An integrated procedure to evaluate hydrological models.

The growing concern for health-related problems deriving from pollutants leaching is driving national and international administrations to support the development of tools for evaluating the effects of alternate management scenarios and identifying vulnerable areas. Cropping systems models are powerful tools for evaluating leachates under different environmental, social, and management conditions. As percolating water is the transport vehicle for pollutants transport in soil, a reliable evaluation of water balance models is a fundamental prerequisite for investigating pesticides and nitrate fate. As specific approaches for the evaluation of multi-layer evolution of state variables are missing, we propose a fuzzy-based, integrated indicator (ISWC: 0, best; 1, worst) for a comprehensive evaluation of soil water content (SWC) simulations. We aggregated error metrics with others quantifying the homogeneity of errors across different soil layers, the capability of models to reproduce complex dynamics function of both time and soil depth, and model complexity. We tested ISWC on a sample dataset where the models CropSyst and CERES-Wheat were used to simulate SWC for winter wheat systems. ISWC revealed that, in the explored conditions, the global assessment of the two models’ performances allowed identification of CropSyst as the best (average ISWC = 0.441, with a value of 0.537 obtained by CERES-Wheat), although each model prevailed for some of the metrics. CropSyst presented the highest accuracy (average agreement module = 0.400), whereas CERES-Wheat’s accuracy was slightly worse, although achieved with a simplified modelling approach (average Akaike Information Criterion = -230.44), thereby favouring largearea applicability. The non-univocal scores achieved by the models for the different metrics support the use of multi-metric evaluation approaches for quantifying the different aspects of water balance model performances.