A method for identifying optimum strategies of measuring soil water contents for calibrating a root water uptake model

Measurements of soil water contents, such as with Time Domain Reflectometry (TDR) are essential to calibrate models that estimate the vertical distribution of root water uptake by simulating the unsaturated flow of soil water. This study presents a method to determine how the number of TDR probes used, their orientation and depth of installation affect uncertainties in calibrating a Root Water Uptake Model (SWIF); thus the suggested method identifies optimal placement of TDR sensors for the calibration of a root water uptake model. Our approach is based on estimating efficiencies of different monitoring strategies with the help of a data set of 3000 simulations with different parameter sets. Parameter sets are accepted if simulations reproduced the ‘measured’ soil water dynamics of a reference run at specific depths within an accuracy of 0.01 m3 m−3; a value representing acceptable error intervals in measured soil water contents. The smaller the number of simulations that meet this criterion, the more information measured soil water dynamics contain to identify model parameters and thus the more effective a measurement strategy for model calibration. The presented method is illustrated with results for a pine stand on a sandy soil with a rooting depth of 1.5 m. For this situation, uncertainties in simulated water contents are already less than 0.01 m3 m−3 if SWIF is calibrated against measurements with a relatively small number (5) of TDR probes installed horizontally. Installation depths are critical when using a lower number of probes. Vertical profiles of measured soil water contents, measured with non-automated techniques, are as well effective for model calibration, especially in combination with measurements from permanently installed TDR probes. These profiles contain most information for model calibration if measured 5 days after a clear rain fall event preceded by a dry period of 20 days or more with a relatively high water demand. Uncertainty intervals in simulated distributions of yearly root water uptake were highest in the topsoil (about 5.0×10−4 m3 m−3 day−1 for z>−0.3 m), regardless of measurement strategy and amounted up to 30% of mean yearly uptake. Yet, in general, these uncertainties were less than 10% of the uncertainties in soil water contents. Additional measurements, like root observations and sap flow measurements that reduce ranges of model parameters, are therefore necessary to confine intervals in simulated uptake distributions. Also, error intervals can be reduced by calibrating parameters one by one, instead of tuning all parameters at the same time. Overall, results for the case study clearly demonstrate the potentials of using model simulations for identifying measurement strategies.