Challenges in unsaturated zone data assimilation
The modelling of water flow in soils is important for agricultural applications, in the context of land surface modelling as part of weather and climate models and for water resources assessment. The governing equation for water flow in soils is highly non-linear and several soil hydraulic parameters are unknown and spatially heterogeneous. This work investigated the estimation of soil hydraulic parameters with Kalman type data assimilation algorithms. In an idealized synthetic study we first showed that the joint estimation of three spatially distributed fields of three different soil hydraulic parameters can be improved via data assimilation with an iterative EnKF. The role of spatial and vertical correlation lengths was investigated. A real-world application was made for the Rollesbroich site in the Eifel mountains in Germany. It was found that the estimation of states in an independent verification period hardly improved compared to open loop simulations. A synthetic experiment, mimicking as good as possible the real-world site, revealed, in part, the reasons for the poor performance: (i) structural model errors; (ii) limited value of the data captured by the sensors for the unsaturated zone.
Prof. Harrie-Jan Hendricks-Franssen Forschungszentrum Julich, Germany