利用遥感数据反演海水CO2分压（pCO2）可获得高时空分辨率及长时间序列的观测数据，有助于减少全球碳收支估算的不确定性。目前大部分的海水pCO2遥感反演算法主要是基于pCO2和遥感参数（如温度、叶绿素等）之间的线性或者多元回归关系获得。但在复杂边缘海区域，pCO2的变化通常是由多种因子共同控制，很难通过遥感可获取的参数直接建立具有显著统计意义的多元回归模型。本项目研究人员基于海水碳化学和遥感科学的交叉研究，提出了基于控制机制分析的海水pCO2半解析概念框架（“Mechanistic-based Semi-Analytic-Algorithm” (MSAA-pCO2)）。算法思路为: 首先理清研究海区海水pCO2的主要控制机制，然后建立各种主控因子的定量化遥感模型，将不同控制因子引起的pCO2改变量进行叠加，最终获得总的海水pCO2。如何建立每种控制机制的解析或者半解析遥感量化模型是该算法的难点。通过CHOICE-I项目2009年8月走航数据验证，该模型具有较好的精度，算法可以很好的反映受陆源影响的近岸高pCO2，以及冲淡水内高浮游植物生产力造成的低pCO2，以及外海受温度热力学控制的pCO2变化。与传统的回归统计方法比较，MSAA-pCO2算法包含了更多的控制机制信息，反演值与观测值的偏差能在一定程度上反映其他未考虑/留意的控制过程变化；算法的进一步完善可以通过增加新的控制因子量化模型完成，无需对原有模型进行较大的改动。

    卫星遥感结果与变化：(a) CHOICE-C I夏季航次（16–31 August. 2009）遥感反演的pCO₂分布，灰线为航次航迹 (b) 遥感反演结果与现场测量的比较

Abstract: While satellite remote sensing has become a very useful tool contributing to assessments of sea surface partial pressure of carbon dioxide (CO2) that subsequently allow quantification of air-sea CO2 flux, the application of empirical approaches in coastal oceans has proven challenging owing to the interaction of multiple controlling factors. We propose a “mechanistic semi-analytic algorithm” (MeSAA) to estimate sea surface CO2 in river-dominated coastal oceans using satellite data. Observed CO2 can be analytically expressed as the sum of individual components controlled by major factors such as thermodynamics (or temperature), mixing, and biology. With marine carbonate system calculations, temperature and mixing effects can be predicted using thermodynamic principles and by assuming conservative two end-member mixing of total dissolved inorganic carbon and total alkalinity (e.g., the Changjiang River and Kuroshio water in the East China Sea, ECS). Next, an integral expression for CO2 drawdown due to biological effects can be parameterized using the chlorophyll  concentration We demonstrate the validity and applicability of the algorithm in the ECS during summertime. Sensitivity analysis shows that errors in empirical coefficients and three input satellite parameters (salinity, SST,  have limited influence on the algorithm, and satellite-derived CO2 is consistent with underway data, even though no in situ CO2 data from the ECS shelves was used to train the algorithm. Our algorithm has more physical and biogeochemical mechanistic meaning than empirical methods, and should be applicable to other similar systems.