Propagating uncertainty through spatial estimation processes for old-growth subalpine forests using sequential Gaussian simulation in GIS

Abstract Based on data from 83 plot locations, the geostatistical Monte Carlo technique of sequential Gaussian simulation (s.G.s.) was used to generate 1000 independent spatially continuous representations of three variables. These were then used in a geographic information system analysis to create maps of relative uncertainty for estimated areas of potential old-growth forest conditions across a 121 hectare first-order subalpine watershed. First, identical selection criteria were applied to each of the 1000 three-layer input sets to determine areas that simultaneously satisfied three old-growth forest conditions for mean stem diameter, percent crown cover, and mean age of overstory stems. This created 1000 equally probable realizations of potential old growth for the study area. An uncertainty image for the potential old-growth forest areas was created by summing these realizations. Cells were selected from the image histogram that indicated the highest proportions of old-growth conditions. Spatially, these results followed those obtained from a similar analysis using kriging. s.G.s. is recommended as a generic spatial Monte Carlo technique that can be used to assess stochastic elements in complex integrated ecological predictions.