CISESS Passive Microwave Emulator Seed Grant Results

(a) GMI observed PMW at 14:40 UTC on 1 February 2020; (a) emulated PMW data; and (c) estimated variance of emulated PMW data.

Emulating Satellite Passive Microwave Brightness Temperature from the GOES Advanced Baseline Imager

CISESS Scientists Veljko Petkovic and Malarvizhi Arulraj, with the help of PhD student, Vesta Gorooh, completed their CISESS Seed Grant designed to determine whether it is feasible for one satellite instrument to provide both high sampling (returning to a location frequently) and information rich content (large amounts of data from each location). Currently, these two goals have been considered a trade off, with some sensors providing high sampling (up to every 30 seconds) and low information content and other sensors offering high information content but low sampling (up to every two weeks).

The instrument they selected to work on for this project was the GOES-R Advanced Baseline Imager (ABI). The ABI is the primary instrument on the GOES-R Series spacecraft for imaging Earth's weather, oceans, and environment. ABI views Earth with 16 different spectral bands, including two visible channels, four near-infrared channels, and ten infrared channels. A geostationare satellite has very high sampling because it remains above the same location. The goal of this project was to achieve highly accurate information on clouds by approximating of passive microwave (PMW) observations. PMW cloud data are currently measured by satellites with extensively long revisit times and are not an ABI product.

Since the ABI does not have a microwave channel, the PMW data was emulated from the measured radiances from ABI using machine learning methods. The ML was trained on radiance data from ABI and the Global Precipitation Measurement (GPM) Microwave Imager (GMI). A Bayesian Deep Neural Network was used to calculate data variance. The end result was to significantly expand the PMW data field compared to the swath of GMI observations.

Emulated PMW observations can be used for hurricane tracking applications, data assimilation and synthetic data retrievals. This project will continue with additional funding from the NOAA High Performance Computing and Communications Program and the UMD Grand Challenges Programs. Petkovic and Arulraj plan to expand the emulated PMW field beyond the ocean surface.

« Back

close (X)