NSF Research Projects
Precipitation Estimation from Multi-Source Information using Advanced Machine Learning
Climate variability and climate change are major sources of uncertainty in freshwater resource planning and management. The extreme precipitation events, flooding by excessive precipitation and drought of long-time scarce precipitation become more frequent and severe. The need for accurate precipitation observations is apparent. Recent innovation in Machine Learning algorithms, along with feature detection and extraction techniques, have extended the capability to harvest critical knowledge and information essential to cloud-precipitation systems from the vast amount of available data. Our goal with this study is to develop and integrate the state-of-the-art computational science methods and Machine Learning techniques to improve the performance and accuracy of global precipitation estimation.
The objectives for this project are to:
- Adapt and improve Geostationary Earth Orbit (GEO) and Low Earth Orbit (LEO) satellite-based precipitation estimation algorithms using state-of-the-art machine learning algorithms that can handle multi-source/multi-dimensional information (e.g. physical conditions related to precipitation generation and ambient meteorological fields).
- Develop a cyber-enabled data driven modeling system that uses the vast earth and environmental observational data to estimate precipitation (the most important element of the global hydrologic cycle), critical for accurate water balance studies and sustainability of freshwater resources.
- Harness computer vision techniques for feature characterization and recognition of meteorological fields and/or storm-system physical features that can help accurately estimate precipitation globally.
- Assess the representation and computational challenges and limitations of current machine learning algorithms when applied to the vast amounts of earth and environmental data required for accurate precipitation estimation.