Data characterizing basin attributes are crucial for enhancing the generalizability of hydrological models. However, few studies have thoroughly investigated the effects of these basin feature data. To address this gap, three key efforts are necessary.

First, it is essential to investigate the relative importance of different basin feature data on the generalizability of hydrological models. Second, certain datasets, such as temporal land use information, present challenges in effective utilization due to their inherent high uncertainty, non-Gaussian distributions, and implicit functions. Specific parameterization frameworks need to be developed to address these challenges. Third, integrating physical information with pure basin feature data, such as Height Above the Nearest Drainage (HAND), should be explored.

In this context, we have implemented sensitivity analyses using a general framework, developed a dedicated parameterization framework, and examined the effects of reanalyzed data based on topographic information. These efforts were evaluated using streamflow data from approximately 3,200 basins via the differentiable HBV hydrological model.

Our results highlight the varying impacts of different basin feature data, the necessity of specialized parameterization strategies, and the importance of incorporating physical information to effectively utilize these basin feature data.