Weather Applicable Reliable Data Initialization System for Forecasting Prediction

The U.S. Navy is seeking development of a machine learning weather prediction (MLWP) system to generate a complete model analysis, initialization, and produce skillful forecasts from an incomplete, non-grided set of observations that may include sparse and irregular spatial and temporal sampling of in situ, remote sensing, satellite, and qualitative forecaster-based information. To address the Navy's need, Intellisense Systems, Inc. (Intellisense) proposes to develop the Weather Applicable Reliable Data Initialization (WARN) system for forecasting prediction, based on a novel graph neural network architecture. The approach leverages a state-of-the-art methodology designed to handle sparse and irregular environmental datasets and represent complex spatial and temporal relationships in weather data, enabling accurate interpolation and prediction even in regions with limited observational coverage. By building on these principles, the WARN system will optimize sparse data initialization for weather prediction models, enhancing the accuracy, stability, and reliability of forecasts in challenging operational environments. In Phase I, Intellisense will study technical limitations of initializing MLWP models with sparse observations and formulating innovative concepts to overcome those challenges. Intellisense will perform a background study of both data assimilation and state-of-the-art forecast analysis methods that will be required to motivate and inform how the proposed effort will address gaps in current processes. We will develop a theory and/or simple method to demonstrate the feasibility of the initialization methods to produce robust analyses, stable forecasts, alignment with Navy-specific needs, and identifying suitable downstream applications. In Phase II, Intellisense plans to develop, demonstrate, and validate an end-to-end prototype MLWP software suite using Phase I results, focusing on novel data ingestion, initialization, and analysis. The toolset will handle diverse observational data, manage data discontinuities, enable restarts with degraded information, and include robust validation methods. Multiple demonstrations will be conducted in coordination with field testing, with regular reporting and program participation.