Great Lakes Regional Aquatic GAP

Anticipated completion date: September 2007

Contact: Jana Stewart, Regional Coordinator

USGS, Middleton, Wisconsin

jsstewar@usgs.gov, (608) 821-3855

Michigan: Stephen S. Aichele, Co-PI

USGS, Lansing, Michigan

saichele@usgs.gov, (517) 887-8918

Dora R. Passino-Reader, Co-PI

USGS, Ann Arbor, Michigan

Dora_Reader@usgs.gov, (734) 214-7229

New York : James E. McKenna, PI

USGS, Cortland, New York

Jim_McKenna@usgs.gov, (607) 753-9391 x21

Ohio: S. Alex Covert, PI

USGS, Columbus, Ohio

sacovert@usgs.gov, (614) 430-7752

Wisconsin: Jana S. Stewart, PI

USGS, Middleton, Wisconsin

jsstewar@usgs.gov, (608) 821-3855

The Great Lakes Aquatic GAP began as a regional project in 2001, to be completed in 2007 in the states of Michigan (MI), New York (NY), and Wisconsin (WI). The Ohio (OH) Aquatic GAP pilot project has been in progress since early 2000 (see separate status report for Ohio in this section). The objectives of the regional project are to develop an Aquatic Gap Analysis for riverine systems in all eight states in the Great Lakes Region by 2009. Projects are planned sequentially, with new projects starting up when existing ones are nearing completion. In addition, a Coastal Pilot project is under way to develop a habitat classification framework, aquatic biota database, and initial gap analysis for near-shore coastal systems of the Great Lakes. Two pilot studies are currently under way in western Lake Erie and eastern Lake Ontario.

Development of a regionally consistent database and spatial data layers, with uniformity across state boundaries, has been a major focus of the Great Lakes Aquatic GAP project. A central relational database is being developed to accommodate stream habitat characteristics, aquatic biota, and habitat affinity data in a consistent manner across all states in the Great Lakes region. Processing of stream habitat characteristics has also been coordinated across state boundaries, so assessments can be completed at the regional, state, and local levels.

Central Database development: The Central Database uses Oracle Discoverer and Oracle 9i software with capabilities for data sharing on the Web through user-client and Web-interface clients and is housed at the USGS Great Lakes Science Center in Ann Arbor, MI. Fish data have been acquired, organized, formatted, and reviewed for MI, NY, OH, and WI, with plans to load all data into the Central Database by June 2004. Habitat characteristics of stream segments for MI and WI will be loaded by July 2004 and those of streams in the Great Lakes drainages of NY by September 2004. Fish life history and habitat affinity information is in the process of being acquired and will be loaded into the Habitat Affinity section of the Central Database for validation of predicted fish distributions and analysis of fish community ecology. Plans are under way to develop additional data tables to store habitat characteristics for the Coastal Pilot studies. Future plans also include acquisition and review of invertebrate and freshwater mussel databases for possible incorporation into the Central Database.

A Web-based map application prototype has been developed to produce dynamic species distribution maps for the WI Aquatic project, with future plans to apply the map interface to the Central database. Using a relational database and spatially enabling the data (Oracle Spatial), a user is dynamically able to query the database via a Web browser through a graphical user map interface.

Stream habitat classification and modeling: The stream habitat classification methods developed and used by the Missouri (MoRAP) and OH Aquatic GAP projects were reviewed and modified for use by the Great Lakes Aquatic GAP project to improve habitat characterization, model predictions and analysis, and regional consistency. The modifications were made to better reflect factors contributing to physical habitat at all scales, including the channel segment, riparian buffer, and entire upstream contributing area of each channel segment. These modifications include ( 1) preservation of interval/ratio data rather than grouping all variables into categories, (2) calculation of habitat characteristics for not only the channel segment but the riparian buffer and watershed for each segment, (3) preservation and attribution of lake and double-line stream features, (4) addition of land cover, (5) and development of temperature and flow models to predict real values of temperature and flow for each stream segment.

Numerous processing scripts were developed by the MI Aquatic GAP team to expedite processing of stream habitat characteristics and provide consistency in methods and results. Streams have been attributed with habitat characteristics in MI and WI and will be 80% complete for the Great Lakes drainages of NY by June 2004. As part of habitat characterization, regression models have been developed to predict stream temperature for every stream segment in MI and WI. Fish sample locations will also be linked to stream segments by June 2004. Initial exploration of modeling methods for fish-environment relationships has begun, and a significant array of example results should be available for evaluation by September 2004.

Coastal GAP pilot project: A conceptual framework for identification and classification of coastal habitat types has been developed and applied to the western Lake Erie pilot study area. Databases of fish distributions in western Lake Erie and eastern Lake Ontario have been acquired. A substantial amount of fieldwork, designed to help assess the efficacy of the classification framework and to collect data from unsampled and important habitat types, was completed.

Outreach and meetings: A USGS fact sheet describing the Great Lakes GAP project was published in June 2003 and is available on the Great Lakes GAP Web page ( http://www.glsc.usgs.gov/GLGAP.htm). Numerous papers describing the Great Lakes Aquatic GAP project were presented in a special session entitled “Biodiversity Conservation in the Great Lakes Region” at the International Association for Great Lakes Research in Chicago, IL, in June 2003. A Great Lakes Aquatic GAP poster was presented at the Society for Conservation Biology meeting in Duluth, MN, in June 2003, and a number of papers and a poster were presented at the National GAP meeting in Fort Collins, CO, in October 2003. Contributed and symposia papers describing the Central Database, preliminary modeling results, and the overall projects have been submitted for presentation at the 134 th Annual Meeting of the American Fisheries Society in Madison, WI, in August 2004.

The Great Lakes Aquatic GAP team has worked closely with an EPA Star Grant group on ecological classification of the rivers of IL, MI, and WI to share expertise and develop regional methods for habitat classification and species modeling. Individual state projects continue to work closely with stakeholder agencies, including the MI Institute of Fisheries Research (MIFR), the New York State Department of Environmental Conservation, and the WI Department of Natural Resources (WDNR).

Hawaii Aquatic GAP

Anticipated completion date: May 2005

Contact: Michael H. Kido

University of Hawaii, Hawaii Stream Research Center, Kapaa

mkido@hawaii.edu, (808) 956-4907

In the past year the HI-GAP project has established a methodology for mapping the distribution of aquatic species. The methodology is unique and accounts for Hawaii’s topographic complexity and climate. The model has been implemented on Kauai and Maui, and preliminary results indicate the methodology is producing an accurate indication of alien and native aquatic species distribution. The model is also providing indicators of stresses on individual watersheds to identify watersheds with degraded stream habitat. The project has also been successful in bringing together stream researchers in Hawaii through interest meetings and positive collaboration among partners.

Analysis: The modeling methodology being implemented utilizes cluster analysis modeling to group watersheds per island into similar entities. The clustered watersheds are then analyzed, and aquatic species distribution along the stream continuum is mapped. The watershed clustering is based on variables derived from the GIS, which include average slope, average elevation, aspect values, watershed size, stream length, percentage of perennial streams, percentage of intermittent streams, and land cover types. Each value is summarized per watershed and used as inputs to cluster the watersheds into similar entities. Initial results indicate between 5 and 8 different watersheds exist per island.

The aquatic species distribution along each stream is then mapped, using GIS attributes derived for all streams. The variables used to define species ranges include slope variation, elevation ranges, waterfall locations, and distance from the mouth of the stream. Based on extensive observation data, each aquatic species is assigned physical attributes that define its desired habitat and range. The assigned habitat parameters are then used to define the distribution of each aquatic species along the stream continuum. The results for the Island of Kauai indicate the methodology is accurate.

Future work: In an effort to protect the remaining native aquatic species and identify the location of alien aquatic species, the Hawaii Gap Analysis Project will be providing the derived data on an ArcIMS server to allow public access to this information for academic research and public awareness. HI-GAP is also planning on conducting several analyses to prioritize streams for future conservation. The overall goal of the HI-GAP project is to approach the assessment of conservation from an integrated system of terrestrial, aquatic, and marine ecosystems.

Lower Missouri River Basin Aquatic GAP

a. Iowa

Anticipated completion date: December 2004

Contact: Kevin Kane

Iowa State University, Ames

kkane@iastate.edu, 515/294-0526

Recent achievements

a) Biological assessment:

• completed draft range maps for all fish species
• completed test models for Eastern Broadleaf Forest subregion using AnswerTree software, including determining habitat variables
• completed the professional review of one third of all species
• submitted an article to the 2003 GAP Bulletin
• presented a poster on the development of the biological inventory database at the Iowa Water Monitoring Conference
• submitted an abstract for the Ecological Society of America annual meeting.
  

b) GIS coverages:
The NHD linework was received from MoRAP in two separate coverages.  MoRAP had processed the IA linework we sent them early last year to produce a variety of attributes that would possibly be used to model fish distributions.  Roughly 25 attributes were created and assigned variables using ARC aml scripts.  The two coverages were modified to be able to be mapjoined, and the resulting coverage was processed to add three additional attributes deemed necessary for modeling.  Not all 28 attributes will be used in modeling; a subset of about 9 variables will be used. 

Several attributes have been identified as necessary already (flow, temperature, size discrepancy, gradient, and downstream link), and the linework is being checked for valid values in those attributes.  Flow is a variable directly from the original NHD, and it has over 900 zero values in the 57 watersheds for Iowa GAP.  Those are being corrected using aerial photography and 24K topo maps.  It is possible several secondary channels that were not part of the attribute calculations at MoRAP will have to be used in our final modeling due to the need to keep the fish samples on those reaches.  The model variable values for those secondary channels will have to be calculated, and the process and source data are being investigated and compiled.

Tasks to complete

We have made excellent progress in many areas thus far, including what has been referred to as one of the premier state historical fish record databases in the country. The spatial data portion of the project has progressed more slowly, however. We just received what we hope is the final river database from MoRAP, our data provider.

We are leveraging Aquatic GAP funding with the Iowa DNR, which continues to fund the development of the complimentary Iowa Rivers Information System (IRIS) this year.

a) GIS databases and analysis:

• Prepare final database
• Categorize variables based on preliminary biologic analysis
• Create analysis variables
• Run all analyses
• Create and print preliminary analysis maps
• Create and print final analysis maps
• Update stewardship layer
• Stewardship/prediction analysis

b) Biological assessment:

• Complete aquatic inventory data set
• Compile aquatic range maps from current and historical data
• Build habitat models
• Professional review of ranges and habitat
• GIS analysis for predictive distributions (three months)
• Report writing

c) Access and Interface:

• Create IA Aquatic GAP Web site for information and data/analysis/map delivery
• Create IMS interface for users (extension to IRIS IMS)

b. Kansas

Anticipated completion date: May 2005

Contact: Keith Gido

Kansas State University, Manhattan

kgido@ksu.edu, (785) 532-6615

The Kansas Aquatic GAP project is currently in its third year. To date we have acquired data on the distribution and abundance of fish and mussel species from over 4,000 localities in Kansas . These data have come from a variety of sources including the Kansas Department of Wildlife and Parks, Kansas Department of Health and Environment, University of Kansas Museum of Natural History, Kansas Natural Heritage Inventory, Sternberg Museum of Natural History, The Nature Conservancy, and various individuals with scientific collections in the state. With the help of MoRAP, we have modified the existing National Hydrology Database in our state to include a suite of environmental variables for over 100,000 individual valley segments. These data sets are being combined to construct predictive models of species occurrences across the state. In addition, we are working with other colleagues involved in Aquatic GAP projects in the Missouri River Basin to standardize efforts among states.

Our data have been compiled, and species distribution maps have been constructed for peer review and are available on our Web page ( www.ksu.edu/aquaticgap). As a pilot study, we have evaluated several modeling approaches and different suites of environmental variables in the Big Blue River basin. Although some species models do not perform well in this region, several models for species of special concern worked quite well and are promising tools for conservation. We are continuing to explore (1) the feasibility of different modeling approaches, (2) the use of different independent variable sets, and (3) what scale of analyses is appropriate for this region.

c. Missouri

Anticipated completion date: December 2005

Contact: Scott Sowa

MoRAP, University of Missouri, Columbia

scott_sowa@usgs.gov, (573) 441-2791

MoRAP is completing the aquatic ecological classification system for the Lower Missouri River Basin, which includes the states of Iowa, Kansas, and Nebraska. This past year we completed a draft version of Aquatic Subregions and Ecological Drainage Units for the lower basin. These draft units will have to be verified and modified through analyses of the biological data sets currently being developed in each of the states. We have completed the Valley Segment Classifications for Iowa and Kansas, and we are currently working on this same coverage for Nebraska. Continued funding from the Gap Analysis Program has allowed us to begin the classification of Aquatic Ecological System Types in all three states. We will also be generating human stressor statistics for each of the AESs and generating local, upstream riparian, and overall watershed ownership statistics (by GAP stewardship category) for each of the stream segments within the lower basin. All of these data sets are scheduled to be completed in December 2005.

Missouri Aquatic GAP

Anticipated completion date: October 2004

Contact: Scott Sowa

MoRAP, University of Missouri , Columbia

scott_sowa@usgs.gov, (573) 441-2791

Aquatic ecological classification: We are currently working on detailed biophysical descriptions for each of the Aquatic Subregions, Ecological Drainage Units, and Aquatic Ecological Systems. A draft version of the specific classification procedures for each level of the hierarchy has been completed.

Species modeling: We developed 517 separate models in order to generate predictive distribution maps for 315 species of fish, mussels, and crayfish. The vast majority of predicted models were developed empirically using Decision Tree Analyses. However, due to data limitations, some models were based on contingency-table analyses combined with habitat-affinity information compiled from existing literature. A hyperdistribution MS Access database has been completed, which spatially links to our statewide valley segment coverage via unique segment identifiers. This relational database also contains taxonomic, ecological, and conservation status information for each species.

We also completed an accuracy assessment of our models using independent data sets for each taxon. Table 1 provides a breakdown of the overall, commission, and omission errors. Omission errors are relatively low, averaging less than 10%, while commission errors are relatively high, averaging over 50%. However, these accuracy statistics are very misleading. There are many problems associated with this accuracy assessment related to spatial and temporal sampling “inadequacies” of the independent data sets and with the inherent difference in what we are trying to predict (i.e., biological potential) versus the fact that most of the stream segments sampled in these independent data sets were degraded to some degree. In fact, some of the sites are highly degraded, and in such instances we would expect very little correspondence between our predicted assemblage and the assemblage that presently occupies the site. Based on a separate evaluation of two fish collection sites, where the data are more temporally and spatially comprehensive, we found our overall accuracy to increase to nearly 70%, mainly due to a significant decrease in commission errors. A proper evaluation of the accuracy of our models will require a separate project that identifies relatively high-quality sites, which are then sampled intensively throughout long stretches of stream during several seasons and over a period of several years.

Habitat-affinity reports were completed for each species. These reports can be viewed at the MoRAP Web site, http://www.cerc.usgs.gov/morap/projects.asp?project_id=1. These are stand-alone reports, which provide images of the species, the predicted distribution map (for species that are not state-listed as either rare, threatened, or endangered), state range description, habitat affinity information extracted from the literature, the predictive model(s), and literature pertaining to that species. A draft version of the methods used to develop the predictive models was also completed.

Human stressors/threats: In addition to assessing multiple forms of public ownership, we must also consider existing human stressors, since even ownership of a substantial portion of a watershed does not ensure effective conservation. To account for this, we generated statistics for nearly 50 individual human stressors (e.g., percent urban, lead mine density, degree of fragmentation) for each Aquatic Ecological System, which are ecologically defined watershed units. We then used correlation analyses to reduce this overall set of metrics into a final set of 11 relatively uncorrelated measures of human disturbance or stress. Relativized rankings (range 1 to 4) were then developed for each of these 11 metrics. A rank of 1 is indicative of relatively low disturbance for that particular metric, while a rank of 4 indicates a relatively high level of disturbance. These rankings were based on information contained within the literature or simply quartiles when no empirical evidence on thresholds was available. For instance, rankings for percent urban were 1: 0-5%, 2: 6-10%, 3: 11-20%, and 4: >20%, based on the collective results of various studies that have examined the effects of urban land cover on the ecological integrity of stream ecosystems. However, existing research for percent agriculture has not identified clear thresholds, suggesting a more or less continual decline in ecological integrity with each added percentage of agriculture in the watershed. For this measure of human stress we simply used quartiles, 1: 0-25%, 2: 26-50%, 3: 51-75%, and 4: >75%.

The relativized rankings for each of these 11 metrics were then combined into a three-number Human Stressor Index. The first number reflects the highest ranking across all 11 metrics (range 1 to 4). The last two numbers reflect the sum of the 11 metrics (range 11 to 44). This index allows you to evaluate both individual and cumulative impacts. For instance, a value of 418 indicates relatively low cumulative impacts (i.e., last two digits = 18 out of a possible 44), however, the first number is a 4, which indicates that one particular land use is potentially severe. This index is an admittedly crude measure of human disturbance, however, it is well suited for a coarse-filter assessment since it does act as a red flag. This purpose of the HSI is to further evaluate those locations that appear to be well represented within the existing matrix of public lands, since ownership does not ensure effective conservation within riverine ecosystems. Using this index, we have found a handful of instances where stream segments flowing within GAP 1 or 2 lands are not being adequately conserved due to human disturbances occurring outside the boundaries of the public lands.

Gap analysis: We found a major error in the STATSGO soil coverage for Missouri , which affected the classification of our Aquatic Ecological Systems and resulted in a trickle-down effect on our already completed gap analysis for Missouri . We have since fixed the problem with the STATSGO coverage, reclassified our AES-types, and are now redoing the gap analysis. Although the errors in the STATSGO coverage were major and will change the specific geographic context of our gap analysis, the general conclusions from the original analysis will not change.

It is quite evident from our original analysis that only a tiny fraction of the stream resources in Missouri are being adequately represented in the current matrix of public lands. This statement does not pertain so much to the amount of the stream resource base currently within public lands as it does to the spatial arrangement of public ownership. From a local ownership perspective, 5% (9,365 km) of total stream length in Missouri (173,074 km) are currently flowing through public lands. Based on our definition of what constitutes a gap (in terms of local ownership), this is nearly twice as much as is minimally required to represent the full spectrum of ecosystem, community, and species diversity within the state. The problem lies in the fact that ownership patterns are highly fragmented and do not holistically represent interacting systems, as well as being overly redundant in terms of their representation of ecosystem and community types.

The benefits of our project to future conservation efforts are not so much related to our ability to document conservation gaps. More importantly, we have developed the data necessary to make informed decisions that incorporate fundamental principles of stream ecology and conservation biology into an overall conservation planning strategy. We are currently working with the Missouri Department of Conservation to use these data to develop a statewide comprehensive conservation plan for conserving riverine biodiversity. For each of the 15 Ecological Drainage Units in Missouri we are identifying a set of focus areas that collectively represent the full spectrum of abiotic and biotic diversity. These focus areas then serve as a geographic template for a variety of conservation actions, including new land acquisitions, changes in management designations, or private land conservation initiatives. Although not completed for the entire state, present results suggest that an idealized network of reserves, which focuses on local ownership of critical stream segments, would require approximately 5,000 to 6,000 km of stream to represent the full spectrum of biological diversity in the state. Certainly, this says nothing about the difficulty of securing enough land to protect the watersheds of these critical segments, which would be a mind-boggling number (millions of acres) for a Midwestern landscape. However, it is our contention that efforts must initially focus on protecting critical stream segments in public lands or through intensive private land conservation, since even the most ambitious watershed protection measures can be circumvented by local disturbances (e.g., channelization, point-source pollution, impoundments, etc.) to the stream segments to be protected.

Ohio Aquatic GAP

Anticipated completion date: March 2005

Contact: S. Alex. Covert, Ohio GAP Coordinator

U.S. Geological Survey, Columbus, Ohio

sacovert@usgs.gov, (614) 430-7752

Animal modeling: The crayfish database was completed in 2003. An expert review of distributions of 92 freshwater mussel and clam species and 20 crayfish species was also completed. An effort to model potential species distributions for freshwater mussels, clams, and crayfish was started in 2003 using the same modeling approach as was used for fish in Ohio, i.e., GARP. An analysis that integrates all aquatic biota will be completed in 2004.

Land stewardship mapping: All digital maps of Ohio conservation lands were obtained and compiled into one map. Each land parcel was attributed with a GAP land-status code. The map was reviewed and finalized in 2003.

Analysis: Potential fish distributions for 148 species were displayed at the USGS 14-digit Hydrologic Unit (HUC) for each of three stream-size classes. The number of unique fish species predicted for each 14-digit HUC (1,790 in Ohio) was calculated and compared to watersheds in the larger, 8-digit HUC (44 in Ohio) that the 14-digit HUC occupied. This effort helped to identify watersheds with high numbers of predicted fish species, as well as to ensure a distributed assessment unit throughout the state. The number of predicted fish species, areas predicted for Ohio’s endangered species, and areas predicted to have high-quality, cold-water fish assemblages will be used to perform a gap analysis in 2004. Areas that have never been sampled were identified to help guide future sampling efforts. Likewise, areas identified as having high species numbers despite having been sampled abundantly with no captures of particular species were identified for further analysis of limiting conditions.

Reporting and data distribution: A data-CD entitled “Fish Distribution and Valley Segment Type Data from Ohio Aquatic Gap Analysis Project (GAP),” was published in 2003. This CD contains Ohio fish sampling data and derived valley segment data that was used to model potential fish species distributions. Ohio Aquatic GAP has presented methods and progress to the Upper Midwest Environmental Science Center in Onalaska, Wisconsin, at the joint annual International Association for Great Lakes Research (IAGLR) and International Lake Environment Committee (ILEC) meeting, at the National GAP meeting in Fort Collins, Colorado, to the Ohio EPA, and at two stakeholder meetings.

Southeast Aquatic GAP

a. Alabama

Anticipated completion date: August 2004

Contact: Elise Irwin

Alabama CFWRU, Auburn University, Auburn

eirwin@acesag.auburn.edu, (334) 844-9190

We are in the final stages of project completion for the entire Tallapoosa River basin. We have recently completed QA/QC of newly delineated 12-digit HUCs, which is our mapping and modeling unit. All landscape-level GIS layers are complete, including geology, hydrology (1:100,000 NHD), LU/LC (Anderson Level I classification; 1992 NLCD) and a small impoundment layer (from DOQs). Watershed-level characters are also complete for each 12-digit HUC. These include but are not limited to drainage density, road density, and physiographic province. Stream reach characters have also been quantified, including site elevation, stream gradient and aspect, link magnitude, and downstream link magnitude. Fauna data include fish collections from over 300 sites in the basin. Crayfish, snail, and mussel data are sparse but available from over 100 sites. We are estimating detection probabilities for species using the program CAPTURE. Detection probabilities will be used as model weights in K-nearest neighbor (KNN) models to account for incomplete detection of species. We concluded that KNN classification was more appropriate than other techniques to determine significant predictors for species occurrence. These empirical models will be valuable for development of decision support systems.

b. Georgia

Completed. Draft data available from state contact.

Contact: James Peterson

Georgia CFWRU, University of Georgia , Athens

peterson@smokey.forestry.uga.edu, (706) 542-6032

Upper Missouri River Basin Aquatic GAP

Anticipated completion date: October 2004

Contacts: Jonathan A. Jenks and Charles R. Berry, Jr.

South Dakota State University, Brookings

jonathan_jenks@sdstate.edu, (605) 688-4783

charles_berry@sdstate.edu, (605) 688-6121

Status: We have acquired all necessary data sets from state, federal, and international agencies for physical habitat, fish distribution, and stewardship. We have completed attributing valley segments with ten physical habitat affinities, which include temperature, stream size, flow regime, channel gradient, size discrepancy, floodplain interaction, surficial geology, elevation, stream connectivity, and groundwater input. We are now performing a quality check of valley segment habitat attributes and are about 75% finished. Fish distribution data is also undergoing quality control. We have edge-matched and merged land cover data from states and provinces and are in the process of merging stewardship layers. We have completed delineation of watersheds similar to 10-digit hydrological units for portions of North Dakota within our study area and are combining these with 10-digit hydrological units provided by the North Dakota Department of Health, Division of Water Quality. We have completed field sampling of fish within the Frenchman River watershed, which flows from Saskatchewan, Canada, into Montana, and also sampled the Sweet Grass Creek watershed within Montana. Field data is being summarized and will be used to test the accuracy of fish distribution models for these areas. We also have collected invertebrate data from sampled stream reaches in these watersheds and will use this information to model invertebrate distributions and biodiversity.

Future plans: We will continue to quality-check our valley segment and fish distribution databases. We plan to use decision tree analysis (Answer Tree, SPSS) to produce fish-habitat models similar to methods used by the Lower Missouri River Basin Aquatic GAP Project. Fish will be modeled for seven regions based upon ecoregions and major drainages. Accuracies of fish-habitat models will be evaluated using field data and also statistical methods (e.g., boot strapping). Fisheries experts from each region will also review models and fish distribution maps.

Upper Tennessee River Basin Aquatic GAP

Anticipated completion date: May 2004

Contacts: Paul Angermeier

Virginia CFWRU, Virginia Tech, Blacksburg

biota@vt.edu, (540) 231-4501

Jeff Waldon

Conservation Management Institute, Virginia Tech, Blacksburg

fwiexchg@vt.edu, (540) 231-7348

In 2001, researchers from the Department of Fisheries and Wildlife Sciences and the Conservation Management Institute of Virginia Tech began an aquatic gap analysis of the upper Tennessee River basin (UTRB), which is shared by Virginia, Tennessee, North Carolina, and Georgia. In 2003 we assembled available GIS coverages on (a) biota (mostly fishes but some data on mussels and crayfishes), (b) land and water use (e.g., dams, roads, effluents, urban areas, row crops, pastures), and (c) physical landscape features (e.g., physiography, elevation, hydrography). We are using these coverages to develop models that (a) predict species occurrence and (b) estimate threat to watershed health.

We are building two types of models to predict species distribution. The type built for a particular species depends on data availability. We identified 524 assemblage samples containing a total of 126 fish species, 71 samples containing a total of 11 crayfish species, and 66 samples containing a total of 5 mollusk species. For species occurring in many samples we are building a suite of logistic regression models, then choosing the best model. Useful predictor variables differ considerably among species. Elevation (73%), state (67%), and stream order (61%) are most frequently useful for predicting fishes. Elevation (81%) was the most frequently useful predictor of crayfish occurrence. Stream order (60%), elevation (60%), and sinuosity (60%) were most frequently useful in predicting mussel occurrence. We are also developing less precise (descriptive) models based on species accounts in recent books on the region’s fish fauna. These models largely reflect known distributions as described by drainage, physiography, and stream size. For poorly sampled species these descriptive models are the only model type available. For well-sampled species the logistic regression models can be used to calibrate the reliability of the descriptive models.

A main research focus is to develop more powerful protocols to assess threats to aquatic biota. We anticipate that the standard stewardship data layer used in gap analyses, which is based on land ownership, will not provide an informative assessment of protective status. Threats to biota vary in scope of origin (nonpoint vs. point source), frequency of occurrence (accidental spill vs. permitted effluent), and severity (heavy metal contamination vs. nutrient enrichment). Moreover, most threats to aquatic biota emanate from outside the aquatic environment. Thus, we have developed an integrative protocol that assesses a wide array of threats to stream biota and converts degree and extent of threat into a numerical form that facilitates ranking among watersheds. We have assembled georeferenced data layers on dams, roads, railroads, pipelines, waste disposal sites, permitted water discharges, agriculture, urban areas, and industrial sites. We ranked each human activity based on its potential impact on flow regime, water quality, habitat quality, energy sources, and biological interactions. Then we estimated the frequency of each activity within individual catchments. Finally, an index based on impact and frequency was computed for each catchment. This protocol enabled us to develop maps of severity for individual threats as well as for cumulative threat, and to identify large-scale patterns of protective status across the entire UTRB. We are in the final stages of map development and are preparing to examine how protective status based on stewardship data compares to protective status based on our new protocol.