating indicator of the threshold effects on stream macroinvertebrate communities than does the simple proportion of developed land, because distance-weighting accounts for the stronger effects of development near an assessment point without ignoring more distant land. The index also can be useful in targeting best management practices and ecological restoration by identifying not only what land use practices should be changed to improve conditions, but also by informing where in a watershed those changes would be most effective.
Source-Specific Mean Riparian Buffer Width
Riparian buffer width has long been considered a key measure for estimating buffer effects on water chemistry and other stream responses. Watershed scale analyses, however, have relied on the proportion of buffer within a fixed distance of streams as the measure of buffering potential. Our calculation of the source-specific mean riparian buffer width metric provides a more functionally based measure that focuses on that portion of the riparian system that actually is connected to a source area, and measures width along paths of likely hydrologic transport.
Source-specific mean riparian buffer width examines the potential of riparian buffers to reduce the effects of a specific land cover on aquatic systems. The source area can be any land cover type (such as cropland or developed land) that can affect stream responses. The metric is calculated from digital land cover, elevation, and stream maps using a GIS. Prior to analysis, digital elevation surfaces first must be modified to align with stream maps. Within a watershed, all surface flow paths leading downhill from source areas to a stream are identified, and then the width of riparian buffer (forest or wetland contiguous with streams) is calculated for each flow path. Mean width is averaged across all flow paths weighted by the source land cover area contributing to a flow path.
Mean riparian buffer width for crop lands was quantified for 503 small (6-48,000 ha) watersheds within 4 major physiographic provinces of the Chesapeake Bay drainage. Source-specific mean widths were compared with a more commonly used measure—the percentage of forest within 100 m of a stream. Source-specific mean buffer width for cropland captured large differences among watersheds that had similar values of forest with 100 m of streams.
Source-specific mean riparian buffer width considers only that part of the riparian system that is likely connected to a source area and then integrates the buffering potential along the likely lines of flow from source areas to streams. This indicator is derived from combining measurements about the characteristics of the riparian buffer along the source-to-stream transects. This functional connectivity provides a more discriminating indicator of buffering potential across whole watersheds than the common method of calculating the percentage of forest within a fixed distance of streams. It could be useful as a planning tool for identifying where the restoration of forested buffers in watersheds might be most effective, thus providing insight into more economically efficient approaches to stream and riparian restoration.
Stream, Wetland, Riparian (SWR) Index
Components of the sampling protocol are based on rapid assessment methods developed and tested by the EPA (e.g., Stream Habitat Assessment) and by the Cooperative Wetlands Center at the Pennsylvania State University (e.g., stressor checklist, riparian buffer score). Metrics used to compute the SWR Index were selected based on a conceptual model of their relationship to aquatic system condition.
An SWR Index was developed to produce simultaneous assessments of condition for these interrelated components of aquatic ecosystems. A GIS was used to select about 20 stream-centered points for 24 small watersheds stratified by mid-Atlantic ecoregions and land-use type. In 2003, aspects of hydrology, soils, vegetation, and topography were measured in one 100 m x 100 m plot per site using a rapidly implemented sampling protocol (< 2 hours). Observations of onsite stressors were recorded. Landscape metrics were computed from 1-km radius circles centered on each point. We combined the floodplain-wetland and stream measurements into an indicator of overall condition for small watersheds (SWR Index) and examined the relationship with the Landscape Index. Comparisons were made to assessments derived from existing biological and chemical data from intensive studies. This indicator is composed of the following floodplain-wetland metrics: buffer width, basal area, number of tree species, abundance of invasives, number of stressors, and these stream condition metrics: Stream Habitat Assessment score, incision ratio, and number of stressors.
The sampling protocol was applied at approximately 20 sites in each of 24 watersheds in the ASC study region, representing a range of land cover types and physiographic regions. Values of the index were compared with IBI values for fish and benthic macroinvertebrates in selected watersheds of the study region. For the most part, the SWR Index agreed well with these biotic indices. Agreement was better for macroinvertebrates at the site level and better for fish at the watershed scale. The SWR index also was compared with a landscape-level (GIS-based) index (Figure 1).
Figure 1. Nested SWR Index and Landscape Index
For sites where we had both Level 2 and Level 3 measurements, we found a highly significant correlation between the SWR Index and the benthic IBI, but the correlation with the fish IBI was weaker, and the link with nitrate was very weak. One would expect benthic invertebrates to be more influenced by site-level conditions than fish (which are more mobile) or nitrate (which integrates over a larger upstream area).
When Level 3 measurements were compared with the average SWR Index in their upstream contributing area, all three Level 3 indices were correlated with the SWR Index. The relationship was strongest for the benthic IBI, followed by the fish IBI, and nitrate. This suggests that looking at multiple sites in the upstream area may give us a broader representation of condition at a point. Although the correlation with the benthic IBI was somewhat weaker than the site-to-site comparison, the relationships of the SWR Index with fish and nitrate were strengthened.
We also compared the average of all SWR points in a HUC-14 watershed with the average of all Level 3 points for a HUC-14 watershed. The correlation was statistically significant for the fish IBI, nearly so for the benthic IBI, but not for the nitrate. The very small sample size made relationships difficult to discern and statistical significance difficult to achieve. The indication is, however, that our sample of 20 SWR points provides a reasonable estimate of biological condition, but not of chemical condition, at the watershed level.
Rapid assessment protocols for riparian and stream condition are important for identifying water and habitat quality problems at watershed scales. More intensive methods that require considerably more time still are necessary prior to deciding on restoration activities for specific projects. The rapid assessment methods, when applied to randomly chosen sites within a watershed, provide an unbiased evaluation at watershed scales. The SWR Index strives to assess all components of these aquatic ecosystems simultaneously, rather than to treat each separately. The SWR Index should complement existing stream, wetland, and floodplain monitoring programs. It is designed to be used with the Landscape Index. Since their development, both the SWR Index and Landscape Index have been incorporated into monitoring programs of specific units of the National Park Service and by the state of North Carolina.
Spot-Sampled Average Stream Nitrate Concentration
Nitrate often is the dominant chemical form of nitrogen lost from watersheds through storm runoff and stream flow, and nitrate is even more predominant over other nitrogen forms in the discharges from croplands, developed lands, and other areas disturbed by human activities. Nitrate is highly soluble, so it easily enters the soil and follows subsurface transport pathways to streams. Therefore, much of the nitrate loss from watersheds occurs in baseflow between storms, and nitrate concentrations are less temporally variable than concentrations of other nitrogen forms transported episodically on particles during storms.
Stream nitrate concentrations can be a direct measure of nitrogen pollution in streams and a potential predictor of aquatic biological responses. Nitrate concentration rises with increasing proportions of agricultural or developed land in a watershed. To measure spot-sampled average stream nitrate concentration, multiple water samples are collected from a stream reach during non-storm conditions, with at least one sample in each season. The samples are analyzed chemically for the concentration of nitrate, and the results are averaged to estimate the mean annual nitrate-nitrogen concentration in baseflow.
For 66 Chesapeake Basin subwatersheds of differing land cover proportions in 4 major physiographic provinces, we compared average spot-sampled nitrate concentration to flow-weighted average nitrate and total nitrogen concentrations measured with 1-3 years of weekly composite flow-proportional sampling at an automated stream sampling station. Average spot-sampled nitrate concentration was a very strong predictor of nitrate (r2 = 0.98) and total nitrogen (r2 = 0.98) concentrations from the much more labor intensive and expensive automated sampling. Spot sampling is a cost-effective approach to gauge nitrate pollution accurately across many sites in broad, regional studies or assessments.
Average spot-sampled nitrate concentration in stream water increased with increasing amounts of developed land or cropland in a watershed, ranging from 5 mg N/L in completely forested watersheds to 20 mg N/L or more in heavily agricultural watersheds. The increase per unit area is steeper for cropland, and the rate of increase for a land-use type can vary with physiographic setting.
The high degree of agreement between intensive automated sampling of streams and spot samples for nitrate in streams gives confidence that this indicator is reliable. Because of the cost savings, the approach can be applied to many more streams and reaches, and thus improve the effectiveness of water quality monitoring programs.
Beaver Impoundment Presence
Beaver (Castor canadensis Kuhl) populations have experienced a resurgence in recent years in the ASC study area. Beaver impoundments increase the surface area of wetlands, depth of flooding, and residence time of water relative to streams. These impoundments are found mostly in low order streams, which constitute the majority of stream networks and generally are closest to sediment and nitrate sources. Nearly everywhere in North American where beaver impoundments have been studied, they reduce the downstream transport of nutrients and sediments and alter the floodplain habitat from one of emergent marsh and forest to one dominated by submerged aquatic vegetation and shallow ponds.
In a study of 13 beaver impoundments in the inner coastal plain of North Carolina, impoundments significantly decreased nitrate and total suspended solids concentrations relative to control reaches. The presence of beaver impoundments should be considered a factor that improves water quality. This indicator, however, is very tentative at this point because the metrics have not been worked out to quantitatively predict the effects of individual impoundments on water quality at watershed scales.
The increase in water depth and wetland area from beaver activities convert forested habitats to other wetland types. This can be perceived as a negative effect by landowners, particularly if timber revenues are lost, agricultural land is flooded, or property is devalued in other ways. The open habitat created by beaver impoundments can lead to the proliferation of invasive and exotic species. These potential negatives must be weighed against the advantages of water quality improvement through reduction in nitrate concentrations and suspended sediments. Currently, it is not possible to evaluate the effect of an individual beaver impoundment on water quality within a watershed. Additional analyses, however, are ongoing.
Journal Articles on this Report : 1 Displayed | Download in RIS Format
Other subproject views: | All 45 publications | 6 publications in selected types | All 6 journal articles |
---|---|---|---|
Other center views: | All 166 publications | 51 publications in selected types | All 44 journal articles |
Type | Citation | ||
---|---|---|---|
|
Baker ME, Weller DE, Jordan TE. Comparison of automated watershed delineations: effects on land cover areas, percentages, and relationships to nutrient discharge. Photogrammetric Engineering & Remote Sensing 2006;72(2):159-168. |
R828684C003 (Final) |
not available |
Supplemental Keywords:
indicators, integrated assessment, wetland, stream, estuary, watershed, biological integrity, decisionmaking, ecosystem, environmental exposure and risk, geographic area, ecology, exposure indicators, bioindicators, land use, mid-Atlantic, hydrology, estuarine ecosystems,, RFA, Scientific Discipline, ENVIRONMENTAL MANAGEMENT, Geographic Area, Water, ECOSYSTEMS, Ecosystem Protection/Environmental Exposure & Risk, estuarine research, Hydrology, Water & Watershed, Ecosystem/Assessment/Indicators, Ecosystem Protection, Economics, Aquatic Ecosystems, Terrestrial Ecosystems, Ecological Monitoring, Mid-Atlantic, Ecological Risk Assessment, Ecology and Ecosystems, Biology, Watersheds, Ecological Indicators, Risk Assessment, ecological exposure, bioindicator, coastal ecosystem, degradation, biogeochemical study, remote sensing, water sheds, aquatic biota , ecosystem assessment, estuaries, optical indicators, nutrients, aquatic habitat, socioeconomics, submerged aquatic vegetation, biomonitoring, ecological assessment, ecosystem indicators, estuarine ecosystems, integrated assessment, Atlantic Slope Consortium, environmental stress, coastal ecosystems, integrative indicators, environmental indicators, water quality, ecology assessment models, watershed assessment, Chesapeake BayProgress and Final Reports:
Original AbstractMain Center Abstract and Reports:
R828684 Center for Integrated Multi‐scale Nutrient Pollution Solutions Subprojects under this Center: (EPA does not fund or establish subprojects; EPA awards and manages the overall grant for this center).
R828684C001 Integrated Assessment of Estuarine Ecosystems
R828684C002 Development of an Optical Indicator of Habitat Suitability for Submersed Aquatic Vegetation
R828684C003 Integrated Assessment of Watersheds
R828684C004 Socioeconomic and Institutional Research
The perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.
Project Research Results
6 journal articles for this subproject
Main Center: R828684
166 publications for this center
44 journal articles for this center