A SPATIAL MODELING AND DECISION SUPPORT SYSTEM FOR CONSERVATION OF BIOLOGICAL DIVERSITY
Monitoring Environmental Change
Nearly all of our IBM-ERP research effort was devoted to mapping and modeling biological patterns for a single time period. However, we recognized from the outset the need to place this work within the context of multi-temporal monitoring of environmental change, and made some progress in this area, specifically in acquisition and processing of multitemporal satellite imagery and in the use of map-guided image classification for change detection and land use/land cover monitoring. We have already described the map-guided image classification approach, which in initial tests has shown promise for a change detection algorithm in which the maps from earlier time periods are used to classify land use/land cover in the current time period, thereby ensuring definitional consistency across time. Here we simply summarize our efforts in the area of image acquisition and processing.
UCSB's Institute for Computational Earth System Science (ICESS) maintains a Terascan receiving station that can download NOAA AVHRR satellite imagery two times a day (night, afternoon). As a test of AVHRR-based monitoring of land use/land cover and wildlife habitats in California, we devised a semi-automated approach for daily downloading, rectification, archiving, and web-based re-distribution of afternoon images. These images are subsequently processed to develop cloud-free composites and applied to map-guided classification and feature extraction, as described in previous sections. For each image and each of 5 spectral channels, data pertaining to the western U.S. are downloaded and re-formatted into a standard image format using commercial Terascan software. They are then mapped into a standard coordinate system using a set of ground control points that are recognizable features in the image. Each scene is stored as a gzipped tar file containing 5 image bands and 3 bands of information on satellite position and look angle. Each tar file also contains files with other scene-specific information. We have made these images publicly available with a Web-based ftp interface (http://www.biogeog.ucsb.edu/projects/ibm/ibm_data.html).