Interdecadal variability of terrestrial water storage since 2003

Earth’s water cycle is changing due to anthropogenic forcing and internal variations in the climate system. These changes are leading to an intensification of the water cycle that manifests as more frequent and stronger droughts in some areas, and pluvials in others. The resulting impacts on terrestrial water storage can be crucial for water availability. However, current understanding is hampered by limitations in observations and models, and the variety of processes that influence terrestrial water storage across temporal scales. Here, we leverage the now two-decades long satellite record from the Gravity Recovery and Climate Experiment and subsequent Follow-On missions to investigate persistent trends in the presence of internal variability. We use cyclostationary empirical orthogonal function analysis to uncover statistical modes of variability that help explain a shift in decadal terrestrial water storage trends that occurred around 2012. The dominant statistical mode suggests an interdecadal periodicity that is also found in the precipitation record. The second leading mode is highly correlated with the Pacific Decadal Oscillation. Isolating these modes points to regions where the magnitude of the terrestrial water storage trend may flatten or reverse in coming decades due to internal climate variability and reduces uncertainty in multidecadal linear trends.

受人为强迫（anthropogenic forcing）与气候系统内部变率的共同影响，地球水循环正发生改变。这些变化正导致水循环加剧，具体表现为部分地区干旱愈发频繁且强度增大，而另一些地区则进入多雨期（pluvials）。由此对陆地水储量（terrestrial water storage）产生的影响，对水资源可利用量至关重要。然而，观测与模型的局限性，以及不同时间尺度下影响陆地水储量的过程多样性，制约了当前对该问题的理解。在此，我们利用重力恢复与气候实验（Gravity Recovery and Climate Experiment, GRACE）及其后续任务（Follow-On missions）长达二十年的卫星观测记录，在考虑内部变率的前提下，研究持续趋势。我们采用循环平稳经验正交函数分析（cyclostationary empirical orthogonal function analysis），揭示有助于解释2012年前后陆地水储量十年尺度趋势转变的统计变率模态。主导统计模态显示出一种年代际周期性（interdecadal periodicity），这在降水记录中同样存在。第二主模态与太平洋年代际振荡（Pacific Decadal Oscillation, PDO）高度相关。分离这些模态，可识别出未来数十年内因气候内部变率导致陆地水储量趋势趋平或逆转的区域，并降低多年代际线性趋势的不确定性。