A fusion approach of the improved Dubois model and best canopy water retrieval models to retrieve soil moisture through all maize growth stages from Radarsat-2 and Landsat-8 data

Soil moisture (SM) retrieval from synthetic aperture radar data in maize fields is a challenging process, as the proportion of surface scattering from underlying soil declines with maize growth. The goal of this study was to develop an SM retrieval algorithm from multi-source fusion data, through the sowing (bare soil), jointing, heading and flowering stages. At the sowing stage, the relationship between backscattering simulations based on an integration equation model and impact factors showed that the influence of surface roughness could be reduced by using the co-polarized difference (CPD). Furthermore, the Dubois model was improved by developing a new CPD model. Comparison of measured and estimated SM contents showed that the improved Dubois model (IDubois) was better than the Dubois model, based on root mean square errors (RMSE_IDubois = 0.0371, RMSE_Dubois = 0.0654). At other growth stages, a variety of vegetation indices were simulated by the PROSAIL model for correlation analysis with the equivalent water thickness of maize leaves. The normalized difference water index was found to be the best vegetation index, and the ideal canopy water (CW) retrieval model could be obtained. The best CW retrieval and IDubois models were subsequently used in the water cloud model (WCM) to retrieve SM content in the maize field. The retrieved SM content agreed well with the measured data (RMSE_HH = 0.0278, 0.1226, 0.0719, RMSE_VV = 0.0346, 0.1809, 0.0723). Overall, these results indicated that the WCM was effective for SM retrieval at some maize growth stages. It was most suitable for estimating SM content at the maize jointing stage.