Biophysical aggregations of a forested landscape using an ecological diagnostic system

We demonstrate a methodology for executing ecosystem models on large regional data sets by organizing and reducing their size prior to executing the model. A knowledge-based (KB) classification method was constructed to aggregate large raster data layers into fewer biophysical landtypes as defined by a climate-soil-vegetation equilibrium. Statistical tests compared simulated seasonal waters stress from raster data simulations to their KB landtype and verified that the KB classifications are ordered along a seasonal water stress gradient (p=0.95). The KB method produced a more distinct water stress classification than an alternative GIS overlay classification. Internal concepts modelled by the KB, such as snowpack persistence, were compared to simulated snowpack depletion dates for a range of sites. The KB exhibited the same sensitivity in direction and magnitude of variation, as simulated snowpack depletion dates from the process model FOREST-BGC. The KB method was 1000 times faster than the optimized versions of the physically based models, justifying the use of a KB as an efficient pre-processor to reduce a large database prior to ecosystem simulations.