Soil moisture retrieval over a site of intensive agricultural production using airborne radiometer data

Abstract This study investigates soil moisture retrievals using airborne passive microwave data at two different resolutions collected during the Soil Moisture Active Passive Validation Experiments in 2012 and 2016 (SMAPVEX12 and SMAPVEX16-MB). Based on the fine-resolution passive data (500 m), we integrate the surface roughness parameters which are traditionally used in the radar backscatter models into the passive emission models. To parameterize the effective roughness Hr in L-band Microwave Emission of the Biosphere (L-MEB) model, we analyze two different functions which include only surface Root Mean Square height (s), as well as both s and the autocorrelation length l (zs = s2/l). For each of the two roughness functions, the b vegetation parameters are optimized for canola, soybean and wheat. The transferability of the b parameters between 2012 and 2016 is also evaluated by comparing the L-MEB model simulated and the measured brightness temperature. Then, the calibrated L-MEB model is applied to the subpixels of the coarse-resolution passive data (1500 m) given the vegetation heterogeneity, to map the soil moisture over the SMAPVEX entire experimental site. The results indicated the Hr model with the zs parameter outperformed that with the s parameter. This suggests that the inclusion of the roughness autocorrelation length in the L-MEB model improved the accuracy of modeling the brightness temperature. The vegetation attenuation on the brightness temperature at V-polarization was stronger than that at H-polarization, due to the dominant vertical structure of the crop canopy. Since the airborne passive observations exhibited remarkable consistency between the 2012 and 2016 measurements, the b parameters obtained in 2012 can be transferred to the 2016. Based on the obtained Hr and b parameters, the soil moisture maps were retrieved using the calibrated L-MEB model applied to the sub-pixel of the coarse-resolution passive data, implying a Root Mean Square Errors (RMSEs) of 0.049–0.058 m3/m3 and correlation coefficients of 0.82–0.87. This paper suggests that the physical roughness zs in the radar domain can be coupled into the L-MEB model to refine the soil moisture retrievals from passive brightness temperature.