ML-Based Streamflow Prediction in the Upper Colorado River Basin Using Climate Variables Time Series Data

Streamflow prediction plays a vital role in water resources planning order to understand the dramatic change of climatic and hydrologic variables over different time scales. In this study, we used machine learning (ML)-based models, including Random Forest Regression (RFR), Long Short-Term Memory (LSTM), Seasonal Auto- Regressive Integrated Moving Average (SARIMA), Facebook Prophet (PROPHET) predict 24 months ahead natural streamflow at Lees Ferry site located bottom part Upper Colorado River Basin (UCRB) US. Firstly, only historic data ahead. Secondly, considered meteorological components such as temperature precipitation additional features. We tested models on monthly test dataset spanning 6 years, where 24-month predictions were repeated 50 times ensure consistency results. Moreover, performed sensitivity analysis identify our best-performing model. Later, analyzed effects considering span window sizes quality made by best Finally, applied model, RFR, two more rivers states UCRB model’s generalizability. evaluated performance predictive using multiple evaluation measures. The multivariate time-series found be accurate, with RMSE less than 0.84 mm per month, R-squared 0.8, MAPE 0.25. Therefore, conclude that increases accuracy predictions. Ultimately, RFR performs among four is generalizable other UCRB.