Global Wet/Dry Patterns and Mechanisms Since the Last Glacial Maximum: A Key to Future Projection

Abstract

Investigating the response mechanisms of long‐term global wet/dry pattern changes to cold/warm periods and climate forcings can provide scientific supports for the projection of future wet/dry patterns in the context of global warming. Here we present a systematic assessment into the response of global wet/dry patterns to cold/warm periods since the Last Glacial Maximum, and test the triggers for global wet/dry status change. Then we conduct an assessment of future global wet/dry patterns based on a thorough analysis of modern observations, paleoclimate simulations and records. Results show that regions following the hypothesis of “wet gets wetter, dry gets drier” (WWDD) account for 22.81% of the world's land area except Antarctica in the future. Regions complied with future “DD” pattern mostly concentrate in the parts of southwestern North America, southwestern South America, Mediterranean, northern Africa, southern Africa and Asia, covering 18.21% of the modern dry regions. Regions complied with future “WW” pattern occupy 25.41% of the modern wet regions, mainly distributing around the parts of northeastern South America, central Africa, western Russia, eastern Asia and northeastern Australia. Besides we investigate global wet/dry patterns during the abrupt climate change and find that global wet/dry patterns generally have the opposite status between abrupt warm and cold periods. If an abrupt cold event appears in the near warm future, it is likely that some parts of southwestern North America and Mediterranean in modern dry regions will be getting wet.

Introduction

The hypothesis that precipitation minus evaporation (P-E) intensifies in the tropics where atmospheric moisture converges, while P-E weakens in the drylands where atmospheric moisture diverges, as the consequences of rising atmospheric humidity under global warming, is now known as the “wet gets wetter, dry gets drier” (WWDD) pattern (Held & Soden, 2006). Subsequently, this pattern has gained wide examinations and assessments from many scientists (Chou et al., 2009, 2013; Roderick et al., 2014; Seager & Vecchi, 2010; Seager et al., 2014, 2020). More than 300 global hydrological data sets from 1948 to 2005 evaluated that only 10.8% of the global land area follows “WWDD” pattern (Greve et al., 2014). Another study based on satellite soil moisture observations between 1979 and 2013 found that 15.12% of the land areas conforms this pattern and 7.77% has experienced the opposite trend (H. H. Feng & Zhang, 2015). However, the “WWDD” pattern presents in approximately 19.5% of all land area from climate projections (Greve & Seneviratne, 2015). When considering seasonality, unforced internal climate variability, and limitations of water on evaporation in dry land areas, Kumar et al. (2015) suggested that “WWDD” theory is still an effective framework for explaining hydrological change, and therefore they questioned the previous assessments about “WWDD” pattern on land. Although there is still a controversy in testing of “WWDD” pattern, the “WWDD” pattern has become a common view.

Previous evaluations of “WWDD” pattern mostly focus on modern observations and future climate simulations of Representative Concentration Pathways scenarios (RCP4.5 and RCP8.5), while the relation between “WWDD” pattern and past climate change has received little attention. Before “WWDD” theory was proposed, paleolake studies of Late Quaternary have long realized that the tropics were drier and mid-latitude deserts were wetter during the Last Glacial Maximum (LGM), which just coincides with the opposite pattern of “WWDD” in the warm modern and future (Quade & Broecker, 2009; Street & Grove, 1979). The latest research also reinforces the importance of studying past climate on predicting future climate, which gives us inspirations on assessing the “WWDD” paradigm from the perspective of paleoclimatology (D’Agostino & Lionello, 2020; Tierney et al., 2020).

Synergistic effect of climate driving forces like orbital forcing, ice sheets, greenhouse gases and others influences the formation of global cold and warm periods to varying degrees. Maximum ice sheets, low greenhouse gases level and low summer solar insolation all contribute to the millennial cold LGM (Clark et al., 2009; Yokoyama et al., 2000), while high greenhouse gases level, retreated ice sheets and low solar summer insolation are responsible for warm modern and near-future periods (IPCC, 2013). In addition, sudden fluctuation in driving forces would also trigger abrupt climate change events characterized by sudden warming or cooling. Under conditions of steady changes in solar insolation and greenhouse gases, disintegration of Laurentide and Fennoscandian ice sheets released icebergs which melted and discharged freshwater in the North Atlantic causes the cold Heinrich Event 1 (H1) (Roche et al., 2004). Abrupt rise in atmospheric CO₂ is highly synchronous with warm Bølling-Allerød (B/A) (Köhler et al., 2011). And if the release of meltwater from the retreating glacial ice sheets during the H1 suddenly stopped, the restored Atlantic Meridional Overturning Circulation (AMOC) can also promote the occurrence of B/A (Liu et al., 2009). Likewise, the dramatic drop of greenhouse gases and negative radiative forcing are also reflected in the cold Younger Dryas (YD) event (Alley et al., 2003; Renssen et al., 2015, 2018). Understanding the relationship between the synergies of climate forces and the formation of cold/warm periods is critical. A fully comprehensive and credible assessment of global wet/dry patterns can be achieved only when studies on the driving mechanism of hydroclimate patterns in different cold/warm periods are conducted.

In the case of similarity in orbital forcing but huge differences in greenhouse gases level and ice sheets size between the modern times and the LGM, the wet/dry patterns of cold LGM can be considered as a reverse analog to warming future “WWDD” pattern (Quade & Broecker, 2009). Like that, taking into account the climate forces of other cold and warm periods in the past, it is possible to reversely infer the wet/dry patterns in a future abrupt climate change under similar or inverse conditions. P-E simulations which are widely accepted are often used to assess past wet/dry patterns. And the regional consistent fluctuations of lake level can indicate the change of regional climate conditions, which can respond well to past wet/dry changes. Using multi-model ensemble participating in the third phase Paleoclimate Modeling Intercomparison Project (PMIP3), a transient climate change model (TraCE-21 ka) and modern observed data sets, here we performed virtual lake simulations and P-E simulations, and then conducted the analysis based on the intersection results of the two simulations. Numerous paleoclimate records were combined to achieve a better evaluation of the global wet/dry patterns and mechanisms in different cold and warm periods, including LGM, B/A, YD, mid-Holocene (MH), Medieval Climate Anomaly (MCA), Little Ice Age (LIA) and pre-industrial (PI). Then we tested the “WWDD” hypothesis globally and made assessments for the wet/dry patterns in the future.

Materials and Methods Modern Meteorological Data Sets and Modern Wet/Dry Regions

The global modern observation data of precipitation and temperature are derived from Climatic Research Unit Time-Series version 4.03 (CRU TS4.03), which is a monthly 0.5° latitude/longitude high-resolution gridded multivariate climate data sets obtained by the monthly climate anomalies interpolation from extensive networks of weather station observations (Harris et al., 2020). Version four of the CRU TS changes the interpolation process to use angular-distance weighting (ADW), of which this implementation provides improved traceability between each gridded value and the input observations. And the global modern evaporation data set estimated by using CRU-NCEP climate forcing are provided by the Community Land Model version 5 (CLM5) which is the latest version of the land component in the Community Earth System Model (CESM), including new and updated processes and parameterizations (Lawrence et al., 2019).

Due to the vast differences of drought conditions in different countries and regions, many different drought indices have been proposed by scientists from different fields. There are as many as 58 drought indices in various countries listed in the Technical Report of the World Meteorological Organization (WMO) alone (WMO, 1975). In the past century, the global climate change is characterized by the consistent warming of global temperature and more temporal and spatial variations of precipitation. In the process of formulating drought index and dividing global arid regions, scientists have gradually reached some consensus that the variable of precipitation should be considered more (Redmond, 2002). Here we comprehensively consider the basis for dividing the arid regions in central Asia with precipitation less than 400 mm (Wu et al., 2014). According to the annual mean precipitation from 1901 to 2018, regions with annual mean precipitation below 400 mm are simply classified as modern perennial dry regions, while regions with annual mean precipitation above 400 mm are classified as modern perennial wet regions. By comparing with the arid areas classified by aridity index, precipitation, Köppen-Geiger climate and surface vegetation types, our results are generally consistent with them (Huang et al., 2016).

PMIP3 and TraCE Data Sets

PMIP3 is a large-scale international research project on the paleoclimate, supported by World Climate Research Program (WCRP) and International Geosphere Biosphere Program (IGBP). This study adopted time slice simulations of PI, MH and LGM from six models participating in PMIP3 (Table 1). And time slice simulations of MCA and LIA come from five models (Table 2). Orbital parameters (precession, obliquity and eccentricity) were set according to A. L. Berger (1978) and values of greenhouse gases (CO₂, CH₄, and N₂O) were specified following protocols of the PMIP3. In view of the different atmosphere resolutions of models, we first interpolated all models into a resolution of 0.5° × 0.5°, and then the median values of models in different periods were extracted by sorting values from minimum to maximum.

Table 1 Details of Models Used in LGM, MH and PI

Model Name	Number of grids (lon)	Number of grids (lat)	Levels	Resolution (lon °)	Resolution (lat °)
CCSM4	288	192	17	1.25	0.9375
CNRM-CM5	256	128	17	1.40625	1.40625
GISS-E2-R	144	90	17	2.5	2
MIROC-ESM	128	64	35	2.8125	2.8125
MPI-ESM-P	192	96	25	1.875	1.875
MRI-CGCM3	320	160	23	1.125	1.125

Table 2 Details of Models Used in MCA and LIA

Model Name	Number of grids (lon)	Number of grids (lat)	Levels	Resolution (lon °)	Resolution (lat °)
GISS-E2-R	144	90	17	2.5	2
MIROC-ESM	128	64	35	2.8125	2.8125
MPI-ESM-P	192	96	25	1.875	1.875
MRI-CGCM3	320	160	23	1.125	1.125
HadCM3	96	73	15	3.75	2.466

Since the data of B/A and YD cannot be obtained from PMIP3, we therefore adopted the data of TraCE model to conducting time slice simulations of B/A and YD. TraCE model is a transient simulation on the climate evolution completed by Community Climate System Model 3 (CCSM3) of the National Center for Atmospheric Research (NCAR). Four-dimensional model data sets with T31_gx3v5 resolution are provided by the TraCE-21 ka project which investigates the coupled atmosphere-ocean-sea ice-land surface mechanisms and feedbacks that explain the evolution of the climate system over the last 21,000 years (Otto-Bliesner et al., 2006; Yeager et al., 2006). And comparison between multi-model ensemble data sets of PMIP3 and TraCE data sets during PI period shows that PMIP3 and TraCE match well, giving us confidence to carry out subsequent simulations (Figure S1 in Supporting Information S1).

Virtual Lake Simulations and P-E Simulations

An assumption about virtual lakes is widely used in paleoclimate simulations, which regards each grid cell as a 1-m deep, freshwater lake. By constructing a virtual lake system, Li and Morrill (2010, 2013) well reproduced the wet/dry patterns of the LGM and Holocene in Asia, and identified the multiple factors causing these patterns. The simulated hydroclimate status around the North and South America performed by Lowry and Morrill (2019) is also highly in agreement with the climate pattern indicated by lake proxy records. Therefore, it is reasonable to believe that this method is an effective way to evaluate the wet/dry patterns. A one-dimensional lake energy balance model after Hostetler and Bartlein (1990) is conducted to calculate lake evaporation. Lake surface energy balance model is calculated as $c_w{\rho }_{w}z\frac{\partial T}{\partial t}={\varphi }_s+{\varphi }_{ld}-{\varphi }_{lu}\pm Q_e\pm Q_h$, where $c_w$ is the specific heat of water (J/kg K), ${\rho }_w$ is the density of water (kg/m³), z is the lake depth (m), T is the lake temperature (K), t is the time (s), ${\varphi }_s$ and ${\varphi }_{ld}$ are shortwave and longwave radiation absorbed by the water surface, ${\varphi }_{lu}$ is longwave radiation emitted by the water surface, $Q_e$ and $Q_h$ are latent heat flux and sensible heat flux, respectively. Then the lake water balance model with steady-state conditions of climate is applied for evaluating regional hydrologial status, and the lake water balance equation is calculated as $D=A_{B}R+A_L\left(P_L-E_L\right)$, where D is discharge from the lake (m³/year), $A_B$ and $A_L$ are watershed area (m²) and lake area (m²), R is runoff from the drainage basin (m/year), $P_L$ is on-lake precipitation (m/year) and $E_L$ is lake evaporation (m/year). For a hypothetical lake, actual values of $A_B$ and $A_L$ cannot be obtained. Therefore, this equation is simplified for grid cells where $P_L-E_L\ge 0$ and grid cells where $P_L-E_L<0$. Grid cells of $P_L-E_L\ge 0$ are regarded as open lakes (D > 0) which adjust the water balance through the discharge of lake water, while grid cells of $P_L-E_L<0$ are considered as closed lakes (D = 0) of which the net loss of lake water is compensated by runoff into the lake. For the former, we use the area of grid cells as values of $A_B$ and $A_L$. While for the latter, these grid cells adjust to water balance changes through changing the ratio of $A_L$to $A_B$, namely $A_L/A_B=R/\left(E_L-P_L\right)$, where $A_L/A_B$ represents lake level. More details of this model are shown in Li and Morrill (2010, 2013). Input values from the PMIP models and TraCE model are precipitation, runoff, air temperature, surface temperature, wind speed, longwave radiation and shortwave radiation. For PMIP3, the running time step of lake models is 1 day, while time step of lake models in TraCE is set to 10 years.

P-E is also a common index to measure past global wet/dry patterns. Here we combine lake simulations and P-E simulations to comprehensively evaluate the regional wet/dry patterns and to refine our research. Data sets of evaporation used in P-E simulations are directly derived from PMIP3 and TraCE. Finally, we will use the intersection results of the two simulations, namely, only select regions with consistent changes of wet/dry patterns.

Reconstruction of the Wet/Dry Status in Different Cold and Warm Periods

Numerous climate records indicating wet/dry conditions have been collected to verify the simulations. Climate records in this paper are chosen according to three criteria: (1) the selected records must be able to indicate wet/dry climate change; (2) records should be available during the study period; (3) the indicators must have reliable chronologies. We uniformly interpolated the records to a 10-year resolution, and standardized all records to a range of 0–1, as well as extracted the mean value of standardized data in different periods. In the reconstruction process, the periods of LGM, B/A, YD, MH, MCA and LIA are defined as 20–18 ka, 15–13 ka, 12.9–11.7 ka, 6.5–5.5 ka, 950–1250 CE and 1450–1850 CE, respectively. And the periods of stable climate change relative to B/A and YD are defined as 16–12 and 13.9–10.7, while for MCA and LIA are defined as 850–1850 CE. The evaluation criterion of wet/dry status is to compare the average hydroclimate status in different cold and warm periods with that over corresponding stable climate change periods.

In addition, we supplemented our paleoclimate records with a part of lake level records from the Global Lake Status Data Base (GLSDB) for the LGM, MH and PI. As a long-standing international project, the GLSDB documents the changes in past regional water balance by compiling the geomorphic and biostratigraphic data for changes in lake status (https://pmip2.lsce.ipsl.fr/synth/lakestatus.shtml) (Harrison et al., 2003; Kohfeld & Harrison, 2000). The climate anomaly corresponding to LGM-MH, LGM-PI, B/A, YD, MCA, and LIA is shown in Tables S1–S6 in Supporting Information S2.

Results Changes of Global Wet/Dry Pattern in Modern Times

A general trend for global climate change can be captured by linear changes of precipitation, evaporation and temperature (1901–2010). Based on the modern meteorological data sets of CRU and CLM5, there is a significant spatial difference in precipitation variation with an upward trend in most regions and a downward trend in northern Africa and Mediterranean, while evaporation (except for some parts of northern Africa) and temperature are on the rise almost globally (Figure 1). From the time series of global mean temperature, the temperature has increased more sharply in recent decades. The comparison of P-E on the decadal scales exhibits that the humidity in southern North America, northeastern and southwestern South America, Mediterranean, southern Africa, central Asia, and central and western Australia decreases in the 1950s relative to 1900s, and other regions become wetter (Figure 2). Then southwestern North America, central South America, Mediterranean, most of Africa, most of Asia, and most of Australia generally experience drier climate, but southeastern North America, southeastern South America, and northern Asia experience wetter climate in the 2000s relative to 1950s. The hydroclimate difference between 2000s and 1900s is similar to that between 2000s and 1950s, except for parts of Africa. On the whole, obvious drying trend from 1900s to 2000s has been sought in observations of southwestern North America, Mediterranean, southern Africa, central Asia, central and western Australia, while southeastern North America, southeastern South America, and northern Asia exhibit prominent wetting trend.

Figure 1. Linear trends (a, b, c) and time series (d) of global precipitation (a), evaporation (b) and temperature (c) from 1901 to 2010 based on the CRU and CLM5 data sets.

Figure 2. Differences in P-E simulations between 1900s, 1950s and 2000s based on the CRU and CLM5 data sets.

Hydroclimate Change in Millennial Cold and Warm Periods

From the intersection results of lake simulations and P-E simulations with a matching degree of 73.27% (Figures 3a, 3c, and 3e), it is fairly evident that the wet climate of southern North America, most of South America, Mediterranean, southern Africa, and parts of Asia appears in the LGM relative to MH, accounting for 16.05% of the world's land area. While climates over parts of northeastern South America, most of northern and central Africa, and most of Asia are characterized by dry conditions at the LGM relative to MH, covering 27.63% of the world's land area. Moreover, paleoclimate records match well with lake simulations, P-E simulations and their intersection results, which generally verifies the climate anomaly between LGM and MH (Table 3). The matching degree of the intersection results of two simulations between LGM and PI is 69.81%. The climate anomaly during the LGM relative to PI revealed by the intersection results shows that southern North America, parts of southwestern and southeastern South America, Mediterranean, most of Africa, and parts of Asia prevail wet climate, however, northeastern South America, parts of central Africa, most of Asia and most of Australia undergo dry climate, which respectively covers 15.63% and 25.99% of the world's land area (Figures 3b, 3d, and 3f). Likewise, the good match between paleoclimate records, lake simulations, P-E simulations and their intersection results also generally validates this pattern (Table 3). It can be seen that wet/dry patterns between LGM and MH are roughly similar to that between LGM and PI, but there are differences in central Africa of which the climate in the LGM is dry relative to MH, but wet relative to PI (Figures 3e and 3f). More detailed descriptions of hydroclimate characteristics for LGM-MH and LGM-PI revealed by paleoclimate records, lake simulations, P-E simulations and their intersection results are respectively shown in Table 3.

Figure 3. Relative changes in the wet/dry patterns for LGM-MH and LGM-PI based on the PMIP3 data sets. (a, b) The difference in virtual lake simulations for LGM-MH (a) and LGM-PI (b). (c, d) The difference in P-E simulations for LGM-MH (c) and LGM-PI (d). (e, f) The difference in intersection results of the two simulations for LGM-MH (e) and LGM-PI (f). Blue and red points respectively represent the wet and dry records at the former time. More details of paleoclimate records for LGM-MH and LGM-PI are shown in Tables S1 and S2 in Supporting Information S2.

Table 3 Summary of Hydroclimate Characteristics in Different Cold and Warm Periods Revealed by Paleoclimate Records, Lake Simulations, P-E Simulations and Their Intersection Results

[Table omitted. See PDF]

Hydroclimate Change in Different Abrupt Cold and Warm Periods

In addition to the fact that LGM and MH are respectively recognized cold and warm periods with a long duration on the millennial scale (Li et al., 2018), several abrupt climate events have occurred since the LGM. Here we selected four typical cold and warm periods, including warm B/A, cold YD, warm MCA and cold LIA to test global wet/dry patterns when an abrupt change occurs in stable climate change. Obviously, when the abrupt B/A and MCA events occur, the global land temperature increases on the whole, while when the abrupt YD and LIA events occur, the global land temperature decreases on the whole (Figure S2 in Supporting Information S1). There is a 71.90% intersection between lake simulations and P-E simulations during the abrupt warm B/A, displaying that most of southern North America, eastern and southwestern South America, Mediterranean, southern Africa, eastern Asia and most of Australia are dominated by dry climate, and other regions are primarily wet climate (Figures 4a–4c). On the contrary, according to the intersection results of two simulations with a matching degree of 75.48% during the abrupt cold YD, the regions with dry and wet climates during the warm B/A transform respectively into the regions with wet and dry climates during the cold YD (Figures 4d–4f). During the abrupt warm MCA, a 64.04% intersection between lake simulations and P-E simulations is captured. The intersection results commonly indicate that the dry climate appears over most of southern North America, parts of northeastern South America, Mediterranean and parts of Asia, while wet climate occurs at other regions (Figures 4g–4i). Analogously, on the basis of the intersection results of two simulations with a matching degree of 63.44% during the abrupt cold LIA, the regions with wet climate during the cold LIA are generally corresponding with the dry climate found in the warm MCA. Besides, the matching degree of paleoclimate records with lake simulations, P-E simulations and their intersection results during the B/A, YD, MCA and LIA is shown in Table 3. It can be detected that the global wet/dry patterns generally have the opposite status between abrupt warm and cold periods. And judging from the wet/dry patterns of the above cold and warm periods (LGM, B/A, YD, MH, MCA, LIA and PI), parts of the southern North America and Mediterranean all present the characteristics of dry climate in the warm periods but wet climate in the cold periods. More detailed descriptions of hydroclimate characteristics for B/A, YD, MCA and LIA revealed by paleoclimate records, lake simulations, P-E simulations and their intersection results are respectively shown in Table 3.

Figure 4. Relative changes in the wet/dry patterns between abrupt climate change periods and their corresponding stable climate change periods based on the PMIP3 and TraCE data sets. (a, d, g, j) The difference in virtual lake simulations between B/A (a), YD (d), MCA (g) and LIA (j) and their corresponding stable climate change periods. (b, e, h, k) The difference in P-E simulations between B/A (b), YD (e), MCA (h) and LIA (k) and their corresponding stable climate change periods. (c, f, i, l) The difference in intersection results of the two simulations between B/A (c), YD (f), MCA (i) and LIA (l) and their corresponding stable climate change periods. Blue and red points respectively represent the wet and dry records at different abrupt cold and warm periods. More details of paleoclimate records of B/A, YD, MCA and LIA are shown in Tables S3–S6 in Supporting Information S2, respectively.

Driving Mechanism of Hydroclimate Response on Cold and Warm Periods Driving Mechanism of Hydroclimate Response on Millennial Cold and Warm Periods

Figure 5 displays the change of different climate forcings since the LGM. Apparently, the changes of various climate forcings can be summarized as follows: the magnitude of global ice sheets gradually decreases since the LGM (Figures 5a and 5b); the content of greenhouse gases gradually increases since the LGM, and increases more drastically in modern times (Figures 5c, 5e, and 5f); and the summer (winter) solar insolation enhances (reduces) from LGM to the early Holocene and then reduces (enhances) (Figure 5d). The lower temperature of global land surface during the LGM (annual mean temperature [Figures 6a and 6b], monthly mean summer temperature [Figures S3a and S3b in Supporting Information S1] and monthly mean winter temperature [Figures S3c and S3d in Supporting Information S1]) is a comprehensive feedback of the large-scale development of the global ice sheets, lower level of greenhouse gases, and the weaker summer solar insolation (Clark et al., 2009). And the lower temperature leads to a general reduction of evaporation during the LGM over global land area. Accordingly, the regional differences in precipitation during the LGM caused by atmospheric circulation contribute to the formation of LGM wet/dry patterns to a large extent.

Figure 5. Summarization of different climate forcings on the centennial and millennial scales. (a) Ice sheets at the LGM worldwide (Ehlers et al., 2011). (b) Ice sheets at the LGM, MH and PI worldwide from ICE-4G model (Peltier, 1994). (c) Global ice volume δ18O proxy (Lisiecki & Raymo, 2005), N2O record in GISP2 and Taylor Dome ice core (Sowers et al., 2003) and CO2 record in Antarctic ice core (Bereiter et al., 2015). (d) Global insolation changes in June and December since the LGM (A. Berger & Loutre, 1991). (e) CH4, CO2 and N2O records from the EPICA Dome C ice core (Antarctica) (Flückiger et al., 2002). (f) Law Dome CH4, CO2 and N2O records (Etheridge et al., 1996, 1998; Ferretti et al., 2005; MacFarling, 2004; MacFarling et al., 2006).

Figure 6. Annual mean temperature (a, b) and annual precipitation (c, d) differences for LGM-MH (a, c) and LGM-PI (b, d) based on the PMIP3 data sets.

Contrasts of wet/dry patterns between LGM-MH and LGM-PI suggest that the regions where LGM is wet relative to MH and PI are southern North America, southwestern South America, Mediterranean, southern Africa, and parts of Asia (Figures 3e and 3f). Due to the existence of ice sheets in the Northern Hemisphere during the LGM, the westerlies are forced to move southward and strengthen, which bring more precipitation to the southern North America and circum-Mediterranean (COHMAP, 1988; Harrison et al., 1996; Laîné et al., 2009; Wang et al., 2018). From our simulation, the increase of precipitation in these regions is not only reflected in the annual precipitation, but also in the precipitation of summer and winter (Figures 6c and 6d; Figures S3e, S3f, S3g, and S3h in Supporting Information S1). Therefore, the combined effect of increased precipitation and decreased evaporation is responsible for the wet climate of southern North America and circum-Mediterranean during the LGM (D’Agostino & Lionello, 2020; Hostetler, 1991; Hostetler & Benson, 1990; Qin & Yu, 1998). And this LGM wet climate belt from southern North America and circum-Mediterranean may even extend into parts of Asia (Li et al., 2020; Yu et al., 2000). A similar explanation can be applied to the LGM wet regions of southwestern South America and southern Africa. Under the general cold conditions of LGM, polar amplification could result in an enhanced meridional temperature gradient (Masson-Delmotte et al., 2006), and therefore provoke the westerly winds of the Southern Hemisphere to move toward the equator under the thermal wind balance (Holton, 1979). At the South America, the equatorward shift in westerly winds of Southern Hemisphere during the Last Glacial period enhances precipitation to the north of the modern-day position of the Southern Hemisphere westerlies (Heusser, 1989; Lamy et al., 1999). The increase of specific humidity gradient between the Pacific Ocean and the continent strengths moisture advection of the westerlies during the LGM, and further induces the wet climate of the Pacific coast of South America (Lowry & Morrill, 2019). Engelbrecht et al. (2019) proposed that southern Africa has been generally wetter during the LGM, which has a close relation with the influence of rainfall. The equatorward expansion or enhancement of the westerlies in the Southern Hemisphere is the reasonable explanation for the wetter west coasts of Southern Hemisphere continents (Kohfeld et al., 2013).

While the regions where LGM is dry relative to MH and PI are eastern and southern Asia, and Asian high-latitudes on the basis of the comparison of wet/dry patterns between LGM-MH and LGM-PI (Figures 3e and 3f). Decreases in LGM precipitation are responsible for the dry climate in these regions (Figures 6c and 6d). A relatively weak Asian summer monsoon during the LGM is revealed by the coherent speleothem δ¹⁸O records, which may bring less precipitation to eastern and southern Asia, resulting the corresponding dry climate (Wang et al., 2001, 2008; Yuan et al., 2004). While for the Asian high-latitudes, the dry climate can be explained primarily by the mean circulation changes, which might be associated with the decreased intensity of the Arctic Oscillation and southward shift of the westerlies (Li & Morrill, 2013; Lü et al., 2010). As mentioned earlier, wet/dry patterns between LGM and MH are roughly the same as that between LGM and PI, but some regional differences in central Africa. According to the moisture index reconstructed by Wang et al. (2017), the climate during the MH in eastern Africa is prominently wetter than that at the LGM, which is mainly driven by the monsoon system at the low-latitudes. And it is widely recognized that the monsoon system at the low-latitudes generally follows the changing trend of low-latitude summer solar insolation (An et al., 2015; Fleitmann et al., 2003; Laskar et al., 2004; Yuan et al., 2004). The relatively strong summer solar insolation in the MH is a reason for the difference of wet/dry patterns between LGM-MH and LGM-PI.

Contrasting the different climate forcings between LGM and PI, it is found that the intensity of orbital-driven summer and winter solar insolation in LGM and PI is similar, while the content of greenhouse gases and magnitude of ice sheets are significantly different (Figure 5). Thus, for the near warm future, the orbital forcing conditions will not change much, but the differences in greenhouse gases level and ice sheets size relative to LGM will be even greater. Under the similar orbital forcing conditions, the future rising temperature caused by the increase in greenhouse gases may lead to the intense evaporation of global land, and the melting of ice sheets may cause changes in the strength and position of global atmospheric circulation systems. From this perspective, it is therefore speculated that LGM can be regarded as a reverse analogue for future to some extent. However, the assessment of the future “WWDD” paradigm should also be taken into account the actual condition during the modern times.

Driving Mechanism of Hydroclimate Response on Different Abrupt Cold and Warm Periods

When climate forcings undergo abrupt change, they will accordingly promote the occurrence of abrupt climate events. Based on the characteristics of millennial scale climate evolution, speculating that future climate evolution will also experience abrupt climate events, with cooling and warming as the main climate characteristics is reasonable. Relevant studies have confirmed that the global abrupt events during the last glacial period occur synchronously (Corrick et al., 2020). An in-depth understanding of the response of the global wet/dry patterns to the past cold/warm periods triggered by abrupt climate change can better predict the future hydrological climate (Alley et al., 2003). Detecting simulated wet/dry patterns from the above different cold and warm periods, the regions in which the wet/dry patterns have a common feature are the parts of the southern North America and Mediterranean, reflecting the dry climate in the warm periods but wet climate in the cold periods (Section 3.3).

As the consequences of warm B/A and cold YD, reconstructed climates over western America indicate the abrupt changes in hydroclimate. Paleoclimate proxies, which are sensitive to water balance change, generally reflect the regional dry climate during the former but regional wet climate during the latter (Benson et al., 2013; MacDonald et al., 2008; Polyak et al., 2004, 2012). The lake status correspondingly indicates that the Mediterranean has a tendency to dry out after 15 kyr BP which is close to the beginning of B/A event (Qin et al., 1997; Yu & Wang, 1998). Besides, a speleothem record of the southwestern America also shows the dry conditions at the B/A and the relative wet conditions soon after the start of the YD (Polyak et al., 2004). And the prominent increased moisture is found by Renssen et al. (2018) over southeastern North America during the YD. Wong et al. (2016) revealed the possible causes for this phenomenon from the paleoclimate model, that is, the changes in the sinuosity of the Pacific winter storm track which is driven by the weak atmospheric pressure centers over the deglaciation related with the retreated continental ice sheets. During the YD, the southward shift of the Pacific winter storm track contributes to the cold temperatures in the North Atlantic region, and maintains the winter precipitation and wet climate over western America (Asmerom et al., 2010; Benson et al., 2013; Wagner et al., 2010). Moreover, changes of meltwater flux to the Atlantic affect the meridional temperature gradient of the Pacific and further regulate storm track intensity at the YD (Wong et al., 2016).

Warm MCA and cool LIA are crucial periods with climate anomaly during the past 2000 years within the late Holocene. Multiproxy evidence of MCA shows that a series of severe droughts last for decades in western North America (S. Feng et al., 2008; Seager et al., 2007; Woodhouse et al., 2010). However, the increased precipitation anomalies occur in the southwestern America in the LIA (Dee et al., 2020). Relevant studies suggest that the combination of the cold tropical Pacific Ocean and the warm North Atlantic Ocean can explain the severity and longevity of the drought during the MCA (S. Feng et al., 2008). More specifically, this hydroclimate pattern is affected by the persistently La Niña-like tropical Pacific, the warm phase of the Atlantic Multidecadal Oscillation (AMO) and the positive North Atlantic Oscillation (NAO) (Seager et al., 2007). Despite global cooling during the LIA, the central-eastern tropical Pacific relative to the MCA warms, which shifts teleconnections eastward and enhances rainfall anomalies of the southwestern America (Dee et al., 2020). For the Mediterranean, positive AMO and positive NAO conditions are at the MCA, however, negative AMO and negative NAO conditions are at the LIA. And the western Mediterranean is generally dry under the positive NAO (Lüning et al., 2019). Speleothem records near the Mediterranean confirm the prevailing dry climate during the MCA but wetter climate during the LIA, which remains under the combined influence of both the NAO and the AMO (Brahim et al., 2017).

Summarily, although there are differences in the driving mechanisms of the wet/dry patterns during cold and warm periods at different time scales, the wet/dry patterns are generally indirectly regulated by the increase or decrease of temperature which can directly influence the magnitude of ice sheets, atmospheric circulation systems, or the ocean oscillations, etc.

Assessment of Future Global Wet/Dry Patterns From Paleoclimate Perspective

The formation of different cold and warm periods is inseparable from the synergistic effect of climate forcings, especially the cold and warm periods on long-term time scale (Figure 5). There is a similar orbital forcing condition but huge difference in greenhouse gases level and ice sheets size between cold LGM and warm future. Accordingly, LGM can be regarded as a reverse analogue for future, namely LGM relative wet regions are more likely to turn dry in the future and vice versa. However, LGM performance as a reverse analog to future projected hydroclimate change is regionally dependent (Lowry & Morrill, 2019). Accordingly, a better assessment of the future “WWDD” paradigm can be achieved only by combining the LGM wet (dry) regions with the modern perennial dry (wet) regions. As defined in Section 2.1, the modern perennial dry regions mostly distribute in northern and southwestern North America, southwestern South America, Mediterranean, northern Africa, southern Africa, Arabian Peninsula, central Asia, northeastern Russia and Australia, and others are modern perennial wet regions. The intersection results of the LGM relative wet regions and the modern perennial arid areas are considered as an area that conforms to the future “DD” pattern, and the intersection results of the LGM relative dry regions and the modern perennial humid areas are regarded as an area that conforms to the future “WW” pattern (Figure 7). An assessment of future wet/dry patterns projects that 18.21% of modern dry regions will be drier, which is centered in the parts of southwestern North America, southwestern South America, Mediterranean, northern Africa, southern Africa, and Asia. And 25.41% of the modern wet regions will be wetter, which is mainly concentrated in the parts of northeastern South America, central Africa, western Russia, eastern Asia and northeastern Australia. It is worth noting that this future trend will occur as future greenhouse gases continue to rise and the ice sheets continue to melt. Moreover, above regions confirm that 22.81% of world's land area follows the previous hypothesis regarding “WWDD.” Based on the hydroclimate responses of different cold and warm periods, it is reasonable to speculate that if an abrupt cold event appears in the near warm future, some parts of southwestern North America and Mediterranean in modern dry regions will be getting wet.

Figure 7. Future projections and the examination about “WWDD” paradigm. The blue shades present the parts of modern wet regions will be wetter in the future. The orange shades present the parts of modern dry regions will be drier in the future. The combined shades represent the regions confirmed future “WWDD” paradigm.

The gradual increase in concentrations of greenhouse gases triggers a significant climate change characterized by global warming (IPCC, 2013). Under global warming, pluvial and drought phenomenon has occurred frequently in recent decades and exerts large socioeconomic impacts around the world. Numerous proxy climate data prove that the mega-drought occurred from 2000 to 2018 over southwestern North America, which may be ongoing (Williams et al., 2020). Scientists even predicted that a combination of increased temperature and worsening aridity will be the key feature of 21st century in southwestern North America (Woodhouse et al., 2010). Similar drought events are also centered in the western Mediterranean, Greece, and the Levant (Cook et al., 2016). As for pluvial events, longest and largest pluvial of Asia occurs over Kazakhstan and western Russia from 2012 to 2016 (He et al., 2020). The spatial drought-pluvial seesaw is detected in China in recent two decades, with North and Northeast regions experience severe droughts but the Yangtze River basin undergoes extreme rainfall/floods events (Ding et al., 2008). Further, most of India in recent years suffer frequent damaging floods triggered by heavy precipitation event (Guhathakurta et al., 2011). Future projections of global pluvial and drought events based on 24 CMIP5 models suggest that more severe pluvials and droughts are most concentrated in the Northern Hemisphere mid-latitudes and the Americas, respectively (Martin, 2018). As forgoing studies mentioned, a lot of evidence provides the support to the assessment of future “WWDD” paradigm we made. In order to better prevent pluvial and drought disasters, how to effectively predict future climate is extremely important. It is expected that the evaluation of future climate change from the perspective of paleoclimatology can provide a scientific basis in coping with pluvial and drought disasters and policy making in the future.

Conclusion

With the implementation of the Future Earth, the focus of Past Global Changes has shifted from the reconstruction of paleoenvironment to the assess current climate and predict future climate change based on paleoclimate data. Given the similarity and difference of climate forcings between cold LGM and warm future, we synthetically evaluated the future wet/dry patterns from the perspective of paleoclimatology, hoping to contribute to the projection of future global climate change. We not only improve the shortage of single index indicating wet/dry patterns, but also validate the reliability of future prediction by adopting comprehensive indexes combining paleoclimate records, P-E simulations, lake simulations and modern observations. Under the future global warming resulted from the human-made greenhouse gases emissions, it would be getting drier in the parts of southwestern North America, southwestern South America, Mediterranean, northern Africa, southern Africa, and Asia, and getting wetter in the parts of northeastern South America, central Africa, western Russia, eastern Asia and northeastern Australia. Despite the controversy surrounding the proposed future “WWDD” paradigm, our results still verify that about one-fifth of the land area fits this paradigm.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant No. 42077415, 41822708); the National Key Research and Development Program of China (No. 2019YFC0507401); the Second Tibetan Plateau Scientific Expedition and Research Program (STEP) (Grant No. 2019QZKK0202); the Strategic Priority Research Program of Chinese Academy of Sciences (Grant No. XDA20100102); the 111 Project (BP0618001).

Conflict of Interest

The authors declare no conflicts of interest relevant to this study.

Data Availability Statement

PMIP3 simulations are available from the Earth System Grid Federation (ESGF) Peer-to-Peer (P2P) enterprise system website https://esgf-node.llnl.gov/projects/esgf-llnl/. CRU TS4.03 data are available from https://crudata.uea.ac.uk/cru/data/hrg/. The TraCE-21 ka data set comes from Climate Data Gateway at National Center for Atmospheric Research (NCAR) website https://www.earthsystemgrid.org/project/trace.html.