This paper explores the application of Long-Short-Term-Memory (LSTM) networks for rainfall-runoff modeling, a key challenge in hydrology. The authors propose a novel data-driven approach using LSTMs, which are capable of learning long-term dependencies essential for modeling storage effects in catchments with snow influence. The study uses 241 catchments from the CAMELS dataset to test the LSTM approach and compares it to the Sacramento Soil Moisture Accounting Model (SAC-SMA) coupled with the Snow-17 snow routine. The results show that the LSTM model achieves better performance than the SAC-SMA + Snow-17 model in over 50% of the catchments, particularly in snow-influenced and higher precipitation areas. The study also investigates the potential of LSTMs as regional hydrological models, where a single model predicts discharge for multiple catchments. The regional models outperform individual LSTM models in catchments with similar hydrological characteristics but perform poorly in catchments with diverse behaviors. The authors suggest that grouping catchments based on attributes like topography and soil properties could improve the performance of regional models. Overall, the study highlights the potential of LSTMs for hydrological modeling, especially in predicting runoff in ungauged basins.This paper explores the application of Long-Short-Term-Memory (LSTM) networks for rainfall-runoff modeling, a key challenge in hydrology. The authors propose a novel data-driven approach using LSTMs, which are capable of learning long-term dependencies essential for modeling storage effects in catchments with snow influence. The study uses 241 catchments from the CAMELS dataset to test the LSTM approach and compares it to the Sacramento Soil Moisture Accounting Model (SAC-SMA) coupled with the Snow-17 snow routine. The results show that the LSTM model achieves better performance than the SAC-SMA + Snow-17 model in over 50% of the catchments, particularly in snow-influenced and higher precipitation areas. The study also investigates the potential of LSTMs as regional hydrological models, where a single model predicts discharge for multiple catchments. The regional models outperform individual LSTM models in catchments with similar hydrological characteristics but perform poorly in catchments with diverse behaviors. The authors suggest that grouping catchments based on attributes like topography and soil properties could improve the performance of regional models. Overall, the study highlights the potential of LSTMs for hydrological modeling, especially in predicting runoff in ungauged basins.