Remote monitoring of spatial and temporal surface soil moisture in fire disturbed boreal forest ecosystems with ERS SAR imagery

摘要

Due to the large volume of carbon currently stored in boreal regions and the high frequency of wildfire, the prospects of a warming climate would have important implications for the ecology of boreal forests which in turn would have significant feedbacks for carbon cycling, fire frequency, and global climate change. Since ecological studies and climate change models require routine information on surface soil moisture, the ability to remotely sense this variable is highly desirable. Toward this end research was conducted on developing methods for the retrieval of spatially and temporally varying patterns of soil moisture from recently burned boreal forest ecosystems of Alaska using C-band satellite radar data. To do this we focused on both individual date and temporal SAR datasets to develop techniques and algorithms which indicate how moisture varies across a recently burned boreal forest. For each of the methods developed we focused on reducing errors of SAR-derived soil moisture estimates due to confounding factors of variations in vegetative biomass and surface roughness. For the individual date soil moisture monitoring, we grouped test sites by a measurable biophysical variable, burn severity, and then developed algorithms relating moisture to SAR backscatter for each burn severity group. The algorithms developed had high coefficients of determination (0.56-0.82) and the moisture maps produced had high accuracy (3.61 rms error) based on the minimal validation conducted. For the seasonal soil moisture mapping we used principal component analysis to capture the time-variant feature of soil moisture and minimize the relatively time-invariant features that confound SAR backscatter. This resulted in good agreement between the drainage maps produced and our limited in situ observations and weather data. However, further validation, with larger sample sizes, is needed. While this study focuses on Alaska, research indicates that the techniques developed should be applicable to boreal forests worldwide.