Investigating the value of AMSR-E soil moisture data assimilation for hydrological modelling using SWAT

A challenge in hydrology is the lack of measured soil moisture data, which limits the potential of precise prediction. To tackle this shortcoming, we assimilate satellite-based soil moisture data into a hydrological model. We explore the value of assimilating the AMSR-E product for hydrological simulation using Soil Water Assessment Tool (SWAT). Three scenarios are defined and evaluated: univariate assimilation of soil moisture data, univariate streamflow data assimilation and bivariate data assimilation of both soil moisture and streamflow data. For comparative analysis, a baseline scenario (BS) is considered using the SUFI-2 algorithm in the SWAT-CUP package. The proposed method is applied to a watershed located in the Mahabad River Basin, northwest of Iran. The results demonstrate that univariate assimilation of soil moisture data significantly improves soil moisture simulation, increasing the Nash–Sutcliffe efficiency (NSE) from –2.73 (in BS) to 0.34, but it leads to a decline in streamflow estimation accuracy. Conversely, univariate assimilation of streamflow observations improves streamflow estimation NSE from 0.73 to 0.77. This scenario severely reduces soil moisture estimation accuracy, with the NSE dropping from −2.73 to −5.48. In comparison, the multivariate assimilation scenario offers a more balanced outcome with improving soil moisture estimation accuracy to an NSE of 0.10 while maintaining streamflow accuracy at a level comparable to the BS.