Thank you for your response.
Resampling the DEM or using a finer grid resolution does indeed result in a significantly larger number of grid cells, ultimately leading to an unmanageable number of HRUs and subbasins. For example, at a 30 × 30 m resolution, the model ends up generating more than 5 million HRUs (even with the Dominant HRU option), which becomes computationally infeasible.
Given this limitation, is there an alternative way to incorporate small waterbodies—those smaller than the grid cell size—into the model? Perhaps by representing them as wetlands or detention ponds without requiring direct stream connectivity?
Regarding the Python error:
The issue only arises when I manually add lake cells that cover less than 50% of a grid cell area. However, I have ensured that the RES field contains only valid integer values.