Integrated sensitivity analysis of a macroscale hydrologic model in the north of the Iberian Peninsula

Process-based hydrologic models allow to identify the behavior of a basin providing a mathematical description of the hydrologic processes underlying the runoff mechanisms that govern the streamflow generation. This study focuses on a macroscale application of the Variable Infiltration Capacity (VIC) model over 31 headwater subwatersheds belonging to the Duero River Basin, located in the Iberian Peninsula, through a three-part approach: (1) the calibration and evaluation of the VIC model performance for all the subwatersheds; (2) an integrated sensitivity analysis concerning the soil parameters chosen for the calibration, and (3) an assessment of equifinality and the efficiency of the calibration algorithm. The temporal evaluation of the model was done for the calibration and the subsequent validation periods, and showed good results for most of the subwatersheds that largely improved the benchmark performance. The spatial performance reflected a high transferability for most parameter combinations, and the least transferable were related to subwatersheds located in the northern mountains. An additional evaluation of the simulated actual evapotranspiration produced satisfactory adjustments to two selected data products. The sensitivity measures were obtained with the Standardized Regression Coefficients method through a post-process of the outputs of a Monte Carlo simulation carried out for 10,000 parameter samples for each subwatershed. This allowed to quantify the sensitivity of the water balance components to the selected parameters for the calibration and understanding the strong dependencies between them. The final assessment of the equifinality hypothesis manifested that there are many parameter samples with performances as good as the optimum, calculated using the Shuffled-Complex-Evolution Algorithm. For almost all the analyzed subwatersheds the calibration algorithm resulted efficient, reaching the optimal fit. Both the Monte Carlo simulation and the use of a calibration algorithm will be of interest for other feasible applications of the VIC model in other river basins.