A combination of DISPATCH downscaling algorithm with CLASS land surface scheme for soil moisture estimation at fine scale during cloudy days

- A new approach to estimate SMOS soil moisture at fine scale for cloudy days was proposed.
- This is a combination of DISPATCH with CLASS surface scheme.
- Data from SMAPVEX12 field campaign were used for validation.
- SMOS soil moisture data and DISPATCH downscaling algorithm were also evaluated.
- The new approach demonstrated good performances compared to ground data.

The main objective of this study is to propose and evaluate a new approach to overcome the major limitation of downscaling methods based on optical/thermal data, particularly the DISaggregation based on Physical And Theoretical scale Change (DISPATCH) algorithm. Data collected over an agricultural site located in Winnipeg (Manitoba, Canada) during the SMAP Validation Experiments 2012 (SMAPVEX12) field campaign were used. At this site, SMOS soil moisture estimates showed a relatively good correlation for both AM and PM overpasses (R ≥ 0.67), but with a significant underestimation (bias ≈ − 0.10 m3/m3), when compared to ground data. SMOS soil moisture data also showed a significant sensitivity to rainfall events. The DISPATCH algorithm was used to downscale bias-corrected SMOS soil moisture data over the study area for the cloud-free days during SMAPVEX12. Compared to ground data, DISPATCH performed well, especially with the soil evaporative efficiency (SEE) linear model (R = 0.81, bias = − 0.01 m3/m3, RMSE = 0.05 m3/m3), which slightly outperformed the SEE non-linear model (R = 0.72, bias = − 0.01 m3/m3, RMSE = 0.06 m3/m3). For both models, the accuracy of the downscaling soil moisture is inversely proportional to the absolute value of soil moisture. For cloudy days, a new operational downscaling approach was proposed. It consists of combining the soil moisture simulations of the Canadian Land Surface Scheme (CLASS) with DISPATCH-downscaled soil moisture during cloud-free days in order to provide estimates of temporally continuous series of soil moisture at 1 km resolution. Compared to ground soil moisture data, the results indicated the high potential of our approach to retrieve soil moisture at 1 km resolution during cloudy days (R = 0.80, bias = − 0.01 m3/m3, RMSE = 0.07 m3/m3).