Remote Sensing of Soil Water Storage Capacity Using the Landsat and MODIS Image Archives

Abstract:

We will present a method for the quantitative assessment of the soil water storage capacity of each pixel in a Landsat or MODIS image using the information available in the historic Landsat and MODIS archives. The soil water storage capacity represents the maximum amount of water that can be stored in the soil and/or bedrock so that it is available for release into the atmosphere through transpiration by vegetation and/or evaporation from the land surface. First, the METRIC algorithm is used to convert 15 images representative for growing seasons in wet, dry and normal years into evaporative fraction maps. The evaporative fraction is an expression of the relative evapotranspiration and is strongly correlated to soil moisture conditions in the root zone: high and low evaporative fractions indicate, respectively, high and low root zone soil water contents. We use an experimental relationship to derive a normalized root zone soil moisture value between 0 (dry) to 1 (saturation) from the evaporative fraction. Then, the wetness score for each pixel is calculated as the sum of its 15 “normalized root zone soil moisture” values; it is a relative measure of the overall wetness of a pixel compared to other pixels with values between 0 and 15. Large and small values for the wetness score indicate, respectively, large and small values for the soil water storage capacity. The challenge is to convert the ranking of the wetness scores for each pixel into a quantitative soil water storage capacity. For this operation we use the hydrological Distributed Parameter Watershed Model (DPWM). After construction of seven physically realistic conversion functions between wetness score rank and soil water storage capacity, we evaluate the seven distributions of the differences between the 15 METRIC observed and DPWM simulated “normalized root zone soil moisture” maps. The conversion function that yields the smallest sum of differences is considered the optimal function and is used for generation of the final map of soil water storage capacities. This map can be used for the parameterization of any hydrological model for simulations of hydrological events either in the past before availability of satellite imagery or in the future. The method is demonstrated with a case study in the San Gabriel Mountains in California.