Regionalizing State-Level Data
Patrick Crist and Mike
An important objective of GAP is to provide the conservation status of biotic elements, not truncated by political boundaries. The potential to analyze the status of an element throughout its entire range is one aspect that makes GAP unique and innovative. This ability will allow data users interested in the status of locally occurring elements to place them in the context of a watershed, ecoregion, national range, or ultimately continental and global range.
Although this capability has not yet been attained, the key to achieving it is the regionalization of the GAP data such that one can use any geographic unit of analysis desired. There are three basic approaches available: 1) merging the tabular results for reporting distribution and conservation status statistics, 2) remapping or remodeling the elements over the larger region using the current GAP coverages in the process, or 3) edge-matching and merging the current coverages as they are. To date, there are few examples of these approaches because only now do we have groups of contiguous states completing GAP projects. Because GAP began in the western states, this is where the approaches have been tested, though over the next few years most state data will be regionalized.
At the University of Idahos Landscape Dynamics Laboratory, the current effort to regionalize western land cover maps used the tabular merge approach (1 above)(Nancy Wright, pers. comm.). The land cover maps have been cross-walked to a common scheme and then resampled to a 1-km grid. The resulting map, not cartographically edge-matched, shows surprisingly good consistency despite the use of a variety of mapping techniques and thematic and spatial resolutions by the individual state projects. After applying the common scheme to all coverages, the tabular results of the regional gap analysisto be reported in a forthcoming publicationwill provide a first approximation conservation assessment without having to physically edge-match the state coverages. We feel this approach is a useful way to achieve the GAP objectives for these early land cover products.
The remapping approach (2 above) was used by Stoms et al. (1997) to create a seamless land cover map and conduct gap analysis of vegetation types in the Intermountain Semi-Desert Province (Bailey 1995). Here too, the first step involved cross-walking the 7-state classification schemes to the National Vegetation Classification Scheme (NVCS, Federal Geographic Data Committee, 1997). The next step was to use the original state coverages to "train" multitemporal AVHRR (NOAA Advanced Very High Resolution Radiometer) imagery to create a new land cover map of the region.
The third approach is the actual edge-matching and merging of the original state GAP data sets. This is a difficult task in the western states because: 1) when they began, there were no proven and standardized methods for creating these types of maps, therefore each one was conducted as a pilot research project; 2) there were few, if any, federal standards or protocols for digital data (the FGDCs first protocols on metadata were largely predicated on metadata work done for GAP by Cogan and Edwards ); and 3) with the exception of Nevada, all western state GAP projects were under way before a consensus was reached on using what was then known as the "TNC/UNESCO" vegetation classification system (Jennings 1993).
Through the three approaches described above, we believe sufficient regionalization can be achieved to produce useful gap analyses of biotic elements for multistate regions. However, for newer state projects, we anticipate a much higher degree of compatibility among data sets that will allow true regionalization using the original data without excessive transformation.
Regionalization of vertebrate distribution maps has received less attention to date. A preliminary workshop on Northwest regional vertebrate modeling was held in September 1996 with GAP teams from Oregon, Washington, and Idaho. As a test, ten bird and mammal species that span all three states were mapped by state according to occurrence within EMAP hexagons. Maps from the three states were then joined to determine if there was continuity of distribution at this coarse spatial level. Guidelines were established for progressing with seamless wildlife habitat relationship maps for the Northwest (Nancy Wright, pers. comm.). Generally, we believe it will be impractical to edge-match the hundreds of species coverages for each state, and instead have proposed a remodeling process in the current vertebrate modeling chapter of the GAP Handbook (1997). That process will require states to first edge-match their range extent maps, preferably using the EPA EMAP hexagons and then to rectify differences among the wildlife habitat relation models (WHRMs). After that, the computational part of the distribution modeling should take relatively short amounts of time. Some additional efforts and ideas for regionalization of both land cover and vertebrates are described below.
The Southwest ReGAP project (see the Southwest breakout session report in this bulletin): The states of NV, UT, CO, NM, and AZ were all initiated within a couple of years and all used TM imagery acquired between 1990 and 1993. The projects, however, have had completion dates ranging from 1994 to early 1998. These projects were all conducted prior to the adoption of standards that could aid regionalization such as the NVCS for land cover classification and use of the EPA EMAP hexagonal grid for use in the vertebrate modeling process. The ReGAP project, slated for initiation in 1998, calls for continuation of the state business model, but in a regional cooperative effort to ensure seamless coverages throughout the region. For land cover mapping, the goal is to allow the mapping staff to focus on fewer numbers of land cover types than occur throughout a state by mapping by TM-derived "mapping zones" (Collin Homer, pers. comm.), e.g., "montane" versus "plains grassland." We believe this approach will eliminate or greatly reduce edge-matching problems for the regional map. For animal modeling, these states will begin the cooperative process by using the EPA EMAP hexagonal grid to delineate species range extents across the region. Next, they will ensure that WHRMs are consistent across the region, or stratified by ecoregions when true habitat association differences occur across a species range. The use of these approaches and consistent, regional, ancillary data on which to base the models should ensure seamless predicted animal distributions.
The Mid America GAP Consortium (http://ulysses.unl.edu/midam/) held their first joint land cover and vertebrate modeling workshop on October 21 and 22, 1997. A primary goal was to discuss ways to achieve regionalization of their data during the production process. Though the states involved in the consortium (IA, KS, MO, ND, NE, OK, SD) initiated GAP over a wide timespan, there is less variation in mapping methods and biota than in the 11 western states, and the prospect for achieving concurrent regionalization is good. One impetus to edge-match land cover is an EPA grant that provided approximately $45,000 per state (IA, KS, MO, NE) to help these projects achieve data consistency. The regionalization effort is beginning with the creation of web pages to share land cover classifications and plans to regionalize ancillary data and methods of vertebrate modeling.
Regionalization is a difficult though critical task. While it has been an objective of GAP from the start, some have seen it as an "unfunded mandate" if left to the individual state projects. Yet we know from experience that without a concerted regional effort by groups of states, later regionalization sacrifices some of the quality of the original products and is made more difficult if conducted by remote labs not involved in the original data creation. We urge all GAP projects to aggressively pursue cooperation with their surrounding states to achieve regionalization during the mapping phase. The Great Plains states have, on their own initiative, instituted a regional cooperative approach for both land cover and vertebrate phases. We applaud their efforts and encourage others to follow. If your state project would like assistance in setting up a cooperative effort, contact Mike Jennings or Patrick Crist.
Bailey, R.G. 1995. Descriptions of the ecoregions of the United States. Misc. Publ. No. 1391 (rev.), Washington, D.C. USDA Forest Service. 108 pp.
Cogan, C., and T.C. Edwards, Jr. 1994. Metadata standards for Gap Analysis, version 1, in: A handbook for Gap Analysis. USGS Gap Analysis Program, Moscow, Idaho. http://www.gap.uidaho.edu/gap/.
Federal Geographic Data Committee, Vegetation Subcommittee. 1997. FGDC vegetation classification and information standards, June 3, 1996 draft. FGDC Secretariat, Reston, Virginia. 35 pp.
Gap Analysis Program. 1997. A handbook for Gap Analysis. USGS Gap Analysis Program, Moscow, Idaho. http://www.gap.uidaho.edu/gap/.
Jennings, M.D. 1993. Natural terrestrial cover classification: Assumptions and definitions. Gap Analysis Technical Bulletin No. 2. National Gap Analysis Program, Moscow, Idaho.
Stoms, D.M., F.W. Davis, K.L. Driese, K.M. Cassidy, and M. Murray (in prep). Gap Analysis of the vegetation of the Intermountain Semi-Desert ecoregion.