STATISTICAL PROPERTIES OF SOIL MOISTURE IMAGES REVISITED

Abstract

Data from passive microwave remote sensing with an ESTAR L-band radiometer, deployed on an aircraft, were used to produce soil moisture images over the area of the Little Washita watershed in Oklahoma in 1992. This area was revisited during the Southern Great Plains 1997 Hydrology Experiment. This offered an opportunity to evaluate the time-specificity of the conclusions, relating to scaling of the surface soil moisture, that have been reported for 1992. The objective of this work was to compare scaling properties of soil moisture fields observed in 1992 and 1997. We analyzed one 1992 data set and three 1997 data sets, each covering several days of continuous drydown. Different resolutions were introduced by aggregating the pixels of original 200-m resolution into bigger square cells. Scaling in dependencies on resolution was observed for the variance of moisture content, for the within-cell variance, and for the first six moments about zero, the latter indicating multiscaling. Parameters of the scaling equations differed among four drying periods studied. However, once a scaling dependency on resolution was established in the beginning of a drying period, its shape was maintained during the drydown both in 1992 and 1997. Slopes of the dependencies changed only slightly, whereas the intercepts decreased as the drying progressed. Having constant slopes and intercepts dependent on average area water contents gives an opportunity to reduce the volume of observations needed to predict scaling of surface soil moisture during drydowns.