Enhance the prediction of complex hydrological models by pseudo-simulators

Complex hydrological models demand significant computational cost for representing a hydrological system due to its greater number of parameters, the involvement of various datasets, selection of samples, objective functions and algorithms. Pseudo-models are the cheap simulators and it is the best alternative to represent the complex hydrological systems with an input-output response. In the present study, Soil Water Assessment Tool (SWAT) hydrological model is scrutinized by developing pseudo-modeling-based optimization (PMO) method for the agriculturally dominated watershed in India. Results conclude that Gaussian Process regression performed as the best pseudo-model with minimal discrepancy quasi-random sampling of 200 initial design and the prediction of streamflow showed as NSE-0.88 & 0.82 and R-2-0.9 & 0.84 in calibration and validation. The proposed model incorporating into future climate data can provide accurate water-related issues like water use, demand and quality with a less computational burden and moreover, it can identify the water-related disasters like floods and droughts.

Automated summary

Complex hydrological models require significant computational cost, while pseudo-models offer a cheaper and effective alternative. In this study, Gaussian Process regression was found to be the best pseudo-model for optimizing the SWAT hydrological model using the PMO method, resulting in accurate streamflow predictions during calibration and validation.