Comparing global sensitivity analysis approaches of a semi-distributed hydrological model in tropical regions. Study case: Middle Magdalena Valley, Colombia

The management of uncertainty in hydrologic modeling is contingent upon two key factors: the ability to select appropriate values for model parameters and the capacity to assess the extent to which their variation affects a simulated response. This study employs sensitivity and uncertainty analyses to assess the influence of pivotal parameters in the distributed watershed model, TopModel, for approximating surface runoff in the Middle Magdalena Valley, Colombia. A variance-based global sensitivity analysis (GSA) was conducted, employing Sobol's indices, in conjunction with a classical decomposition of variance and other recently developed indices that estimate the contribution of each model parameter. This was done for the purpose of measuring the contribution of parameters to the mean, variance, skewness, and kurtosis of simulation outputs. The analysis, based on 150,000 model simulations over a 12-year period, incorporated random parameter variations within ranges supported by existing literature and initial model calibration. The findings identified a reduced set of sensitive parameters, including the recession curve, maximum root zone storage deficit, and initial subsurface flow, which were found to significantly influence model performance.