Using Vegetation to Assess Wetland Condition: a multimetric approach for temporarily and seasonally flooded depressional wetlands and herbaceous-dominated intermittent and ephemeral riverine wetlands in the northwestern glaciated plains ecoregion, Montana Prepared for: Montana Department of Environmental Quality and U.S. Environmental Protection Agency By: W. Marc Jones Montana Natural Heritage Program Natural Resource Information System Montana State Library February 2004 MONTANA Natural Heritage Ftogtam Using Vegetation to Assess Wetland Condition: a multimetric approach for temporarily and seasonally flooded depressional wetlands and herbaceous-dominated intermittent and ephemeral riverine wetlands in the northwestern glaciated plains ecoregion, Montana Prepared for: Montana Department of Environmental Quality and U.S. Environmental Protection Agency DEQ Contract Number: 203025 By: W. Marc Jones MONTANA Natuial Heritage Ptt^jtam ^It ^-^ MUSTAFA j^/^Ma y^WTiHA ^T^itate If jIV Natuial Kesouice ^ Library ^^JjjP IniomiatJQn System © 2004 Montana Natural Heritage Program P.O. Box 201800 • 1515 East Sixth Avenue • Helena, MT 59620-1800 • 406-444-5354 This document should be cited as follows: Jones, W.M. 2004. Using Vegetation to Assess Wetland Condition: a multimetric approach for temporarily and seasonally flooded depressional wetlands and herbaceous-dominated intermittent and ephemeral riverine wetlands in the northwestern glaciated plains ecoregion, Montana. Report to the Montana Department of Environmental Quality and the U.S. Environmental Protection Agency. Montana Natural Heritage Program, Helena, MT. 34 pp. plus appendices. Summary The Montana Department of Envi- ronmental Quality is implementing a com- prehensive wetland monitoring and assess- ment program to evaluate the condition of the state's wetlands. As part of this effort, the Montana Natural Heritage Program is developing site-level numerical vegetation biocriteria for wetlands. Assessing wetland condition by measuring the response of the biological community has been successfully demonstrated in many wetland systems us- ing a variety of taxa. This study attempted to evaluate wetlands by measuring vegeta- tion response to anthropogenic stressors for temporarily and seasonally flooded depres- sional and herbaceous-dominated intermit- tent and ephemeral riverine wetlands in the northwestern glaciated plains ecoregion in north-central Montana. Sample wetlands were ranked along a human disturbance gra- dient based on site- and local landscape- level factors. Vegetation attributes that changed predictably along this gradient were identified and combined into a multimetric index for each wetland type. These indices were significantly related to wetland condi- tion, as measured by the human disturbance gradient, for both depressional and riverine wetlands. When wetlands were divided into three disturbance categories (reference con- dition, moderately disturbed, and severely disturbed), vegetation metrics were able to correctly classify 73% of depressional and 86% of riverine wetlands sampled. The multimetric index for depressional wetlands responded primarily to on-site agricultural disturbance and was comprised of four met- rics: the floristic quality index and relative cover of native perennials, species with a coefficient of conservatism > 4, and exotic species. The riverine multimetric index re- sponded primarily to on-site grazing inten- sity and included the richness of native per- ennials, Simpson diversity index, propor- tionate richness of tolerant species, relative cover of intolerant species, and floristic quality index. Another aspect of this study was to evaluate the effectiveness of classifying wa- tersheds (5^^-level U.S. Geological Survey hydrological units) into disturbance catego- ries based on land use patterns. Watershed- scale disturbance categories showed no cor- relation with either wetland condition as measured by site-level and smaller-scale dis- turbance measures or vegetation metrics. Smaller-scale disturbance factors appear to be more important in determining condition of sampled wetlands in this study. Ill IV Table of Contents Summary iii Introduction 1 Methods 2 Study Area 2 Geology and Climate 2 Depressional Wetlands 2 Riverine Wetlands 3 Upland Vegetation 4 Study Design 4 Watershed Ranking 5 Targeted Sampling 6 Data Collection 6 Data Analysis 8 Human Disturbance Parameters 8 Multimetric Methods 10 Multivariate Methods 15 Results 16 Watershed Disturbance Categories 16 Depressional Wetlands 16 Metrics and the Multimetric Index 16 Vegetation Community Response 19 Riverine Wetlands 20 Metrics and the Multimetric Index 22 Vegetation Community Response 24 Discussion 24 Acknowledgments 30 Literature Cited 30 Appendix A. Species Lists for Depressional and Riverine Wetlands A-1 Appendix B. Photographs of Sites Representative of Reference Condition, Moderately Disturbed, and Severely Disturbed Wetlands B-1 List of Tables Table 1 . Human disturbance factors used to rank 5*-level watersheds in the Middle Milk sub- basin 7 Table 2. Functional weights assigned to land cover types 7 Table 3. Disturbance factors used to develop human disturbance indices for depressional and riverine wetlands 12 Table 4. Vegetation attributes initially considered as potential metrics for depressional and riverine wetlands and their predicted response to increasing human disturbance 13 Table 5. Ranges of attribute values for metric scoring categories for temporarily and seasonally flooded depressional wetlands 21 Table 6. Species indicative of severely disturbed and reference condition temporarily or seasonally flooded depressional wetlands 24 Table 7. Ranges of attribute values for metric scoring categories for herbaceous-dominated intermittent and ephemeral riverine wetlands 26 List of Figures Figure 1. Study area and locations of sample wetlands 3 Figure 2. Extent of national wetland inventory coverage in the study area 6 Figure 3 . Selected 5^^-level watersheds and their relative disturbance categories 8 Figure 4. Schematic illustrating how temporarily and seasonally flooded depressional wetlands were ranked along a human disturbance gradient 10 Figure 5. Schematic illustrating how herbaceous-dominated intermittent and ephemeral riverine wetlands were ranked along a human disturbance gradient 1 1 Figure 6. Bar graphs of disturbance measures by watershed disturbance categories for temporarily and seasonally flooded depressional wetlands 17 Figure 7. Bar graphs of disturbance measures by watershed disturbance categories for herbaceous-dominated intermittent and ephemeral riverine wetlands 17 Figure 8. Bar graphs of vegetation metrics by watershed disturbance categories for temporarily and seasonally flooded depressional wetlands 18 Figure 9. Bar graphs of vegetation metrics by watershed disturbance categories for herbaceous- dominated intermittent and ephemeral riverine wetlands 19 Figure 10. Scatter plots of attribute values against site disturbance index for temporarily and seasonally flooded depressional wetlands 20 Figure 1 1 . Relationship between multimetric index and site disturbance index for temporarily and seasonally flooded depressional wetlands (n = 30) 21 Figure 12. The predicted membership of temporarily and seasonally flooded depressional wetlands to disturbance categories compared with actual group membership 22 Figure 13. Graphical representation of the NMS ordination of sampled temporarily and seasonally flooded depressional wetlands 23 Figure 14. Scatter plots of attribute values against site disturbance index for herbaceous- dominated intermittent and ephemeral riverine wetlands 25 Figure 15. Relationship between multimetric index and site disturbance index for herbaceous- dominated intermittent and ephemeral riverine wetlands (n = 22) 26 Figure 16. The predicted membership of herbaceous-dominated intermittent and ephemeral riverine wetlands to disturbance categories compared with actual group membership 27 Figure 17. Graphical representation of the NMS ordination of sampled herbaceous-dominated intermittent and ephemeral riverine wetlands 28 VI Introduction Wetlands are critically important systems that provide numerous biological and economic benefits, including groundwa- ter recharge, filtration and storage of sedi- ments, nutrients, and pollutants, and flood- water storage and attenuation, as well as providing habitat to numerous species across a broad array of taxa (Brinson et al. 1981, Keddy 2000, Hauer et al. 2002a, b). They are essential to the maintenance of regional biodiversity, and, in the case of riparian habitats, provide structural habitat diversity otherwise lacking in semi-arid areas (Patten 1998). Consequently, the significance of wetlands is disproportionate to their physical extent on the landscape, especially in semi- arid regions such as Montana (Finch and Ruggiero 1993, Patten 1998). Yet despite their importance to both humans and wild- life, an estimated 25% of Montana's wet- lands have been lost since 1780 (Dahl 1990). Although in the last 30 years the number of regulatory and incentive-based wetland con- servation programs has increased considera- bly, wetlands continue to be lost and de- graded nationwide (U.S. EPA 1994, Dahl 2000, National Research Council 2001). To improve wetland conservation in Montana, the Montana Department of Envi- ronmental Quality has initiated a compre- hensive statewide wetland monitoring and assessment program. The goals of this pro- gram are to characterize the condition and extent of Montana's wetlands and to identify and document which anthropogenic stressors are most limiting to wetland health statewide and within regional watersheds. The im- plementation of this program will help pri- oritize statewide wetland conservation and restoration efforts. A component of this overall effort is to develop site-level nu- merical biocriteria for different wetland types within broad ecoregions across the state. Biological assessments can be accu- rate and cost-effective tools to assess wet- land condition and measure impairment (Karr and Chu 1999). Because biota inte- grate multiple physical and chemical pa- rameters, directly measuring the biotic community's response to anthropogenic stressors can be the most direct means to evaluate the effect of those stressors on wet- land condition and function (Danielson 2002). The utility of using biota to assess wetlands and streams has been demonstrated for various taxa, including fish, diatoms, benthic and terrestrial macroinvertebrates, birds, and vegetation (Karr 1991, DeKeyser 2000, Helgen and Gernes 2001, Kimberling et al. 2001, Mack 2001, Bryce et al. 2002, Fore and Grafe 2002, but see Heino et al. 2003, Tangen et al. 2003). The effective- ness of this method has already been dem- onstrated for perennial wetlands in Montana by Apfelbeck (2001), who developed biocri- teria for diatoms and macroinvertebrates. The purpose of this study was to use a multimetric approach (Karr and Chu 1999) to develop numerical vegetation biocriteria for depressional and riverine wetlands in the northwestern glaciated plains ecoregion in Montana. Many of the depressional and riv- erine wetlands in this ecoregion have a rela- tively brief inundation period in the spring and may be dry through most of the growing season. Therefore, plants were chosen as response taxa because vegetation is a good indicator of wetland condition and may be especially useful in wetlands that are only seasonally or ephemerally flooded. Plant species are diverse, have rapid growth rates, and respond directly to environmental change; additionally, it is easy to quantify shifts in plant community composition (Fennessy et al. 2002). Vegetation is also an important habitat variable for numerous in- vertebrate and vertebrate animals. Vegeta- tion metrics, therefore, have the possibility of integrating wetland condition and wildlife habitat assessments. Methods Study Area Geology and Climate The study area encompassed the Middle Milk sub-basin (4*-level U.S. Geo- logical Survey hydrologic unit code 10050004) and adjoining areas (Figure 1). This region lies within Blaine, Hill, Phillips, and Valley Counties in north-central Mon- tana. The area is part of the northwestern glaciated plains ecoregion (Woods et al. 1999) and is characterized by plains, ter- races, floodplains, and morainal landforms formed in glacial till, gravel deposits, and alluvium over shale, clay shale, sandstone, and siltstone (Nesser et al. 1997). Most of the study area is underlain by the marine- origin shale and clay-shale of the Bearpaw and Claggett formations. Other geologic substrates include sandstone and sandy shale in the breaks along the Milk River and Qua- ternary-age alluvium in the valley bottom of the Milk River and its larger tributaries. The region's climate is semi-arid and continental, with cold winters and warm summers. Mean temperatures range from -9,TC in January to 20. FC in July at Havre and from -14.2^C in January to 18.6^C in July at Opheim; mean annual precipitation at these stations is 291 mm and 303 mm, re- spectively (Western Regional Climate Cen- ter 2004). Most precipitation falls in late spring and early summer and occurs as steady, soaking frontal system rains. Sum- mer rainfall comes mainly from convection thunderstorms that typically deliver bursts of intense rain in scattered locations. These storms are often accompanied by large- diameter hail and flashfloods. Where rain- fall exceeds evapotranspiration, conditions are suitable for agriculture, particularly ce- real grains. The landscape is rolling prairie char- acterized by modest vertical relief. Eleva- tions range from 600 m a.s.l. along the Milk River at Glasgow to 915 m a.s.l. near Opheim. The region's gentle topography is the product of past glacial scour and deposi- tion. Much of the area is mantled by depos- its of glacial till, outwash, and drift up to 30 m thick (Nesser et al. 1997). Another aspect of semi-arid, conti- nental climates is extreme year-to-year vari- ability in precipitation. Severe drought con- ditions occur on average in two out of every ten years. Climate data from Redstone, Montana, which is comparable to the study area, indicate that one year in ten will have a total annual precipitation of less than 200 mm or more than 450 mm (Richardson and Hanson 1977). Depressional Wetlands Depressional wetlands, known as prairie potholes, occur throughout the study area but are most abundant north of the Milk River on gently rolling prairie terrain. Prai- rie potholes form in small, shallow depres- sions. These are primarily of glacial origin and many potholes were created when stranded ice blocks melted following glacia- tion. In the study area, these wetlands aver- age less than 0.5 ha in size and are often only ephemerally flooded. Potholes flood seasonally in the spring to early summer and are sometimes inundated for as little as a few weeks in spring (Kantrud et al. 1989). Vegetation in these wetlands is primarily structured along a hydrological gradient. Plant communities occur as concentric zonal bands, depending on each zone's relative period of inundation (Johnson et al. 1987, van der Valk and Welling 1988). Drier, temporarily flooded potholes are dominated by western wheatgrass (Pascopyrum smithii) and needle spikerush {Eleocharis acicu- Legend Middle Milk Syb-basirt Sampling Locations o Depression a I A Riverine Figure 1. Study area and locations of sample wetlands. laris). As the inundation period increases and wetlands become seasonally flooded, foxtail barley (Hordeum jubatum) and common spikerush {Eleocharis palustris) become dominant. Prairie potholes receiv- ing saline groundwater inputs or that occur in more alkaline soils often support salt- tolerant species, such as Nuttall's alkaligrass (Puccinellia nuttalliana), saltgrass {Dis- tichlis spicata), bearded sprangletop {Lep- tochloa fusca ssp. fascicularis), and com- mon threesquare {Schoenoplectus pungens). Semipermanently flooded wetlands, which retain water into late summer and support hydrophytic vegetation, such as broadleaf cattail (Typha latifolia) and hardstem bul- rush {Schoenoplectus acutus), are rare in the study area. Riverine Wetlands Riverine wetlands are extremely di- verse across the study area. These systems encompass a wide range of natural variabil- ity and differ considerably depending on hy- drology, geomorphology, and time since last flood disturbance. Riparian habitats range from oxbow marshes and cottonwood gal- lery forests along the Milk River and its lar- ger perennial tributaries to mesic herba- ceous-dominated communities along small, ephemeral drainages. Study wetlands were confined to Milk River tributaries, which are generally small and often intermittent or ephemeral. Vegetation is often shrub- or herbaceous- dominated, although narrow, discontinuous bands of plains cottonwood (Populus del- toides), box-elder {Acer negundo), and green ash {Fraxinus pennsylvanica) occur sporadi- cally along floodplains and terraces. De- pending on a site's hydrologic potential, channels and floodplains may be dominated by hydrophytic vegetation, such as Nebraska sedge (Carex nebrascensis), water sedge (Carex aquatilis), or common spikerush, or by mesic vegetation, such as western snow- berry {Symphoricarpos occidentalis). Woods rose {Rosa woodsii), clustered field sedge {Carex prae gracilis), tufted hairgrass {Deschampsia caespitosa), or western wheatgrass. Saltgrass, common three- square, and black greasewood {Sarcobatus vermiculatus) are common along more alka- line streams. Terraces often support com- munities of silver sage {Artemisia cana) and western wheatgrass. Upland Vegetation The native upland vegetation is a mix of short- and mid-grass prairie commu- nities intermixed with shrub steppe. Steppe vegetation is the result of a semi-arid conti- nental climate: the highly variable precipita- tion favors shallow-rooted herbaceous per- ennial grasses and deep-rooted shrubs over forests or woodlands. Shrub steppe vegeta- tion is characterized by open stands of silver sagebrush or Wyoming big sagebrush {Ar- temisia tridentata ssp. wyomingensis) over an herbaceous layer dominated by western wheatgrass, blue grama {Bouteloua gracilis), or needle-and-thread {Hesperostipa comata). The co-occurrence of short- and mid-grass prairies is due to climatic variability. Shorter, drought-resistant grasses such as blue grama increase in abundance during times of drought, whereas mid-grasses, such as the rhizomatous western wheatgrass and the bunch-forming prairie junegrass {Koel- eria macrantha) and needle-and-thread, in- crease under more favorable moisture condi- tions. Study Design Bioassessments and multimetric in- dices, such as indices of biological integrity (IBIs), are designed to detect the response of a biological community to human distur- bance. Central to developing a multimetric index is sampling the target population (in this case depressional and riverine wetlands) across a human disturbance gradient (Teels and Adamus 2002). Two basic sample de- signs are available: probabilistic (e.g., stratified random) and targeted. Probabilis- tic designs are more powerful in that they allow inferences to be made from the sample population (i.e., sampled depressional wet- lands) to the larger population of concern (i.e., all temporarily or seasonally flooded depressional wetlands within the study area). Thus probabilistic designs allow wetland condition to be characterized at a watershed scale and the proportion of wetlands that meet minimum aquatic life uses to be de- termined. Because of the greater inferential power of a probabilistic sample design, wet- lands were initially sampled using the strati- fied random procedure described under the Watershed Ranking section below. Unfortunately, a potential shortcom- ing in probabilistic designs is that the ex- tremes of the human disturbance gradient will be under sampled (Danielson 2002). This is a severe limitation to the develop- ment of an IBI, which depends on the com- parison of least- to most-disturbed wetlands (Karr and Chu 1999, Teels and Adamus 2002). Because of this, the U.S. EPA has recommended using targeted sampling to develop IBIs (Danielson 2002, Teels and Adamus 2002). Indeed, an examination of site data collected in the first field season of this project revealed that reference condition wetlands and, in the case of depressional wetlands, highly disturbed wetlands, had not been adequately sampled. Additional wet- lands were therefore inventoried as de- scribed under Targeted Sampling below. Two wetland classes, depressional and riverine, were sampled. The depres- sional class was restricted to temporarily and seasonally flooded wetlands as defined by Cowardin et al. (1979) and mapped by the National Wetland Inventory (NWI). Tem- porarily and seasonally fiooded wetlands make up the vast majority of prairie potholes in the study area. According to NWI cover- age in the Middle Milk sub-basin (Figure 2), 68% of potholes are classified as temporar- ily fiooded, 30% as seasonally fiooded, and 2%) as semipermanently fiooded. Riverine sampling locations were chosen from Milk River tributaries that ranged from 0-2%) val- ley slope. Initial criteria for site selection are presented in the following section. Watershed Ranking Wetlands were initially sampled us- ing a stratified random design. The 4^^-level Middle Milk sub-basin is comprised of 24 5*-level watersheds. These watersheds were ranked based on 12 factors that repre- sent landscape-scale surrogates of human disturbance (Table 1). Factor values were calculated for each watershed using a geo- graphical information system (GIS; Arc- View 3.2, ESRI, Redlands, California 92373). To minimize scaling issues among factors, factor values were rounded to the nearest integer, and watersheds were ranked based on those rounded values with ties re- ceiving the same rank. For example, three watersheds with road densities of 1.98, 2.29, and 3.15 (rounded values of 2, 2, and 3) would be ranked 1, 1, and 2, respectively. Watersheds were ranked based on ascending values, except for percent federal land, wil- derness, and land cover, which were ranked by decreasing values. Sample watersheds were then selected based on their overall mean rank. To evaluate the land cover category, each cover class was assigned a weight be- tween 0 and 1 (Table 2). This weighting scheme, based on Hauer et al. (2002a, b), represents the degree to which land cover types affect wetland functionality. Land cover types weighted 1 are natural habitats that provide the same functional value as reference conditions. Decreasing weights indicate an increasing departure from refer- ence conditions and a resulting loss of func- tional value. Land cover scores were calcu- lated by multiplying the percent of the wa- tershed in each land cover type by that type's functional weight and then summing and multiplying by 100. Thus a watershed with only natural vegetation would score 100, while lower scores represent conver- sions to human land uses. Assuming that the condition of a wa- tershed's population of wetlands was corre- lated with watershed rank, the three least impacted (highest ranked) watersheds, three most impacted (lowest ranked) watersheds, and three moderately impacted (middle ranked) watersheds were selected. Because some of the initially selected watersheds oc- curred primarily on Tribal land, where ac- cess was limited, alternatives (next ranked watersheds) were selected. Selected water- sheds are shown in Figure 3. Individual sampling points were then randomly chosen within each selected wa- tershed. Twenty-seven wetlands were sam- pled from each class (three depressional and three riverine wetlands sampled in each of the nine watersheds). If a wetland could not be sampled because access to private land was not granted, another wetland was ran- domly selected. Depressional wetlands were selected using NWI coverage. The sample popula- tion was considered to be all wetlands clas- sified as temporarily or seasonally fiooded palustrine emergent (PEMA and PEMC, Cowardin et al. 1979). The riverine sample Legend Middle Milk Sub-basin NWI Coverage ^^H Palustrine Emergent Wetland Figure 2. Extent of national wetland inventory coverage in the study area. population was defined as stream reaches of Milk River tributaries that had a valley gra- dient from 0.0001-2.0% and a minimum length of 1 00 m. Reaches were identified by combining the 2001 National Elevation Dataset (30-m digital elevation model) with the 1999 1:100,000 National Hydrography Dataset. Reaches were then broken into 1 km segments (minimum segment length of 100 m) with the 100-m reach at the segment midpoint being the sample unit. Targeted Sampling Fifty-six sites were sampled in 2002: 27 depressional and 29 riverine wetlands (27 randomly chosen and two targeted samples that appeared to represent reference condi- tions). During the course of this field work, it appeared that the probabilistic sampling strategy was not representing the full range of wetland condition. Primarily, reference condition wetlands were not being ade- quately sampled. Therefore, additional wet- lands were targeted in 2003. Reference wet- lands were identified in consultation with federal and state resource agency personnel and local experts in Montana and Sas- katchewan. Based on this consultation, an additional 11 wetlands were sampled in 2003. These consisted of six depressional wetlands, four in reference condition and two highly disturbed, and five reference condition riverine wetlands. Data Collection Sample wetlands were stratified by hydrological and geomorphological parame- Table 1. Human disturbance factors Middle Milk sub-basin. used to rank 5^^-level watersheds in the Human Disturbance Factor Unit of Measurement Water Rights Irrigation Population Density Corps 404 Stream/Wetland Permits S303d Listed Streams Road Density Well Density Mine Density Discharge Permit Density Road/Stream Crossings Federal Land Wilderness Land Cover Percent Persons per Square Mile Permits per 100 Square Miles Meters per Square Mile Miles per Square Mile Wells per Square Mile Mines per Square Mile Permits per Square Mile Crossings per 10 Square Miles Percent Percent Percent ters. Depressional wetlands were stratified by inundation period using the zones de- scribed by Stewart and Kantrud (1971) (i.e., wet meadow and shallow marsh, corre- sponding to temporarily and seasonally flooded, respectively). Riverine wetlands were stratified by geomorphology (i.e., de- positional bar, channel shelf, floodplain (sensu Hupp and Osterkamp 1985)). Vege- tation was sampled from each fluvial surface or inundation zone and was characterized using randomly placed 1.0-m x 0.5-m quad- rats. Abundance of each vascular plant spe- cies was estimated as percent canopy cover within quadrats. Random quadrat samples were repeated until no new species were found and then one more quadrat was sam- pled. For multimetric and multivariate Table 2. Functional weights assigned to land cover types. Land Cover Functional Weight Forest Grassland Shrubland Snow Water Wetland Pasture Barren Orchards Residential (low) Crops Mines/Quarries Residential (high) Commerce/Industrial 1.0 1.0 1.0 1.0 1.0 1.0 0.6 0.5 0.5 0.4 0.2 0.2 0.2 0.1 Legend Middle Milk Sub-basin Selected Watersheds Distuft)an€B Rank ^^H Low ^^1 Medium ^H High Jrr^=^^^^^ Figure 3. Selected 5^^-level watersheds and their relative disturbance categories. analyses, quadrat data were aggregated by zone/fluvial surface and site. Nomenclature follows Kartesz (1999), which forms the ba- sis for the national naming standard for vas- cular plants (U.S. Department of Agriculture 2004). Environmental factors recorded at each site included hydrologic and geomor- phic modifications (e.g., presence and extent of ditches, dikes, tiles, revetment, slumped or unstable banks), physical site distur- bances (e.g., presence and extent of pugging and hummocking), and land use within the wetland and adjacent uplands. Data Analysis Human Disturbance Parameters Watershed disturbance categories, based on ranking 5*-level hydrologic unit watersheds, were initially used to sample wetlands across a putative human distur- bance gradient. However, the correspon- dence between watershed ranks and human disturbance at a particular site was un- known. To test what relationship, if any, watershed ranks had to site condition, addi- tional human disturbance parameters were measured for each site. These parameters spanned multiple spatial scales: within the wetland or sample reach, within a 500-m buffer around the site, and within the site's upstream catchment (riverine wetlands only). The buffer width of 500 m was cho- sen in part because it included an area suffi- cient to encompass the catchments of most depressional wetlands. At the local scale (within the depres- sional wetland or riverine sample reach), grazing intensity and previous agricultural use were considered. Grazing intensity was defined as low, medium, or high (low/medium and high for depressional wet- lands) based on bank stability and the extent of ground disturbance, such as pugging or hummocking. Previous agricultural use was binary (none, site previously tilled; depres- sional wetlands only). Human disturbance at the buffer and catchment scale were char- acterized using a GIS (ArcGIS 8.3, ESRI, Redlands, California 92373). Catchments upstream from riverine sample locations were delimited using the hydrology model- ing extension in ArcGIS 8.3. This routine uses a sink-filled digital elevation model (DEM) to define catchments. The base DEM used was from the 30-m raster Na- tional Elevation Dataset. The extent of landscape-scale human disturbance within buffers and catchments was characterized by measuring land cover, road density, and the number of dams (catchment-scale only). Land cover was determined from the National Land Cover Database (30-m raster data). Land cover types (e.g., grass- land/herbaceous, shrubland, row crop, fal- low, pasture/hay) were grouped into two classes, native vegetation and agricultural, and the proportion of the buffer or catch- ment in each class was measured. Other human-modified land covers, such as devel- oped areas, did not occur within buffers or catchments examined. Buffer and catch- ment road density was calculated from 2000 U.S. Census Bureau TIGER 1:100,000 line files. The number of dams in a catchment was determined from the Montana Dams Database (a compilation of the U.S. Army Corps of Engineers National Inventory of Dams and the U.S. Geological Survey Geo- graphic Names Information System that is maintained by the Montana Department of Fish, Wildlife & Parks). As an alternative to watershed dis- turbance categories, I used a rule-based dis- turbance hierarchy to construct numerical disturbance indices for depressional and riv- erine wetlands, similar to Lopez and Fen- nessy (2002). These indices integrated dis- turbance factors across spatial scales. The relative importance of landscape vs. site- level disturbance factors to the biological community varies by system and taxa (Bis- son et al. 2002, Seabloom and van der Valk 2003, Wright et al. 2003). Based on previ- ous research and personal observation, I as- sumed that vegetation was responding pri- marily to on-site disturbances. Therefore, for depressional wetlands, the disturbance hierarchy included on-site agricultural dis- turbance, on-site grazing disturbance, and road density within a 500-m buffer (Figure 4). These factors have all been shown to influence wetland vegetation, faunal assem- blages, and functionality of prairie potholes (Kantrud et al. 1989, Euliss and Mushet 1996, Kantrud and Newton 1996, Euliss and Mushet 1999, Freeland and Richardson 1999, Euliss et al. 2001) or wetlands gener- ally (Findlay and Houlahan 1997, Trombu- lak and Frissell 2000, Houlahan and Findlay 2003). The disturbance hierarchy for river- ine wetlands used on-site grazing intensity and hydrological modification, as measured by the number of dams in the upstream catchment (Figure 5). Both these factors can greatly influence riparian vegetation and wetland function (Kauffman and Krueger 1984, Schulz and Leininger 1990, Boggs and Weaver 1994, Scott et al. 1997, Auble and Scott 1998, Friedman et al. 1998, Scott et al. 2003). Table 3 lists the disturbance factors used and how each factor was scored. Graz- ing intensity and agricultural use were cate- gorical variables. To place values for road density or number of dams into disturbance categories, quantile plots were examined for these variables. Break points for distur- bance categories were determined by trisect- ing the value range of each variable. Depressional Wetland No Agricultural Disturbance Previously Tilled Low/Medium Grazing Intensity High Grazing Intensity Buffer Road Density Low Med High IT IT IT Buffer Road Density Low Med 8 High Buffer Road Density Low Med High IT IT IT 1 Low < High Disturbance Rank Figure 4. Schematic illustrating how temporarily and seasonally flooded depressional wetlands were ranked along a human disturbance gradient. Relationships between watershed disturbance categories and other disturbance measures were analyzed using Kruskal- Wallace tests (SYSTAT 2002). The Kruskal- Wallace procedure is a non- parametric analog to one-way analysis of variance and was used because data did not meet assumptions of homogeneity of vari- ance (Levene 1960). In addition, I exam- ined the relationship between disturbance categories and vegetation response variables selected as metrics (see Multimetric Meth- ods below), also using Kruskal- Wallace tests. Multimetric Methods Multimetric analysis seeks to deter- mine the health of a site, such as a water- body or wetland, by directly measuring the condition of one or more components of its biota, such as vegetation or macroinverte- brates (Danielson 2002). This method is based on defining a relatively homogeneous study environment (e.g., high- or low- gradient streams, depressional wetlands) and measuring the response of target biota across a gradient of human disturbance (Karr and Chu 1999). The response is calculated by assessing measurable attributes of the bio- logical system. Attributes that increase or 10 Low Grazing Intensity Hydrologic Modification Low Med High Riverine Wetland Medium Grazing Intensity Hydrologic Modification Low Med High irlrlr iririr High Grazing Intensity Hydrologic Modification Low Med High irlrlr Low < High Disturbance Rank Figure 5. Schematic illustrating how herbaceous-dominated intermittent and ephemeral riverine wetlands were ranked along a human disturbance gradient. decrease predictably with increasing human disturbance, are sensitive to a range of bio- logical stresses, discriminate between hu- man-caused perturbations and natural vari- ability, and are easy to measure and interpret can be successfully used as metrics (Karr and Chu 1999). Multimetric approaches, such as indices of biological integrity, com- bine metrics reflecting diverse biotic re- sponses to anthropogenic stressors into an integrative measure of biological condition (Karr and Chu 1999, Teels and Adamus 2002). Attributes can be divided into several categories: species richness and composi- tion, tolerance/intolerance to human distur- bances, trophic composition, and population characteristics (Teels and Adamus 2002). In regards to vegetation, these attribute catego- ries can be refined to those representing community-based metrics, metrics based on plant functional groups, and species-specific metrics (Fennessy et al. 2002). Examples of metrics include changes in species richness and dominance (community-based metrics), changes in the number of perennials, annu- als, or intolerant species (plant functional group metrics), and dominance of individual species (species-specific metrics). In this study, two wetland classes were sampled, temporarily and seasonally flooded depressional wetlands and intermit- tent and ephemeral riverine wetlands, and human disturbance was measured along nu- merical disturbance indices. Thirty-five vegetation attributes were examined. Many of these attributes were chosen because they had been proven to be successful vegetation metrics for wetlands in western Montana (Borth 1998), North Dakota (DeKeyser et al. 11 Table 3. Disturbance factors used to develop human disturbance indices for depressional and riverine wetlands. Disturbance Factor Scale Wetland Type Criteria Agricultural Use Local Grazing Intensity Local Depressional Depressional/ Riverine Road Density Hydrological Modification Buffer Depressional Catchment Riverine Low: No evidence that wetland was previ- ously tilled High: Evidence of past tillage, such as plow lines or rock piles Low: Banks stable with little or no slump- ing, little to no pugging or hum- mocking Med: Moderate or localized bank erosion or slumping, some pugging or hum- mocking present High: Extensive bank erosion or slumping over channel length, extensive pug- ging or hummocking Low: < 5 m/ha Med: 5-63 m/ha High: > 63 m/ha Low: 0 dams/1,000 ha Med: 0.01-0.3 dams/1,000 ha High: > 0.3 dams/1,000 ha 2003), Ohio (Mack et al. 2000, Mack 2001), or Minnesota (Helgen and Gernes 2001). Potential metrics and their predicted re- sponse to human disturbance are listed in Table 4. The relationship between attributes and site disturbance index was examined graphically and with Spearman rank-order correlation coefficients (rs). Metrics show- ing a strong linear or curvilinear response to disturbance and that differentiated between least and most disturbed wetlands were cho- sen for inclusion into the multimetric index. Where two or more metrics that conveyed a similar biological response had a robust re- sponse to disturbance (e.g.. Shannon and Simpson diversity indices), the metric with the higher rs value or greater ecological rele- vance was chosen. To combine individual metrics into a multimetric index, metric data was con- verted into a common scoring base. The scoring base used was that recommended by Karr and Chu (1999), where metric values that represent reference conditions are scored 5, those that deviate somewhat from reference condition are scored 3, and those that strongly deviate from reference condi- tion are scored 1 . For metrics with a linear response to disturbance, quantile plots were used to determine value ranges for scoring categories. Break points were calculated for the 67* and 33^^ quantiles, except where modified due to natural breaks in values or ties. Value ranges for metrics with a curvi- linear response were determined by graphi- cal fitting. Total multimetric scores were calculated by summing metric scores and dividing by the number of metrics. Thus the multimetric index developed for each wet- land class would have common values, even if the number of individual metrics differed. 12 Table 4. Vegetation attributes initially considered as potential metrics for depressional and river- ine wetlands and their predicted response to increasing human disturbance, dep = depressional wetlands, riv = riverine wetlands Vegetation Attribute Predicted Response Richness of native genera Shannon diversity index (H) Simpson's diversity index (D) Total species richness Richness of native perennials Richness of native graminoids Richness of intolerant graminoids (C-value > 6) Richness of Carices Richness of dicots Richness of annual or biennial species Richness of exotic species Richness of species with C-value > 4 Richness of species with C-value > 5 Richness of intolerant species (C-value > 6) Richness of tolerant species (C-value < 2) Richness of species with a wetland indicator status of FAC or wetter Richness of hydrophytes (wetland indicator status of FAC W or wetter) Proportion of total species richness comprised of annual or biennial species Proportion of total species richness comprised of exotic species Proportion of total species richness comprised of tolerant species (C-value < 2) Relative cover of native perennials Relative cover of native graminoids Relative cover of intolerant graminoids (C-value > 6) Relative cover of Carices Relative cover of dicots Relative cover of annual or biennial species Relative cover of exotic species Relative cover of species with C-value > 4 Relative cover of species with C-value > 5 Relative cover of intolerant species (C-value > 6) Relative cover of tolerant species (C-value < 2) Relative cover of species with a wetland indicator status of FAC or wetter Relative cover of hydrophytes (wetland indicator status of FAC W or wetter) Floristic quality index Average C-value decrease increase (dep)/ decrease (riv) increase (dep)/ decrease (riv) decrease decrease decrease decrease decrease increase (dep)/ decrease (riv) increase increase decrease decrease decrease increase decrease decrease increase increase increase decrease decrease decrease decrease increase (dep)/ decrease (riv) increase increase decrease decrease decrease increase decrease decrease decrease decrease 13 How well multimetric indices re- flected site condition (as measured by the disturbance index) was assessed using linear regression. Assumptions of linear regres- sion were tested by examining normal prob- ability plots of residuals (assumption that errors are normally distributed) and scatter plots of studentized residuals against esti- mated values (assumptions that errors have constant variance and are independent) (SYSTAT 2002). Goodness-of-fit of mul- timetric indices was also assessed using the multivariate methods described in the fol- lowing section. Floristic Quality Index One of the vegetation attributes con- sidered as a metric was the floristic quality index (FQI). Floristic quality assessments were initially developed by Swink and Wilhelm (1979) for plant communities in the Chicago area. This method has since been expanded for use in other areas of the Mid- west. Its usefulness as a vegetation metric has been demonstrated by DeKeyser (2000) and Mushet et al. (2002) for prairie potholes in North Dakota and by Lopez and Fennessy (2002) for wetlands in Ohio. (However, see Matthews (2003) concerning problems of comparing FQI scores across wetland types.) This method assigns a coefficient of conser- vatism (C) to all native species that occur in a specified region. This coefficient, which ranges from 0 to 10, represents a species' relative tolerance to disturbance. The inter- pretation of coefficient values is as follows (from Fennessy et al. 1998): Value Interpretation 0 alien taxa and those natives that are opportunistic invaders or common components of ruderal communi- ties; 1-3 widespread taxa that are found in a variety of communities, including disturbed sites; 4-6 taxa that display fidelity to a par- ticular community, but tolerate moderate disturbance to that com- munity; 7-8 taxa that are typical of well estab- lished communities which have sus- tained only minor disturbance; 9-10 taxa that exhibit high degrees of fidelity to a narrow set of ecological conditions. In this study, I used C values established for native species in the Dakotas (Northern Great Plains Floristic Quality Assessment Panel 2001). The floristic quality index is calculated as FQI = 2 C / V n where FQI = floristic quality index, C = co- efficient of conservatism, and n = richness of native species. In addition to the FQI itself, several other vegetation attributes based on C- values, such as richness and relative cover of species with certain C values and the aver- age C-value of a sample unit, were calcu- lated (Table 4). Diversity Measures Two diversity measures were con- sidered, the Shannon and Simpson diversity indices. Both of these indices are related to and based partially on species richness. However, these measures also incorporate the equitability of species' abundance as well. For example, for both indices a plot containing three species where one species is dominant would be rated as being less di- verse than a three-species plot where the species occurred with equal abundance. The Shannon diversity index is cal- culated as H' = -2 p, log p, 14 where H' = Shannon diversity index and p/ = the relative cover of the /* species within a sample unit. The Simpson diversity index is cal- culated as D = 1 - 2 p,^ where D = Simpson diversity index and p/ = the relative cover of the f^ species within a sample unit. These two measures are similar but vary in their sensitivity to rare species: Shannon diversity is intermediate between species richness and Simpson diversity in its sensitivity to rare species. Multivariate Methods Multivariate analyses were per- formed to validate the multimetric approach and to assess the response of the entire vege- tation community to human disturbance. These analyses included parametric and non-parametric comparisons of group differ- ences and indirect ordination. First, group differences were assessed by assigning sites to disturbance categories. These categories were created by combining sites into groups representing reference condition wetlands (disturbance index scores of 7-9), moder- ately disturbed wetlands (disturbance index scores of 4-6), and severely disturbed wet- lands (disturbance index scores of 1-3). The ability of metrics to correctly classify sites into these three groups was assessed using discriminant analysis. Discriminant analysis is an eigenanalysis technique that finds lin- ear functions that best separate cases into predefined groups and was used to predict group membership based on metric values. Predicted and actual group memberships were compared to determine how well met- rics discriminated among disturbance cate- gories. Linear discriminant analysis was performed using the complete estimation method (SYSTAT 2002). Assumptions of multivariate normality and homogeneity of within-group variance were not met in all cases; however, this was not considered critical as the analysis was exploratory (McCune and Grace 2002). Multi-response permutation proce- dure (MRPP, Biondini et al. 1988) was used to test whether plant community composi- tion for all species sampled at a site differed among disturbance categories (PC-ORD, McCune and Mefford 1999). In addition to a P-value, MRPP describes group tightness with A, a statistic that compares the within- group heterogeneity to that expected by chance (A = 1 when items are identical within groups, A = 0 when heterogeneity within groups equals that expected by chance, and A < 0 when heterogeneity within groups is greater than that expected by chance) (McCune and Mefford 1999). To improve the correspondence of MRPP results with non-metric multidimensional scaling (see below), MRPP was based on a rank-transformed S0rensen distance matrix (McCune and Grace 2002). Where commu- nity composition differed significantly with disturbance, associations between species and groups was examined using indicator species analysis (Dufrene and Legendre 1997). This method assigns each species an indicator value for a particular group that ranges from 0 (no indication) to 100 (perfect indication). The statistical significance of indicator values was tested using a Monte Carlo randomization procedure with 10,000 iterations. To examine relationships among species and between species and environ- mental factors, sample sites were ordinated in species space using non-metric multidi- mensional scaling (NMS, Kruskal 1964, Mather 1976). Ordination is a data reduc- tion method that attempts to describe under- lying patterns of species composition by graphically summarizing complex relation- ships (McCune and Grace 2002). NMS is 15 an indirect ordination technique that works without assuming that a species responds to environmental gradients in a Unear or uni- modal fashion and is robust to large num- bers of zero values. It therefore avoids many of the distortions of eigenvector-based ordination methods, such as detrended cor- respondence analysis (Kenkel and Orloci 1986, Minchin 1987). NMS is an iterative method that attempts to reduce differences between the ranked distances in the original multidimensional species space and ranked distances in the reduced dimensions of the ordination. These differences, termed stress, are measured as the degree of departure from monotonicity in the original space and the reduced space (McCune and Grace 2002). Dimensionality was determined by running NMS on PC-ORD's autopilot mode for 40 runs with real data and 50 runs with randomized data in each of six dimensions (McCune and Mefford 1999). Dimensional- ity was chosen by selecting the highest number of dimensions that appreciably re- duced stress and where the final stress for real data was significantly lower than that for randomized data. Additional parameters included the use of the quantitative version of the S0rensen distance measure, the global form of NMS, and an instability criterion of 0.00001 to be achieved after 500 iterations or 50 continuous iterations within the crite- rion. To reduce beta diversity {(5^, composi- tional heterogeneity among sample units (Whittaker 1972)) and improve the inter- pretability of results, species occurring in fewer that 5% of sample units were omitted from the analysis. Results Watershed Disturbance Categories Watershed disturbance categories showed no statistically significant relation- ship with other measures of human distur- bance (Kruskal-Wallace tests, a = 0.05). This lack of correspondence was observed for both depressional and riverine datasets and held true for disturbance factors meas- ured within a 500-m buffer around sites and within the upstream catchment (riverine sites only) as well as for an integrative site disturbance index (Figures 6 and 7). Vege- tation metrics were also compared among watershed disturbance categories (Kruskal- Wallace tests, a = 0.05). These comparisons also showed no statistical relationship, ex- cept for two metrics for depressional wet- lands, the relative cover of native perennials and the relative cover of exotic species (Fig- ures 8 and 9). Both these metrics strongly respond to whether or not a site has been previously tilled, and the significant results are the product of the clustering of agricul- turally disturbed sites within one watershed ranked at medium disturbance. Depressional Wetlands Metrics and the Multimetric Index Seven of the 35 attributes examined showed a robust response to the human dis- turbance gradient (Figure 10). The response of two of these attributes, the Shannon and Simpson diversity indices, was observed within the shallow marsh zone of seasonally fiooded wetlands only. One of the precepts of the multimetric method is that individual metrics will represent different aspects of the biological response to human distur- bance. Therefore, biologically redundant metrics should be avoided (Kimberling et al. 2001, Karr and Kimberling 2003). Of the seven potential metrics, two pairs of attrib- utes were biologically redundant. These were the relative cover of native perennials and relative cover of native graminoids and the Shannon and Simpson diversity indices. Sixty -two percent of native perennials were also native graminoids and the two values for the two metrics were highly correlated 16 Low Med High Watershed Disturbance Category Low IVIed High Watershed Disturbance Category Low IVIed High Watershed Disturbance Category Figure 6. Bar graphs of disturbance measures by watershed disturbance categories for temporar- ily and seasonally flooded depressional wetlands. Bars are mean values ± 1 SE. Comparisons among disturbance categories were non-significant for all factors (Kruskal- Wallace tests). Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Figure 7. Bar graphs of disturbance measures by watershed disturbance categories for herba- ceous-dominated intermittent and ephemeral riverine wetlands. Bars are mean values ± 1 SE. Comparisons among disturbance categories were non-significant for all factors (Kruskal- Wallace tests). (rs = 0.899); similarly, Shannon and Simp- son diversity were highly correlated (rs = 0.944). The relative cover of native peren- nials was selected over the relative cover of native graminoids because it was inclusive of more species (34 vs. 23) and had a stronger correlation to the disturbance index (rs = 0.747 vs. rs = 0.709). Although the cor- relation between Shannon diversity and site disturbance was greater than that for Simp- 17 Low Med High Watershed Disturbance Category Low IVIed High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Figure 8. Bar graphs of vegetation metrics by watershed disturbance categories for temporarily and seasonally flooded depressional wetlands. Bars are mean values ± 1 SE. Graphs with aster- isks indicate that the metric was significantly different among disturbance categories (Kruskal- Wallace tests, P < 0.05). son diversity (rs = -0.598 vs. rs = -0.469), Simpson diversity was chosen because its intrinsic properties made it a more robust measure when comparing sites/inundation zones that had been sampled with varying intensity. The vegetation sampling protocol did not specify a uniform number of quad- rats per site or per area. The Simpson diver- sity index ameliorates problems associated with uneven sampling intensity because it is little affected by rare species and is therefore relatively stable with sample size (McCune and Grace 2002). Value ranges for selected metrics were calculated based on quantile plots for attributes with a linear response (relative cover of species with C value > 4, floristic quality index, and Simpson diversity index) and by graphical fitting for attributes with a curvilinear response (relative cover of native perennials and relative cover of exotic spe- cies). Value breaks for quantile plots were determined at the 33^^ and 67^^ quantiles, with some variability based on natural breaks in the data or tied values. Value ranges are presented in Table 5. The result- ing multimetric index showed a significant response to human disturbance, as measured by the site disturbance index (Fi 28= 47.505, R^ = 0.629, P < 0.001, Simpson diversity metric not included; Figure 11). The Simpson diversity metric was only observed to be valid within the season- ally flooded zone, which is naturally less diverse than the adjoining temporarily flooded zone (personal observation). As the goal of this study was to produce metrics that are valid for both temporarily and sea- sonally flooded depressional wetlands, I tested whether the Simpson diversity metric 18 Low Med High Watershed Disturbance Category Low IVIed High Watershed Disturbance Category Low IVIed High Watershed Disturbance Category Low Med High Watershed Disturbance Category Low Med High Watershed Disturbance Category Figure 9. Bar graphs of vegetation metrics by watershed disturbance categories for herbaceous- dominated intermittent and ephemeral riverine wetlands. Bars are mean values ± 1 SE. Com- parisons among disturbance categories were non-significant for all factors (Kruskal- Wallace tests). added appreciably to the multimetric index by comparing linear regressions of the mul- timetric index with and without the diversity metric for the 1 8 seasonally flooded depres- sional wetlands sampled. The linear regres- sion without the diversity metric produced higher F and R^ statistics (with D: Fi^ i6 = 49.755, R^ = 0.757, P < 0.001, without D: Fi, 16 = 57.129, R^ = 0.781, P < 0.001). The Simpson diversity index was therefore dropped as a metric. When sites were divided into distur- bance categories (reference condition, mod- erately disturbed, and severely disturbed), vegetation metrics correctly classified a site's disturbance category membership for 5 out of 5 severely disturbed sites, 5 out of 7 moderately disturbed sites, and 12 out of 18 reference condition sites for an overall clas- sification accuracy of 73% (results based on discriminant analysis; Figure 12). Vegetation Community Response In addition to the selected metrics, the whole vegetation community in sampled depressional wetlands also responded to human disturbance. Plant community com- position and abundance was significantly different among disturbance categories (se- verely disturbed, moderately disturbed, and reference condition) (MRPP, A = 0.189, P < 0.001). This is a fairly robust finding, as A- values in community ecology are commonly less than 0.1, even when the test is signifi- cant (McCune and Grace 2002). Vegetation differences among disturbance categories are graphically displayed in the results from the NMS ordination (three-dimensional so- lution, final stress = 8.197, instability = 19 c, 100 § 90 CD I 80 CD I 60 o O 50 CD '^ 40 CD ^ 30 r, = 0.747 3 5 7 Disturbance Index .E 90 E CD ■^ 70 CC 1 60 > 0 50 CD 1 40 CD ^ 30 r, = 0.709 3 5 7 Disturbance Index 3 5 7 Disturbance Index 70 0 60 CD f 50 1 40 ^ 30 o O 0 20 1 10 DC 0 -i 1 1 r r^ = -0.772 i I L 3 5 7 9 Disturbance Index 5 I 1 • i_ 0 h = 0.573 • • • • • • • • • • • • • • • • • • 5 • • • 0 "i 1 i~ 13 5 7 Disturbance Index 1.6 f 1.4 CD T3 S 1.2 w ^ 1.0 Q § 0.8 ^ 0.6 0.4 1 1 1 1 i_ • "• . ~ • • • • •- ~ • • •- rs = -0.598 • "i 1 1 1 i~ 13 5 7 Disturbance Index • 0.7 0.6 0.5 0.4 ' 0.3 0.2 . r, = -0.469 J I i_ 13 5 7 Disturbance Index Figure 10. Scatter plots of attribute values against site disturbance index for temporarily and seasonally flooded depressional wetlands. Shown are vegetation attributes that had a linear or curvilinear response to human disturbance and differentiated between least and most disturbed sites. Disturbance index ranges from 1 (most disturbed) to 9 (least disturbed), rs = Spearman rank-order correlation coefficient. Data for Shannon and Simpson diversity indices are for the shallow marsh zone of seasonally flooded wetlands only. 0.00001, 72 iterations; Figure 13). Species that were signiflcantly associated with se- verely disturbed and reference condition wetlands are presented in Table 6 (indicator species analysis; no species were signifl- cantly associated with moderately disturbed wetlands). Riverine Wetlands In contrast to depressional wetlands, riverine wetlands are naturally highly vari- able. This is especially true of small-order ephemeral streams in arid and semi-arid en- vironments (Friedman and Lee 2002, Eby et 20 Table 5. Ranges of attribute values for metric scoring categories for temporarily and seasonally flooded depressional wetlands. Relative cover values are expressed as percentages. Value Range Value Range Value Range Metric for 1 for 3 for 5 Relative cover of native perennials Relative cover of species with C-value > 4 Relative cover of exotic species Floristic quality index Simpson diversity index* * metric is for seasonally flooded wetlands only <75 75-90 >90 <40 40-64 >64 >10 10-5 <5 <6.7 6.7-10.1 >10.1 >0.68 0.68-0.55 <0.55 al. 2003). Thirty-four riverine wetlands were sampled in the course of this study. These sites encompassed a wide range of environmental heterogeneity, including per- ennial and ephemeral streams, alkaline and non-alkaline systems, and vegetation domi- nated by herbaceous and woody species. When these sites were considered together, the "noise" of this natural variability made it difficult to detect the "signal" from anthro- pogenic stressors. To restore a workable level of environmental homogeneity, sites that sampled alkaline or perennial streams were removed from the analysis. The one site dominated by woody vegetation was also removed, as it was both a mathematical and ecological outlier. These types of sys- tems are ecologically different from the pri- mary target population, and this is reflected by the vegetation attributes of these sites. The resulting metrics, therefore, are derived from herbaceous-dominated intermittent or ephemeral riverine wetlands. 05 c g C/) CD Q_ CD Q I X CD TD o 'B 2 CD E 5- 4- 3- 1 - _ 1 1 1 J • • l/_ Fi, 28 =47.505 R^ = 0.629 • • ^^ _P< 0.001 •• • • • • • • •" _ ^^^^ • _ • • • -.• 1 1 1 " 3 5 7 Disturbance Index Figure 1 1 . Relationship between multimetric index and site disturbance index for temporarily and seasonally flooded depressional wetlands (n = 30). Metrics include the relative cover of na- tive perennials, relative cover of species with C-value > 4, relative cover of exotic species, and floristic quality index. 21 03 c o ■(/) if) CD L_ Q- CD Q I x: CD "D C o S 2 CD 5- 4- 3- I II I • I I • • i« I I «. • I - ■•! ^* I ' I • • I I- . - I I I I A I I - ! ! -A . . — Predicted Membership ^ Severely Disturbed ■ IVIoderately Disturbed • Reference Condition 13 5 7 9 Severe Moderate Reference Disturbance Category Figure 12. The predicted membership of temporarily and seasonally flooded depressional wet- lands to disturbance categories compared with actual group membership. Predicted membership is based on discriminant analysis of vegetation metrics. Metrics and the Multimetric Index Thirteen of the 35 attributes exam- ined exhibited a predictable response to hu- man disturbance (Figure 14). Unfortunately, many of these attributes were biologically redundant. The first group of redundant at- tributes was total species richness, richness of native perennials, and richness of dicots. Of the 147 species sampled, 118 were native perennials and 100 were dicots; therefore, the species pools of these groups substan- tially overlapped. Additionally, these attrib- utes were highly inter-correlated (total rich- ness-richness of native perennials, rs = 0.848; total richness-richness of dicots, rs = 1.000; richness of native perennials-richness of dicots, rs = 0.848). The richness of native perennials was the metric selected, as it was both the most highly correlated with distur- bance and the most ecologically relevant. The second redundant group included the two diversity measures. Shannon and Simp- son diversity. Although Shannon diversity was more strongly correlated with distur- bance, Simpson diversity was chosen, as its properties were more appropriate to the vegetation sampling methodology used (see Results under Depressional Wetlands). Many attributes were derived from func- tional groups based on a species' C-value. These included richness of species with C- value > 4, richness of species with C-value > 5, and richness of intolerant species (C- value > 6), as well as the relative cover of intolerant species and intolerant graminoids. All of these attributes were highly inter- correlated and shared many species in com- mon. Although it was not the most highly correlated with disturbance, the relative cover of intolerant species was selected as a metric. It was chosen over the richness of species with C-value > 4 and richness of species with a C-value > 5, both of which had higher rs values, because it was re- stricted to more disturbance-intolerant spe- cies and because cover is a more sensitive measure of species response than is pres- ence-absence (Rahel 1990). The last redun- dant group was FQI and average C-value (rs = 0.932). FQI was selected based on its higher correlation with disturbance. 22 A X C\J .cn • Disturbance Category ,\ < A Severely Disturbed 1 IVI ode rate ly Disturbed • Reference Condition ^ A Al . >^ ^A - . - ^ • • ■ • • • • • y FQI • • X ni Axis 1 C_EXO • ■ ■ ■ C_C4 C_NATPER • • ■ 1 • ^ B A A A ELAC CADUe/ DES02 , / \ //ALGE2 POPA2 \ // THAR5 2^^r>^V Dl POPR ^^-^ SOARU PASM Figure 13. Graphical representation of the NMS ordination of sampled temporarily and season- ally flooded depressional wetlands. Points represent species cover and composition data for quadrats aggregated by site. Distance between points is proportional to dissimilarity between samples (i.e., samples with similar species composition are plotted closer together). Axis 1 represents 68.2% of the variation in the data while Axis 2 accounts for 14.1% (total variation ex- plained = 82.3%)). Axes were rotated such that Axis 1 corresponds to the human disturbance gradient. Vectors are joint plots of variables correlated with ordination scores. Vector lengths represent the strength of the correlation; all variables have an R^ > 0.20. Vectors in the main graph represent vegetation metrics and the disturbance index; vectors in Inset B represent highly correlated indicator species. Labels are: (main graph) CNATPER = relative cover of native perennials, C_C4 = relative cover of species with C-value > 4, CEXO = relative cover of exotic species, FQI = floristic quality index, DI = disturbance index; (Inset B) ALGE2 = Alopecurus geniculatus, CADU6 = Carex duriuscula, DES02 = Descurainia sophia, ELAC = Eleocharis acicularis, PASM = Pascopyrum smithii, P0PA2 = Poa palustris, POPR = Poa pratensis, SOARU = Sonchus arvensis ssp. uliginosus, TAOF = Taraxacum officinale, THAR5 = Thlaspi arvense, DI = disturbance index. DI values range from 1 (most disturbed) to 9 (least disturbed). Inset A shows the importance of inundation period to plant community composition. Symbols represent site hydrology: A = seasonally flooded wetlands, x = temporarily flooded wetlands. 23 Table 6. Species indicative of severely disturbed and reference condition temporarily or season- ally flooded depressional wetlands. Indicator and P-values were determined using indicator spe- cies analysis; species shown had an indicator value > 20 and a P-value < 0.1. Species Disturbance Category^ Indicator Value P-value Poa palustris Poa pratensis Thlaspi arvense Rumex salicifolius Convolvulus arvensis Taraxacum officinale Descurainia sophia Sonchus arvensis ssp. uliginosus Rumex crispus Alopecurus geniculatus Cryptantha torreyana Pascopyrum smithii Eleocharis acicularis Veronica peregrina Carex duriuscula D 60.0 0.002 D 60.0 0.002 D 59.5 0.002 D 69.6 0.009 D 40.0 0.021 D 73.8 0.026 D 34.3 0.055 D 36.6 0.063 D 31.0 0.093 R 51.4 0.033 R 48.5 0.036 R 49.6 0.044 R 50.3 0.061 R 41.7 0.062 R 44.2 0.072 D = severely disturbed, R = reference condition Value ranges for selected metrics were calculated based on quantile plots. Value breaks for quantile plots were deter- mined at the 33^^ and 67* quantiles, with some variability based on natural breaks in the data or tied values. Metric value ranges are presented in Table 7. The resulting mul- timetric index was significantly related to human disturbance (Fi^ 20 = 77.511, R = 0.795, P< 0.001; Figure 15). Similar to depressional wetlands, the disturbance index was divided into three dis- turbance categories. The overall ability of vegetation metrics to correctly classify sites with regard to disturbance category was 86%, with 7 out of 8 severely disturbed sites correctly classified, 7 out of 9 moderately disturbed sites correctly classified, and 5 out of 5 reference condition sites correctly clas- sified (results based on discriminant analy- sis; Figure 16). Vegetation Community Response In contrast to the many vegetation at- tributes that showed a strong response to human disturbance, composition and abun- dance of the whole plant community sam- pled at a site was not predictably associated with disturbance category (MRPP, A = 0.016, P = 0.310). This lack of correspon- dence is graphically displayed in Figure 17 (NMS, three-dimensional solution, final stress = 12.373, instability = 0.00001, 139 iterations). Discussion Vegetation metrics were successful in assessing condition of both depressional and riverine wetlands. Human disturbance, as measured by an integrative disturbance index, explained 63% of the variation in the multimetric index for depressional wetlands and 80% of the multimetric variation for riverine wetlands. Vegetation metrics were 24 3 5 7 Disturbance Index 3 5 7 Disturbance Index 3 5 7 Disturbance Index 13 5 7 Disturbance Index 3 5 7 Disturbance Index 13 5 7 Disturbance Index 15 rs = 0.691 A 1 '' > O • • • • _§ 5 . • • - • • • • • • 0 "* 1 1 , i" Disturbance Index Disturbance Index Disturbance Index 3 5 7 Disturbance Index T 1 1 1- ■ rs = 0.761 I • 3 5 7 Disturbance Index 3 5 7 Disturbance Index T 1 1 1 r ■ rs = 0.727 • • Disturbance Index Figure 14. Scatter plots of attribute values against site disturbance index for herbaceous- dominated intermittent and ephemeral riverine wetlands. Shown are vegetation attributes that had a linear or curvilinear response to human disturbance and differentiated between least and most disturbed sites. Disturbance index ranges from 1 (most disturbed) to 9 (least disturbed), rs = Spearman rank-order correlation coefficient. 25 Table 7. Ranges of attribute values for metric scoring categories for herbaceous- dominated intermittent and ephemeral riverine wetlands. Relative cover and proportionate rich- ness values are expressed as percentages. Metric Value Range fori Value Range for 3 Value Range for 5 Richness of native perennials Simpson diversity index Relative cover of intolerant species Proportionate richness of tolerant species Floristic quality index <11 11-14 >15 < 0.710 0.710-0.819 > 0.820 <1.00 1.00-11.99 > 12.00 > 30.00 18.00-29.99 < 18.00 < 10.00 10.00-15.77 > 15.78 also able to correctly classify a wetland as either being severely disturbed, moderately disturbed, or in reference condition for 73% of depressional and 86% of riverine wet- lands sampled. The multimetric index for depres- sional wetlands was less robust than the riv- erine index, largely because it did not clearly differentiate between reference condition and moderately disturbed sites. While 100%) of severely disturbed sites were correctly classified, only 71%) of moderately disturbed and 67%) of reference condition sites were correctly identified. This is likely due in part to the underlying response of some of the metrics to human disturbance. Two of the four metrics used, relative cover of spe- cies with C-value > 4 and relative cover of exotic species, strongly responded to previ- ous agricultural use in a wetland, but showed no response to other disturbance factors. In fact, only the FQI metric had a clearly linear response along the entire dis- turbance gradient. The sensitivity of the multimetric in- dex to agricultural disturbance is consistent with previous studies that have identified direct (i.e., wetland cropping) and indirect (i.e., adjacent land use) agricultural distur- bances as important stressors of depressional CD c CD > X Q) "D _C o CD E 3 5 7 Disturbance Index Figure 15. Relationship between multimetric index and site disturbance index for herbaceous- dominated intermittent and ephemeral riverine wetlands (n = 22). Metrics include the richness of native perennials, Simpson diversity index, relative cover of intolerant species, proportionate richness of tolerant species, and floristic quality index. 26 0, 5 c 0) > 4 DC ^ X CD _E 3 _ 1 1 • ■ 1 1 1 ■ 1 - ■ J • 1 o ▲ ■ ■ 1 ▲ 1 2 - ■ 1 - Predicted Membership Multi 1 A 1 1 1 1 1 " ^ Severely Disturbed ■ IVIoderately Disturbed • Reference Condition 13 5 7 9 Severe Moderate Reference Disturbance Category Figure 16. The predicted membership of herbaceous-dominated intermittent and ephemeral riv- erine wetlands to disturbance categories compared with actual group membership. Predicted membership is based on discriminant analysis of vegetation metrics. wetlands in the prairie pothole region (Kan- trud et al. 1989, Euliss and Mushet 1996, Kantrud and Newton 1996, Euliss and Mushet 1999, Freeland and Richardson 1999, Seabloom and van der Valk 2003). Unlike other disturbances, such as grazing, fire, or drought, wetland tilling can com- pletely alter the species composition in a wetland. The effects of this conversion can persist even after cropping ceases and any concomitant hydrological alterations to the wetland have been restored (Galatowitsch and van der Valk 1996, Seabloom and van der Valk 2003). Given the magnitude of the disturbance, it is unsurprising that sites that had been previously tilled were so well dif- ferentiated by both metrics and NMS ordina- tion. All species indicative of severely dis- turbed (i.e., previously tilled) sites were ei- ther invasive or weedy exotic species or na- tive ruderals. In contrast to agricultural distur- bances, vegetation did not strongly respond to grazing-related stress. This may reflect vegetation adaptation to grazing pressure or inadequate sampling of heavily grazed sites. Prairie pothole vegetation developed in con- junction with American bison (Bison bison) and was subjected to intense short-term grazing pressure. Cattle, in contrast, tend to preferentially use wetter and more produc- tive areas, leading to long duration, heavy use of these areas. Unlike riverine wetlands, the effects of this grazing pattern may be ameliorated in temporarily and seasonally flooded depressional wetlands due to their brief inundation periods. The riverine multimetric model was better at assessing site condition across the entire human disturbance gradient, and was especially good at identifying the most and least disturbed sites (classiflcation accuracy was 88% for severely disturbed wetlands and 100% for reference condition wetlands). Yet in contrast to the success of the mul- timetric index and unlike the results for the depressional dataset, whole-community analysis showed no relationship between vegetation and disturbance categories. The results of the multivariate analyses, how- ever, were inconclusive not because vegeta- tion did not change along a human distur- 27 ■ A A C\J w - Disturbance Category < A Severely Disturbed p IVI ode rate ly Disturbed • Reference Condition ■ • ■ A ■ Axis 1 A ■ A • • • A Figure 17. Graphical representation of the NMS ordination of sampled herbaceous-dominated intermittent and ephemeral riverine wetlands. Points represent species cover and composition for quadrats aggregated by site. Distance between points is proportional to dissimilarity between samples (i.e., samples with similar species composition are plotted closer together). Axis 1 represents 45.5% of the variation in the data and axis 2 accounts for 24.9% (total variation ex- plained = 70.4%)). Neither vegetation metrics nor the disturbance index were sufficiently corre- lated with ordination scores to be displayed as joint plots. bance gradient, but because vegetation within disturbance categories was so vari- able. Even though the riverine wetland "ref- erence domain" was restricted over the course of this study (i.e., the removal of per- ennial, alkaline, and woody-dominated wet- lands), these wetlands still encompassed considerable environmental heterogeneity . Variation in species composition related to this environmental heterogeneity is reflected by the multiple vegetation assemblages rep- resentative of both reference and highly dis- turbed sties. For example, reference condi- tion wetlands included sites along ephemeral streams dominated by tufted hairgrass and western wheatgrass as well as wetter sites along intermittent streams dominated by Nebraska sedge, common threesquare, and softstem bulrush (Schoenoplectus tabernae- montani). In contrast, depressional wetlands were consistently occupied by a relatively small suite of dominant species. The superior ability of metrics to ac- curately classify wetland condition is due to the attributes used. While multivariate analyses examined the response of the whole plant community, vegetation metrics, with the exception of Shannon diversity, were 28 based on plant functional groups. Func- tional groups classify plant species based on common attributes, adaptations, or responses to environmental factors (Runkiaer 1934, Grime 1977, 1988, Mclntyre and Lavorel 1994, Lavorel et al. 1997). For example, classifications of wetland plants have been proposed based on shared functional or life- history traits, such as plant height, total leaf area, life span, propagule longevity and es- tablishment requirements, and relative growth rate (van der Valk 1981, Boutin and Keddy 1993, Hills et al. 1994). In this study, two types of functional groups were used, native perennials and species assemblages based on coefficient of conservatism values. Both of these factors have been shown to be responsive to distur- bance. The index used to quantify human disturbance for riverine wetlands was pri- marily a measure of grazing intensity, and perennial species have been shown to re- spond negatively to grazing intensity in temperate grasslands in Australia (Mclntyre et al. 1995). The C-values used in this study were subjectively assigned by an expert panel and represent a collective best profes- sional judgment of a species' tolerance to disturbance and fidelity to habitat integrity (Northern Great Plains Floristic Quality As- sessment Panel 2001). As such, C-values are an integrative measure of species re- sponse to a broad array of anthropogenic stressors. These panel-assigned C-values were good indicators of species response and fidelity and gave comparable results when compared to C-values derived from an independent dataset by Mushet et al. (2002) for prairie pothole wetlands in North Da- kota. C-values and the resulting floristic quality index have been shown to be robust metrics in various wetland settings by An- dreas and Lichvar (1995), DeKeyser (2000), and Lopez and Fennessy (2002). Although the riverine multimetric index is relatively robust, it could be im- proved by defining additional functional groups based on shared species response to dominant stressors. For example, in the riv- erine systems evaluated, grazing is an im- portant disturbance factor. Using metrics based on a grazing-sensitive functional group (for examples see Mclntyre et al. 1995, Lavorel et al. 1997, Pausas and La- vorel 2003) could improve the biological sensitivity and overall accuracy of the mul- timetric index. The multimetric indices developed for depressional and riverine wetlands are based on limited datasets. To validate and refine these tools, as well as evaluate their applicability to other ecoregions in the Great Plains, they should be tested in additional watersheds and in adjacent ecoregions. Prairie potholes are found primarily on gla- cial landforms and so are largely limited to the northwestern glaciated plains ecoregion of northern Montana. In contrast, the river- ine multimetric model may have broad ap- plicability to intermittent and ephemeral streams for much of eastern Montana. Watershed disturbance rankings did not correlate with either small-scale distur- bance measures or vegetation metrics, and this finding argues against the use of large- scale land use patterns as a surrogate for site-level measures of disturbance. In con- trast, meso-scale land use variables, applied to a buffer area around a wetland or its up- stream catchment, were shown to be mean- ingful components of a spatially integrative site disturbance index. Unfortunately, the effectiveness of the watershed ranking model was hampered by methodological problems: high multicollinearity and an un- standardized scoring base among variables. A refined ranking procedure may provide a more accurate measure of human distur- bance at a 5^^-level watershed scale. Still, the applicability of such a model will be lim- ited by the correspondence between large- scale land use measures and the dominant 29 wetland stressor. For example, site-level grazing intensity is less likely to be corre- lated with watershed-scale land uses than is agricultural disturbance. Thus, a refined wa- tershed disturbance rank procedure would likely be more useful as an indicator of wet- land condition for depressional wetlands than riverine wetlands in the study area. Acknowledgments This study was made possible through a U.S. Environmental Protection Agency wetland protection grant adminis- tered by the Montana Department of Envi- ronmental Quality. I would first like to thank Lynda Saul and Randy Apfelbeck at DEQ. Lynda for her tireless efforts as the DEQ Wetland Coordinator and her vision for wetland conservation in Montana, and Randy for his work in developing and fur- thering a statewide wetland monitoring and assessment program. I would also like to thank Bryce Maxell and Brad Cook at the University of Montana for help with water- shed ranking, and Greg Kudray and Coburn Currier at the Montana Natural Heritage Program for assistance with editing and formatting this report. For field assistance, thanks to Coburn Currier, Randy Apfelbeck, Lynda Saul, and Erin Fehringer (DEQ). Many people helped me identify and pro- vided access to reference sites. These are Dennis Lingohr and Dave Waller (Bureau of Land Management), Fritz Prelwitz (U.S. Fish and Wildlife Service), Rick Northrup (Montana Department of Fish, Wildlife & Parks), Pat Fargey (Parks Canada), John Carlson and Steve Cooper (Montana Natural Heritage Program), and Linda Poole (The Nature Conservancy). Finally, my sincere thanks to Brad Lloyd for the gift of 20 liters of gas one late Sunday afternoon when every gas station within 100 km of Val Marie was closed. Literature Cited Andreas, B. K., and R. W. Lichvar. 1995. Floris- tic index for assessment standards: a case study for northern Ohio. Wetland Research Program Technical Report WRP-DE-8, U.S. Army Corps of Engineers Waterways Ex- periment Station, Vicksburg, Mississippi. Apfelbeck, R. S. 2001. Development of biocrite- ria for wetlands in Montana. Pages 141-166 in R B. Rader, D. P. Batzer, and S. A. Wissinger, editors. Bioassessment and man- agement of North American freshwater wet- lands. John Wiley & Sons, Inc., New York. Auble, G. T., and M. L. Scott. 1998. Fluvial dis- turbance patches and cottonwood recruit- ment along the upper Missouri River, Mon- tana. Wetlands 18:546-556. Biondini, M. E., P. W. Mielke, Jr., and K. J. Berry. 1988. Data-dependent permutation techniques for the analysis of ecological data. Vegetatio 75:161-168. Bisson, P. A., M. G. Raphael, A. D. Foster, and L. L. C. Jones. 2002. Influence of site and landscape features on vertebrate assem- blages in small streams. Pages 61-72 in A. C. Johnson, R. W. Haynes, and R. A. Mon- serud, editors. Congruent management of multiple resources: proceedings from the Wood Compatibility Initiative workshop. General Technical Report PNW-GTR-563, U.S. Department of Agriculture, Forest Ser- vice, Pacific Northwest Research Station, Portland, Oregon. Boggs, K., and T. Weaver. 1994. Changes in vegetation and nutrient pools during riparian succession. Wetlands 14:98-109. Borth, C. S. 1998. Effects of land use on vegeta- tion in glaciated depressional wetlands in western Montana. Masters Thesis. Montana State University, Bozeman, Montana. Boutin, C, and P. A. Keddy. 1993. A functional classification of wetland plants. Journal of Vegetation Science 4:591-600. Brinson, M. M., B. L. Swift, R C. Plantico, and J. S. Barclay. 1981. Riparian ecosystems: their ecology and status. U.S. Fish and Wild- life Service Biological Report 81, U.S. Gov- ernment Printing Office, Washington, D.C. Bryce, S. A., R. M. Hughes, and P. R. Kauf- mann. 2002. Development of a bird integrity 30 index: using bird assemblages as indicators of riparian condition. Environmental Management 30:294-3 1 0. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe. 1979. Classification of wetlands and deepwater habitats of the United States. FWS/OBS-79/31, U.S. Department of Inte- rior, Fish and Wildlife Service, Washington, D.C. Dahl, T. E. 1990. Wetland losses in the United States 1780's to 1980's. U.S. Department of Interior, Fish and Wildlife Service, Wash- ington, D.C. Dahl, T. E. 2000. Status and trends of wetlands in the conterminous United States 1986 to 1997. U.S. Department of Interior, Fish and Wildlife Service, Washington, D.C. Danielson, T. J. 2002. Methods for evaluating wetland condition: introduction to wetland biological assessment. EPA-822-R-02-014, Office of Water, U.S. Environmental Protec- tion Agency, Washington, D.C. DeKeyser, E. S. 2000. A vegetative classifica- tion of seasonal and temporary wetlands across a disturbance gradient using a mul- timetric approach. Ph.D. Dissertation, North Dakota State University, Fargo, North Da- kota. DeKeyser, E. S., D. R. Kirby, and M. J. Ell. 2003. An index of plant community integ- rity: development of the methodology for assessing prairie wetland plant communities. Ecological Indicators 3:119-133. Dufrene, M., and P. Legendre. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Eco- logical Monographs 67:345-366. Eby, L. A., W. F. Fagan, and W. L. Minckley. 2003. Variability and dynamics of a desert stream community. Ecological Applications 13:1566-1579. Euliss, N. H., Jr., and D. M. Mushet. 1996. Wa- ter-level fluctuation in wetlands as a func- tion of landscape condition in the prairie pothole region. Wetlands 16:587-593. Euliss, N. H., Jr., and D. M. Mushet. 1999. In- fluence of agriculture on aquatic inverte- brate communities of temporary wetlands in the prairie pothole region of North Dakota, USA. Wetlands 19:578-583. Euliss, N. H., Jr., D. M. Mushet, and D. H. Johnson. 2001. Use of macro invertebrates to identify cultivated wetlands in the prairie pothole region. Wetlands 21:223-231. Fennessy, M. S., R. Geho, B. Elifritz, and R. D. Lopez. 1998. Testing the floristic quality as- sessment index as an indicator of riparian wetland disturbance. Ohio Environmental Protection Agency, Division of Surface Wa- ters, Wetland Ecology Group, Columbus, Ohio. Fennessy, S., M. Gernes, J. Mack, and D. H. Wardrop. 2002. Methods for evaluating wet- land condition: using vegetation to assess environmental conditions in wetlands. EPA- 822-R-02-020, Office of Water, U.S. Envi- ronmental Protection Agency, Washington, D.C. Finch, D. M., and L. F. Ruggiero. 1993. Wildlife and biological diversity in the Rocky Moun- tains and northern Great Plains. Natural Ar- eas Journal 13:191-203. Findlay, C. S., and J. Houlahan. 1997. Anthro- pogenic correlates of species richness in southeastern Ontario wetlands. Conservation Biology 11:1000-1009. Fore, L. S., and C. Grafe. 2002. Using diatoms to assess the biological condition of large rivers in Idaho (USA). Freshwater Biology 47:2015-2037. Freeland, J. A., and J. L. Richardson. 1999. Soil indicators of agricultural impacts on north- ern prairie wetlands: Cottonwood Lake Re- search Area, North Dakota, USA. Wetlands 19:56-64. Friedman, J. M., and V. J. Lee. 2002. Extreme floods, channel change, and riparian forests along ephemeral streams. Ecological Mono- graphs 72:409-425. Friedman, J. M., W. R. Osterkamp, M. L. Scott, and G. T. Auble. 1998. Downstream effects of dams on channel geometry and bottom- land vegetation: regional patterns in the Great Plains. Wetlands 18:619-633. Galatowitsch, S. M., and A. G. van der Valk. 1996. The vegetation of restored and natural prairie wetlands. Ecological Applications 6:102-112. Grime, J. P. 1977. Evidence for the existence of three primary strategies in plants and its 31 relevance to ecological and evolutionary theory. American Naturalist 111:1169-1194. Grime, J. P. 1988. The C-S-R model of primary plant strategies - origins, implications and tests. Pages 371-393 in L. D. Gottlieb and S. K. Jain, editors. Plant evolutionary biology. Chapman & Hall, London. Hauer, F. R., B. J. Cook, M. C. Gilbert, E. C. Clairain, Jr., and R. D. Smith. 2002a. A re- gional guidebook for applying the hydro- geomorphic approach to assessing wetland functions of intermontane prairie pothole wetlands in the Northern Rocky Mountains. Technical Report ERDC/EL TR-02-7, U.S. Army Corps of Engineers Research and De- velopment Center, Vicksburg, Mississippi. Hauer, F. R., B. J. Cook, M. C. Gilbert, E. C. Clairain, Jr., and R. D. Smith. 2002b. A re- gional guidebook for applying the hydro- geomorphic approach to assessing wetland functions of riverine floodplains in the Northern Rocky Mountains. Technical Re- port ERDC/EL TR-02-21, U.S. Army Corps of Engineers Research and Development Center, Vicksburg, Mississippi. Heino, J., T. Muotka, H. Mykra, R. Paavola, H. Hamalainen, and E. Koskenniemi. 2003. De- fining macro invertebrate assemblage types of headwater streams: implications for bio- assessment and conservation. Ecological Applications 13:842-852. Helgen, J. C, and M. C. Gernes. 2001. Monitor- ing the condition of wetlands: indexes of biological integrity using invertebrates and vegetation. Pages 167-185 in R. B. Rader, D. P. Batzer, and S. A. Wissinger, editors. Bioassessment and management of North American freshwater wetlands. John Wiley & Sons, Inc., New York. Hills, J. M., K. J. Murphy, L D. Pulford, and T. H. Flowers. 1994. A method for classifying European riverine wetland ecosystems using functional vegetation groups. Functional Ecology 8:242-252. Houlahan, J. E., and C. S. Findlay. 2003. The effects of adjacent land use on wetland am- phibian species richness and community composition. Canadian Journal Fisheries and Aquatic Sciences 60:1078-1094. Hupp, C. R., and W. R. Osterkamp. 1985. Bot- tomland vegetation distribution along Pas- sage Creek, Virginia, in relation to fluvial landforms. Ecology 66:670-681. Johnson, W. C, T. L. Sharik, R. A. Mayes, and E. P. Smith. 1987. Nature and cause of zona- tion discreteness around glacial prairie marshes. Canadian Journal of Botany 65:1622-1632. Kantrud, H. A., J. B. Millar, and A. G. van der Valk. 1989. Vegetation of wetlands of the prairie pothole region. Pages 132-187 in A. G. van der Valk, editor. Northern prairie wetlands. Iowa State University Press, Ames, Iowa. Kantrud, H. A., and W. E. Newton. 1996. A test of vegetation-related indicators of wetland quality in the prairie pothole region. Journal of Aquatic Ecosystem Health Management 5:177-191. Karr, J. R. 1991. Biological integrity: a long ne- glected aspect of water resource manage- ment. Ecological Applications 1:66-84. Karr, J. R., and E. W. Chu. 1999. Restoring life in running waters: better biological monitor- ing. Island Press, Washington, D.C. Karr, J. R., and D. N. Kimberling. 2003. A ter- restrial arthropod index of biological integ- rity for shrub-steppe landscapes. Northwest Science 77:202-213. Kartesz, J. T. 1999. A synonymized checklist and atlas with biological attributes for the vascular flora of the United States, Canada, and Greenland. In J. T. Kartesz and C. A. Meacham, editors. Synthesis of the North American Flora, version 1.0. North Carolina Botanical Garden, Chapel Hill, North Caro- lina. Kauffman, J. B., and W. C. Krueger. 1984. Livestock impacts on riparian ecosystems and streamside management implications... a review. Journal of Range Management 37:430-438. Keddy, P. A. 2000. Wetland ecology: principles and conservation. Cambridge University Press, Cambridge. Kenkel, N. C, and L. Orloci. 1986. Applying metric and nonmetric multidimensional scal- ing to ecological studies: some new results. Ecology 67:919-928. Kimberling, D. N., J. R. Karr, and L. S. Fore. 2001. Measuring human disturbance using terrestrial invertebrates in the shrub-steppe 32 of eastern Washington (USA). Ecological Indicators 1:63-81. Kruskal, J. B. 1964. Nonmetric multidimen- sional scaling: a numerical method. Psy- chometrika 29:1 15-129. Lavorel, S., S. Mclntyre, J. Landsberg, and T. D. A. Forbes. 1997. Plant functional classifica- tions: from general groups to specific groups based on response to disturbance. Trends in Ecology and Evolution 12:474-478. Levene, H. 1960. Robust tests for equality of variance. Pages 278-292 in I. Olkin, editor. Contributions to probability and statistics. Stanford University Press, Palo Alto, Cali- fornia. Lopez, R. D., and M. S. Fennessy. 2002. Testing the floristic quality assessment index as an indicator of wetland condition. Ecological Applications 12:487-497. Mack, J. J. 2001. Vegetation index of biotic in- tegrity (VIBI) for wetlands: ecoregional, hy- drogeomorphic, and plant community com- parisons with preliminary wetland aquatic life use designations. Ohio Environmental Protection Agency, Division of Surface Wa- ters, Wetland Ecology Group, Columbus, Ohio. Mack, J. J., M. Micacchion, L. D. Augusta, and G. R. Sablak. 2000. Vegetation indices of biotic integrity (VIBI) for wetlands and cali- bration of the Ohio rapid assessment method for wetlands, v. 5.0. Ohio Environmental Protection Agency, Division of Surface Wa- ter, Wetland Ecology Unit, Columbus, Ohio. Mather, P. M. 1976. Computational methods of multivariate analysis in physical geography. J. Wiley & Sons, London. Matthews, J. W. 2003. Assessment of the floris- tic quality index for use in Illinois, USA, wetlands. Natural Areas Journal 23:53-60. McCune, B., and J. B. Grace. 2002. Analysis of ecological communities. MjM Software De- sign, Gleneden Beach, Oregon. McCune, B., and M. J. Mefford. 1999. Multi- variate analysis of ecological data, version 4.26. MjM Software Design, Gleneden Beach, Oregon. Mclntyre, S., and S. Lavorel. 1994. Predicting richness of native, rare, and exotic plants in response to habitat and disturbance variables across a variegated landscape. Conservation Biology 8:521-531. Mclntyre, S., S. Lavorel, and R. M. Tremont. 1995. Plant life-history attributes: their rela- tionship to disturbance response in herba- ceous vegetation. Journal of Ecology 83:31- 44. Minchin, P. R. 1987. An evaluation of the rela- tive robustness of techniques for ecological ordination. Vegetatio 69:97-110. Mushet, D. M., N. H. EuHss, Jr., and T. L. Shaffer. 2002. Floristic quality assessment of one natural and three restored wetland complexes in North Dakota, USA. Wetlands 22:126-138. National Research Council. 2001. Compensating for wetland losses under the Clean Water Act. Committee on Mitigating Wetland Losses, Board on Environmental Studies and Toxicology, National Academy Press, Washington, D.C. Nesser, J. A., G. L. Ford, M. C. Lee, and D. S. Page-Dumroese. 1997. Ecological units of the Northern Region: subsections. General Technical Report INT-GTR-369, U.S. De- partment of Agriculture, Forest Service, Intermountain Research Station, Ogden, NorUhfeth. Great Plains Floristic Quality Assess- ment Panel. 2001. Floristic quality assess- ment for plant communities of North Da- kota, South Dakota (excluding the Black Hills), and adjacent grasslands. Northern Prairie Wildlife Resource Center, James- town, North Dakota. Retrieved October 5, 2002, from http://www.npwrc.usgs.gov/ resource/200 1/fqa/fqa.htm. Patten, D. T. 1998. Riparian ecosystems of semi- arid North America: diversity and human impacts. Wetlands 18:498-512. Pausas, J. G., and S. Lavorel. 2003. A hierarchi- cal deductive approach for functional types in disturbed ecosystems. Journal of Vegeta- tion Science 14:409-416. Rahel, F. J. 1990. The hierarchical nature of community persistence: a problem of scale. American Naturalist 136:328-344. Richardson, R. E., and L. T. Hanson. 1977. Soil survey of Sheridan County, Montana. U. S. Department of Agriculture, Soil Conserva- tion Service, Bozeman, Montana. 33 Runkiaer, C. 1934. The life forms of plants and statistical plant geography. Clarendon Press, Oxford. Schulz, T. T., and W. C. Leininger. 1990. Dif- ferences in riparian vegetation structure be- tween grazed areas and exclosures. Journal of Range Management 43:295-299. Scott, M. L., G. T. Auble, and J. M. Friedman. 1997. Flood dependency of cottonwood es- tablishment along the Missouri River, Mon- tana, USA. Ecological Applications 7:677- 690. Scott, M. L., S. K. Skagen, and M. F. Merigli- ano. 2003. Relating geomorphic change and grazing to avian communities in riparian forests. Conservation Biology 17:284-296. Seabloom, E. W., and A. G. van der Valk. 2003. Plant diversity, composition, and the inva- sion of restored and natural prairie pothole wetlands: implications for restoration. Wet- lands 23:1-12. Stewart, R. E., and H. A. Kantrud. 1971. Classi- fication of natural ponds in the glaciated prairie region. Resource Publication 92, U.S. Fish and Wildlife Service, Washington, D.C. Swink, F. A., and G. S. Wilhem. 1979. Plants of the Chicago region: a checklist of the vascu- lar flora of the Chicago region, with keys, notes on local distribution, ecology, and taxonomy, and a system for evaluation of plant communities. Morton Arboretum, Lisle, Illinois. SYSTAT. 2002. SYSTAT for Windows, Ver- sion 10.2. SYSTAT Software Inc., Rich- mond, California. Tangen, B. A., M. G. Butler, and M. J. Ell. 2003. Weak correspondence between macroinvertebrate assemblages and land use in prairie pothole region wetlands, USA. Wetlands 23:104-115. Teels, B. M., and P. Adamus. 2002. Methods for evaluating wetland condition: developing metrics and indices of biological integrity. EPA-822-R-02-016, Office of Water, U.S. Environmental Protection Agency, Washington, D.C. Trombulak, S. C, and C. A. Frissell. 2000. Re- view of ecological effects of roads on terres- trial and aquatic communities. Conservation Biology 14:18-30. U.S. Department of Agriculture. 2004. The PLANTS database. Version 3.5 (http://plants.usda.gov). National Plant Data Center, Baton Rouge, Louisiana. U.S. EPA. 1994. National water quality inven- tory. 1992 report to Congress, EPA 841-R- 94-001, U.S. Environmental Protection Agency, Washington, D.C. van der Valk, A. G. 1981. Succession in wet- lands: a Gleasonian approach. Ecology 62:688-696. van der Valk, A. G., and C. H. Welling. 1988. The development of zonation in freshwater wetlands: an experimental approach. Pages 145-158 in H. During, M. Werger, and J. Willems, editors. Diversity and pattern in plant communities. Academic Publishing, The Hague. Western Regional Climate Center. 2004. Mon- tana climate summaries. National Climatic Data Center, 1971-2000 climate normals. National Oceanic and Atmospheric Admini- stration. Retrieved January 19, 2004, from http://www.wrcc. dri.edu/summary/climsmm t.html. Whittaker, R. J. 1972. Evolution and measure- ment of species diversity. Taxon 21:213- 251. Woods, A. J., J. M. Omernik, J. A. Nesser, J. Shelden, and S. H. Azevedo. 1999. Ecore- gions of Montana (1:1,500,000 map). U.S. Geological Survey, Reston, Virginia. Wright, J. P., A. S. Flecker, and C. G. Jones. 2003. Local vs. landscape controls on plant species richness in beaver meadows. Ecol- ogy 84:3162-3173. 34 Appendix A. Species Lists for Depressional and Riverine Wetlands Table A. List of vascular plant species observed in temporarily and seasonally flooded depres- sional wetlands. Growth Scientific Name Common Name Form^ Duration^ Nativity^ Achillea millefolium common yarrow F P N Agropyron cristatum crested wheatgrass G P E Agrostis scabra rough bentgrass G P N Agrostis stolonifera creeping bentgrass G P E Alopecurus geniculatus water foxtail G P E Alopecurus pratensis meadow foxtail G P E Argentina anserina silverweed cinquefoil F P N Arnica fulgens foothill arnica F P N Artemisia cana silver sagebrush S P N Artemisia ludoviciana white sagebrush F P N Atriplex argentea silverscale saltbush F A/B N Beckmannia syzigachne American sloughgrass G A/B N Bouteloua gracilis blue grama G P N Bromus hordeaceus ssp. hor- soft brome G A/B E deaceus Bromus inermis smooth brome G P E Bromus japonicus Japanese brome G A/B E Carex atherodes wheat sedge G P N Carex duriuscula needleleaf sedge G P N Carex pellita woolly sedge G P N Cerastium nutans nodding chickweed F A/B N Chenopodium album lambsquarters F A/B E Chenopodium spp. goosefoot F A/B E/N Cirsium arvense Canada thistle F P E Collomia linearis tiny trumpet F A/B N Convolvulus arvensis field bindweed F P E Cryptantha torreyana Torrey's cryptantha F A/B N Deschampsia caespitosa tufted hairgrass G P N Descurainia incana mountain tansymustard F A/B N Descurainia sophia herb sophia F A/B E Distichlis spicata inland saltgrass G P N Eleocharis acicularis needle spikerush G P N Eleocharis palustris common spikerush G P N Elymus repens quackgrass G P E Festuca spp. fescue G P N Glaux maritima sea milkwort F P N Gnaphalium palustre western marsh cudweed F A/B N Grindelia squarrosa curlycup gumweed F A/B N Hordeum jubatum foxtail barley G P N Juncus balticus Baltic rush G P N A-1 Growth Scientific Name Common Name Form^ Duration^ Nativity^ Koeleria macrantha prairie junegrass G P N Lepidium densiflorum common pepperweed F A/B N Leptochloa fusca ssp. fascicularis bearded sprangletop G A/B N Medicago saliva alfalfa F P E Mentha arvensis wild mint F P N Muhlenbergia richardsonis mat muhly G P N Myosurus apetalus bristly mousetail F A/B N Navarretia intertexta needleleaf navarretia F A/B N Opuntia polyacantha plains pricklypear S P N Pascopyrum smithii western wheatgrass G P N Penstemon spp. beardtongue F P N Plagiobothrys scouleri sleeping popcornflower F A/B N Plantago elongata prairie plantain F A/B N Poa palustris fowl bluegrass G P N Poa pratensis Kentucky bluegrass G P E Poa secunda Sandberg bluegrass G P N Polygonum spp. knotweed F P N Polygonum ramosissimum bushy knotweed F A/B N Puccinellia nuttalliana Nuttall's alkaligrass G P N Ratibida columnifera prairie coneflower F P N Rumex crispus curly dock F P E Rumex spp. dock F P E/N Rumex salicifolius willow dock F P N Schedonnardus paniculatus tumblegrass G P N Schoenoplectus pungens common threesquare G P N Sonchus arvensis ssp. uliginosus moist sowthistle F P E Taraxacum officinale common dandelion F P E Thlaspi arvense field pennycress F A/B E Tragopogon dubius yellow salsify F A/B E Trifolium spp. clover F P E Veronica peregrina neckweed F A/B N Vicia americana American vetch F P N Vulpia octoflora sixweeks fescue G A/B N F = forb, G = graminoid, S = shrub A/B = annual/biennial, P = perennial E = exotic, N = native A-2 Table B. List of vascular plant species observed in intermittent and ephemeral riverine wetlands. Growth Scientific Name Common Name Form^ Duration^ Nativity^ Achillea millefolium common yarrow F P N Achnatherum nelsonii Columbia needlegrass G P N Agoseris glauca pale agoseris F P N Agropyron cristatum crested wheatgrass G P E Agrostis scabra rough bentgrass G P N Agrostis stolonifera creeping bentgrass G P E Allium geyeri Geyer's onion F P N Allium textile textile onion F P N Alopecurus geniculatus water foxtail G P E Antennaria microphylla littleleaf pussytoes F P N Apocynum cannabinum Indianhemp F P N Argentina anserina silverweed cinquefoil F P N Artemisia cana silver sagebrush S P N A rtemisia frigida prairie sagewort s P N Artemisia ludoviciana white sagebrush F P N Atriplex argentea silverscale saltbush F A/B N Beckmannia syzigachne American sloughgrass G A/B N Bouteloua gracilis blue grama G P N Bromus hordeaceus ssp. hor- soft brome G A/B E deaceus Bromus japonicus Japanese brome G A/B E Calamagrostis stricta northern reedgrass G P N Calamovilfa longifolia prairie sandreed G P N Calystegia sepium hedge false bindweed F P N Carex aquatilis water sedge G P N Carex spp. sedge G P N Carex nebrascensis Nebraska sedge G P N Carex pellita woolly sedge G P N Carex praegracilis clustered field sedge G P N Chamaesyce serpyllifolia thymeleaf sandmat F A/B N Chenopodium album lambsquarters F A/B E Chenopodium spp. goosefoot F A/B E/N Chenopodium pratericola desert goosefoot F A/B N Cicuta douglasii western water hemlock F P N Cirsium arvense Canada thistle F P E Cirsium vulgare bull thistle F A/B E Collomia linearis tiny trumpet F A/B N Convolvulus arvensis field bindweed F P E Conyza canadensis Canadian horseweed F A/B N Danthonia spp. oatgrass G P N Deschampsia caespitosa tufted hairgrass G P N Descurainia sophia herb sophia F A/B E A-3 Growth Scientific Name Common Name Form^ Duration^ Nativity^ Distichlis spicata inland saltgrass G P N Downingia laeta Great Basin calicoflower F A/B N Echinacea angustifolia blacksamson echinacea F P N Echinochloa crus-galli barnyardgrass G A/B E Eleocharis acicularis needle spikerush G P N Eleocharis palustris common spikerush G P N Elymus elymoides squirreltail G P N Elymus repens quackgrass G P E Elymus trachycaulus slender wheatgrass G P N Epilobium spp. willowherb F P N Epilobium pygmaeum smooth spike-primrose F A/B N Equisetum arvense field horsetail F P N Erigeron spp. fleabane F P N Euphorbia esula leafy spurge F P E Gaillardia aristata common gaillardia F P N Galium boreale northern bedstraw F P N Glaux maritima sea milkwort F P N Glycyrrhiza lepidota American licorice F P N Gnaphalium palustre western marsh cudweed F A/B N Grindelia squarrosa curlycup gumweed F A/B N Hackelia deflexa nodding stickseed F A/B N Helianthella quinquenervis fivenerve helianthella F P N Helianthella uniflora oneflower helianthella F P N Helianthus annuus common sunflower F A/B N Helianthus nuttallii Nuttall's sunflower F P N Hesperostipa comata needle and thread G P N Heterotheca villosa hairy false goldenaster F P N Hieracium spp. hawkweed F P N Hordeum jubatum foxtail barley G P N Juncus balticus Baltic rush G P N Juncus spp. rush G P N Koeleria macrantha prairie Junegrass G P N Lactuca serriola prickly lettuce F A/B E Lemna minor common duckweed F P N Lepidium densiflorum common pepperweed F A/B N Lesquerella arenosa Great Plains bladderpod F A/B N Linum lewisii prairie flax F P N Lomatium spp. desertparsley F P N Lycopus asper rough bugleweed F P N Maianthemum stellatum starry false lily of the valley F P N Medicago sativa alfalfa F P E Mentha arvensis wild mint F P N Muhlenbergia asperifolia scratchgrass G P N A-4 Growth Scientific Name Common Name Form^ Duration^ Nativity^ Muhlenbergia richardsonis mat muhly G P N Muhlenbergia spp. muhly G P N Nassella viridula green needlegrass G P N Navarretia intertexta needleleaf navarretia F A/B N Opuntia polyacantha plains pricklypear S P N Pascopyrum smithii western wheatgrass G P N Pediomelum argophyllum silverleaf Indian bread- root F P N Plantago elongata prairie plantain F A/B N Plantago spp. plantain F P E Plantago major common plantain F P E Poa arida plains bluegrass G P N Poa spp. bluegrass G P N Poa palustris fowl bluegrass G P N Poa pratensis Kentucky bluegrass G P E Poa secunda Sandberg bluegrass G P N Polygonum aviculare prostrate knotweed F A/B N Polygonum e rectum erect knotweed F A/B N Polygonum spp. knotweed F P N Polygonum ramosissimum bushy knotweed F A/B N Potentilla gracilis slender cinquefoil F P N Puccinellia nuttalliana Nuttall's alkaligrass G P N Ranunculus cymbalaria alkali buttercup F P N Ranunculus spp. buttercup F A/B N Ratibida columnifera prairie coneflower F P N Rhus trilobata skunkbush sumac S P N Ribes aureum golden currant s P N Rosa woodsii Woods' rose s P N Rumex aquaticus western dock F P N Rumex crispus curly dock F P E Rumex spp. dock F P E/N Rumex salicifolius willow dock F P N Salix amygdaloides peachleaf willow T P N Salix exigua narrowleaf willow S P N Salix spp. willow s P N Salsola tragus prickly Russian thistle F A/B E Schoenoplectus maritimus cosmopolitan bulrush G P N Schoenoplectus pungens common threesquare G P N Schoenoplectus tabernaemontani softstem bulrush G P N Selaginella densa lesser spikemoss F P N Solidago spp. goldenrod F P N Sonchus arvensis ssp. uliginosus moist sowthistle F P E Spartina gracilis alkali cordgrass G P N Spartina pectinata prairie cordgrass G P N A-5 Growth Scientific Name Common Name Form^ Duration^ Nativity^ Stellaria spp. starwort F A/B N Suaeda calceoliformis Pursh seepweed F A/B N Symphoricarpos occidentalis western snowberry S P N Symphyotrichum falcatum white prairie aster F P N Symphyotrichum spathulatum western mountain aster F P N Taraxacum officinale common dandelion F P E Thermopsis rhombifolia prairie thermopsis F P N Thlaspi arvense field pennycress F A/B E Tragopogon dubius yellow salsify F A/B E Trifolium spp. clover F P E Trifolium repens white clover F P E Triglochin maritimum seaside arrowgrass G P N Typha latifolia broadleaf cattail F P N Urtica dioica stinging nettle F P N Veronica peregrina neckweed F A/B N Vicia americana American vetch F P N Viola spp. violet F P N Xanthium strumarium rough cockleburr F A/B N Zigadenus elegans mountain deathcamas F P N i F = forb, G = graminoid, S = shrub, T = tree A/B = annual/biennial, P = perennial E = exotic, N = native A-6 Appendix B. Photographs of Sites Representative of Reference Condition, Mod- erately Disturbed, and Severely Disturbed Wetlands B Figure A. Examples of (A) reference condition, (B) moderately disturbed, and (C) severely dis- turbed temporarily and seasonally flooded depressional wetlands. B-1 Figure B. Examples of (A) reference condition, (B) moderately disturbed, and (C) severely dis- turbed ephemeral riverine wetlands. B-2 Figure C. Examples of (A) reference condition, (B) moderately disturbed, and (C) severely dis- turbed intermittent riverine wetlands. B-3