Download (PDF 459 KB): https://research.fs.usda.gov/download/treesearch/4524.pdf
Abstract
A reduction in variance for estimates of forest area and volume in the state of Connecticut was accomplished by stratifying FIA ground plots using raw, transformed and classified Landsat Thematic Mapper (TM) imagery. A US Geological Survey (USGS) Multi-Resolution Landscape Characterization (MRLC) vegetation cover map for Connecticut was used to produce a forest/non-forest map derived from a classified 1993 TM image. A 1996 TM image was used to provide spectral reflectance variables for each pixel, including the values for all 6 raw TM bands and several transformed layers: normalized difference vegetation index (NDVI) and Tasseled Cap brightness, greenness, wetness, and "fourth." Each pixel in the map was assigned a value indicating how many surrounding pixels within a 3X3 or 5X5 window were forested. These same windows were used to calculate, for each pixel, a mean, maximum, minimum, and standard deviation of the raw and transformed layers. FIA ground plots (1996) were split into percent timberland classes using a "decision tree" algorithm that recursively determines the most significant variable and the most significant split of that variable. The final set of grouping criteria was used to statistically stratify a set of FIA ground plots. Results were compared with aerial photo based stratification as well as TM derived forest/non-forest stratification.
Citation
Hoppus, Michael; Arner, Stan; Lister, Andrew. 2001. Stratifying FIA Ground Plots Using A 3-Year Old MRLC Forest Cover Map and Current TM Derived Variables Selected By "Decision Tree" Classification. In: Reams, Gregory A.; McRoberts, Ronald E.; Van Deusen, Paul C., eds. 2001. Proceedings of the second annual Forest Inventory and Analysis symposium; 2000 October 17-18; Salt Lake City, UT. Gen. Tech. Rep. SRS-47. Asheville, NC: U.S. Department of Agriculture, Forest Service, Southern Research Station. pp. 19-24
