Retrieval of soil moisture using airborne synthetic- and real- aperture radar data at different spatial scales

For the studies of surface soil moisture, radar observations offer high spatial resolution through aperture synthesis. The spatial resolution may be tens of meters from airborne platform to a few kilometers for global mapping from spaceborne platforms such as the L-band Soil Moisture Active Passive (SMAP) mission. A suite of sophisticated radar scattering forward models 'data-cubes' were developed for 17 land cover classes that simulate the radar response to soil moisture, surface roughness, and diverse vegetation. The inversion of the forward models to estimate soil moisture becomes a well-constrained problem through time-series analysis and parameterization of vegetation structure through an empirical allometric relationship. The retrieval approach has been developed for application using SMAP radar observations. This paper studies the applicability of the approach to radar observations over various spatial scales. The high-resolution airborne L-band radar data were collected by the Uninhabited Aerial Vehicle Synthetic Aperture Radar during the SMAP Validation Experiment 2012 conducted near Winnipeg, Canada in 2012 over agriculture fields. During the six-week campaign period, there were three major drying cycles of surface soil moisture that occurred while the agricultural vegetation grew from bare condition to full canopy. The volumetric soil moisture and radar backscattering generally showed a positive relationship when the vegetation effect was not strong, such as over pasture fields, or when the vegetation effect did not change substantially over time, such as for wheat fields. However, the significant temporal changes in corn and soybean canopies from bare to full growth result in a reverse of this relationship over these fields. The data-cube retrieval approach was able to systematically correct for the effect of the changing vegetation. Errors in soil moisture retrieval ranged from 0.037 to 0.086 cm3/cm3, when evaluated over all the available fields. In addition, the Passive Active L-band Sensor (PALS) was flown during the campaign. This instrument collected real-aperture radar data over the same spatial and temporal domains as those of the UAVSAR. The PALS spatial resolution is about 650m and often includes more than one crop types within a footprint. The data-cube retrieval algorithm will be applied to the PALS radar data and its ability to perform with the heterogeneous vegetation conditions will be evaluated. The scatterometer onboard the Aquarius/SAC-D satellite has provided the global 70km-resolution radar data for 3 years so far. The radar forward model predictions match the Aquarius data with the mean difference of about 2 dB and a standard deviation of about 3 dB (one sigma) globally. The data-cube retrieval will also be tested on a global scale using the Aquarius data.