The timing of unprecedented hydrological drought under climate change

Abstract

Droughts that exceed the magnitudes of historical variation ranges could occur increasingly frequently under future climate conditions. However, the time of the emergence of unprecedented drought conditions under climate change has rarely been examined. Here, using multimodel hydrological simulations, we investigate the changes in the frequency of hydrological drought (defined as abnormally low river discharge) under high and low greenhouse gas concentration scenarios and existing water resource management measures and estimate the time of the first emergence of unprecedented regional drought conditions centered on the low-flow season. The times are detected for several subcontinental-scale regions, and three regions, namely, Southwestern South America, Mediterranean Europe, and Northern Africa, exhibit particularly robust results under the high-emission scenario. These three regions are expected to confront unprecedented conditions within the next 30 years with a high likelihood regardless of the emission scenarios. In addition, the results obtained herein demonstrate the benefits of the lower-emission pathway in reducing the likelihood of emergence. The Paris Agreement goals are shown to be effective in reducing the likelihood to the unlikely level in most regions. However, appropriate and prior adaptation measures are considered indispensable when facing unprecedented drought conditions. The results of this study underscore the importance of improving drought preparedness within the considered time horizons.

Fig. 1. Projected spatiotemporal changes in the frequency of drought days (FDD) under climate change (during the low-flow season).

a The maps show the ensemble median values of the climatological percent changes derived for the FDDs in the mid-21^st century (2036–2065) under RCP2.6 and RCP8.5 compared to the historical period (1971–2005). The results obtained for the low-flow season are presented. The colors indicate the direction and magnitude of the change [%]. Grids with nonsignificant changes between two periods (derived according to a two-sided Kolmogorov–Smirnov test (confidence level 0.05)) are marked with gray, and grids in which agreement in the sign of change among ensemble members is lower than 60% are also shown in gray (Supplementary Fig. 8). Additionally, Greenland is masked out in gray. b The plots present time series of the regional average FDDs derived during the low-flow season [%] from 1865 to 2099 under RCP2.6 and RCP8.5 in the nine selected regions. These regions show robust median TFE₅ values under either or both RCP2.6 or/and RCP8.5 (Fig. 2a). The lines present the ensemble median time series, and the shading shows the uncertainty in terms of the interquartile range across ensemble members. The region names and numbers are listed in Supplementary Fig. 3. The time series estimated for the rest of the regions are presented in Supplementary Fig. 5.

Fig. 2. Timing of the first emergence (TFE) of consecutive unprecedented regional drought conditions (during the low-flow season).

a Timing of the first onset of consecutive exceedance for equal to or more than five years compared to the historical maximum value (TFE₅) under RCP2.6 and RCP8.5 in the 59 regions. The ensemble median results derived from the resampled time series are presented. Only regions in which the ensemble median TFE₅ obtained during the 21st century is statistically robust at the $\pm$5% level by the bootstrap test are shown in color (see Methods). Otherwise, regions are shown in gray. The hatched areas indicate robust TFE₅ signals; in these regions, more than 95% of the bootstrap ensemble members showed TFE₅ during the 21st century. b The cumulative distribution functions (CDFs) of TFE₅ occurrences under the two considered RCPs as a function of time, i.e., the likelihood of TFE₅ occurrence over time, in three regions with particularly robust TFE₅ signals concerning RCP8.5. The CDFs shown as solid lines are estimated from the entire resampled results. Considering the internal variabilities and original ensemble member spreads, the shading represents the uncertainty in the cumulative probability of TFE₅ estimated from resampled ensemble member subsets (see Methods). The cumulative probabilities of TFE₅ occurrence by 2050 and by the end of the 21st century are given in Supplementary Table 1. The same CDF plots for all regions with the median TFE₅ shown in color in a are presented in Supplementary Fig. 11. The regional definitions were derived following the HydroBASINS level-2 product (Supplementary Fig. 3).

Fig. 3. Total numbers of years in which unprecedented conditions are expected from 2010 to 2099.

The error bar shows the 5–95% confidence interval in terms of the ensemble medians (see Methods). The asterisks next to region names indicate that the difference between the two scenarios is statistically significant with regard to the median derived for that region.

Fig. 4. Global mean temperature rise corresponding to regional TFE5 values derived under RCP8.5.

a The global mean temperature rise (∆GMTs [°C]) above the preindustrial level (1850–1900) corresponds to each regional median TFE5 presented in Fig. 2a. The ∆GMT values are derived from each GCM containing the resampled ensemble members that constitute the overall median TFE5. b Cumulative probability functions of TFE5 as a function of ∆GMT, corresponding to those presented in Fig. 2b. Note that the maximum ∆GMT derived under RCP8.5 is approximately 6.0 °C.