What's New
Research & Projects Publications
People
Data
Links
A SPATIAL MODELING AND DECISION SUPPORT SYSTEM FOR CONSERVATION OF BIOLOGICAL DIVERSITY
Frank W. Davis, David M. Stoms, Allan D. Hollander, Michael J. Bueno, Richard L. Church, W. J. Okin, Ross A. Gerrard
Institute for Computational Earth System Science and Department of Geography, University of California Santa Barbara, CA 93106 Phone: 805-893-3438 Final Report to IBM Environmental Research Program
Full report Report Date: September 30, 1997
EXECUTIVE SUMMARY The earth is experiencing a mass extinction of species that is unparalleled in its history. Installation of an effective reserve network to minimize future loss of biodiversity will require coordinated conservation assessments at international, national, regional and local levels. Such assessments already rely heavily on advanced mapping technologies and computing systems for spatial data analysis and display.
The goal of this project was to design and test a prototype Spatial Modeling and Decision Support System for Conservation of Biological Diversity. The project is closely tied to the USGS-Biological Resources Division's Gap Analysis Program, and to related efforts at multi-species conservation planning in southern California and the Sierra Nevada. Our project objectives and results are summarized below:
Design and enable a prototype "regional" computing facility for storage, analysis and visualization of biodiversity data.
A prototype regional computing facility was designed and implemented in the UCSB Biogeography Lab. The facility is centered around a 58H workstation file server and a 39H compute server. These servers support a network of RS6000 workstations and Xterminals, connected by a local fddi network. The network runs a suite of commercial and public-domain geographic information system, remote sensing, statistical and analytical software packages needed for regional conservation assessment and planning.
As originally conceived, the project was designed to develop entirely a new, fully integrated spatial analysis and decision support system. Emerging technology progressed so rapidly, however, that it seemed more prudent to take advantage of widely available new technology such as the World Wide Web (WWW) and spatial analysis and visualization packages and to focus our effort on developing specific applications and linkages between poorly integrated systems. We instituted a WWW site for sharing data and information via the Internet (http://www.biogeog.ucsb.edu/). We also took advantage of WWW tools to reconstruct our cataloging tool interface using HTML and used PERL CGI scripts to access the database. In addition to offering users a now-familiar WWW browser interface, the new version provides more display functions and automatic filling of fields whose values can be obtained from file description and header information. We now run our own httpd server and are providing many of our datasets via interactive text and graphical interfaces.
New applications for plant community and habitat type classification, image compositing for cloud removal, image classification, species distribution modeling, and visualization were programmed to support the Gap Analysis Program. For example, a new technique was developed for compositing daily AVHRR satellite images to obtain cloud-free coverage of large regions. A "map-guided classification" technique was developed to use these composited images, or other multispectral data, to generate a thematic land-cover map from existing maps from either another time period or from a set of maps of varying resolution. Two programs were developed with graphical user interfaces to integrate a non-spatial database of wildlife habitat preferences to a spatial database of the distribution of habitats to predict the distribution of the wildlife species. Also, a custom interface was written using the ARCVIEW scripting language (AVENUE) to improve the accessibility of the complex California GAP database for visualization and query. This interface and the database will be published on CD-ROM, linking the spatial data with the textual and graphical report and analysis.
GAP databases for the individual states manifested obvious differences in spatial resolution and pattern caused by use of different mapping techniques and classification systems. This raised concerns about whether these databases could be consolidated into a consistent product for regionwide (i.e., multi-state) gap analysis. We used our image compositing strategy to develop time-series images for the Intermountain Semi-Desert Ecoregion (portions of Washington, Oregon, California, Idaho, Nevada, Utah, Montana, Wyoming, and Colorado) to provide a data set with consistent spatial, temporal, and spectral properties over one entire growing season. The map-guided classification technique was then used to integrate the original statewide GAP land-cover maps into a more consistent regional map of standardized cover classes. The nation's first regional gap analysis was conducted using this new version of the land-cover map.
Our research in environmental monitoring was directed towards developing an operational system for retrieval and storage of NOAA-AVHRR data from a local receiving station, and on the comparisons of different compositing strategies for cloud-free imagery described above.
We demonstrated that the reserve selection problem described in the conservation biology literature can be reformulated as a classic maximal covering location problem (MCLP) described in the operations research and regional science literature twenty years ago. We solved the MCLP model for a real application using vertebrate distribution data prepared for the gap analysis of southwestern California. Using IBM's Optimization Subroutine Library on a RS6000 workstation, the MCLP could be solved in seconds. Previous researchers had used non-optimal heuristics to solve the problem and found it took hours on a supercomputer to find optimal solutions using naive exhaustive enumeration methods. We then reformulated the MCLP approach with three significant refinements, including weighting species (to emphasize protection for rare species in a multi-objective problem), weighting sites (based on their suitability for biodiversity management), and integration of the model into a commercial geographic information system.
In support of the USDA Forest Service Sierra Nevada Ecosystem Project, we developed a related reserve selection tool to explore alternative strategies for locating core biodiversity management areas in that region to represent native biodiversity. The new Biodiversity Management Area Selection (BMAS) model requires that specific acreage targets be specified for each biodiversity element (rather than simple representation or covering in the reserve solution). Second, suitability of the sites for biodiversity management was incorporated into the objective function of the model. Third, management class definitions were refined by incorporating data on grazing and timber management from existing land use plans. The BMAS model has also been used to explore alternative strategies for The Nature Conservancy to conserve plant communities and rare species in the Columbia Plateau portion of the Intermountain Semi-Desert.
At a more local scale, a GIS-based model was developed for designing core reserves and linking corridors using a combination of spatial analysis techniques. This model was used to help design a reserve network in the imperiled coastal sage scrub habitat in western San Diego County in southern California.
These and other accomplishments are described in more detail in the main body of this report and in the appendices. Our specific objectives and activities evolved over the course of the four-year project in response to rapid changes in technology, unexpected opportunities for expanding the scope of our scientific research and collaboration, and changes in our approaches to specific data and analytical problems. Nevertheless, the entire enterprise began with, and continues to be built upon, high-performance IBM technology. Virtually every significant phase of the project has depended critically upon IBM hardware and software to manage, analyze, and visualize very large and complex sets of geospatial data. In summary, the project has been successful and gone well beyond our original expectations in developing new tools and solutions for conserving biological diversity, in integrating those tools to support conservation planning in several regions of the western U.S., and in demonstrating the potential of modern computing technologies in addressing and mitigating human-caused losses of biodiversity.
IBM-ERP Project Home Page