Enhancing the Autocalibration of SWAT-Modflow Using the Melody Search Algorithm: Fid-Bau Portal

Enhancing the Autocalibration of SWAT-Modflow Using the Melody Search Algorithm

Dariane, A. B. / Rahmani, A.

Model calibration is a crucial aspect of effective hydrologic model development. This paper comparatively investigates the application of the Melody Search (MSe) algorithm, a recently developed robust heuristic method, in calibrating the soil and water assessment tool (SWAT) and SWAT–modular finite-difference flow model (SWAT-Modflow) models. The comparison is made with the top two calibration techniques of SWAT-CUP, namely, sequential uncertainty fitting 2 (SUFI2) and particle swarm optimization (PSO) using single and two-phase procedures based on the speed and accuracy factors. For this purpose, a batch file is developed using FORTRAN programming via the SWAT-Edit interface. The findings suggest that both the Melody and SUFI2 algorithms, as well as their combination with different orders, outperform other individual and coupled methods when applied to both the SWAT and SWAT-Modflow models. Melody showed impressive accuracy and speed, whereas SUFI2 displayed high accuracy with relatively lower speed. On the other hand, the PSO demonstrated high speed but lower accuracy in calibrating both models. Melody demonstrated superior performance in calibrating the SWAT model, achieving a final Nash–Sutcliffe efficiency index (NSE) of 0.72, surpassing SUFI2 and PSO with NSE values of 0.69 and 0.68, respectively. In the SWAT-Modflow model, SUFI2 with an NSE of 0.76 performed marginally better than Melody ( $NSE = 0.75$ ), and both outperformed PSO with an NSE equal to 0.70. Furthermore, a combination of the Melody-SUFI2 demonstrated superior performance in the SWAT model, achieving an NSE of 0.77 compared to the best single algorithm (i.e., Melody with $NSE = 0.72$ ). Additionally, utilizing the SUFI2-Melody combination order substantially boosted the SWAT-Modflow NSE value to 0.80, as compared to 0.76 of SUFI2 as the best single algorithm, establishing it as the most effective combination method. Thus, it seems there is a great opportunity to improve the calibration of hydrological models through more robust algorithms such as the MSe and combination approaches.

The application of the Melody Search algorithm in calibrating hydrological models such as SWAT and SWAT-Modflow offers significant practical benefits. By achieving superior accuracy and speed compared to traditional methods like SUFI2 and PSO, Melody Search enhances the efficiency of model calibration processes. This improvement is crucial for optimizing water resource management decisions, such as assessing the impacts of land use changes, predicting water availability, and planning sustainable water use strategies. Additionally, its ability to handle large-scale data sets and complex interactions within watershed systems makes it a valuable tool for researchers and practitioners aiming to achieve higher precision and reliability in hydrological modeling. Integrating Melody Search into such applications enhances the robustness and reliability of hydrological predictions, thereby supporting informed decision making for water resource management and planning.