Harnessing Machine Learning and Principal Component Analysis for Atmospheric and Glint Correction of Ocean Color Measurements Using Hyperspectral Satellite Observations Such As the Geostationary TEMPO (Tropospheric Emissions: Monitoring of POllution)

Retrievals of ocean color from space are important for understanding the ocean ecosystem and its biogeochemistry. The launch of atmospheric geostationary hyperspectral sensors such as NASA’s Tropospheric Emissions: Monitoring of Pollution (TEMPO) provides a unique opportunity to examine the diurnal variability of ocean color. TEMPO does not have the high spatial resolution or full spectral coverage of planned geostationary coastal ocean sensors such as the NASA Geosynchronous Littoral Imaging and Monitoring Radiometer (GLIMR) or the NOAA GeoXO Ocean Color instrument (OCX). However, its hourly measurements provide coverage of regions such as Lake Erie and the Gulf of Mexico at spatial scales of approximately 5 km. These data are useful for testing new algorithms and providing users with a glimpse of the temporal information that geostationary hyperspectral sensors will provide. We apply our newly developed machine learning-based atmospheric correction approach for ocean color retrievals to TEMPO data. We first decompose TEMPO hyperspectral radiances into principal components that represent spectral features associated with scattering and absorption in the atmosphere and ocean. The coefficients of the principal components are then used as input features to train a neural network to predict target ocean properties derived from collocated MODIS/VIIRS physically-based retrievals. The targets include chlorophyll concentration, remote sensing reflectance, and chlorophyll fluorescence line height. Because this machine learning approach relies on MODIS or VIIRS ocean color data for training, it can adjust for possible calibration errors or other artifacts in the hyperspectral TEMPO data. We have applied our approach using blue and ultraviolet observations from the Ozone Monitoring Instrument (OMI) and TROPOspheric Monitoring Instrument (TROPOMI) in Low Earth Orbit (LEO) to show that ocean color properties can be retrieved in less-than-ideal conditions such as lightly to moderately cloudy conditions as well as sun glint. Our approach then improves the spatial coverage of ocean color measurements. TEMPO provides an opportunity to advance this approach in two important ways: 1) from its geostationary orbit, it provides more frequent measurements; and 2) it provides measurements at green and red wavelengths that are important particularly for coastal waters but are not available from OMI and TROPOMI. Other advantages of our technique are that it can be applied early in the mission, and it has potential to rapidly produce coastal ocean products that are important for monitoring harmful algal blooms and other oceanic phenomena.