Experiencing System Outages
Please note: Due to storm impacts on systems at NOAA's National Centers for Environmental Information in Asheville, NC, certain maps and statistics on Drought.gov may be experiencing issues. We apologize for any inconvenience at this time. Read more.
Improving Hydroclimate Forecasts by Multi-Model Combination Approaches for Enhanced Reservoir Operations on the Colorado River
Ensemble Streamflow Prediction (ESP) forecasts from the NOAA National Weather Service’s Colorado Basin River Forecast Center are used as inputs to models that forecast reservoir levels, such as the Colorado River Mid-term Modeling System, which supports decision-making and hydrological operational planning efforts in the Colorado River Basin. In 2020, water management decision-makers in the Colorado River Basin commissioned the Colorado River Basin Climate and Hydrology: State of the Science Report. This report identified a key problem in the basin: the limited skill and probabilistic nature of climate forecasts—like the ESP—may not mesh well with decision frameworks, so water managers are unable to extract value from the forecast information. This project aims to help solve this critical research-to-operations problem, by extracting value from NOAA climate forecasts.
Specifically, this project aims to:
- Develop multi-model combination methods that combine North American Multi-Model Ensemble (NMME) statistical models and ESP traces using machine learning and trace weighting to enhance skillful prediction of streamflows over the entire river network at 0 to 24 months’ lead time.
- Assess the impact of more skillful streamflow forecasts on water resources management variables using the Colorado River Basin Operational Prediction Testbed (CRBOPT).
This project is part of the MAPP/NIDIS-supported Drought Task Force V.
Research Snapshot
Rajagopalan Balaji, Department of Civil, Environmental and Architectural Engineering & Cooperative Institute for Research in Environmental Sciences (CIRES), CU Boulder
Emerson LaJoie and Matthew Rosencrans, NOAA's Climate Prediction Center
What to expect from this research
- Calibrated and validated NMME forecasts for the Colorado River Basin.
- A multi-model combination model using Random Forest (machine learning technique) for ensemble flow forecasts at all the decision points in the Colorado River Basin. (See Woodson et al. 2024 for a machine learning model for short-term forecasting.)
- Stakeholder engagement on the potential for using this combined framework of flow and decision variable ensemble forecasts.
- A web application that contains flow forecast ensembles from the hindcast period, derivative CRBOPT decision variables, and associated performance metrics. The web portal will also have, were available, future flow forecasts from the various approaches.
