Use of Satellite Observations for Analog Ensemble Predictions to Contribute to Decision Advantage

Despite significant advances in high-resolution numerical weather prediction (NWP) and satellite remote sensing, multiple factors, including errors in initial conditions, quality of model initialization, model physics, and parameterizations, influence model forecast uncertainty which impacts Army mission preparation and planning. Further, NWP deterministic forecasts do not present uncertainty estimates, whereas NWP ensembles require significant computational overhead to generate uncertainty estimates. To support the Army mission, Areté and the National Center for Atmospheric Research (NCAR) propose the Analog Ensemble Generative Environmental Applications System (AEGEAS). AEGEAS integrates the Analog Ensemble (AnEn) technique, which has been developed at NCAR over the past 10 plus years, with Areté’s Prompt Realistic Inference of Nowcasted Cloud Environments (PRINCE) system, a novel artificial intelligence (AI) framework producing high-resolution deterministic observation forecasts from NWP forecast fields. The AnEn technique offers an ensemble prediction of a target variable by approximating its probability density function through a Monte Carlo approach. Leveraging historical observations and deterministic predictions from NWP, AI techniques, or a combination thereof, AnEn constructs an ensemble set of forecasts for the variable of interest at each forecast lead time. AEGEAS, therefore, leverages the power of AI and low-computational-cost analog uncertainty estimates to produce high-resolution probabilistic environmental forecasts that can be integrated into Army’s geospatial software to support Army mission needs. The framework also aims to provide more robust and probabilistic forecasts for the target variable than current NWP. The Phase I technical objectives include a proof-of-concept demonstration of AEGEAS, delivery of the deterministic and probabilistic prototype product, and a physical validation and verification to quantify performance.