YIP Using Deep Learning to Improve Geostationary Satellite Remote Sensing of Ocean Precipitation
Abstract
Precipitation plays a key role in understanding the complex atmospheric processes, air-sea interactions, and their variability and impacts on atmospheric prediction. Accurate precipitation measurements or estimates are also instrumental in driving various scientific and operational applications ranging from long-term weather, climate, and water resources studies to real-time monitoring of natural disasters such as severe convective storms. Reliable quantitative precipitation estimation and short-term prediction over oceans are also vital for Naval research and maritime operations such as aviation and real-time mission planning. However, obtaining accurate precipitation estimates over oceans at high spatial and temporal resolutions remains a formidable challenge due to the limited coverage and sampling capabilities of existing sensor such as rain gauges, weather radars, and low Earth orbit satellites. In contrast, the latest generation of the geostationary operational environmental satellite, i.e., GOES-R series, has unique advantages in continuously monitoring atmospheric conditions at large scales over land and oceans at a high spatiotemporal resolution, although the conventional parametric approaches for GOES-based precipitation retrievals are insufficient to represent the complex precipitation distribution and variability.The goal of this project is to revolutionize geostationary satellite remote sensing of ocean precipitation using artificial intelligence (AI) and deep learning techniques. Specific research objectives are to 1) Design a transfer learning framework for geostationary satellite retrievals of ocean precipitation; 2) Develop and implement a deep learning model for ocean precipitation nowcasting; 3) Investigate the explainability and generalization capability of the developed deep learning models to understand the controlling factors on ocean storm initiation, growth, and decay. These research objectives will be accomplished through data collection, the development of physics-guided, data- and model-centric deep learning approaches, as well as academic activities that involve research communication, networking, and AI powered knowledge discovery. The outcomes from this project can improve the understanding of marine meteorology and ocean precipitation processes and their variability and dynamics. The high-resolution, high-performance precipitation estimation and nowcasting products can serve as a unique resource to support real-time Naval operations and mission planning.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 14, 2024
- Source ID
- N000142512025
Entities
People
- Haonan Chen
Organizations
- Colorado State University
- Office of Naval Research
- United States Navy