Spatial interpolation of forest conditions using co-conditional geostatistical simulation

Abstract

In recent work the author used the geostatistical Monte Carlo technique of sequential Gaussian simulation (s.G.s.) to investigate uncertainty in a GIS analysis of potential old-growth forest areas. The current study compares this earlier technique to that of co-conditional simulation, wherein the spatial cross-correlations between variables are included. As in the earlier study, uncertainties were assessed across 500 independent spatial Monte Carlo realizations for each of three variables of interest (quadratic mean stand diameter; age of dominant and co-dominant trees, and percent canopy cover). Potential old-growth for the study area was estimated for each set of these perturbed realizations using a simple GIS analysis. An uncertainty histogram was created by adding the 500 realizations on a cell-by-cell basis. Results were compared using empirical confidence legions from the upper percentiles of the histogram for each study. For uncertainty assessment using these particular co-located spatial data, co-conditional simulation creates intuitively less desirable results, and does not appear to provide any advantage over independent realizations using s.G.s.. For other data sets, where all variables are not measured at each location, improvements may result.

Parent Publication

• Integrated tools for natural resources inventories in the 21st century

Citation

Mowrer, H. Todd. 2000. Spatial interpolation of forest conditions using co-conditional geostatistical simulation. In: Hansen, Mark; Burk, Tom, eds. Integrated tools for natural resources inventories in the 21st century. Gen. Tech. Rep. NC-212. St. Paul, MN: U.S. Dept. of Agriculture, Forest Service, North Central Forest Experiment Station. 214-220.