ACIX-Aqua: A global assessment of atmospheric correction methods for Landsat-8 and Sentinel-2 over lakes, rivers, and coastal waters

Atmospheric correction over inland and coastal waters is one of the major remaining challenges in aquatic remote sensing, often hindering the quantitative retrieval of biogeochemical variables and analysis of their spatial and temporal variability within aquatic environments. The Atmospheric Correction Intercomparison Exercise (ACIX-Aqua), a joint NASA – ESA activity, was initiated to enable a thorough evaluation of eight state-of-the-art atmospheric correction (AC) processors available for Landsat-8 and Sentinel-2 data processing. Over 1000 radiometric matchups from both freshwaters (rivers, lakes, reservoirs) and coastal waters were utilized to examine the quality of derived aquatic reflectances (ρ̂w). This dataset originated from two sources: Data gathered from the international scientific community (henceforth called Community Validation Database, CVD), which captured predominantly inland water observations, and the Ocean Color component of AERONET measurements (AERONET-OC), representing primarily coastal ocean environments. This volume of data permitted the evaluation of the AC processors individually (using all the matchups) and comparatively (across seven different Optical Water Types, OWTs) using common matchups. We found that the performance of the AC processors differed for CVD and AERONET-OC matchups, likely reflecting inherent variability in aquatic and atmospheric properties between the two datasets. For the former, the median errors in ρ̂w560 and ρ̂w664 were found to range from 20 to 30% for best-performing processors. Using the AERONET-OC matchups, our performance assessments showed that median errors within the 15–30% range in these spectral bands may be achieved. The largest uncertainties were associated with the blue bands (25 to 60%) for best-performing processors considering both CVD and AERONET-OC assessments. We further assessed uncertainty propagation to the downstream products such as near-surface concentration of chlorophyll-a (Chla) and Total Suspended Solids (TSS). Using satellite matchups from the CVD along with in situ Chla and TSS, we found that 20–30% uncertainties in ρ̂w490≤λ≤743nm yielded 25–70% uncertainties in derived Chla and TSS products for top-performing AC processors. We summarize our results using performance matrices guiding the satellite user community through the OWT-specific relative performance of AC processors. Our analysis stresses the need for better representation of aerosols, particularly absorbing ones, and improvements in corrections for sky- (or sun-) glint and adjacency effects, in order to achieve higher quality downstream products in freshwater and coastal ecosystems.