The impact of calibration strategies on future evapotranspiration projections: a SWAT-T comparison of three hydrological modeling approaches in West Africa

Abstract. Actual evapotranspiration (AET) is pivotal for the assessment of current and future water availability, particularly for sub humid and AET dominant regions such as West Africa. In this region, climate change is projected to be substantial, which will catalyze hydrological changes. In the climate-hydrological modeling chain for impact assessment, multiple sources of uncertainty are embedded. While the uncertainties inherent in general circulation models (GCM) are difficult to reduce, minimizing uncertainties from hydrological modeling remains a critical focus for researchers and practitioners. Hence, the present study investigates the impact calibration strategies can have on future hydrological changes in West Africa. Given the key role of AET in West Africa, the study particularly evaluates how calibration shapes its future dynamics. In addition, we test whether a specific plant growth modeling, attributed as leaf area index (LAI), can be used as a proxy to predict AET. The Bétérou Catchment in Benin is selected as a demonstration case along hydrological modeling with the eco-hydrological SWAT-T model. To investigate calibration impacts, we apply three strategies, which range from simple (discharge (Q) only) to more comprehensive (Q and LAI; Q, LAI, and AET) approaches. We use the Robust Parameter Estimation algorithm in each calibration strategy to address parameter equinfinality. We use the standardized future climate data from ISIMIP3b (CMIP6) with five GCMs and three emission scenarios and evaluate changes for the near (2031–2050) and far (2070–2099) future periods. The findings show that the amount of future annual AET depends on the calibration strategy, where the change signal for all strategies indicates AET increases. The approach including AET calibration (Q, LAI, AET) shows high future changes, with e.g., multi-model mean changes for SSP5–8.5 of ΔE_near = 5.8 % and ΔE_far = 8.4 %. The results moreover demonstrate that the combined "Q + LAI" can be used as a proxy to predict AET rates. For discharge, the change signal mostly indicates future decreases across all calibration strategies with multi-model mean changes for SSP5–8.5 of ∆Q_far = −7.0 % (Q, LAI, AET) to ΔQ_far = −1.6 % (Q only). Yet, contrasting predictions of future changes depending on single GCMs are simulated. The present study under-20 scores the relevance of uncertainty integration in climate-hydrological modeling and contributes to an improved understanding of water availability assessment in West Africa.