River runoff is estimated as a water budget residual using Gravity Recovery and Climate Experiment (GRACE) terrestrial water storage time series, ERA5 reanalysis data, and precipitation observations for January 2003 through December 2015 for the Obidos upstream drainage basin and for the entire Amazon basin. Estimated runoff based on the water budget agrees remarkably well with in situ gauge observations at Obidos, especially at seasonal time scales, with nearly perfect phase agreement but slightly larger seasonal amplitude. The discrepancy in the seasonal amplitude may be attributed to underestimation of river gauge runoff during the wet season when water overflows the riverbanks. The ERA5 model appears to overestimate long‐term mean evapotranspiration in the Amazon by ~2 cm/month based on comparisons with precipitation and runoff observations. Using precipitation data based on satellites and gauge observations relative to gauge observations alone improved agreement between water budget runoff estimates and in situ runoff observations. Seasonal variations in ERA5‐simulated runoff are about twice as large as those from in situ observations and show a large phase lag as well. Water budget‐based runoff for the entire Amazon (~7,200 km³ averaged over the 13‐year period) is significantly larger than observed runoff (~5,700 km³) at Obidos and notably larger than previous estimates for the entire Amazon. These differences may be partly related to submarine runoff from the Amazon basin that cannot be captured by surface gauges.

中文翻译：

---

GRACE重力卫星，气候模型和原位观测的流域径流估算：以亚马逊河流域为例

使用重力恢复和气候实验（GRACE）地面水存储时间序列，ERA5再分析数据以及Obidos上游流域和整个亚马逊盆地2003年1月至2015年12月的降水观测值，将河流径流估算为水预算残差。根据水预算估算的径流量与奥比都斯的实地观测资料非常吻合，尤其是在季节性时标上，相位吻合度近乎完美，但季节性振幅却稍大。季节幅度的差异可能归因于在雨季当水溢出河岸时河道径流量的低估。根据与降水和径流观测值的比较，ERA5模型似乎将亚马逊地区的长期平均蒸散量高估了约2厘米/月。使用基于卫星和量具观测值的降水数据（相对于量具观测值），可以改善水预算径流估算值与实地径流观测值之间的一致性。ERA5模拟的径流的季节性变化约为原位观测值的两倍，并且也显示出较大的相位滞后。整个亚马逊地区的水预算径流（约7,200公里在13年期间的平均值³）显着大于在Obidos观测到的径流（〜5,700 km ³），并且显着大于先前整个亚马逊地区的估计值。这些差异可能与亚马逊流域的海底径流有关，而海面径流无法捕获这些差异。