The Pacific Northwest Regional Gap Analysis Project Final Report on Land Cover Mapping Methods: Map Zones 8 and 9 1 May, 2006 Institute for Natural Resources Oregon State University, Corvallis, OR, USA USDA-PNW Forest Sciences Laboratory, Corvallis, OR, USA This report represents the land cover portion of the final project report for Map Zones 8 and 9 of the Pacific Northwest Regional Gap Analysis Project i Final Report on Land Cover Mapping Methods: Map Zones 8 and 9, Pacific Northwest ReGAP Authors: James S. Kagan.a* Janet A. Ohmannb* Mathew J. Gregoryc Claudine Tobalsked John C. Hakd Jeremy Friede a Institute for Natural Resources, Oregon State University, Corvallis, OR, USA b Pacific Northwest Research Station, USDA Forest Service, Corvallis, Oregon c Forest Science Department, Oregon State University, Corvallis, Oregon d Oregon Natural Heritage Information Center, Oregon State University, Portland, Oregon e Pacific Northwest Research Station, USDA Forest Service, Portland, Oregon Recommended Citation: Kagan, J.S., J.A. Ohmann, M.J. Gregory, C. Tobalske, J.C. Hak, and J. Fried. 2006. Final Report on Land Cover Mapping Methods, Map Zones 8 and 9, PNW ReGAP. Institute for Natural Resources, Oregon State University, Corvallis, OR. Acknowledgements Much of this report is taken directly from the Southwest Regional Gap Analysis final report, by J.H. Lowry et al. (2006). The authors greatly appreciate their willingness to share their work. The work here was based on previous projects developed with other key individuals. Steve Knick and Steve Hanser of the USGS Snake River Field Station in Boise, both managed and facilitated the SageMap project, which provided the foundation of most of the vegetation data incorporated. SageMap relied heavily on the 15 year ESI project coordinated by the Burns District of the BLM, vegetation mapping by all the eastern Oregon and Washington National Forests, and shrub steppe mapping done by the Washington Department of Fish and Wildlife. Finally, Rex Crawford of the Washington Natural Heritage Program had a critical part in assuring the Washington portion of the map was not terrible. We gratefully acknowledge the financial support of the USGS BRD, Gap Analysis Program, without which completion of this project could not have been possible, and in particular, would also like to thank Jocelyn Aycrigg and Todd Sajwaj for their patience and support. ii Abstract For more than a decade the USGS Gap Analysis Program has focused considerable effort on mapping land cover to assist in the modeling of wildlife habitat and biodiversity for large geographic areas. The GAP Analysis Program has been traditionally state-centered; each state having the responsibility of implementing a project design for the geographic area within their state boundaries. The Northwest Regional Gap Analysis Project (NW ReGAP) is the third formal GAP project designed at a regional, multi-state scale, building off the work developed by the Southwest Regional Gap Analysis Project (SW ReGAP). A land cover map was generated for USGS Map Zones 8 and 9, covering most of Eastern Washington and Eastern Oregon, parts of western Idaho, and most of northern Nevada. The map was derived from two primary components. The first was a combination of two large regional datasets: SageMap covering eastern Oregon and Washington, and southern Idaho and SW ReGAP, covering the northern Nevada portion of the Map Zone 9. These used regionally consistent geospatial data (Landsat ETM+ imagery and DEM derivatives), similar field data collection protocols, a standardized land cover legend, and a common modeling approach (decision tree classifier). The second was a Gradient Nearest Neighbor (GNN) modeling effort developed for the forests, based on the network of forest vegetation plots in the region. This report presents an overview of the process and methodologies used to create the land cover dataset and results and lessons learned from the different methodologies used. iii Table of Contents Introduction..................................................................................................................................... 1 Land Cover Map Development....................................................................................................... 2 Land Cover Mapping and GNN Modeling Results ...................................................................... 28 Results........................................................................................................................................... 34 Land Cover Map Validation ......................................................................................................... 38 Discussion..................................................................................................................................... 42 Literature Cited ............................................................................................................................. 50 Tables Table 1. Hierarchical structure of the U.S. NVC with ecological systems..................................... 7 Table 2. Number of forest condition plots used in GNN modeling in map zones 8 and 9........... 20 Table 3. Number of forest condition plots by Ecological Systems and GNN modeling regions.21 Table 4. Environmental and spectral variables used in GNN gradient models of forest composition................................................................................................................................... 22 Table 5. Area in hectares of Systems for Map Zone 8 sorted by Abundance.............................. 34 Table 6. Area in hectares of Systems for Map Zone 9 sorted by Abundance.............................. 36 Table 7. Error matrix for forest Ecological Systems* in map zones 8 and 9. ............................. 40 Table 8. Kappa statistics for forest Ecological Systems* in GNN Mapping Region 1 ............... 41 Table 9. Accuracy Assessement and Sample Size by Region for SageMap and PNW ReGAP . 42 Figures Figure 1. Mapping zone boundaries for SWReGAP land cover mapping effort........................... 3 Figure 2. Mapping zone boundaries for SageMap and NWGAP land cover mapping effort........ 4 Figure 3. SageMap Plot Samples: Collected: 5,000, Existing: 16,000, Total: 21,000. ................. 6 Figure 4. SWReGAP area showing Landsat ETM+ scenes........................................................... 9 Figure 5. NW Re GAP and SageMap area showing Landsat ETM+ scenes ............................... 11 Figure 6. SageMap area showing Landsat ETM+ scenes............................................................ 12 Figure 7. Overview of the NW ReGAP Mapping Process .......................................................... 18 Figure 8. Gradient Nearest Neighbor (GNN) modeling regions in map zones 8 and 9............... 25 Figure 9. Box Plot of 6 classes of overall shrub cover ............................................................... 31 Appendices Appendix A. Ecological Systems of Map Zones 8 and 9 Appendix B. Description of Core Set of Vegetation Variables for Forests Imputation Appendix C. Key to Classification of Plots to Ecological Systems Appendix D. Summary of Mapping Results iv Introduction In its "coarse filter" approach to conservation biology (Jenkins 1985, Noss 1987) gap analysis relies on maps of dominant land cover as the most fundamental spatial component of the analysis for terrestrial environments (Scott et al. 1993). For the purposes of GAP, most of the land cover of interest can be characterized as natural or semi-natural vegetation defined by the dominant plant species. Vegetation patterns are an integrated reflection of physical and chemical factors that shape the environment of a given land area (Whittaker 1965). Often vegetation patterns are determinants for overall biological diversity patterns (Franklin 1993, Levin 1981, Noss 1990) which can be used to delineate habitat types in conservation evaluations (Specht 1975, Austin 1991). As such, dominant vegetation types need to be recognized over their entire range of distribution (Bourgeron et al. 1994) for beta-scale analysis (sensu Whittaker 1960, 1977). Various methods may be used to map vegetation patterns on the landscape, the appropriate method depending on the scale and scope of the project. Projects focusing on smaller regions, such as national parks, may rely on aerial photo interpretation (USGS-NPS 1994). Mapping vegetation over larger regions has commonly been done using digital imagery obtained from satellites, and may be referred to as land cover mapping (Lins and Kleckner 1996). Generally, land cover mapping is done by segmenting the landscape into areas of relative homogeneity that correspond to land cover classes from an adopted or developed land cover legend. Technical methods to partition the landscape using digital imagery-based methods vary. Unsupervised approaches involve computer-assisted delineation of homogeneity in the imagery and ancillary data, followed by the analyst assigning land cover labels to the homogenous clusters of pixels (Jensen 2005). Supervised approaches utilize representative samples of each land cover class to partition the imagery and ancillary data into clusters of pixels representing each land cover class. Supervised clustering algorithms assign membership of each pixel to a land cover class based on some rule of highest likelihood (Jensen 2005). Supervised- unsupervised hybrid approaches are common and often offer advantages over both approaches (Lillesand and