Space-Time Sea Surface pCO 2 Estimation in the North Atlantic Based on CatBoost

• Published in: Remote Sensing
• Main Authors: Hongwei Sun, Junyu He, Yihui Chen, Boyu Zhao
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI AG 2021
• Subjects: sea surface pCO 2; ocean color remote sensing; CatBoost algorithm; temporal and spatial distribution; Science; Q; North Atlantic; Ocean acidification
• Online Access: https://doi.org/10.3390/rs13142805; https://doaj.org/article/1d75dda80e494223a9eb1b1b84445ec4

Sea surface partial pressure of CO 2 (pCO 2 ) is a critical parameter in the quantification of air–sea CO 2 flux, which plays an important role in calculating the global carbon budget and ocean acidification. In this study, we used chlorophyll-a concentration (Chla), sea surface temperature (SST), dissolved and particulate detrital matter absorption coefficient (Adg), the diffuse attenuation coefficient of downwelling irradiance at 490 nm (Kd) and mixed layer depth (MLD) as input data for retrieving the sea surface pCO 2 in the North Atlantic based on a remote sensing empirical approach with the Categorical Boosting (CatBoost) algorithm. The results showed that the root mean square error (RMSE) is 8.25 μatm, the mean bias error (MAE) is 4.92 μatm and the coefficient of determination (R 2 ) can reach 0.946 in the validation set. Subsequently, the proposed algorithm was applied to the sea surface pCO 2 in the North Atlantic Ocean during 2003–2020. It can be found that the North Atlantic sea surface pCO 2 has a clear trend with latitude variations and have strong seasonal changes. Furthermore, through variance analysis and EOF (empirical orthogonal function) analysis, the sea surface pCO 2 in this area is mainly affected by sea temperature and salinity, while it can also be influenced by biological activities in some sub-regions.