Accurate assessment of global river flows and stores is critical for informing water management practices, but current estimates of global river flows exhibit substantial spread and estimates of river stores remain sparse. Estimates of river flows and stores are hampered by uncertainties in land runoff, an unobserved quantity that provides water input to rivers. Here we leverage global river flow observations and an ensemble of land surface models to generate a globally gauge-corrected monthly river flow and storage dataset. We estimate a global river storage mean (± monthly variability) of 2,246 ± 505 km³ and a global continental flow of 37,411 ± 7,816 km³ yr⁻¹. Our global river water storage time series demonstrates that flow wave residence time is a fundamental driver that can double or halve river water stores and their variability. We also reconcile the wide range in previous estimates of monthly variability in global river flows. We identify previously underappreciated freshwater sources to the ocean from the Maritime Continent (Indonesia, Malaysia and Papua New Guinea) amounting to 1.6 times the Congo River and illustrate our capability of detecting severe anthropogenic water withdrawals.

准确评估全球河流流量和蓄水量对于为水管理实践提供信息至关重要，但目前对全球河流流量的估计显示出很大的分散性，而对河流蓄水量的估计仍然很少。对河流流量和蓄水量的估计受到土地径流不确定性的影响，土地径流是向河流提供水量的未被观测到的量。在这里，我们利用全球河流流量观测和陆地表面模型集合来生成全球仪表校正的每月河流流量和存储数据集。我们估计全球河流蓄水量平均值（±每月变化）为2,246 ± 505 km ³，全球大陆流量为37,411 ± 7,816 km ³  yr ^-1。我们的全球河流水储存时间序列表明，流波停留时间是一个基本驱动因素，可以使河流水储存及其变异性增加或减半。我们还协调了先前对全球河流流量每月变化的估计的广泛范围。我们确定了以前未被充分认识的来自海洋大陆（印度尼西亚、马来西亚和巴布亚新几内亚）的海洋淡水源，其数量是刚果河的 1.6 倍，并说明了我们检测严重人为取水的能力。