Bayesian atmospheric correction over land: Sentinel-2/MSI and Landsat 8/OLI

Abstract. Mitigating the impact of atmospheric effects on optical remote sensing data is critical for monitoring intrinsic land processes and developing Analysis Ready Data (ARD). This work develops an approach to this NERC NCEO medium resolution ARD Landsat 8 (L8) Sentinel 2 (S2) products, called Sensor Invariant Atmospheric Correction (SIAC). The contribution phrase solve that problem within a probabilistic (Bayesian) framework multispectral sensors S2/MSI L8/OLI provide per-pixel uncertainty estimates traceable from assumed top-of-atmosphere (TOA) measurement uncertainty, making progress towards important aspect CEOS target requirements. A set observational priori constraints are developed in SIAC constrain estimate coarse (500 m) aerosol thickness (AOT) total column water vapour (TCWV), along with associated uncertainty. then used (10–60 surface reflectance given 5 % TOA reflectance. MODIS MCD43 BRDF/Albedo product, giving constraint 500 m reflectance, Copernicus Atmosphere Monitoring Service (CAMS) operational forecasts AOT TCWV, providing state at core 40 km spatial resolution, correlation model. mapping scale between observations coarser achieved using calibrated effective point spread function MCD43. Efficient approximations (emulators) outputs 6S radiative transfer code parameters correction stage. demonstrated global S2 L8 images covering AERONET RadCalNet sites. retrievals show very high (correlation coefficient around 0.86, RMSE 0.07 both sensors), although small bias AOT. TCWV accurately retrieved over 0.96, <0.32 g cm−2). Comparisons situ measurements network provides accurate across entire spectrum, mismatches reference 0.01 0.02 units L8. For near-simultaneous acquisitions, there tight relationship 0.95 all common bands) sensors, negligible biases. Uncertainty assessed through discrepancy analysis found viable TCWV. they give conservative suggesting lower might be appropriate.