全球1km分辨率大气二氧化碳浓度数据集（2003-2023）

持续增加的人为CO₂排放导致了全球变暖和气候变化，进而引发了全球范围的重大环境、经济和健康损失，基于卫星遥感数据准确连续地监测大气CO₂变化对于理解全球碳循环、评估碳源和碳汇的分布以及制定有效的减排政策至关重要。大气CO2柱浓度（XCO2）指从地表到大气顶层干燥空气柱中CO2的平均体积比，是用来表征大气中CO2分子含量的物理量。当前已公开发表的全球无缝XCO2产品存在无法同时提供长时间跨度和高时空分辨率的问题，限制了其更为广泛的科学应用。本数据集基于来自SCIAMACHY、GOSAT 和 OCO-2 三颗卫星/传感器的XCO2观测数据进行二次研发，以卫星XCO2观测数据为训练标签，与 CO₂ 排放、吸收和传输相关的多源因素为解释变量，利用整合了U-Net网络和ConvLSTM网络的深度学习算法构建预测模型，生成了国际首套2003-2023年全球时空连续1公里分辨率逐日XCO2数据集。经全球27个TCCON地面观测站点的验证，结果表明该产品具有较好的精度（决定系数R2为0.989，均方根误差RMSE为1.021ppm）。本数据集为深化对全球碳循环的理解、评估减排政策以及应对气候变化挑战提供了重要的基础数据。

Growing anthropogenic CO₂ emissions have caused global warming and climate change, leading to severe environmental, economic, and health losses worldwide. Accurate and continuous monitoring of atmospheric CO₂ variations using satellite remote sensing data is critical for understanding the global carbon cycle, evaluating the distribution of carbon sources and sinks, and formulating effective emission reduction policies. Atmospheric CO₂ column abundance (XCO₂) refers to the average volume mixing ratio of CO₂ in a dry air column from the Earth's surface to the top of the atmosphere, which is a physical quantity used to characterize the content of CO₂ molecules in the atmosphere. Currently, publicly available global seamless XCO₂ products suffer from the limitation that they cannot simultaneously provide long-time-span and high spatiotemporal resolution data, which restricts their broader scientific applications. This dataset was secondarily developed based on XCO₂ observation data from three satellites/sensors: SCIAMACHY, GOSAT, and OCO-2. Taking satellite XCO₂ observation data as training labels and multi-source factors related to CO₂ emission, absorption, and transport as explanatory variables, we constructed a predictive model using a deep learning algorithm integrating U-Net and ConvLSTM networks, and generated the world's first globally spatiotemporally continuous daily XCO₂ dataset with 1 km resolution spanning 2003 to 2023. Validated against data from 27 global TCCON ground-based observation stations, the results show that this product has excellent accuracy, with a coefficient of determination R² of 0.989 and a root mean square error (RMSE) of 1.021 ppm. This dataset provides important foundational data for deepening the understanding of the global carbon cycle, evaluating emission reduction policies, and addressing the challenges of climate change.

搜集汇总

背景与挑战

背景概述

该数据集是全球首套2003-2023年时空连续的1公里分辨率逐日大气二氧化碳柱浓度（XCO2）产品，基于多卫星观测数据和深度学习算法生成，填补了高时空分辨率与长时间跨度兼顾的空白。数据精度高（R²=0.989，RMSE=1.021 ppm），为全球碳循环研究、减排政策评估和气候变化应对提供了关键基础数据。