Impact of irrigation forcing on parameter estimation of 1-D Richards equation

In irrigated agricultural regions, remotely sensed soil moisture and evapotranspiration data are widely used to calibrate unsaturated zone models, specifically those employing the Richards equation and van Genuchten-Mualem (VG) relationships. However, this often leads to a critical forcing-observation mismatch. While remote sensing products capture both rainfall and irrigation signatures, standard meteorological datasets typically include only rainfall. Calibrating models against irrigation-influenced observations without accounting for irrigation as an input flux is likely to introduce significant parameter bias. The present study attempts to analyze this effect for an experimental site at IIT Kanpur during a wheat growing season. Subplot specific leaf area index, root zone depth, irrigation amounts, and rainfall were recorded separately for four subplots. Soil moisture and water retention curves were measured at 10, 25, 50, and 80 cm depths covering root zone of these subplots. Meteorological variables from an onsite automatic weather station were used to estimate crop evapotranspiration. For analysis, two calibration schemes that minimize root zone soil moisture simulation errors were formulated, a) RET: considers rainfall as the input flux (ignoring irrigation) along with evapotranspiration, and b) RIET: considers both rainfall and irrigation fluxes along with evapotranspiration. Four VG-parameters (θ_s, α, n, and K_s) were calibrated using mean soil moisture (µ_θ) and evapotranspiration data within a genetic algorithm framework. The analysis was further extended to (µ_θ-σ) and (µ_θ+σ) dataset to analyze the performance of the proposed framework in identifying the parameters with drier and wetter soil moisture data, respectively. The RIET scheme yields substantially lower relative errors than RET for K_s (~36% compared to ~66%), n (~4.7% compared to ~5.9%), and θ_s(~9.2% compared to ~17.7%), whereas both schemes achieve comparable high accuracy for α (~1–2% relative error). Soil moisture estimates obtained using the optimal parameters from the RIET scheme exhibit 2 to 3 times lower RMSE compared to those from the RET scheme. These findings underscore the need for considering irrigation in model forcing during calibration for reliable parameter estimation.