A Köppen–Geiger classification derivative tailored for numerical modeling of ecosystems within watershed dynamics

The dynamics of watersheds result from the interplay of multiple, coupled hydrological and ecological processes, which are strongly modulated by climatic conditions. Addressing these processes through numerical modeling often requires the generation of climate scenarios that accurately reflect short-term climatic variability in key parameters such as air temperature, precipitation, solar radiation, wind speed, relative humidity and cloud cover. In this study, we introduce a new framework to generate climate scenarios that capture the specific climatic features of the regions represented by the Köppen-Geiger classification. To this end, clustering methods were applied to identify climatic curve patterns, leading to the subdivision of each Köppen-Geiger class into subclasses defined by distinct temporal patterns, resulting in a total of 108 subclasses. A complementary analysis was then conducted to explore whether these pattern differences could be related to partial climatic signatures or specific regional contexts. This led to the reorganization of these subclasses into 10 groups, for which such characteristic signatures were proposed. Climatic scenarios can be directly derived from the newly generated subclasses, providing ready-to-use climate forcings for hydro-environmental numerical models. The approach was illustrated through a case study on Lake Mendota using one of the generated scenarios. Result quality is evaluated by comparing the errors between vertical profiles simulated using the climatic scenario and those simulated using observed climatic data. Finally, the simulated primary productivity remains close to the reference value, with an approximate difference of 6 %. This proof-of-concept experiment, which calls for further applications, shows that the framework reproduces the key environmental drivers required for hydro-ecological and environmental modeling. Its extension across diverse climatic contexts therefore holds strong potential for predictive purposes.