Progress in adapting k-NN methods for forest mapping and estimation using the new annual Forest Inventory and Analysis data

The k-nearest neighbor (k-NN) method has been undergoing development and testing for applications with USDA Forest Service Forest Inventory and Analysis (FIA) data in Minnesota since 1997. Research began using the 1987-1990 FIA inventory of the state, the then standard 10-point cluster plots, and Landsat TM imagery. In the past year, research has moved to examine potentials for improving cover type and volume mapping and estimation with the new annual FIA data, notably the new four-subplot cluster plot, and Landsat ETM+. Major findings to date point to the difficulty of choosing the number of neighbors (k). A value of k between 1 and 3 seems appropriate for mapping. A larger number of neighbors reduces the overall estimation error, but it also leads to a reduction in the producer's accuracy. Additionally, using multiple image dates for an area typically improves results considerably. Recent results with the new four-subplot cluster plot data show that stratification of the data into upland/lowland strata, use of thermal bands, and a plot location optimization all improve mapping and estimation results. Finally, segmentation algorithms show potential for improving mapping and the k-NN estimation process. A C-language program package for applying the k-NN method to forest inventory has also been developed.