Prediction of soil properties using a process-based forest growth model to match satellite-derived estimates of leaf area index

Without better estimates of soil properties than are currently available from coarse-scale maps, it is difficult to predict forest productivity accurately across broad regions. While soil properties are not directly available from remote sensing data, they can potentially be inferred from linking these properties to observations that are readily available from satellite data, such as vegetation leaf area. We took advantage of the direct link that exists between above-ground productivity and maximum leaf area index (LAImax) to derive and map soil fertility (FR) and available soil water storage capacity (ASWC) at 1 km resolution across forested areas in western North America. Initially, we generated estimates of LAImax with a process-based growth model (3-PG), holding soil properties constant (FR = 50% of maximum, ASWC = 200 mm). To derive more realistic estimates of soil properties we inverted the model to infer FR and ASWC from iterative non-linear adjustments of the two soil properties so that model-predicted LAImax values corresponded closely with MODIS-derived observations. We parameterized 3-PG for the most widely distributed tree species in the region, Douglas-fir. The resulting maps were notably more detailed than those derived from the globally available Harmonized World Soil Database. Among 51, level III ecoregions, and the ranges in the two soil properties tended to increase in parallel with LAImax. Further improvements in the approach are envisioned by combining MODIS and LiDAR observations to extend the range and accuracy of LAImax observations.