An improved calibration technique to address high dimensionality and non-linearity in integrated groundwater and surface water models

The calibration of integrated groundwater-surface water models is often associated with high dimensionality and stagnation around local optimum solutions. Since these models are computationally demanding and also non-linear, finding their global optimum solution requires efficient optimization techniques. Here, we introduce the Multi-Memory Particle Swarm Optimization (MMPSO) algorithm. The swarm cognitive capacity is enhanced to minimize the number of local optimums and calibrate the model based on sub-objective functions. We used the MMPSO to simultaneously calibrate groundwater head, streamflow, baseflow, and nitrate loads in the SWAT-MODFLOW-RT3D model with 78 sensitive parameters. The results demonstrate that enhancing the cognitive capacity led to a marked improvement in discovering the global optimum solution. Furthermore, we evaluated the calibrated model's performance to quantify groundwater nitrate loads to streams and characterize the shallow surficial aquifer under intensive fertilizer land use. The results show the effectiveness of the MMPSO algorithm for calibrating complex hydrogeochemical models for large-scale applications. •Calibrating integrated models based on geological sections minimizes equifinality.•A new algorithm effectively calibrate a non-linear integrated model with high dimensionality.•The calibrated model quantified NO3 loads to the Great Barrier Reef.