https://doi.org/10.1109/TGRS.2026.3676135

Shahvaran, A. R., Chegoonian, A. M., Kheyrollah Pour, H., Reshadi, M. A. M., & Van Cappellen, P. (2026). Multisource assessment of machine learning models and atmospheric correction products for chlorophyll-a retrieval in meso-eutrophic waters: Application to Hamilton Harbour, an area of concern in Lake Ontario, Canada. IEEE Transactions on Geoscience and Remote Sensing, 64, Article 4205318, 1–18. https://doi.org/10.1109/TGRS.2026.3676135

Freely available multispectral satellite imagery enables frequent monitoring of lake Chlorophyll-a (Chl-a), but retrieval accuracy depends strongly on both atmospheric correction (AC) and model choice. Using more than 600 pixel–in situ matchups (2000–2023) from more than 240 Landsat 5, 7, and 8 and Sentinel-2 images in Western Lake Ontario (±4-day window; 3×3 pixel means), we benchmarked four radiometric products (Level-1, Level-2, DOS, ACOLITE) with four machine-learning (ML) models: least absolute shrinkage and selection operator (LASSO), mixture density network (MDN), support vector regression (SVR), XGBoost). ACOLITE paired with XGBoost was consistently best on the held-out test sets: root-mean-squared logarithmic error (RMSLE) = 0.34 (L5), 0.43 (L7), 0.56 (L8), and 0.42 (S2), with slopes ~0.7, 0.6, 0.4, and 0.7, respectively. Minimally corrected products (Level-1, DOS) performed poorly (test RMSLE 0.6; slopes 0.1). For Landsat 8, the provisional aquatic reflectance (AR) generalized better than LaSRC (test slope ≈0.6 versus 0.1; RMSLE 0.59 versus 0.68), but ACOLITE remained strongest overall. Against semi-empirical indices, ML—especially XGBoost—reduced errors and produced fits closer to unity across sensors. We then mapped seasonal Chl-a in Hamilton Harbour (2000–2024) to showcase model application. Monthly means rose from late spring to midsummer (May 13.2 μ g L−1; June 11.5; July 19.2; August 20.2), with persistent nearshore hotspots near the Royal Botanical Gardens (RBGs) and Windermere Basin. Mann–Kendall trend analyses during bloom months indicated little monotonic change: mean Sen’s slopes were near zero ( | mean |≤2×10−4μ g L−1 day−1) and <15% of pixels showed significant trends per month. Thus, interannual variability dominates over simple linear trends. Overall, harmonizing radiometry with ACOLITE and deploying XGBoost provides a practical, cross-sensor pathway for robust, long-term Chl-a mapping, demonstrated here for a meso-eutrophic embayment of Western Lake Ontario.

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11449305