A hybrid approach to enhance streamflow simulation in data-constrained Himalayan basins: Combining the Glacio-Hydrological Degree-Day Model and Recurrent Neural Networks

The Glacio-hydrological Degree-day Model (GDM) is a distributed model but prone to uncertainties due to its conceptual nature, parameter estimation, and limited data in the Himalayan basins. To enhance accuracy without sacrificing interpretability, we propose a hybrid model, GDM-RNNs, combining GDM with Recurrent Neural Networks (RNNs). Three RNN types (Simple RNN, GRU, LSTM) are integrated with the GDM. Rather than directly predicting streamflow, RNNs forecast GDM's residual errors. We assessed performance across different data availability scenarios, with promising results. In limited data conditions (one year), GDM-RNNs (GDM-SimpleRNN, GDM-LSTM, GDM-GRU) outperformed standalone GDM and machine learning models. For GDM-SimpleRNN, NSE, R2, and PBIAS were 0.85, 0.82, and -6.21, for GDM-LSTM 0.86, 0.79, and -6.37, and for GDM-GRU 0.85, 0.8, and -5.64, compared to GDM's 0.80, 0.63, and -4.78, respectively. Machine learning models yielded similar results, with SimpleRNN at 0.81, 0.7, and -16.6, LSTM at 0.79, 0.65, and -21.42, and GRU at 0.82, 0.75, and -12.29, respectively. Our study highlights the potential of machine learning in enhancing streamflow predictions in data-scarce Himalayan basins while preserving physical stream flow mechanisms.