Seasonal forecasts tell water managers when to fill or release the Colorado River’s big reservoirs. The official forecast, based on the Bureau of Reclamation’s in-house Ensemble Streamflow Prediction (ESP) traces from the Colorado Basin River Forecasting Center (CBRFC), becomes unreliable after just a few months. We blended those ESP traces with cutting-edge climate-model predictions and machine-learning tools. Testing 40 years of data, our Gradient-Boosting model stayed accurate out to two years, while the current approach lost skill by month six. This longer warning window can help operators prepare for very wet or very dry years, improving drought planning, power generation, and overall river management.