Assessment of Data Importance and Monitoring Network Optimization for HYDRUS-3D Model Calibration with a Linear Statistical Inference Method

The objective of this study was to assess the importance of existing and new monitoring data for calibration of HYDRUS-3D model. A tracer experiment was carried out at USDA-ARS OPE3 field site. A pulse of KCL solution was applied on 13x14 m irrigation plot and the Cl concentrations were monitored for 131 days at three depths in12 observations wells installed within, and at distances of 7 m and 14 m from the irrigation plot. Distributions of soil material were obtained from soil cores and hydraulic properties for each material were estimated using the ROSETTA software. HYDRUS-3D model was manually calibrated on Cl time series measured in the observation wells. Local sensitivity analysis was conducted for the saturated hydraulic conductivity by varying values of Ksat at each soil material. The sensitivity indices (Si) were computed for 8 soil materials in 256 simulation runs. The importance of existing and new observation locations for the HYDRUS-3D calibration was evaluated using the Observation-Prediction (OPR) statistics. Averaged over time the Si values obtained in the HYDRUS-3D simulations were used to calculate the matrix of sensitivities for the OPR method, while the observation weights in the OPR were represented by the proportion of simulation days with nonzero Si values in the total number of simulated days. The results showed that the number of depths for Cl monitoring in the existing wells can be reduced to one. To reduce the uncertainty in Si values new observation wells should be installed in the zone where transition from relatively high to low concentrations occurs. The outcome of this study can provide the information for future data collection and monitoring efforts to improve reliability of 3D model calibrations.