This study, conducted as part of the French national CIPRHES research project, aims to improve low-flow forecasting systems by investigating rainfall-runoff model behavior and dynamics in simulation mode within the context of the Moselle River catchment. This catchment is characterized by a warming climate and decreasing low flows. A tailored modeling approach using the daily conceptual rainfall-runoff model PRESAGES-G over 40 sub-catchments was designed and tested for the reference period 2001-2020. A comparative assessment of three optimization algorithms used for model calibration-Fmincon, Genetics Algorithms, and Shuffled Complex Evolution-University of Arizona (SCE-UA)-was also conducted. Fmincon emerged as the best solution because of its MATLAB integration, faster execution times, and overall performance. Selecting the right objective function is also a critical point for accurate model calibration. We evaluated five objective functions derived from the Kling-Gupta Efficiency (KGE) metric. We found that the KGE applied to Q 0.2 produces the smallest errors, especially during low flow months. Finally, our modeling framework is used to parameterize a semi-distributed version of PRESAGES-G called PRESAGES-SD. PRESAGES-SD appears to be more competitive than PRESAGES-G in most sub-catchments, emphasizing the advantages of incorporating spatial variability into our modeling framework. Additionally, our study highlights how model calibration can be challenging across varying time periods, especially when abrupt climate and hydrological change occurs.