Estimating spatial and temporal variation in ocean surface pCO2 in the Gulf of Mexico using remote sensing and machine learning techniques

• DOI: 10.1016/j.scitotenv.2020.140965
• Publication Date: 2020-07-19T05:30:41Z
• Authors: Zhiyi Fu; Linshu Hu; Zhende Chen; Feng Zhang; Zhou Shi; Bifeng Hu; Zhenhong Du; Renyi Liu
• Keywords: [SDE] Environmental Sciences; Gulf of Mexico; 13. Climate action; [SDE]Environmental Sciences; Surface pCO2; 14. Life underwater; Remote sensing; 551; Data mining; 01 natural sciences; 0105 earth and related environmental sciences

Research on the carbon cycle of coastal marine systems has been wide concern recently. Accurate knowledge temporal and spatial distributions sea-surface partial pressure (pCO2) can reflect seasonal heterogeneity CO2 flux is, therefore, essential for quantifying ocean's role in cycling. However, it is difficult to use one model estimate pCO2 determine its controlling variables an entire region due prominent spatiotemporal areas. Cubist a commonly-used zoning; thus, be applied estimation regional analysis Gulf Mexico (GOM). A cubist integrated with satellite images was used here GOM, river-dominated area, using products, including chlorophyll-a concentration (Chl-a), temperature (SST) salinity (SSS), diffuse attenuation coefficient at 490 nm (Kd-490). The based semi-mechanistic high-accuracy advantages machine learning methods. overall performance showed root mean square error (RMSE) 8.42 μatm determination (R2) 0.87. Based environmental factors, GOM area divided into 6 sub-regions, consisting estuaries, near-shores, open seas, reflecting gradient distribution pCO2. Factor importance correlation analyses that salinity, chlorophyll-a, are main pCO2, corresponding both biological physical effects. Seasonal changes were also analyzed explained by variables. Therefore, considering high accuracy interpretability, cubist-based ideal method analysis.