OMB number: 4040-0019 expiration date: 06/30/2028 project abstract summary this project abstract summary form must be submitted or the application will be considered incomplete

The National Oceanic and Atmospheric Administration awarded a Cooperative Agreement to the Regents of the University of Colorado for $297,788 on September 1, 2026, under the Cooperative Institute (Inter-Agency Funded Activities) program (CFDA 11.405).

The award funds enhancement of quantitative precipitation forecasting (QPF) guidance through artificial intelligence and machine learning techniques and the use of satellite observations in post-processing. The work addresses uncertainties in QPF that persist despite advances in NOAA's numerical weather prediction model suites, with the goal of improving the precision of operational forecasts that support National Weather Service Weather Prediction Center products used in short- to medium-range flash flood forecasting. The project responds to 2021 data showing flash flooding as the deadliest and most damaging weather phenomenon in the United States, causing 134 deaths and more than $2.4 billion in damage.

Work is performed in Boulder, Colorado. The period of performance runs from September 1, 2026, through August 31, 2028. The Cooperative Institute program facilitates cooperation between government, university, and nonprofit research institutions to increase the effectiveness of research and develop innovative approaches in the environmental sciences.

Ensure the project abstract field succinctly describes the project in plain language that the public can understand and use without the full proposal. Use 4,000 characters or less. Do not include personally identifiable, sensitive or proprietary information. Refer to agency instructions for any additional project abstract field requirements. If the application is funded, your project abstract information (as submitted) will be made available to public websites and/or databases including usaspending.GOV. Funding opportunity number NOAA-OAR-ciao-2026-32241 assistance listing number(s): 11.405 applicant name timothy thielke descriptive title of applicant's project enhancing QPF guidance through AI/ML and the use of satellite observations in post-processing project abstract in 2021, the deadliest and most damaging weather phenomenon was flash flooding, causing 134 deaths and more than $2.4 billion in damage according to the united states' 2021 natural hazard statistics summary. The national weather service (NWS) weather prediction center (WPC) uses quantitative precipitation forecasts (QPF) to produce several operational products that support its mission to protect lives and property during heavy rain and flash flood events. The WPC issues two well-known products that support short- to medium-range flash flood forecasting: the mesoscale precipitation discussion for the short-term (0-6-hour) period and the excessive rainfall outlook for flash-flood impacts out to 5 days. These products are informed using deterministic QPF and probabilistic QPF (pqpf). Despite advancements in NOAA's numerical weather prediction (NWP) model suites, QPF uncertainties remain, even at short lead times, challenging the precision of operational forecasts. Skillful uncertainty guidance is essential to providing accurate, actionable forecasts of heavy rainfall, but improving the underlying NWP models alone has not closed that GAP. Machine learning (ML) offers a complementary path forward. Rather than waiting for incremental improvements in NWP physics, an ML post-processor can learn the systematic relationship between a model's forecast fields and observed rainfall, thereby translating raw model output into better-calibrated precipitation probabilities. This project builds on the existing 1-hour pqpf tool, which uses a u-net ML architecture to map operational high- resolution NWP forecast fields to hourly quantitative precipitation estimates (QPE) observations, producing probability fields for the continental united states (CONUS) out to 24 hours. The project will improve the tool's handling of heavy precipitation, develop a temporally smoother 6-hour pqpf, generate ensembles to communicate forecast uncertainty, and extend the products into the medium range (4-7 days). The performance of each product will be evaluated through verification and sensitivity studies to ensure that reliable short- and medium-term rainfall uncertainty is effectively communicated to WPC forecasters and reflected in the operational decision- support products on which emergency managers and the public depend.