Navigating noise when comparing satellite and acoustic remote sensing data

Abstract

Using a novel combination of NASA satellite products and acoustic remote sensing measurements, we aim to examine the link between the surface expression and vertical structure of ocean productivity and biomass in the California Current System. Water column sonar data collected by the National Marine Fisheries Service for fisheries management have been archived at the NOAA National Centers for Environmental Information. These data are used to evaluate the variability of the marine biogeography. To identify patterns of acoustic reflectance of marine organisms, the data must first be free of noise and seafloor acoustic returns. However, removing the variety of noise present has proved to be a challenge due to need for manual tuning, a resource prohibitive step, and the size and complexity of the acoustic data. Collaboration with the University of Colorado’s Earth Lab has led to the development of machine learning models to automatically remove noise from this large dataset. An analysis of the fully automated, semi-manual, and machine learning quality control processes will be presented. The cleaned acoustic data are then compared to the interannual variability of the distribution of surface chlorophyll concentration and temperature from satellite ocean color measurements.