Presentation | NG21A: Advances in Data Assimilation, Data Fusion, Machine Learning, Predictability, and Uncertainty Quantification in the Geosciences I Oral
Oral
[ONLINE] NG21A-06: Bayesian Hierarchical Network Model for Disaggregation of Spring Seasonal Streamflow
Water managers on the Colorado River are issued a single four-month runoff total each spring, but they need month-by-month numbers at several upstream checkpoints to set reservoir releases and protect fish habitat. We built a new statistical model that unpacks the seasonal total into April, May, June, and July flows while making sure no water is lost or created along the way. Tested on 40 years of historical, naturalized records, the model keeps annual water balance within about ±7 million acre-feet 97% of the time and correctly captures landmark years such as the big 2011 flood. The next step—now under way—is to feed real-time climate forecasts and the Bureau of Reclamation’s ensemble projections into the same framework so operators can see reliable monthly volumes months in advance.