Assessing Wetland Condition on a Watershed Basis in the Mid-Atlantic Region Using Synoptic Land-Cover Maps

Home , Deer Creek (Maryland), Little Fishing Creek, White Deer Creek

ASSESSING WETLAND CONDITION ON A WATERSHED BASIS IN THE MID-ATLANTIC REGION USING SYNOPTIC LAND-COVER MAPS

ROBERT P. BROOKS*, DENICE H. WARDROP, and JOSEPH A. BISHOP Penn State Cooperative Wetlands Center, 302 Walker Building, Pennsylvania State University, University Park, PA 16802 USA (*author for correspondence, phone: 814-863-1596, fax: 814-863-7943, e-mail:[email protected])

Abstract. We developed a series of tools to address three integrated tasks needed to effectively manage wetlands on a watershed basis: inventory, assessment, and restoration. Depending on the objectives of an assessment, availability of resources, and degree of confidence required in the results, there are three levels of effort available to address these three tasks. This paper describes the development and use of synoptic land-cover maps (Level 1) to assess wetland condition for a watershed. The other two levels are a rapid assessment using ground reconnaissance (Level 2) and intensive field assessment (Level 3). To illustrate the application of this method, seven watersheds in Pennsylvania were investigated representing a range of areas (89–777 km2), land uses, and ecoregions found in the Mid-Atlantic Region. Level 1 disturbance scores were based on land cover in 1-km radius circles centered on randomly-selected wetlands in each watershed. On a standardized, 100-point, human-disturbance scale, with 100 being severely degraded and 1 being the most ecologically intact, the range of scores for the seven watersheds was a relatively pristine score of 4 to a moderately degraded score of 66. This entire process can be conducted in a geographic information system (GIS)-capable office with readily available data and without engaging in extensive field investigations. We recommend that agencies and organizations begin the process of assessing wetlands by adopting this approach as a first step toward determining the condition of wetlands on a watershed basis.

Keywords: wetlands, watersheds, land cover, condition assessment, Mid-Atlantic Region

1. Introduction The need for strategic restoration of aquatic ecosystems was strongly supported by the National Research Council (NRC) (1995) which called for integrated approaches. A more recent report by the NRC (2001) provided advice on how to implement and improve the wetlands mitigation process on a watershed basis, a process that includes restoration as one element of many options. To conduct watershed-based management of aquatic ecosystems, or “waters” as defined by the Clean Water Act, one must integrate information obtained at various scales, from site-level assessments of stream, lake, and wetland conditions to landscape-level land uses. The objectives of a project conducted by the Penn State Cooperative Wetlands Center (CWC) were to develop, evaluate, and integrate a series

Environmental Monitoring and Assessment 94: 9–22, 2004. c 2004 Kluwer Academic Publishers. Printed in the Netherlands. 10 BROOKS ET AL. of tools for use in inventorying, assessing, and restoring the ecological condition of wetlands and associated riparian areas. To comprehensively manage wetlands on a watershed basis, these three integrated stages need to be addressed (Brooks et al., 2002). If assessment of wetland condition as a function of human disturbance is considered in a watershed context, restoration potential of those same wetlands also can be considered because the proportion of wetlands falling below an acceptable level of condition, and their spatial location can be determined. For the purposes of this paper, we define condition as the state or quality of the wetland or watershed being studied as a function of physical, chemical, or biological parameters. Depending on the objectives of the assessment, the availability of resources, and the degree of confidence required in the results, there are three levels of effort available to address these tasks. In developing these assessment tools, the CWC has formulated a matrix that shows how con- ducting the three tasks over three levels of assessment is an integrated process (Figure 1). That is, an effort is made to relate information across tasks at an appropriate level of effort. In this paper we confine our discussion to assessing the condition of wetlands at Level 1, which is designed to produce a coarse assessment of wetland condition across a set of watersheds from remote-sensing imagery. An assessment conducted only from remote-sensing data is appropriate primarily for planning purposes (Figure 1). If one applies the collective Level 1 and 2 assessments and detects a problem or irregular “signal” within a specific area relative to an established reference condition, then an intensive Level 3 assessment using Hydrogeomorphic (HGM) Func- tional Models (Brinson, 1993; Smith et al., 1995) and Indices of Biologi- cal Integrity (IBIs) can be used to diagnose specific stressors (Karr and Chu, 1997) and provide design guidance for specific restoration projects. The use of reference sites has become increasingly more common as ecologists and managers search for reasonable and scientifically based methods to measure and describe the inherent variability in natural aquatic systems (Hughes et al., 1986; Kentula et al., 1992). The primary reasons to include reference sites in a regional assessment and restoration effort are the need to compare impacted or degraded sites to a least impaired set of attributes or benchmarks. The primary criterion for selecting reference sites involves choosing sites that represent ideal, near-pristine conditions represented by the least disturbed sites available, which is common for ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 11 stream assessments (Karr and Chu, 1997), or choosing sites that represent the best attainable conditions for a particular region even though they may not be pristine (Smith et al., 1995). Although reference sites often represent areas of minimal human disturbance (i.e., reference standards in HGM parlance; Smith et al., 1995), in many instances it is more useful to represent a range of environmental conditions across a landscape. It is also possible, as we propose in this paper, to designate reference standard watersheds based on similar criteria. Whether individual wetlands or entire watersheds are being considered, we define a reference set as a gradient of conditions, not just the least impacted elements.

INVENTORY ASSESSMENT RESTORATION

Level 1 Use existing map Map land uses in watershed; Produce synoptic watershed resources (NWI) of compute landscape metrics map of restoration potential wetlands

Level 2 Enhance inventory using Rapid site visit and stressor Select sites for restoration; landscape-based decision checklist; preliminary examine levels of threat from rules condition assessment surroundings

Level 3 Map wetland zones Apply HGM and IBI models Map specific sites for abundance using verified to selected sites for condition restoration; design projects inventory based on reference with reference data sets

Figure 1. Integrated tasks for wetland monitoring by watershed at three levels of effort.

2. Methods A synoptic map provides an overall visual representation of the watershed and can be used to help interpret decisions regarding site selection for sampling, protection, and restoration. We have modified the synoptic approach developed by Leibowitz et al. (1992) where geographic information system (GIS) data were used to locate watersheds or wetlands with specific characteristics. We were not able to acquire the necessary data layers (e.g., county-level digital soils data, sufficient hydrologic gauging stations) to apply this approach. We recommend that synoptic maps display at a minimum the most current land-use and land-cover data available. Although land-use patterns do not completely describe disturbance levels, they are usually highly correlated with landscape and wetland condition (O’Connell et al., 1998; Wardrop et al., 1998). During discussions with state environmental managers early on in this project, land use was identified as a preferred factor to portray in watershed maps used for assessing wetland condition. 12 BROOKS ET AL. A synoptic map provides a set of baseline conditions for comparing long-term changes, whether these changes involve degradation or restoration. The map can help identify potential landscape-level threats to parts of the watershed. Targeting of major projects, such as mitigation banks can be facilitated. Using a digital database for creating a synoptic map, a set of metrics for spatial analysis can be generated from GIS software programs to characterize the patterns of the landscape (e.g., proportional land cover, connectivity, Miller et al., 1997). Recommended resources for developing synoptic maps include the listed parameters, although only the first three items were required for the analyses presented here. The other parameters can be useful in other geographic regions and for Levels 2 and 3.

• current land use and land cover from Thematic Mapper (TM) satellite imagery • stream network (digitized 1:24,000 blue line database) • wetlands and water bodies (National Wetlands Inventory [NWI] digitized 1:24,000 base maps) • road network (digitized 1:24,000 database) • topography (Digital Line Graph [DLG] database) • hydric and non-hydric soils (digitized county soil surveys as available, STATSGO) • trends data (indicators of expected change, such as land-use conversion rates, population growth rates, intensity of landscape use)

Assessment at Level 1 serves as a screening tool to focus on broad areas of concern among watersheds in a region or within portions of the watershed. The entire process can be conducted in an office without engaging in field investigations. We have confirmed that the relative rank- ing of watersheds remains the same using the more intensive data from Levels 2 and 3 (Brooks, unpublished). The overall process begins with construction of a synoptic watershed map containing the best available wetlands inventory and land cover. Typically, the inventory stage involves using digitized NWI data. NWI data are now available for all U.S. Geological Survey (USGS) 7.5 min quadrangle maps in Pennsylvania on the NWI (http://wetlands.fws.gov) or Pennsylvania State University (http:/ /www.pasda.psu.edu) websites. Although NWI inventories typically do not show all wetlands present, they are the only digital source available for most of the United States. ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 13 Wetlands and attribute tables from each NWI quad must be tiled together in a GIS to produce the inventory for the designated watershed. The outer perimeter boundary of the watershed polygon is used to identify wetlands that are found within the watershed. Land-cover data was derived from satellite imagery that was classified into selected land-use types. In Pennsylvania, where forest is the predominant reference land cover, we are able to use percent forest as the primary metric to characterize the landscape surrounding each wetland sampled. Other regions may require multiple metrics to express human disturbance adequately. Although only coarse land-cover data are needed to conduct a Level 1 assessment, it may be prudent to compile the other GIS layers useful for assessments at Levels 2 and 3 at the same time. The appropriate number of points with wetlands to sample is not fixed, but depends on the number available from the digital NWI or enhanced inventory, the statistical confidence supporting a decision, the accessibil- ity of wetlands on private lands, and how to integrate data from field reconnaissance which is necessary in Levels 2 and 3. Based on discussions with statisticians involved with U.S. Environmental Protection Agency’s (U.S. EPA) Environmental Monitoring and Assessment Program (EMAP) (Don Stevens, personal communication), we chose 50 as the preferred number of sampling points for any watershed within the range of areas used in this study. This amount of sampling provides a reasonable level of statistical rigor while also allowing a reasonable expenditure of resources to conduct the remote-sensing and ground-based assessments. Sampling <20 wetlands per watershed would seem to be unadvisable, while sampling >100 wetlands per watershed may prove to be unwieldy. Due to the small size of some of our watersheds and the incompleteness of NWI coverage of actual wetlands, the Level 1 assessment was based on less than 50 points for two of our watersheds. The simplified steps necessary to complete the Level 1 process are listed below:

Level 1 – Inventory 1) Designate a watershed of interest for assessment (usually between 100–1,000 km2). 2) Use a GIS to compile land cover and NWI wetlands spatial data (stream and road layers will be needed for Level 2 assessments). 3) Compute the area of wetland by type for the watershed. 4) Assign each NWI wetland a unique number for use in a random sampling process. 14 BROOKS ET AL. Level 1 – Assessment 1) Using the watershed data from Level 1 inventory, select a random sample of NWI wetlands from the watershed (a minimum of 20 wetlands, 50 preferred). 2) Using the approximate center point (latitude, longitude) of each wetland, create a landscape circle of 1-km radius within the GIS file (Brooks et al., 1999). 3) Determine the percentage of land cover for each category of interest within the circle (percent forest is required for a Level 1 assessment in Pennsylvania, others are optional; see Table I for suggested list of landscape metrics. 4) Construct a graph of the Level 1 score (100 – percent forest) of these circles (x-axis) arrayed in order from lowest to highest (left to right) (e.g., Figure 4). 5) Compute an average Level 1 disturbance score for the watershed using all points sampled (e.g., average of all of the 100 minus percent forest measures for each wetland). 6) Determine the number of wetlands above and below the threshold of impairment. Note that the precise level of impairment has not yet been determined by federal or state regulatory agencies. At this time, only relative comparisons among wetlands or watersheds are possible, although based on our investigations at Levels 2 and 3, we suggest that impairment occurs above a Level 1 score of about 60 or higher. Level 1 – Restoration 1) Based on the results of the Condition Assessment, rank watersheds according to their level of wetland impairment at this level of effort. 2) Combine results from the wetland assessment with condition assessments of other waters (at this time independent assessments of other waters must be combined with those of wetlands using best professional judgment by examining whether there is spatial concurrence among impaired or degraded waters). 3) Determine whether all or portions of the designated watershed are in need of restoration (additional detailed information will be needed before restoration strategies, plans, or projects can be developed). ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 15

Table I. Level 1 score, watershed area, wetland area, stream length, and land-use composition of seven Pennsylvania watersheds. ) 2 ) 2

WATERSHED Level 1 Score (km Area NWI Wetland (km NWI Wetland Stream Length (km) Stream % Water % Forest % Transitional Herb % Perennial Annual Herb % % Barren Suburban % Urban %

Brandywine 56 776.60 18.71 846.81 0.46 43.10 6.94 20.67 18.14 0.87 7.07 2.75

Bushkill 13 389.88 47.69 314.52 2.61 84.61 4.01 2.05 0.21 0.91 5.20 0.42 Little Fishing 63 113.83 0.33 88.51 0.01 72.06 6.55 5.64 14.12 0.27 1.13 0.22 Shavers 45 163.17 1.24 190.08 0.25 63.14 11.51 11.54 13.12 0.28 0.11 0.06 Spring 66 378.22 0.95 328.09 0.09 38.03 9.91 11.85 24.84 0.70 10.07 4.50 White Deer 4 88.89 0.11 92.90 0.00 95.81 3.42 0.28 0.21 0.29 0.00 0.00 Yellow Breeches 63 571.11 8.77 475.75 0.20 54.49 7.45 10.29 18.83 1.71 2.11 4.93

Variables in Table 1. (Column headings in brackets) Derived from Land Cover data (1 km circles) [WATERSHED] - Name of study watershed All land cover values are derived for the reclassed [AREA] - Area of study watershed (km2). Stratified Suburban/Urban Land Cover data layer that was [NWI WETLAND] - Area of NWI wetland (km2). created for the PA GAP Analysis Project. [STREAM LENGTH] - Length of stream (km). [WATER %] - Percent of Water [FOREST %] - Percent of Forest [TRANSITIONAL %] - Percent of Transitional Vegetation [PERENNIAL %] - Percent of Perennial Herbaceous [ANNUAL %] - Percent of Annual Herbaceous [BARREN %] - Percent of Barren lands [SUBURAN %] - Percent of Suburban Vegetated [URBAN %] - Percent of Suburban Barren & Urban

Levels 2 and 3 – Assessment of wetlands Level 2 builds upon information collected and analysis conducted during Level 1 by combining Level 1 with ground reconnaissance of stressors, such as sedimentation, eutrophication, and habitat fragmentation (Adamus and Brandt, 1990). Once the sites or wetlands have been selected, and permission granted for access, a brief ground reconnaissance visit is conducted. During that visit, the classification of the wetland is confirmed or modified, information is collected on the number and type of stressors occurring at the site, and the buffer characteristics are described. A one- page sheet is used to record data on stressors and the extent and type of buffer surrounding the wetland. These data, plus the Level 1 assessment 16 BROOKS ET AL. data are combined into a Level 2 human disturbance score which allows comparisons among individual wetlands and the compilation of a more diagnostic overall average condition assessment score for wetlands in the watershed of interest. If one applies this assessment method and detects a problem or irregular “signal” within a specific area relative to an established or expected reference condition, then rapid ground reconnaissance (Level 2) or intensive assessments (Level 3) using HGM Functional Models and/or IBIs can be used to diagnose specific stressors.

3. Results We applied these methods on seven watersheds that varied by ecoregion, area, and impacts. Characteristics of all seven watersheds are shown in Table I. For two contrasting watersheds, we present two figures: 1) a synoptic map showing land cover and landscape circle plots (Figures 2 and 3); and 2) a graph of the condition of each wetland based on Level 1 assessments (Figure 4). The relative disturbance scores for wetlands in each watershed are displayed in Figure 5. After examining the watershed scores presented below, we labeled Bushkill Creek and White Deer Creek as reference watersheds because the majority of wetlands evaluated were surrounded by forested landscapes. Bushkill Creek – The 390 km2 Bushkill Creek watershed is located in the glaciated Pocono Region. It flows generally southeast directly into the Delaware River. The land cover is primarily forested with some residen- tial subdivisions (Figure 2). We located 50 points in this watershed. The average Level 1 disturbance score was a relatively low 13 (Figure 4), and therefore, this watershed is considered to be in reference condition based on best professional judgment and unpublished data. Spring Creek – The Spring Creek watershed (378 km2) is located in Centre County, PA in the vicinity of State College and Bellefonte. It is located in the Ridge and Valley ecoregion, flowing north into Bald Eagle Creek and the West Branch of the Susquehanna. The land cover is mixed, with agricultural (i.e., annual herbaceous cover class, Figure 3) and ur- banizing (vegetated suburban and urban cover classes, Figure 3) activi- ties located in the valleys in proximity to many of the wetlands and forests on the ridges (Table I, Figure 3). A total of 50 points were assessed. For Spring Creek, 93% (n=46) of the wetlands assessed are located within 100 m of a stream, and are considered to be associated with the riverine ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 17

Figure 2. Land cover for Bushkill Creek, Pike County, Pennsylvania with wetland and stream sample locations. system, although all of these may not be subject to flooding. About 7% (n=4) are isolated from the stream network. The average Level 1 score for wetlands in Spring Creek was 66 (Figure 4). White Deer Creek – The White Deer Creek watershed, although relatively small (89 km2), is among the least disturbed watersheds of this size found in central Pennsylvania (Table I), and thus, qualifies as a reference watershed. Because of its small size, only 15 points were available from NWI wetlands, although others are known to exist. White Deer Creek is located primarily in the Bald Eagle State Forest in the Ridge and Valley ecoregion, and flows east directly into the Susquehanna River near Milton. The watershed is primarily forested with minor disturbances from gravel roads and cabins. The only major disturbance is the Interstate-80 high- way corridor located along the northern edge. The average Level 1 disturbance score was low, 4, as expected due to the high proportion of mature forest cover (Figure 5). Little Fishing Creek – This watershed is also relatively small and was chosen because it has been paired with White Deer Creek in previous 18 BROOKS ET AL.

Figure 3. Land cover for Spring Creek, Centre County, Pennsylvania with wetland and stream sample locations. studies (e.g., Miller et al., 1997). This 114 km2 Ridge and Valley watershed has forested headwaters partially located in another section of Bald Eagle State Forest and intensively farmed valleys (Table I). The mainstem flows generally north into Fishing Creek and then into the West Branch of the Susquehanna. We located 50 points in Little Fishing Creek. The average Level 1 disturbance score was 63, reflecting the agricultural influence on wetlands, particularly along the mainstem in the valley (Figure 5). Shavers Creek – The 163 km2 Shavers Creek watershed has a mixed land use, with significant amounts of forest cover, particularly on the ridges, with farming and small towns dominating the valleys (Table I). It too is located in the Ridge and Valley ecoregion, flowing south into the Juniata River. We located 40 points in this watershed. The average Level 1 disturbance score was 45 (Figure 5). Yellow Breeches – This 571 km2 watershed flows east directly into the Susquehanna River south of Harrisburg. It is located in the Ridge and Valley ecoregion. The watershed’s tributaries drain forested areas from the south, located partially in Michaux State Forest. Very few tributaries flow from the agricultural lands to the north which are dominated by karst ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 19

100.00

90.00 Spring Creek Bushkill Creek

80.00

70.00

60.00

50.00

40.00 Level 1 Score 30.00

20.00

10.00

0.00 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49

Rank Order of Sites by Decreasing % Forest

Figure 4. Level 1 scores for wetlands in the Bushkill Creek and Spring Creek watersheds.

100

30 Mean Level 1 Score Level Mean 20

0 White Deer Bushkill Shavers Brandywine Yellow Little Spring Creek Creek Breeches Fishing Creek Creek

Figure 5. Mean Level 1 scores for wetlands in seven watersheds in Pennsylvania.

geology (Table I). We located 49 points in this watershed. The average Level 1 disturbance score was 63 (Figure 5). Brandywine Creek – The Brandywine Creek watershed is located in the Piedmont ecoregion of southeastern Pennsylvania. This 777 km2 watershed flows south into the Delaware River, southwest of Philadelphia. Land use is mixed, with extensive agriculture, increasing urbanization, 20 BROOKS ET AL. and some forest patches (Table I). We located 49 points in this watershed. The average Level 1 disturbance score was 56 (Figure 5).

4. Discussion and Conclusions To illustrate the application of this method of wetland assessment, the seven watersheds investigated were chosen to represent a range of size classes, land uses, and ecoregions. Level 1 disturbance scores, based on percent forest in 1 km radius landscape circles, ranged from a relatively pristine score of 4 to a moderately disturbed score of 66 (Figure 5). For Spring Creek, the Level 1 score of 66 was only slightly less than the Level 2 score of 70, suggesting that there can be relatively close correspondence between assessments using the remote-sensing data versus those that add rapid field reconnaissance for improved stressor identification (Brooks et al., 2002). Additional unpublished data from other watersheds studied by the CWC support this assertion. The seven watersheds investigated varied widely in area by an order of magnitude, from 89–777 km2. The smaller watersheds with areas of about 100 km2, are typical of watersheds in the 14-digit hydrologic unit code (HUC) category compiled by the USGS. Two of these three watersheds, White Deer Creek and Shavers Creek, had substantially less than 50 wetlands identified from NWI maps. Thus, 14-digit HUC watersheds would appear to be the smallest sampling unit appropriate for this method. An aggregation of 14-digit HUC watersheds to the 11-digit HUC category may be the most manageable size for Level 1 assessments. This level of aggregation is represented by the other watersheds studied. The synoptic land-use map produced during a Level 1 wetlands assessment provides an overall visual representation of a watershed (Figures 2 and 3). We propose that reference watersheds, such as Bushkill Creek and White Deer Creek, can be identified based on these analyses. A synoptic map also provides a set of baseline conditions for comparing long-term changes, whether these changes involve degradation or restoration. The map can help identify potential landscape-level threats to parts of a watershed. Targeting of major restoration projects, such as mitigation banks can be facilitated by identifying portions of a watershed where wetlands are most likely degraded based on surrounding land use. This approach has sufficient rigor to be used as a planning tool to help priori- tize watersheds for further investigations, but should not be used without confirming ground reconnaissance for locating or designing specific ASSESSING WETLAND CONDITION USING SYNOPTIC LAND-COVER MAPS 21 restoration or mitigation projects. This approach is simple and cost- effective to implement, such that every 14- to 11-digit HUC watershed in a state or region could be assessed within a relatively short period of time and for modest expenses. Using these methods, watersheds can be priori- tized as to what is an appropriate level of protection or restoration necessary for wetlands within each watershed. Once an individual watershed has been targeted for further investigation, protocols for Levels 2 and 3 can be implemented to confirm these findings and provide wetland- specific, quantitative data for identifying primary stressors and selecting restorative practices.

Acknowledgements This project was supported by contract CD993569, Watershed-Based Protection for Wetlands in Pennsylvania, from the Pennsylvania Department of Environmental Protection, Division of Waterways, Wetlands, and Erosion Control, through funding provided by Region 3 of the U.S. Environmental Protection Agency, Wetlands Program, State/ Tribal Development Grants Program. The Penn State Cooperative Wetlands Center is administered jointly through the Penn State Institutes of the Environment and the School of Forest Resources. We appreciate the assistance of K. Reisinger, K. Heffner, S. Freyermuth, A. Brunner, R. Eyerly, J. Rubbo, K. Saacke-Blunk, D. Vetter, R. Sumner, M. Kentula, and D. Stevens.

References

Adamus, P. R., and Brandt, K.: 1990, ‘Impacts on quality of inland wetlands of the United States: A survey of indicators, techniques and applications of community-level biomonitoring data’, EPA/600/3-90/073, U.S. Environmental Protection Agency, Environmental Research Laboratory, Corvallis, OR, USA, p. 406. Brinson, M. M.: 1993, ‘A hydrogeomorphic classification for wetlands’ Technical Report WRP- DE-4, U.S. Army Corps of Engineers, Waterways Experiment Station, Wetlands Research Program, Washington, DC, USA, p. 79+app. Brooks, R. P., Wardrop, D. H. and Perot, J. K.: 1999, ‘Development and application of assessment protocols for determining the ecological condition of wetlands in the Juniata River Watershed’, EPA/600/R-98-181, U.S. Environmental Protection Agency, Corvallis, OR, USA, p. 30. Brooks, R. P., Wardrop, D. H., Cole, C. A. and Reisinger K. R.: 2002, ‘Using reference wetlands for integrating wetland inventory, assessment, and restoration for watersheds,’ R.W. Tiner (compiler), Watershed-based wetland planning and evaluation, A collection of papers from the Wetland Millennium Event, Quebec City, Quebec, Canada, August 2000, Distributed by Association of State Wetland Managers, Inc., Berne, NY, USA, p. 141. 22 BROOKS ET AL.

Hughes, R. M., Larsen, D. P., and Omernik, J. M.: 1986, ‘Regional reference sites: a method for assessing stream potentials’, Environ. Manage. 10 (5), 629–635. Karr, J. R., and Chu, E. W.: 1997, ‘Biological monitoring and assessment: using multimetric indexes effectively’, EPA 235-R97-001, University of Washington, Seattle , WA. USA, p. 149. Kentula, M.E., Brooks, R.P., Gwin, S.E., Holland, C.C., Sherman, A.D. and Sifneos, J.C.: 1992, Wetlands. An Approach to Improving Decision Making in Wetland Restoration and Creation, Island Press, Washington, DC, USA, p. 151. Leibowitz, S.G., Abbruzzese, B., Adamus, P.R., Hughes, L.E. and Irish, J.T.: 1992, ‘A synoptic approach to cumulative impact assessment: A proposed methodology’, EPA/600/R-92/ 167, U.S. Environmental Protection Agency, Environmental Research Laboratory, Corvallis, OR, USA, p. 127. Miller, J.N., Brooks, R.P. and Croonquist, M.J.: 1997, ‘Effects of landscape patterns on biotic communities’, Landscape Ecol. 12, 137–153. NRC: 1995, Wetlands, Characteristics and Boundarie, National Academy Press, Washington, DC, USA, p. 307. NRC: 2001, Compensating for wetland losses under the Clean Water Act, National Academy Press, Washington, DC, 348 pp. O’Connell, T.J., Jackson, L.E. and Brooks, R.P.: 1998, ‘The bird community index: a tool for assessing biotic integrity in the Mid-Atlantic Highlands’, Rep. No. 98-4, Penn State Cooperative Wetlands Center, prepared for the U.S. Environmental Protection Agency, Washington, DC, USA, p. 57. Smith, R.D., Ammann, A., Bartoldus, C. and Brinson M.M.: 1995, ‘An approach for assessing wetland functions using hydrogeomorphic classification, reference wetlands, and functional indices’, Technical Report WRP-DE-9, U.S. Army Corps of Engineers, Waterways Experiment Station, Wetlands Research Program Washington, DC, USA, p. 79. Wardrop, D.H., and Brooks, R.P.: 1998, ‘The occurrence and impact of sedimentation in central Pennsylvania wetlands’, Environ. Monit. Assess. 51,119–130.