This study discusses the reliability of snow cover products in water resources management usingdeep learning techniques to simulate monthly streamflow in poorly gauged basins with 10 years of MODIS satellite snow cover data and hydroclimatic data. In this work, a powerful deep learning tool, Long-short time memory (LSTM) has been adopted to simulate streamflow on a monthly scale due to its flexible structure and solid results in the hydrology process. In order to avoid our model from overfitting, we split the dataset into training and validation; the training set was to fit the LSTM model, while the validation set was used during training to verify how strong the model is generalizing. The model was evaluated using three main criteria: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and correlation coefficient (R²) during the validation period. As first results, considering one feature, rainfall, the model was quite able to simulate streamflow with respective MAE, RMSE, and R² values of 4.99 (m3/s), 7.45 (m3/s), and 0.44, but when we add SCA, the performance of the model increases, and we get MAE = 4.291 (m3/s), RMSE = 6.859 (m3/s), and R²=0.62. The current work concluded that the deep learning model LSTM improves its ability to produce a reasonable result, showing the effect of SCA on the behavior of the output that plays an important role in streamflow modeling. However, in this study, monthly streamflow prediction with the LSTM in a semi-arid region could be improved by the fusion of remotely sensed snow cover data and hydroclimatic data.