NIES-ML3 ensemble product of surface ocean CO2 concentrations and air-sea CO2 fluxes reconstructed by using three machine learning models with new CO2 trends

This product is the result of our new study on ocean CO2 trends using Random Forest, Gradient Boost Machine, and Feedforward Neural Network. Using the time-dependent trends for ocean CO2 reconstruction substantially reduced the error of using a constant trend and therefore improved the oceanic sink estimate.

本数据集为我们基于随机森林（Random Forest）、梯度提升机（Gradient Boost Machine）与前馈神经网络（Feedforward Neural Network）开展的海洋二氧化碳（CO₂）趋势新研究成果。采用随时间变化的趋势进行海洋二氧化碳重建，可大幅降低使用恒定趋势时的误差，进而提升海洋碳汇（oceanic sink）的估算精度。