A Machine Learning Approach to Estimate Surface Chlorophyll <i>a</i> Concentrations in Global Oceans From Satellite Measurements

Various approaches have been proposed to estimate surface ocean chlorophyll a concentrations (Chl, mg m -3 ) from spectral reflectance measured either in the field or space, each with its own strengths and limitations. Here, we develop machine learning approach reduce impact of noise improve algorithm performance at global scale for multiple satellite sensors. Among several candidates, support vector regression (SVR) was found yield best as gauged by statistical measures against field-measured Chl. While statistically SVR is slightly worse than empirical color index (CI) Hu et al. (2012) Chl <; 0.25 , applicability waters much extended, CIs 0.01-0.25 (about 75% oceans) 0.01-1 [about 96% oceans according Sea-viewing Wide Field-of-view Sensor (SeaWiFS) statistics]. Within this range, not only does show improved over traditional band-ratio OC x approaches, but leads reduced image cross-sensor consistency between SeaWiFS Moderate Resolution Spectroradiometer (MODIS)/Aqua MODIS/Aqua Visible Infrared Imaging Radiometer Suite (VIIRS). Furthermore, compared hybrid Ocean CI (OCI) currently used U.S. NASA default all mainstream sensors, avoids need merge two different algorithms intermediate (band subtraction band ratio x), thus may serve an alternative data processing.