Ensemble streamflow prediction (ESP) forecasts prepared by the national oceanic and atmospheric administration's (NOAA) colorado basin river forecast center (CBRFC) are used as inputs to models that…

Ensemble streamflow prediction (ESP) forecasts prepared by the national oceanic and atmospheric administration's (NOAA) colorado basin river forecast center (CBRFC) are used as inputs to models that forecast reservoir levels, such as the colorado river mid-term modeling system (CRMMS), which supports decision-making and other hydrological operational planning efforts in the colorado river basin (CRB). Among the most impactful decisions derived from these forecasts are the decisions to cut annual water allocations to the lower colorado river basin states of arizona, california, and nevada and the republic of mexico and the decisions to release water downstream from the upper colorado river basin states. The seven colorado river basin states and some of the largest water utilities in the seven states partnered with the U.S. Bureau of RECLAMATION (RECLAMATION) to commission the colorado river basin climate and hydrology: state of the science report, which identified a key prediction challenge regarding how streamflow forecasts may not be able to capitalize on the skill that does exist in sub-seasonal and seasonal climate forecasts in the basin. The problem with this prediction challenge is that beyond 3-to-7- months, ESP is generally no more skillful than climatology. IT could be argued that lack of such skillful long lead forecasts beyond a season, likely contributes to sub-optimal management during the current prolonged two decades drought. The report further revealed an opportunity to support research into post-processing of CBRFC forecasts to generate climate-forecast-informed, usespecific streamflow forecasts. Our prior work, with partial funds from NOAA mapp, showed that using dynamical climate model forecasts (e.g., north american multi-model ensemble, nmme) to weight ESP traces improved inputs to CRMMS and improvements to reservoir projections out to 12 months. We also developed a colorado river basin operational prediction testbed (CRBOPT) which mimics the crmm but can take in ensemble streamflow projections to produce probabilistic projections of water management. Machine learning techniques such as 'random forests' offers encouraging prospects to further enhance forecast skill on timescales beyond 1 year. Building on these prior efforts, the objectives of this proposal are to (I) develop multimodel combination methods that combine nmme statistical models, and ESP traces in a machine learning and trace weighting framework to enhance skillful prediction of streamflows over the entire river network at 0 to 24 months lead time, (II) assess the impact of more skillful streamflow forecasts to water resources management variables using the CRBOPT. This work would apply to the stated goals in NOAA-OAR-CPO-2024-2008268 mappnidis: science for the 21ST century western U.S. Hydroclimate by forging partnerships among two NOAA offices (CBRFC and the climate prediction center, CPC), researchers with the cooperative institute for research in environmental studies (CIRES), RECLAMATION, and colorado river basin decision-makers. Additionally, this proposal would work toward the three priority areas in the intermountain west dews strategic action plan for 2021-2025, creating a direct benefit to nidis goals and stakeholders. This proposal was developed in direct response to the opportunities identified by the decision-making community and achieves noaas leadership and service goals by adopting NOAA modeling to enhance interdisciplinary decision support tools of societal relevance. In a broader sense, this work supports NOAA-cpos goals by advancing capabilities to predict the hydrologic system in a region critical to millions of people, including historically underserved populations