Drylands face potential threat of robust drought in the CMIP6 SSPs scenarios

2.1.1. Observation-based data

2.1.2. Climate model data

2.2.1. Drought indicator

The SPEI was first proposed by Vicente-Serrano et al (2010) to describe conditions based on the deviation of the differences between precipitation and PET from the mean state (Vicente-Serrano et al 2010). In this study, SPEI values on a 12 month time-scale are used to characterize drought in different climatic regions around the world. The steps used in the calculation are given below.

First, the FAO Penman–Monteith equation was employed to calculate PET for the period 1950–2100 (Allen et al 1998), as follows:

$$\begin{equation}{\text{PET}} = \frac{{0.408\nabla \left( {{R_{\text{n}}} - G} \right) + \gamma \frac{{900}}{{T + 273}}{U_2}\left( {{e_{\text{s}}} - {e_{\text{a}}}} \right)}}{{\nabla + \gamma \left( {1 + 0.34{U_2}} \right)}}.\end{equation} \tag{ 1 }$$

Next, the difference between month-by-month precipitation and evapotranspiration was calculated by the formula:

$$\begin{equation}{D_i} = {P_i} - PE{T_i}.\end{equation} \tag{ 2 }$$

Based on the PET and precipitation from GCM data, the log-logistic probability distribution was used to fit the difference between precipitation and PET, and the SPEI value corresponding to each D_i value was calculated. By applying the same parameters of distribution function established in the historical period 1961–2015, the SPEI dryness/wetness conditions in future can be assessed. When the cumulative probability P < 0.5, the SPEI is calculated as:

$$\begin{equation}w = \sqrt { - 2{\text{ln}}\left( P \right)} \end{equation} \tag{ 3 }$$

$$\begin{equation}{\text{SPEI}} = w - \frac{{{c_0} + {c_1}w + {c_2}{w^2}}}{{1 + {d_1}w + {d_2}{w^2} + {d_3}{w^3}}}\end{equation} \tag{ 4 }$$

where: d₁ = 1.432788; d₂ = 0.189269; d₃ = 0.001308; c₀ = 2.551517; c₁ = 0.802853; and c₂ = 0.010328.

However, when P > 0.5, the SPEI is calculated as:

$$\begin{equation}{\text{SPEI}} = - \left( {w - \frac{{{c_0} + {c_1}w + {c_2}{w^2}}}{{1 + {d_1}w + {d_2}{w^2} + {d_3}{w^3}}}} \right).\end{equation} \tag{ 5 }$$

2.2.2. Identification of drought events

The Taylor diagram is an effective method for model evaluation and testing. We use Taylor diagrams to compare models visually. From figure 1 we can see that GCM can simulate changes in temperature very well. Further, the correlation coefficients of all nine GCMs are greater than 0.99, which is quite close to 1. The standard deviation indicates the ability of the GCM to simulate the center amplitude. The standard deviation of the nine GCMs is uniformly distributed around 1 and they have excellent performance in the simulation of the center amplitude. The root mean square error is used to measure the deviation of the observations from the GCM and is less than 0.13 for all nine GCMs.

Zoom In Zoom Out Reset image size

The simulation of GCMs for precipitation is not as competently as that for temperature, Other than for FIO-ESM-2-0, the correlation coefficients between GCMs and measured data are distributed between 0.72 and 0.89. Regarding the ability to simulate the central amplitude, the standard deviation of GCM is generally small overall, with a standard deviation of less than 1. This indicates a weak simulation of the precipitation amplitude. Among them, ACCESS-CM2 has the best performance with a standard deviation value of 1.04. In contrast, FIO-ESM-2-0 has the smallest standard deviation of 0.40. Regarding deviations between observations and GCMs, the root means square error values of the nine GCMs were distributed between 0.47 and 0.85, with the smallest deviation for IPSL-CM6A-LR and the largest deviation for FIO-ESM-2-0.

The GCMs performed well for the simulation of PET, which illustrates the precision of the Penman–Monteith formula for calculating PET. The correlation coefficients in all nine GCMs were greater than 0.98, with high correlation. The overall performance is good in terms of the ability to simulate the center amplitude, and it can simulate the amplitude accurately. Except for the standard deviation of GFDL-ESM4, which is less than 1, the standard deviation of other GCMs is slightly larger than 1. Regarding the deviation between observations and GCMs, the root mean square error was less than 0.25 for all nine GCMs.

We analyzed the spatial distribution characteristics of global drought change trends based on Sen's slope of drought trends and MK significance test with a 0.05 significance level. In terms of spatial distribution, the spatial pattern of future drought change in the lower SSP126 is consistent with the medium SSP245 and the higher SSP585, while the spatial distribution of dry and wet change trends is relatively concentrated and well-defined. The distribution is also relatively concentrated and well-defined. In figure 2, the regions of intensifying drought are mainly concentrated in Australia, Middle East, South Africa, the Amazon basin, North Africa, Europe and Central Asia. These are mostly arid regions, indicating that in the context of climate change, the decrease in precipitation and the increase in potential evapotranspiration in the global arid regions in the future may lead to further intensification of drought in arid areas.

Zoom In Zoom Out Reset image size

In the case of SSP126, the magnitude of change is smaller and less stable. The percentage of global land area that exceed the 0.05 significance test under the SSP126 was 47.4%, the percentage of global land area that exceed the 0.05 significance test and showed drought intensification was 36.2%, in arid and semi-arid regions, 67.0% of the regions indicated droughts intensify significantly. In the case of the high scenario (SSP585), the variation is large and shows substantial drought intensification. Globally, 85.7% of the land surface exceed the 0.05 significance test, the percentage of the global land regions indicated droughts intensify significantly was 68.2%, in arid and semi-arid regions, 93.2% of the regions indicated droughts intensify significantly. This suggests that some changes in future drought can be mitigated by reducing greenhouse gas (GHG) emissions, especially when the lower emission SSP is compared to the medium emission scenario SSP245. In the latter scenario, 68.3% of the global land surface exceed the 0.05 significance test, the model data exceed the 0.05 significance test, and the percentage of global land area exhibiting increased drought is 53.3%, in arid and semi-arid regions, 90.5% of the regions indicated droughts intensify significantly. It is clear that the magnitude of change is skewed toward future changes in higher SSP, which is further strong evidence of the need to control GHG emissions. However, even at lower emissions, strong changes are expected in many regions, especially in Australia, the Amazon, and Central Asia.

The results indicate an overall decreasing trend in global wet and dry variability for the period 2015–2100, with droughts intensifying over time. Meanwhile, in figure 3, the SPEI values for the three future SSPs show dramatic differences in development after 2060. The risk of drought intensification is higher in Australia, South Africa, the Amazon basin, North Africa, Europe, and Central Asia.

Zoom In Zoom Out Reset image size

3.3.1. Duration of drought events

The spatial pattern of drought events duration changes relative to the historical period in two different study time periods and three future SSPs is shown in figure 4. As can be seen, the increase in drought events duration is greater in the dry zone under the high-emissions shared socioeconomic path. In contrast, the drought events duration in the wet zone exhibits a slight increase, with the wet zone with a larger value of drought events duration increase mainly located in the Amazon River basin. It is worth noting that, while under the high emission SSPs, European regions also exhibit an increase in drought events duration.

Zoom In Zoom Out Reset image size

This increase is also evidenced by the mean drought events duration in the global terrestrial humid and semi-humid regions and in the arid and semi-arid regions. The average drought events duration over global land is estimated to be 4.4 months, 5.7 months, and 8.6 months (average of all SSP scenarios) corresponding to the historical periods 1960–2000, 2021–2060, and 2061–2100, respectively. In arid and semi-arid areas, the estimated durations are 6.5, 9.5, and 15.3 months, respectively, whereas in humid and semi-humid areas, there is a slightly increase in the trend of drought events duration to 3.6, 4.3, and 5.7 months, respectively. It should be emphasized that drought events with longer durations are expected to be particularly prominent during 2061–2100, especially in arid and semi-arid regions. For comparisons between shared SSPs in terms of emission effects, longer drought events duration corresponds to higher emission SSPs. This correspondence is mostly concentrated in drought regions in the latter two periods and is more pronounced in 2061–2100. In the Amazon and European regions, this phenomenon is also more pronounced during the same period, with estimated future durations of 3.7, 5.3, and 9.8 months, respectively, in the European region, and distribution corresponding to SSP126, SSP245, and SSP585. In contrast, in the Amazon basin, the estimated future durations are 3.5, 4.8, and 9.1 months, respectively. Overall, these two regions face a high risk of drought under high emission scenarios. Notably, when considering all pathways, regions and future periods, the high emission scenario SSP585 consistently exhibits a longer duration.

3.3.2. Frequency of drought events

The drought events frequencies were calculated for each grid for the reference and future periods (frequencies are events per 40 years). Relative to the reference period, changes in the spatial pattern of drought events frequencies are expected to be manifested in different periods and scenarios primarily in the United States, the Amazon basin, the Sahara, Australia, South Africa, Europe, and Central Asia (figure 5). The decreasing trend in drought events frequency which occurs mainly in the Sahara and the South Asian peninsula is caused by the shift from drought events to long-term drought events in the Sahara.

Zoom In Zoom Out Reset image size

The average global drought events frequency for all regions is estimated to be about 12.7 and 11.8 events (average of all scenarios), respectively, which is higher than the reference cycle of 9.1 events for the future periods 2021–2060 and 2061–2100. Therefore, the frequency of future drought events will be higher relative to the reference period.

The average frequency of drought events in arid and semi-arid regions is 15.5 and 11.9, respectively, and the frequency of long-term drought events in arid regions will decrease during 2021–2060 and 2061–2100. However, the duration of drought events will increase with the intensity of drought, which indicates that drought events in arid regions will turn into stronger drought. Unlike the arid and semi-arid regions, the change in drought events frequency in humid and semi-humid regions will increase insignificantly 11.3 and 11.8 times during 2021–2060 and 2061–2100, respectively.

When comparing SSPs in terms of emission effects, lower DF corresponds to higher emission SSPs. For the next two periods, this correspondence is mostly concentrated in drought regions, with drought frequencies of 16.5, 15.5, and 14.2 during 2021–2060 and distribution corresponding to SSP126, SSP245 and SSP585. This relationship is more obvious during 2061–2100, at 14.4, 12, and 9.3 times, respectively. Combined with the above analysis of drought ephemeris, this further indicates that drought events in the arid zone will transform into long-term drought events under high emission SSPs.

3.3.3. Intensity of drought events

Drought event intensity was calculated as the 40 year average of all event-average SPEI values at each grid cell for each of the periods of interest (averaged over all models). Figure 6 shows the absolute change in drought events intensity between the future period and the reference year. Relative to the reference period, nearly all mid- and low-latitude regions of the globe exhibit an increase in drought events intensity for almost all SSP scenarios and future periods. The projected results indicate that relatively large increases in drought events intensity are concentrated in western United States, Patagonia, Sahara, Middle East, South Africa, Central Asia, and Australia, and that drought events intensity and spatial patterns will be very similar in the three climate scenarios during the future period 2021–2060. Drought events intensity will increase significantly in European region and Amazon River basin during 2061–2100 under SSP245 and SSP585. Moreover, the intensity and duration of droughts will be decreased in the higher latitudes of the northern hemisphere, such as Iceland and northern Siberia.

Zoom In Zoom Out Reset image size

The regional average drought events intensity over the global land surface during the reference period is estimated to be −0.62. However, severe drought events are expected to occur in the future according to the model. Specifically, in 2021–2060 and 2061–2100, the intensity is anticipated to be −0.75 and −0.97 (all SSP averages), respectively.

In both humid and arid areas, although the spatial pattern of drought events intensity change is consistent (as shown in figure 6), the average expected drought events intensity in arid areas is more significant than that in arid and semi-humid areas. In 2021–2060, the drought events intensity in arid areas is expected to be −1.1, while that in wet areas is expected to be −0.58. Meanwhile, in 2061–2100, the intensity is estimated to be −1.5 in arid areas and −0.67 in wet ones. The difference between arid and wet areas is further widened in the late 21st century, when arid areas suffer from even more severe drought consequences. Furthermore, in terms of changes in intensity levels under climate scenarios, the projections show that more severe droughts events are likelier to occur globally under higher emission scenarios, especially in the late 21st century, with global drought events intensity reaching −1.21 under the SSP585 emission scenario and even −2.0 (extreme drought) in arid regions.

By the end of this century, arid areas are expected to cover half of the global land surface, with increased warming and population growth, drylands in developing countries will experience degradation and desertification (Huang et al 2016). Arid areas (drylands) are more sensitive and vulnerable to climate change than wet areas (wetlands), and there are large asymmetries in response. Previous studies have reported that the magnitude of climate change in global drylands will exceed the global average (Huang et al 2017). Due to global warming, some borders of wetlands will be transformed into drylands, and that more severe droughts will occur in these new drylands (Feng and Fu 2013, Sherwood and Fu 2014, Berg et al 2016).

Our study is based on drylands as defined by historical data, investigated regional differences in future changes of drought, arid areas face more severe drought threats compared to wet areas, and wet areas will be transformed into arid areas, which in return will negatively affect more people's lives. Drought will pose greater challenges to agriculture, water management, and human health in arid regions. The findings indicated that by the late 21st century, drought events duration will be 2.7 times longer and drought events intensity 2.2 times more severe in arid areas than in wet ones under the SSP585 path. The cause of this difference may be due to plant transpiration against photosynthesis (Liu et al 2006), where the physiological response of plants under conditions of high CO₂ reduces the prediction of future drought stress (Swann et al 2016). Other factors include the effect of clouds and aerosols on energy balance (Huang et al 2017). Therefore, when conducting global warming studies, anticipated future changes in arid regions and regional differences cannot be ignored.

Based on the three typical SSPs in CMIP6, this study used the Penman–Monteith formula with SPEI indicators to predict future drought conditions globally. The results showed that the magnitude and extent of drought increased significantly with increasing SSPs and warming, which suggests that future drought risk in hotspots can be mitigated by reducing GHG emissions.

Our model projections indicate a general trend towards drier global wet and dry variability, with droughts intensifying over time. The spatial distribution shows a higher risk of increased drought in Australia, South Africa, the Amazon Basin, North Africa, Europe and Central Asia. Our results further validate the risk of future drought. The global warming will increase the risk and severity of droughts in most subtropical and mid-latitude regions (Cook et al 2018). Severe and widespread droughts will occur worldwide over the next 30–90 years due to reduced precipitation and/or increased evaporation (Dai 2013), with drier average conditions, the risk of the most extreme drought events in history has increased with warming, by 200%–300% in some areas (Cook et al 2020). Specifically, the regional average drought event intensity over the reference period is estimated to be −0.62 over global land, while under the SSP585 scenario for 2061–2100, the intensity is projected to reach −1.19. Moreover, if contemporary rates of warming continue, water supplies and demand deficits could increase fivefold in Africa, Australia, southern Europe, south-central United States, Central America, the Caribbean, and parts of northwestern and southern China (Naumann et al 2018). Agricultural focus areas (Chile, Central Europe, Eastern United States and parts of China) and densely populated areas (e.g. Mediterranean. Southern Africa, Western Africa and southern North America) will suffer from more severe droughts, threatening water and food security (Ukkola et al 2020).

Europe and the Amazon basin region face more severe drought risks under the path of high emission SSPs, and future droughts in these two regions will break the balance of global socio-economic, agricultural production, and ecosystems, and further aggravating global warming (Webber et al 2018).

In the SSP585 scenario for 2061–2100, in the European region, the estimated future duration is 9.8 months and the intensity of drought is 2.8 times higher than in the historical period. Meanwhile, in the Amazon basin, the estimated future duration is 9.1 months and the drought intensity is 2.5 times higher than in the historical period. Overall, these two regions face a high risk of drought under the high emissions scenario. More than 82% of Europe's population is expected to live in urban areas by 2050 (UN Habitat 2011). The concentration of population, assets and activities makes it difficult for cities to adapt to increased droughts due to global warming (Shi et al 2016). At the same time, Central Europe is one of the major global food producers and exporters, and increased droughts could also threaten agricultural production in Europe.

Similarly, forest cover and evapotranspiration in the Amazon basin are crucial for the water cycle in tropical South America. Much of the basin will experience reduced rainfall in the future, especially in the eastern and southern parts of the Amazon, which will be dry in the 21st century (Parsons 2020). In addition, future changes in the region are particularly vulnerable to anthropogenic impacts on the region such as deforestation, land management, and forest fires (Nepstad et al 2014). Climate change will double the area burned by wildfires in the Amazon, and by 2050, 16% of the region's forests will be affected (Brando et al 2020). Drought and deforestation feedbacks will hinder efforts to curb drought and deforestation (Staal et al 2020). The combined stress-driven forest declines will in return have significant impacts on regional carbon sequestration (regional carbon sequestration), land degradation risk, biodiversity, and the global climate system. Therefore, to meet the challenges posed by future climate change, international cooperation is urgently needed for disaster management from the perspective of sustainable development of global human society.