Uncertainty assessment of the agro-hydrological SWAP model application at field scale: A case study in a dry region

•The agro-hydrological SWAP model is combined with the uncertainty analysis GLUE approach.•The predictive uncertainty of simulated soil moisture dynamic is quantified.•Uncertain soil hydraulic parameters have different effects on soil water fluxes (i.e. evapotranspiration and deep percolation). Uncertainty analysis can provide useful insights into the sources and effects of uncertainty for decision makers to achieve the goals of reliability and sustainability in water management. This study presents parameters uncertainty of a physically based soil–water–atmosphere–plant (SWAP) model and its effect on model prediction within the generalized likelihood uncertainty estimation (GLUE) framework for two irrigated agricultural fields in a dry region of Iran. To simulate soil water dynamics of the two fields, the SWAP model is calibrated using soil moisture observation data. The results demonstrate that predictive uncertainty in soil moisture during the growing season in both fields is relatively small and a good model performance is achieved. Parameter uncertainty analysis of soil hydraulic parameters showed that in spite of similarity of soil texture in both the fields, the estimated parameters (i.e. posterior distribution) exhibit different behaviors. This was because of the dynamics of soil structure which varies considerably within cultivated fields during the growing season. Moreover, the simulated water balance fluxes (actual evapotranspiration and deep percolation) indicate that in irrigated agricultural fields in dry regions, the precision of actual evapotranspiration predicted by the SWAP model is high (i.e. a high degree of model reliability is achieved). However, deep percolation fluxes show higher variation (lower precision) and are more sensitive to soil hydraulic conductivity parameterization. Finally, this study reveals the importance of uncertainty analysis to estimate the degree of reliability associated with model predictions as an important first step for providing decision makers with realistic information about the models outputs.