An Accuracy Comparison of Six Spatial Interpolation Methods for Modelling Forest Stand Structure on the Fraser Experimental Forest, Colorado

摘要

This study compared four geostatistical methods of interpolation (ordinary kriging, universal kriging, cokriging, and disjunctive kriging) with two traditional estimation methods (polygon mapping, and inverse distance weighting). The six techniques were used to spatially interpolate the number of stems, basal area, and number of seedlings on 82 plots in a 121-hectare first-order forest watershed at the USDA Forest Service, Fraser Experimental Forest, Fraser, Colorado, USA. Secondary variables used for cokriging included elevation, a combined value for slope and aspect, and the normalized difference vegetation index (NDVI) from Landsat-TM satellite imagery. The comparison criterion was the mean square error (MSE) calculated by cross validation. The performance of the estimation techniques was different between variables. Overall, however, cokriging performed the best, followed by polygonal mapping. Universal kriging with a first or second degree trend surface yielded, in general, better results than ordinary kriging. Inverse distance weighting was generally outperformed by the linear kriging methods. The nonlinear kriging method (disjunctive kriging) performed least well. These results indicate that the additional accuracy from spatially cross-correlated variables substantially improves the estimation capability of cokriging, as compared to the other methods for these data.