Research Summaries

Back Deep Learning Uncertainties of GOES Advanced Baseline Imager Products Ingested by GeoIPS for Assimilation into Navy Global Models

Fiscal Year 2021
Division Graduate School of Operational & Information Sciences
Department Computer Science
Investigator(s) Orescanin, Marko
Sponsor Office of Naval Research (Navy)
Summary Visible and infrared radiances observed by geostationary orbiting platforms like GOES and Himawari are readily available and observe the same parts of Earth continuously, with data collected every 15 minutes across a full disk. In contrast, low Earth orbiters observe passive microwave (PMW) radiances at a single location much less frequently. Data assimilation systems, such as the Joint Effort for Data assimilation Integration (JEDI) system, ingest such satellite data to help constrain initial conditions in numerical models of the atmosphere and produce the best possible forecast given observations available and limitations in physical parameterization. Previous work, both by the current investigators and in existing literature demonstrate the ability to use machine learning (ML) to predict PMW brightness temperatures using infrared and visible radiances collected from geostationary orbit. The proposed work will develop a neural network to predict PMW brightness temperatures at frequencies that are currently assimilated by the NRL Atmospheric Variational Data Analysis System (NAVDAS). The machine learning framework developed will be novel in that it will also predict the uncertainty in the synthetic observations. The ML model produced will be integrated into the Geo-located Information Processing System (GeoIPS), giving that system the ability to automatically ingest required data and convert it to synthetic PMW data and uncertainties with high temporal resolution and large spatial coverage. The new products in machine-readable format will be accessible by JEDI. Finally, various experiments will be executed to quantify the benefit (in terms of forecast improvement) of assimilating synthetic PMW data into the Navy Environmental Prediction System using the NUMA Core (NEPTUNE) instead of or in addition to observed PMW data.
Keywords
Publications Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal
Data Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal