TITLE:
Hydrological Modeling of the Diamniadio Urban Area (PUD) Watersheds Using SWAT: Sensitivity Analysis, Calibration, and Validation Using Historical Data
AUTHORS:
Essomanam Marius Pignandi, Louis Eugene Victor Sambou, Coumba Ndiaye, Christelle Musique Mendy, Saïdou Ndao
KEYWORDS:
Hydrological Modeling, SWAT, SWAT-CUP, SUFI-2, Calibration, Validation, Diamniadio
JOURNAL NAME:
Open Journal of Modern Hydrology,
Vol.16 No.4,
September
28,
2026
ABSTRACT: This study presents a historical baseline analysis of the hydrological behavior of the Diamniadio Urban Area (PUD) using the SWAT model, coupled with SWAT-CUP for sensitivity analysis, calibration, and validation. The modeling covers the three watersheds that make up the PUD. The data used include the digital elevation model (DEM), the FAO soil map, land cover derived from 1992 Landsat images, and daily climate data covering the period 1981-2024. The 1992 land-use data were selected as a historical reference because they correspond to the available conditions closest to the observed discharge time series used for calibration and validation. The sensitivity analysis performed using the SUFI-2 algorithm shows that parameter CN2 is the most sensitive across all three watersheds. This result highlights the importance of surface runoff, land cover, and impervious surfaces, as well as the role of groundwater-related parameters in the hydrological response. Model performance was evaluated using the R2, PBIAS, NSE, RSR, RMSE, p-factor, and r-factor metrics. The values obtained indicate generally satisfactory to excellent performance, with R2 values ranging from 0.83 to 0.98 and NSE values ranging from 0.82 to 0.97. The SWAT model accurately reproduces the dynamics of regionalized runoff in WS1 and WS2. Despite satisfactory performance, WS3 has limited coverage of observations within the 95PPU uncertainty band and is therefore subject to greater uncertainty. Its results should be used as exploratory trends. The study thus provides a calibrated and validated historical hydrological database, with satisfactory performance for WS1 and WS2, but greater uncertainty for WS3.