PARENT SESSION
Oral Session 7: GIS / Remote Sensing I.
Presiding: E Ellis and P Valko
Monday, August 2, 8:00 AM to 11:30 AM, Meeting Room B 116.
National vegetation mapping for fire management and other environmental applications.
Zhu, Zhiliang*,1, Vogelmann, James2, Rollins, Matthew3, 1 U.S. Geological Survey, Sioux Falls, SD, USA2 SAIC/EROS Data Center, Sioux Falls, SD, USA3 USDA Forest Service, Missoula, MT, USA
ABSTRACT- The 1999 1-km historical natural fire regime and fire regime condition class maps, developed by the Forest Service using baseline data produced by the USGS scientists for general applications, have been widely used for national fire management planning purposes. However, the use and misuse of the data since 2000 have also demonstrated the need for a dedicated vegetation mapping effort as a component of the LANDFIRE project at a mapping scale suitable for supporting multi-scale management applications from watershed management to national policy implementation. LANDFIRE requirements for the vegetation mapping methodology include consistency, repeatability, accuracy, and capability to map detailed (in terms of spatial and information depth) vegetation types and structure variables at 30-meter resolution. Such an effort, for national wall-to-wall coverage, is unprecedented. In this paper, we discuss a strategy for achieving the LANDFIRE vegetation mapping objectives within a five-year repeat cycle. The methodology is based on four essential design features: 1) reliance on Landsat data acquisition and processing by the USGS land cover program, 2) access to a large quantity of field reference data, 3) applications of environmental gradient layers and mapped potential vegetation types in the mapping process, and 4) the use of nontraditional and flexible classifiers that integrate field plot data with a large volume of predictor variable layers. Results show that the vegetation types and structure characteristics can be developed as a consistent, practical, mid-scale national classification suitable for the Federal Geographic Data Committee applications. Further, the target accuracy of 60% or higher nationwide can be attained.
Key words: structure mapping, gradient modeling, vegetation classification, fire ecology