A Bayesian approach for remote sensing of chlorophyll-a and associated retrieval uncertainty in oligotrophic and mesotrophic lakes

Satellite remote sensing of chlorophyll-a concentration (chla) in oligotrophic and mesotrophic lakes faces uncertainties from sources such as atmospheric correction, complex inherent optical property compositions, imperfect algorithmic retrieval. To improve chla estimation oligo- lakes, we developed Bayesian probabilistic neural networks (BNNs) for the Sentinel-3 Ocean Land Cover Instrument (OLCI) Sentinel-2 MultiSpectral Imager (MSI). The BNNs were built using an situ dataset water bodies (1755 observations 178 systems; median chla: 5.11 mg m−3, standard deviation: 10.76 m−3) provide a per-pixel uncertainty percentage associated with retrieved chla. Shifts systems into eutrophic regime, characterised by higher biomass levels, are widespread. account phytoplankton fluctuation, set (167 31 systems) included this study (maximum 68 m−3). evaluated through five assessments including single day time series match-ups OLCI MSI. BNN accuracy gains >25% MSI >15% achieved when compared to reference algorithms waters (chla ≤ 8 In comparison algorithms, decreased trophic levels increased. measure quality provided estimate, calculated prediction interval coverage probability (PICP), Sharpness mean absolute calibration difference (MACD) metrics. estimate values most (PICP ≥ 75%) across different performance assessments. Further analysis showed that was not constantly well-calibrated evaluation strategies (Sharpness 1.7–6, MACD 0.04–0.25). used test two improvement strategies: 1) identifying filtering uncertain estimates scene-specific thresholds, 2) selecting accurate prior correction algorithm per individual satellite observation retain lowest uncertainty. Both increased result demonstrated significance estimation. This serves research on machine learning visualisation retrieval develop harmonised products small large lakes.