记者10日从中国科学院空天信息创新研究院（以下简称空天院）获悉，该院曾江源研究员团队提出 了一种融合机器学习与插值方法的新型技术框架，有效解决了全球卫星土壤水分产品中常见的大范围数 据缺失问题，显著提高了数据的完整性和实用性。相关研究成果发表于《环境遥感》杂志。 作为反映地球生态健康状况的核心指标，土壤水分对农业灌溉、干旱预警、气候变化分析等具有重 要价值。目前，全球土壤水分数据主要依赖卫星遥感获取，然而受卫星轨道、地表复杂地形、人为信号 干扰等多种因素影响，原始数据常出现大量缺失，限制了其在实际科研与应用中的使用效果。 "当前填补缺失数据主要有两类方法：一类是传统插值法，依赖已知数据推测缺失区域，适用于小 范围缺失，但在大片空白区容易失效；另一类是基于大数据分析的机器学习方法，能够通过分析全球数 据，寻找土壤水分与降雨、植被等因素间的联系来进行预测，但结果容易趋向'平均'，难以准确反映特 别干旱或湿润地区的真实情况。"曾江源介绍。 针对上述问题，研究团队创新采用"优势互补"的思路，将两类方法深度融合。他们运用"堆叠"异质 集成技术，先分别利用插值和机器学习生成初步填补结果，再通过智能算法优化整合，形成同时兼 ...

Core Viewpoint - The article highlights significant advancements in the precise monitoring and accounting of carbon emissions from large coal-fired power plants, which are crucial for achieving global carbon neutrality goals [1][2]. Group 1: Research Breakthroughs - The Chinese Academy of Sciences has achieved a breakthrough in remote sensing and carbon emission estimation, enabling high-precision dynamic quantification and mapping of CO2 emissions from large coal-fired power plants [2]. - The research team developed a new satellite remote sensing approach, marking the first time high-precision dynamic quantification of CO2 emissions from large coal-fired power plants has been realized internationally [2][4]. Group 2: Importance of Carbon Emission Monitoring - Coal-fired power plants account for approximately 50% of global carbon emissions from fossil fuel combustion, making accurate carbon accounting essential for global carbon assessments and the electricity sector [3]. - Traditional methods of calculating emissions rely heavily on self-reported data from power plants, which can lead to discrepancies and lack of comparability due to the absence of unified international accounting standards [3]. Group 3: Methodological Innovations - The research team introduced an innovative optimization algorithm that significantly enhances the efficiency of identifying background carbon emission levels and improves the accuracy of smoke plume trajectory inversion [4]. - The study successfully quantified CO2 emissions from 14 large coal-fired power plants, with emissions ranging from 21.54 thousand tons to 82.3 thousand tons per day, demonstrating a marked improvement in inversion accuracy [4].