A Statistical Analysis of Residual Errors in Satellite Remote Sensing Reflectance Data From Oligotrophic Open Oceans

We present a statistical analysis of how residual error from satellite remote sensing reflectance (
$R_{\mathrm {rs}}$
) depends on the environmental factors. Our results indicate that image coverage affects the data quality, and the residual error in high-quality data correlates significantly with the residual error in data heavily contaminated with stray light. Due to imperfectly corrected bidirectional reflectance, we found many residual error anomalies in the covarying relationship between residual error and illumination–observation geometry even though high-quality
$R_{\mathrm {rs}}$
data appeared more homogeneous than low-quality data. Additionally, we found underestimates of
$R_{\mathrm {rs}}$
in very oligotrophic tropical waters, and these residual errors positively covaried with sensor zenith angles at the edge of the scans because of strong atmospheric multi-scattering effects. Furthermore, we found that due to their strong correlation with
$R_{\mathrm {rs}}$
, the spectral relationship of residual errors should be dynamically reinitialized with the image, which is key to an inherent optical properties (IOPs) data processing system (such as IDAS) removing the residual errors from the
$R_{\mathrm {rs}}$
data. Doing so, and applying the IDAS algorithm, stray-light-contaminated
$R_{\mathrm {rs}}$
was recovered, obtaining comparable results to those achieved considering high-quality measurements. This permits to relax the corresponding quality check for generation of the Level-3 global area coverage
$R_{\mathrm {rs}}$
, largely improving the satellite spatiotemporal coverage and consequently the product accuracy.