Determining soil moisture by assimilating soil temperature measurements using the Ensemble Kalman Filter

• We show the potential of inferring soil moisture from soil temperatures using EnKF.
• We highlight the physical link between soil temperature and moisture.
• We discuss the impacts of different observation strategies on soil moisture estimates.

This study investigates the potential to estimate the vertical profile of soil moisture by assimilating temperature observations at a limited number of depths into a coupled heat and moisture transport model (Hydrus-1D). The method is developed with a view to assimilating temperature data from distributed temperature sensing (DTS) to estimate soil moisture at high resolution over large areas. The correlation between temperature and soil moisture in the shallow soil (top ∼ 50 cm) ensures that soil moisture can be estimated using just soil temperature observations. Synthetic tests across a range of soil textures show that with data assimilation both modeled temperature and the moisture profile are improved considerably compared to the ensemble open loop model simulations. In addition, employing data assimilation provides a means to quantitatively account for different sources of uncertainty. This is particularly relevant in the context of DTS given the influence of spatial variability in soil texture and its impact on estimation error. The data assimilation approach could also be used to determine, the number of temperature observations required and the depths at which they should be made. Results suggest that temperature observed at two depths is typically sufficient to estimate soil moisture using this approach. The root mean square error (RMSE) in soil moisture was reduced by up to 75% in the top 20 cm. Furthermore, this approach solves many of the challenges identified in the application of an inversion approach to estimate soil moisture from DTS.