Abstract: Monitoring and modelling tools may improve irrigation strategies in precision agriculture. We used non-invasive soil moisture monitoring, a crop growth and a soil hydrological model to predict soil water content fluctuations and crop yield in a heterogeneous sandy grassland soil under supplementary irrigation. The sensitivity of the soil hydrological model to hydraulic parameters, water stress, crop yield and lower boundary conditions was assessed after integrating models. Free drainage and incremental constant head conditions were implemented in a lower boundary sensitivity analysis. A time-dependent sensitivity analysis of the hydraulic parameters showed that changes in soil water content are mainly affected by the soil saturated hydraulic conductivity K-s and the Mualem-van Genuchten retention curve shape parameters n and alpha. Results further showed that different parameter optimization strategies (two-, three-, four- or six-parameter optimizations) did not affect the calculated water stress and water content as significantly as does the bottom boundary. In this case, a two-parameter scenario, where K-s was optimized for each layer under the condition of a constant groundwater depth at 135-140 cm, performed best. A larger yield reduction, and a larger number and longer duration of stress conditions occurred in the free drainage condition as compared to constant boundary conditions. Numerical results showed that optimal irrigation scheduling using the aforementioned water stress calculations can save up to 12-22 % irrigation water as compared to the current irrigation regime. This resulted in a yield increase of 4.5-6.5 %, simulated by the crop growth model.

Abstract: Field water flow processes can be precisely delineated with proper sets of soil hydraulic properties derived from in situ and/or laboratory experiments. In this study we analyzed and compared soil hydraulic properties obtained by traditional laboratory experiments and inverse optimization tension infiltrometer data along the vertical direction within two typical Podzol profiles with sand texture in a potato field. The main goal was to identify proper sets of hydraulic parameters and to evaluate their relevance on hydrological model performance for irrigation management purposes. Tension disc infiltration experiments were carried out at four and five different depths for both profiles at consecutive negative pressure heads of 12, 6, 3 and 0.1 cm. At the same locations and depths undisturbed samples were taken to determine Mualem-van Genuchten (MVG) hydraulic parameters (theta(r), residual water content, theta(s), saturated water content, alpha and n, shape parameters and K-ls, saturated hydraulic conductivity) in the laboratory. Results demonstrated horizontal differences and vertical variability of hydraulic properties. The tension disc infiltration data fitted well in inverse modeling using Hydrus 2D/3D in combination with final water content at the end of the experiment, theta(f). Four MVG parameters (theta(s), alpha, n and field saturated hydraulic conductivity K-fs) were estimated (theta(r) set to zero), with estimated K-ls and alpha values being relatively similar to values from Wooding's solution which used as initial value and estimated theta(s) corresponded to (effective) field saturated water content, theta(f). The laboratory measurement of K-ls yielded 2-30 times higher values than the field method K-fs from top to subsoil layers, while there was a significant correlation between both K-s values (r = 0.75). We found significant differences of MVG parameters theta(s), n and alpha values between laboratory and field measurements, but again a significant correlation was observed between laboratory and field MVG parameters namely K-s, n, theta(s) (r >= 0.59). Assessment of the parameter relevance in 1-D model simulations, illustrated that the model over predicted and under predicted top soil-water content using laboratory and field experiments data sets respectively. The field MVG parameter data set resulted in better agreement to observed soil-water content as compared to the laboratory data set at nodes 10 and 20 cm. However, better simulation results were achieved using the laboratory data set at 30-60 cm depths. Results of our study do not confirm whether laboratory or field experiments data sets are most appropriate to predict soil water fluctuations in a complete soil profile, while field experiments are preferred in many studies. (C) 2016 Elsevier B.V. All rights reserved.