<p>Accurate landslide hazard prediction relies on well-calibrated models. Traditional single-objective calibration focuses on optimizing one model aspect—typically the final landslide condition—which can limit the model’s accuracy in representing overall landslide dynamics. In contrast, multi-objective calibration simultaneously optimizes multiple model aspects, being able to capture both antecedent and final conditions to enhance model performance. This study compares both calibration approaches within a physically based slope stability model applied to the Vall d’Aran region. Several calibration scenarios were proposed using an automatic calibration module. While the single-objective approach performed well overall, it struggled to accurately represent soil water dynamics. In contrast, the best multi-objective calibration scenario improved both the model’s accuracy and its physical representativeness, achieving accuracies of 91% and 72% for antecedent and final landslide conditions, respectively. Findings reveal that antecedent effective recharge and drainage area size significantly influence landslide susceptibility, with rainfall infiltration identified as the primary trigger. Additionally, groundwater response analysis indicated a response time of ~ 100 days, meaning that the effects of a single large rainfall event can persist for several months, leading to the generation of lateral flow and ultimately causing slope failure. The multi-objective calibration approach introduced here provides a novel, efficient approach, adaptable to data-scarce regions and applicable to similar models.</p>