Retrieving Chlorophyll Concentration from GOES-16 Using AI

Figure: Comparison of Chlorophyll a ([Chl-a]) derived from GOES-16 using deep learning with [Chl-a] derived from other sensor data. (A) GOES-16 ABI composite; (B) SNPP VIIRS product; (C) NOAA gap-filled product; and (D) OC-CCI datasets.

CISESS Scientist Guangming Zheng gave a presentation on January 21st, 2021 as part of the 2nd NOAA Workshop on Leveraging AI in Environmental Sciences. Zheng shared his recent work on retrieving chlorophyll a concentrations ([Chl-a]) from the Geostationary Operational Environmental Satellite R-series (GOES-R) Advanced Baseline Imager (ABI) using deep learning techniques.

Zheng demonstrated the proof-of-concept of using deep learning to retrieve [Chl-a] for the open oceans from GOES-16 ABI data, which was previously considered unfit for ocean color applications owing to the lack of a green band. The figure below, from Zheng’s slides, shows that the deep learning ABI composite (A) agrees well with the Suomi National Polar-Orbiting Partnership (SNPP) Visible/Infrared Imager Radiometer Suite (VIIRS) product (B) and the Ocean Color-Climate Change Initiative (OC-CCI) datasets (D).The ABI composite shows more detail than the NOAA gap-filled product (C).

Zheng summarized that the deep learning model did well at frontal feature detection even though the input radiance data were not processed with any atmospheric correction. This suggests that deep learning can recognize subtle patterns barely perceptible to the human eye. Deep learning is a powerful tool to take into account a diverse set of input variables that are difficult for humans to handle simultaneously.

The presentation slides and a recording of his talk are available on the workshop website:

Zheng, Guangming, Retrieving Chlorophyll concentration from GOES-16 ABI using Deep LearningTechniques, 2nd NOAA Workshop on Leveraging AI in Environmental Sciences (Virtual, 21 Jan 2021),​20AI/presentations/202101/20210121_Zheng.pdf


« Back

close (X)