能量收支平衡在气候变化中起着关键作用。能量平衡可以在大气顶和地表分 别进行估计，大气顶辐射通量可以被卫星直接观测，其中卫星云和地球辐射能量 系统(CERES，Clouds and Earths Radiant Energy System)的观测结果被国际所广泛 接受，地表通量则需要卫星获取关键参数，再利用经验模型或者物理模型进行反 演。地表通量估算的差异较大，过去的研究往往需要人为调整某些通量才能使能 量闭合。本文通过卫星在大气顶的直接观测评估了最新大气再分析资料在大气顶 的量化表现，基于地基观测筛选出地表辐射通量估算的最优产品，并利用最新的 全球降水测量计划(GPM，Global Precipitation Measurement)新一代卫星降水产品 估计地表蒸散和潜热通量，得到一套无需人为调整的闭合方案。 论文主要结论如下： （1）再分析资料对大气顶反射太阳辐射和出射长波辐射都存在较大高估，最 为显著的高估发生在热带海洋上。大气顶净辐射方面，仅第五代欧洲中期天气预 报中心大气再分析(ERA5，ECMWF Reanalysis V5) 的模拟与观测(1.1 W m−2 )吻 合较好。再分析资料能较好捕捉大气顶除反射太阳辐射外的辐射通量年际变率， ERA5 最接近观测。相对于全球基准地表辐射观测网络(BSRN，Baseline Surface Radiation Network)观测结果，地表入射太阳辐射和向下长波辐射的最优估计分别 是 CERES 和 ERA5 辐射产品，分别是 187 W m−2 和 342 W m−2 。相对于最新的 GPM 降水产品，原有观测和降水融合产品 GPCP(Global Precipitation Climatology Project)全球降水量少 8.4%，主要原因是低估海洋降水。根据 GPM 全球降水量 估算的全球地表潜热通量为 83 W m−2 ，陆地和海洋分别为 41 和 101 W m−2 。 （2）进一步评估 CMIP6 模式在 2000-2014 年量化全球能量平衡的表现，相 对于BSRN地基观测，以往被低估的向下长波辐射在CMIP6却出现了略微高估。 CMIP6 模式集合平均的大气顶辐射通量与观测有很好的一致性，可能是由于模 式调整的原因；地表入射太阳辐射几乎与参考值相当，改善了之前高估的情况。 模式间的不确定范围依旧很大，几乎所有通量在陆地尺度都具有更大离散度。

The energy balance plays a crucial role in climate change. The energy balance can be estimated separately at the top of the atmosphere (TOA) and the Earth's surface. The TOA radiative flux can be directly observed by satellites, with the observations from the Clouds and Earth's Radiant Energy System (CERES) widely accepted internationally. Estimating surface flux requires satellite-derived key parameters and the use of empirical or physical models for inversion. There are significant differences in surface flux estimations, and previous studies often required manual adjustments to certain fluxes to achieve energy closure. This study evaluates the quantitative performance of the latest atmospheric reanalysis data at the TOA through direct satellite observations. It selects the optimal product for estimating surface radiation flux based on ground-based observations and uses the latest satellite precipitation product from the Global Precipitation Measurement (GPM) to estimate surface latent heat and evapotranspiration fluxes, providing a closure solution that does not require manual adjustments. (1) Reanalysis data tend to overestimate both reflected solar radiation and outgoing longwave radiation at TOA, with the most significant overestimation occurring over tropical oceans. Regarding net radiation at the TOA, only the fifth-generation European Centre for Medium-Range Weather Forecasts atmospheric reanalysis (ERA5) simulation shows good agreement with observations (1.1 W m−2 ). Reanalysis data can capture the interannual variability of the TOA radiative fluxes, excluding reflected solar radiation, with ERA5 being the closest to observations. Compared to the observations from the Baseline Surface Radiation Network (BSRN), the optimal estimates for surface incoming solar radiation and downward longwave radiation are the CERES and ERA5 radiation products, respectively, at 187 W m−2 and 342 W m−2 . Compared to the latest GPM precipitation product, the original observation and precipitation fusion product Global Precipitation Climatology Project (GPCP) has 8.4% less global precipitation, primarily due to the underestimation of oceanic precipitation. According to the global surface latent heat flux estimated based on GPM global precipitation, it is 83 W m−2 , with land and ocean contributing 41 W m−2 and 101 W m−2 , respectively. (2) Further evaluation of the performance of CMIP6 models in quantifying global energy balance from 2000 to 2014 reveals a slight overestimation of downward longwave radiation compared to BSRN ground-based observations, contrary to the previous underestimation. The ensemble mean of CMIP6 models shows good consistency with observations in terms of radiative fluxes at the TOA, possibly due to model adjustments. Surface incoming solar radiation is nearly equivalent to the reference values, improving the previous overestimation. However, there is still a large range of uncertainty among the models, with almost all fluxes exhibiting greater variability at the land scale.