Existing Vegetation

DEVELOPMENT OF THE AVHRR FOREST TYPE MAP

The AVHRR (Advanced Very High Resolution Radiometer) forest type map was produced by the Forest Service FIA (Forest Inventory and Analysis) research unit of the Southeastern Forest Experiment Station during 1991 and 1992, and was published as part of the 1993 RPA (Resources Planning Act) report (Powell et al. 1993).

AVHRR data were collected from polar-orbiting satellites operated by NOAA (National Oceanic and Atmospheric Administration). These satellites image nearly all of Earth twice daily, allowing observation of information regarding changes in land cover conditions during a growing season. The AVHRR forest type mapping project utilized selected composite images previously prepared by the U.S. Geological Survey EROS Data Center (EDC) at Sioux Falls, South Dakota. EDC prepared these images as part of their mission to develop a land-cover characteristics database for the U.S., with intended applications in a variety of mapping, research, and monitoring activities (Loveland et al. 1991). The composite images display selected pixels from two weeks of satellite data. Pixels were selected based on their normalized difference vegetation index (NDVI), a measure of the relative greenness of vegetation. Increasingly green vegetation results in higher positive numbers in the index, while negative values are indicators of non-vegetated areas and were used to separate vegetated and non-vegetated areas of the U.S. Composites were made up of pixels with the maximum NVDI value from the two-week period, so that the images displayed the maximum "greenness", an indicator of the level of photosynthetic activity, and also contained fewer cloud-obscured pixels. In addition to NDVI, the data set contains values from the five AVHRR channels that measure reflectance or emittance of electromagnetic energy at different thermal wavelengths, and also information about viewing angles and dates of observation, for each pixel (Eidenshink 1992).

SO-FIA used a test area of seven States comprising the "midsouth" portion of the U.S. to develop the forest type classification procedure (Zhu and Evans 1992). The area was stratified into six physiographic regions based on Hammond's land-surface form map, and unsupervised computer classifications of pixels were conducted using the AVHRR spectral channel data. The classes were identified by comparison with FIA plot data, and other existing maps and photographs.

For the U.S forest type mapping project, SO-FIA utilized nine of EDC's composite images, representing 18 weeks of maximum NVDI during 1990's growing season (early March to mid-November). The U.S. was divided into 15 physiographic regions, and the classification procedure developed in the test "midsouth" area was applied in several iterations. Identification of forest types was based on reference data; for eastern forests, the researchers found that FIA field survey plot data were a key reference. Verification utilized comparisons with photointerpreted data from Louisiana, and FIA state forest area estimates (Zhu and Evans 1994).

Notes about this forest type map:

- The forest type classifications are limited by pixel resolution, which is approximately one square kilometer (about 255 acres). A classification at this resolution will always produce slightly different results than a finer-resolution classification or field-measured data.

- The map does not show any forest type for those pixels with a percent cover of 25% or less; they were identified in a separate density mapping effort described by Zhu (1994).

- The physiographic region that SO-FIA used for stratifying the Lake States' forest type classification is similar to the analysis area used for the Great Lakes Assessment: Province 212 in the Lake States.

- The map data was obtained via Internet from http://www.epa.gov/docs/grd/forest_inventory in July, 1995. The map was clipped, reprojected, and analyzed for the Great Lakes Assessment, USDA-Forest Service, North Central Forest Experiment Station, in cooperation with the Land Information and Computer Graphics Facility, University of Wisconsin-Madison.

REFERENCES

Eidenshink, J.C. 1992. The 1990 conterminous U.S. AVHRR data set. Photogrammetric Engineering and Remote Sensing 58(6):809-813.

Loveland, T.R.; Merchant, J.W.; Ohlen, D.O.; Brown, J.F. 1991. Development of a land-cover characteristics database for the conterminous U.S. Photogrammetric Eng. & Remote Sensing 57(11):1453-1463.

Powell, D.S.; Faulkner, J.L.; Darr, D.R.; Zhu, Z.; MacCleery, D.W. 1993. Forest resources of the United States, 1992. General Technical Report RM-234. Fort Collins, CO; U.S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. 132 p. + map. (Revised, June 1994).

Zhu, Z. 1994. Forest density mapping in the lower 48 States: a regression procedure. Res. paper SO-280. New Orleans, LA: U.S. Dept. Agric, Forest Service, Southern Forest Expt. Station. 11 p.

Zhu, Z,.; Evans, D.L. 1992. Mapping midsouth forest distributions. Jour. Forestry 90(12):27-30.

Zhu, Z.; Evans, D.L. 1994. U.S. forest types and predicted percent forest cover from AVHRR data. Photogrammetric Eng. & Remote Sensing 60(5):525-531.

Return to Existing Vegetation Theme