Estimating erosion risk on forest lands using improved methods of discriminant analysis

Abstract

A population of 638 timber harvest areas in northwestern California was sampled for data related to the occurrence of critical amounts of erosion (>153 m3 within 0.81 ha). Separate analyses were done for forest roads and logged areas. Linear discriminant functions were computed in each analysis to contrast site conditions at critical plots with randomly selected controls. Bootstrapping was used extensively in the development and testing of the equations, in estimating prediction bias, and in placing confidence limits around parameters and posterior probabilities. The resulting three-variable equations had classification accuracy, corrected for prediction bias, of 77.7% for road plots and 69.2% for logged area plots. The use of linear discriminant functions facilitates the explicit consideration of erosion risk when planning land-disturbing activities.

Citation

Lewis, J.; Rice, R. M. 1990. Estimating erosion risk on forest lands using improved methods of discriminant analysis. Water Resources Research 26(8): 1721-1733.