Predicting 3D Atmospheric Structure from Geostationary Satellites

The growing volume and quality of Earth observations offer improvements in data assimilation and weather forecasting that can only be fully realized by advances in AI. 3D atmospheric profiles, including temperature, humidity, and winds, are key missing observables in NASAs Earth Observation System (EOS) and presents large initial condition uncertainties in numerical weather prediction (NWP).Observations from radiosondes and microwave/infrared sounders as well as derived products like atmospheric motion vectors (AMVs) provide key inputs to data assimilation (DA). However, these observations are sparse and the spatio-temporal resolution of modern global DA systems have plateaued largely due to computational requirements. This stands in contrast with advances in artificial intelligence (AI) and computer vision (CV) that enable scalable and data efficient processing with the ability to synthetically generate variables not directly observed. Our work aims to fill EOS gaps and provide an alternative to traditional DA utilizing high-temporal resolution GOES-16/17 geostationary (GEO) satellites operated by NOAA/NASA and radiosonde observations with a probabilistic generative modeling approach. Variational autoencoders (VAEs) are used to independently compress GEO infrared bands and radiosonde profiles into latent representations.This enables us to learn a low-dimensional function between the latent representations to reconstruct temperature and humidity profiles on a pixel-wise basis. Our WindFlow model is applied to track the movement of humidity across sequences of frames to produce wind speed and direction.Lastly, a Neural Ordinary Difference Equation model is used to post-process the derived as a novel approach data-driven DA. The output of this proposal will include Zeus-Analysis, a novel 3D atmospheric dataset, with comprehensive evaluation and user access development through an application programming interface. The growing volume and quality of Earth observations offer improvements in data assimilation and weather forecasting that can only be fully realized by advances in AI. Earth observations from space are growing exponentially as investments from the public and private sectors are leading to technological advancements.  The wide variety of sensors (optical, hyperspectral, radar, sounders, Lidar, SAR, etc.) provide valuable information for environmental monitoring and forecasting. During Phase I of our project, we developed an AI model that generates atmospheric temperature, humidity, and winds at 18 vertical levels with a 2km, 10-minute spatio-temporal resolution dependent only on geostationary satellite imagery. This Phase II proposal aims to continuing developing our AI system with increased performance and efficiency as well as expanding into commercial energy and aviation markets. Spatially and temporally consistent 3D atmospheric structure data can both initialize and potentially replace current operational DA. This project aims to deliver a machine learning technology producing a real-time 3D atmospheric data product and demonstrated with a web-interface. 3D atmospheric profiles will be extracted from global geostationary satellite imagery by learning vertical profiles from radiosonde observations. Specific objectives include: (1) Development of an end-to-end data and inference pipeline, (2) AI modeling advancements in generative learning and neural ordinary difference equations, (3) Evaluation and comparisons against competing datasets, (4) Case study demonstrations, and (5) Application programming interface for data access.  These objectives enable a complete pipeline from raw observations and training data to customer access and technology transfer. An internal database will be developed to help scale evaluation and efficiency identify extreme events.

Applications of 3D atmospheric profiles are numerous throughout NASA Earth science research and development. Assimilation of our data with systems operated at the NASA's Global Modeling and Assimilation Office (GMAO) has the potential to improve analysis and forecast products, including short-term and sub-seasonal. Dense atmospheric winds will also have implications to wildfire monitoring and subsequent air quality issues.   Commercially the developed technology has applications to renewable energy, aviation, and finance/insurance. Renewable energy markets are largely powered by weather conditions and must be accurately estimated for stable operation of the power grid. Forecasts help prevent flight diversions and can cause dozens of downstream flight delays. 

This is early/mid-stage (TRL 4) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.