An improved Kalman filtering approach for the estimation of unsaturated flow parameters by assimilating photographic imaging data

As a non-invasive method, photographic imaging techniques offer some interesting potentials for characterization of soil moisture content in unsaturated porous media, enabling mapping at very fine resolutions in both space and time. Although less explored, the wealth of soil moisture data provided by photographic imaging is also appealing for the estimation of unsaturated soil hydraulic parameters through inverse modeling. However, imaging data have some unique characteristics, including high susceptibility to noise, which can negatively affect the parameter estimation process. In this study a sequential data assimilation approach is developed to simultaneously update soil moisture content and soil hydraulic parameters using photographic imaging data. The study combines numerical modeling, polynomial chaos expansions, and a constrained restart dual ensemble Kalman filter (CRD-EnKF), to formulate a novel method that permits direct assimilation of imaging data. The proposed approach is validated using lab-scale data obtained from taking images of a two-dimensional plexiglass flow tank in a drainage/imbibition cycle. The study demonstrates the effectiveness of the proposed approach, and shows that in some cases, image-based hydraulic parameter estimations can outperform estimations obtained by using sparse direct, contact-based measurements of water content. We also propose a CRD-EnKF approach to estimate the conversion function that relates photographic light intensities to soil moisture content, without the use of direct measurements of the soil water content. The approach uses numerical model outputs instead of direct measurements, and can be useful when a good estimate of model input parameters is already available.