Recent progress in NN (neural network) accelerated improvements in the performance of catchment modeling. Yet flood modeling remains a very difficult task. Focusing on two headwater streams, we developed N-HiTS (Neural Hierarchical Interpolation for Time Series Forecasting) and N-BEATS (Network-Based Expansion Analysis for Interpretable Time Series Forecasting) models and benchmarked them with LSTM (long short-term memory) to predict flooding. N-HiTS and N-BEATS outperformed LSTM for flood predictions. We demonstrated how the proposed models can be augmented with an uncertainty approach to predict flooding that is interpretable without considerable loss in accuracy.