Downscaling of coarse-resolution radiometer brightness temperature by high-resolution radar backscatter

Given the importance of soil moisture for hydrological applications, such as weather and flood forecasting, passive microwave remote sensing is a promising approach for retrieving soil moisture due to its high sensitivity to near-surface soil moisture, applicability to all weather conditions, direct relationship with the soil dielectric constant, and reduced effects from vegetation and roughness. However, passive microwave (radiometer) observations suffer from being relatively low spatial resolution, on the order of 36 km. It is proposed that this scale issue may be overcome by using active microwave (radar) observations, which have much higher resolution when using Synthetic Aperture Radar (SAR) techniques (<3km), and this is the approach being taken by NASA’s Soil Moisture Active Passive (SMAP) mission, with a scheduled launch in late 2014. The rationale behind SMAP is to use the synergy between active and passive observations in a downscaling approach to overcome the individual limitations of each observation type, and ultimately provide a merged soil moisture data set at intermediate resolution (~9 km). The objective of this study is to test the proposed baseline downscaling approach for the SMAP mission using airborne data, thus assessing its viability for future application to SMAP data. The approach is based on the hypothesis of a near-linear relationship between radiometer brightness temperature (Tb) and SAR backscatter (σ), and has thus far received very limited testing. The experimental dataset used in this study was collected during the Soil Moisture Active Passive Experiment (SMAPEx) field campaigns over a study site in south-eastern Australia approximately 38km × 36km in size, equivalent to a SMAP radiometer pixel. This research focuses on the brightness temperature downscaling algorithm; according to the SMAP Algorithm Theoretical Basis Documents these downscaled brightness temperatures will subsequently be converted to soil moisture at fine resolution through the traditional passive microwave retrieval algorithm. The baseline downscaling algorithm was applied to high resolution data from SMAPEx, which include 1km resolution brightness temperature collected by the Polarimetric L-band Multibeam Radiometer (PLMR) and ~ 10m resolution backscatters collected by the Polarimetric L-band Imaging Synthetic aperture radar (PLIS). To minimize noise in the radar data and to approximate the SMAP radiometer/radar pixel ratios (36km Tb to 9km resolution using 3km σ) the PLIS data were aggregated to 250m resolution, so as to downscale 1km Tb to 250m resolution, thus keeping the same ratio of radiometer/SAR spatial resolution as the SMAP mission. Results showed that the Root-Mean-Square Error (RMSE) in Tb downscaled at 100m resolution was around 10K at h-polarization and 8K at v-polarization over a cropping area. This RMSE was reduced to 9K and 7K respectively when downscaling to 250m resolution, due to a decreased spatial heterogeneity during averaging. It was also noted that results at v-polarization were slightly better than those at h-polarization, since the backscatter is more linearly related to Tb at v-polarization than Tb at h-polarization. The accuracy of the downscaling over grassland sites was improved by approximately 3K with respect to the cropping area. This was attributed to the more heterogeneous conditions in cropping areas, compared to the relatively uniform conditions in the grassland area. However, one limitation of this study was the availability of only three days of data for estimating the linearity between radar and radiometer observations.