IOCAS全球大洋二氧化碳分压格点数据v2025

在通过逐步前反馈神经网络拟合法构建的2022版基础上，使用集成学习结构的多个前反馈神经网络组合优化了数据构建算法，将数据时空范围扩展到从1992年1月到2024年12月每月的全球大洋表层海水0.25°×0.25°。与观测数据集SOCATv2025间的平均误差为11.32 μatm，标准误差为17.80 μatm.

Based on the 2022 version constructed via stepwise feedforward neural network fitting method, the data construction algorithm was optimized by using an ensemble learning structure composed of multiple feedforward neural networks. The spatio-temporal scope of the dataset was expanded to global ocean surface seawater data at a 0.25°×0.25° resolution on a monthly basis, covering the period from January 1992 to December 2024. The average error against the observational dataset SOCATv2025 is 11.32 μatm, with a standard error of 17.80 μatm.