资源环境科技发展态势分析平台

In this study, we analyze the nonstationarity in meteorological droughts at the multidecadal scale in different parts of the contiguous United States during 1901-2017. We develop metrics to compare the drought risk calculated under the assumptions of stationarity and nonstationarity and identify their spatial and temporal patterns. By analyzing the variability of drought risk in the past and exploring its ongoing patterns, we evaluate in which regions of the contiguous United States the assumption of stationarity can be safely used for drought risk planning and management. We find statistically significant interdecadal changes in the probability distribution functions of drought severity in parts of the Northwest, upper Midwest, the Northeast, eastern parts of Great Plains and in parts of Arizona, New Mexico, Utah, and Nevada in the Southwest. In these regions, the nonstationary risk has been significantly higher than the stationary estimate of risk in the past, which shows that the assumption of stationarity can lead to the underestimation of drought risk in these regions. The multidecadal drought risk shows low variability in California, parts of northern and western Great Plains, Ohio Valley, and in the Southeast, since the statistical properties of droughts have not changed significantly in these regions during 1901-2017. However, the meteorological drought risk has increased in California and the Southeast in the recent decades due to the influence of global warming and hence the assumption of stationarity for risk estimation may lead to underestimation of drought risk in future in these regions if this effect of global warming persists.

Plain Language Summary Traditionally, statistical approaches adopted by water resource managers for planning and design of water resource systems and infrastructure are based on the assumption of stationarity; that is, it is assumed that the probabilistic characteristics of the hydrological and meteorological processes do not change with time, and hence, the planning and designs for future can be based on the past observations. In this paper, we have evaluated the validity of the stationarity assumption for meteorological drought risk estimation at the multidecadal scale by comparing drought risk calculated under the assumptions of stationarity and nonstationarity, respectively, in different parts of the continental United States. We find statistically significant nonstationarity in meteorological droughts in the Northwest, upper Midwest, the Northeast, eastern Great Plains and in parts of Nevada, Utah, Arizona, and New Mexico in the Southwest United States, which results in high interdecadal variability of drought risk in these regions. This result demonstrates that the assumption of stationarity can lead to underestimation of drought risk in these regions, thereby exposing water resource systems to failure under severe droughts.