Megadroughts in the Common Era and the Anthropocene

Exceptional drought events, known as megadroughts, have occurred on every continent outside Antarctica over the past ~2,000 years, causing major ecological and societal disturbances. In this Review, we discuss shared causes and features of Common Era (Year 1–present) and future megadroughts. Decadal variations in sea surface temperatures are the primary driver of megadroughts, with secondary contributions from radiative forcing and land–atmosphere interactions. Anthropogenic climate change has intensified ongoing megadroughts in south-western North America and across Chile and Argentina. Future megadroughts will be substantially warmer than past events, with this warming driving projected increases in megadrought risk and severity across many regions, including western North America, Central America, Europe and the Mediterranean, extratropical South America, and Australia. However, several knowledge gaps currently undermine confidence in understanding past and future megadroughts. These gaps include a paucity of high-resolution palaeoclimate information over Africa, tropical South America and other regions; incomplete representations of internal variability and land surface processes in climate models; and the undetermined capacity of water-resource management systems to mitigate megadrought impacts. Addressing these deficiencies will be crucial for increasing confidence in projections of future megadrought risk and for resiliency planning. Megadroughts can be defined as persistent, multi-year droughts that are exceptional compared with other regional events during the Common Era. This Review discusses palaeo reconstructions of megadroughts over the past 2,000 years, and outlines the impact of anthropogenic forcing on the severity and frequency of observed and projected events. The term ‘megadrought’ is often used to refer to droughts that exceed the length of most droughts in the instrumental record, the period of climate observations largely serving as the basis for modern water-resource management and infrastructure. Although developing a more quantitative megadrought definition is challenging, it is suggested that the term be reserved for “persistent, multi-year drought events that are exceptional in terms of severity, duration, or spatial extent when compared to other regional droughts during the instrumental period or the Common Era”. Past megadroughts caused major ecological and societal disturbances over the last two millennia and were forced primarily by persistent ocean states, with possible secondary contributions from internal atmospheric variability, volcanic and solar forcing, and land–atmosphere interactions. Some of the most active megadrought regions in the past are also areas where anthropogenic climate change is projected to increase future drought risk through declines in precipitation, increases in evaporative demand, and/or changes in plant water use. Megadroughts have the potential to substantially strain modern water-management systems, although understanding of the risks of such events, and their ultimate impacts, is still limited by imperfect knowledge of past and future megadrought dynamics. The term ‘megadrought’ is often used to refer to droughts that exceed the length of most droughts in the instrumental record, the period of climate observations largely serving as the basis for modern water-resource management and infrastructure. Although developing a more quantitative megadrought definition is challenging, it is suggested that the term be reserved for “persistent, multi-year drought events that are exceptional in terms of severity, duration, or spatial extent when compared to other regional droughts during the instrumental period or the Common Era”. Past megadroughts caused major ecological and societal disturbances over the last two millennia and were forced primarily by persistent ocean states, with possible secondary contributions from internal atmospheric variability, volcanic and solar forcing, and land–atmosphere interactions. Some of the most active megadrought regions in the past are also areas where anthropogenic climate change is projected to increase future drought risk through declines in precipitation, increases in evaporative demand, and/or changes in plant water use. Megadroughts have the potential to substantially strain modern water-management systems, although understanding of the risks of such events, and their ultimate impacts, is still limited by imperfect knowledge of past and future megadrought dynamics.


0123456789();:
becomes persistent enough to represent a shift in the mean state (aridification) rather than a discrete transient event (drought). Moreover, although several reviews of megadroughts exist [34][35][36] , these have overwhelmingly focused on North America and predate more recent advances in understanding of natural drought variability and the role of anthropogenic climate change (ACC) in contemporary and future events.
In this Review, we synthesize advances in understand ing of megadrought dynamics over the past 2,000 years and into the future from a global perspective, leveraging state of the art palaeoclimate reconstructions, detection and attribution studies, and climate model simulations. We suggest a formal definition for megadroughts, and demon strate that regions on all continents outside Antarctica have experienced drought intervals in the past that meet the proposed criteria. We summarize evidence that strengthens previous hypotheses that ocean-atmosphere interactions forced many past megadroughts, and discuss how ACC is likely to increase drought risk and severity in many megadrought prone regions. Finally, we dis cuss extant uncertainties and knowledge gaps that must be addressed to improve understanding of past mega droughts and increase confidence in projections of future risks and impacts.

Key points
• The term 'megadrought' is often used to refer to droughts that exceed the length of most droughts in the instrumental record, the period of climate observations largely serving as the basis for modern water-resource management and infrastructure. • Although developing a more quantitative megadrought definition is challenging, it is suggested that the term be reserved for "persistent, multi-year drought events that are exceptional in terms of severity, duration, or spatial extent when compared to other regional droughts during the instrumental period or the Common era". • Past megadroughts caused major ecological and societal disturbances over the last two millennia and were forced primarily by persistent ocean states, with possible secondary contributions from internal atmospheric variability, volcanic and solar forcing, and land-atmosphere interactions. • Some of the most active megadrought regions in the past are also areas where anthropogenic climate change is projected to increase future drought risk through declines in precipitation, increases in evaporative demand, and/or changes in plant water use. • megadroughts have the potential to substantially strain modern water-management systems, although understanding of the risks of such events, and their ultimate impacts, is still limited by imperfect knowledge of past and future megadrought dynamics. 0123456789();: Combining updated and existing versions of the drought atlases into a pseudo global data set enables investigation of the inherent temporal scales of drought variability over the last 600 years (1400-2000 CE) (Fig. 1). Despite differences in regional climate dynamics, sum mer soil moisture exhibits strong year to year persis tence, indicating a tendency for soil moisture anomalies (deficits or surpluses) to carry forward from one year to the next. Positive autocorrelation (unitless) at a 1 year lag is >0.5 for many regions (Fig. 1a), and extends to >0.3 into year 2 in some areas (Fig. 1b), including the North American megadrought regions of the Central Plains, Southwest, and Mexico. One major exception to this global tendency towards strong persistence is the west coast of North America, namely California, where autocorrelation is weak or even negative owing to large inter annual variability and high frequency, quasi cyclic variations in cool season precipitation 50,51 .  The strong persistence of summer soil moisture anomalies and droughts is also evident in power spec tra slopes [52][53][54] (1 / year 2 ) (Fig. 1c,d). Slopes of the power spectra in the drought atlases filtered for periodicities of 2-50 years are positive across most regions (Fig. 1c), indi cating higher proportional variance at lower frequencies and longer timescales. This pattern is consistent with the strongly positive autocorrelation and suggests that extended periods of persistent wetter or drier soil mois ture states are common in much of the world. When the power spectra are filtered to remove sub decadal time scales (focusing on periodicities of 10-50 years), slopes remain positive in some regions, indicating higher power at multi decadal bands relative to decadal bands (Fig. 1d). These areas include the Central Plains of North America, south western North America and Central Europe, regions with some of the longest documented mega droughts in the palaeoclimate record. Consistent with the autocorrelation results, California and the west coast of North America stand out with flat or negative slopes, indicative of variability dominated by higher frequencies.
These regional differences in low frequency (dec adal to multi decadal) drought variability underscore the difficulty of establishing a universal megadrought definition. For example, although a decade long drought in California would be highly abnormal given the typical high frequency (inter annual) variability in the region, such an event would not stand out as exceptional in Mexico and the south western USA where decadal vari ability is larger. Any useful megadrought definition must therefore be contextualized relative to the background drought variability of the region. This cautionary note extends to analyses of megadroughts in climate model simulations, which have the added complication that cli mate models might have characteristic variability that diverges substantially from the observations.
Defining megadroughts is further complicated by the variety of methods used to determine when an event, whether a drought or megadrought, begins and ends 55,56 (Box 1). Methods include criteria based on consecutive dry or wet years 57,58 ; multi year to multi decadal average drought anomalies 32,33,59,60 ; methodologies that combine both perspectives 61,62 ; or joint criteria that incorporate duration and spatial extent 63 . Thus, analytical choices can strongly affect how megadrought events are defined and discussed, resulting in little consistency across analyses, regions, and time periods.
In recognition of this methodological diversity, and informed by analyses of drought variability, we suggest that megadroughts be defined as persistent, multi year drought events that are exceptional in terms of severity, duration, or spatial extent when compared with other regional droughts during the instrumental period or the CE. This definition is flexible enough to recognize that different methodologies can be used to define drought or quantify drought characteristics, but also empha sizes that for an event to be considered a megadrought, it must be explicitly compared with other droughts in the available instrumental or palaeoclimate records using the same metrics. Such comparisons are critical for establishing the exceptional nature of a megadrought relative to a long term baseline and ensuring that the term megadrought refers to the most extreme events, whether they occur in the distant past, the instrumental era, or the future.
Common Era megadroughts Using our definition, megadroughts can be found on every continent outside Antarctica over the last 2,000 years in the palaeoclimate record (Fig. 2). The evi dence for megadrought activity during the CE before the twentieth century is now discussed.
North America. Partly owing to its rich availability of drought sensitive palaeoclimate archives 2 , North America is the most investigated region for mega droughts. Research began in the early twentieth cen tury with the first description of the thirteenth century megadrought (the 'Great Drought') in the south western USA 64,65 . Subsequently, multi centennial periods of low run off and streamflow in California's Sierra Nevada Mountains were documented from the mid 800s to the mid 1000s, and from the early 1100s to the late 1200s 5 . Palaeo drought research then advanced rapidly, with evidence of multi decadal megadroughts across nearly every part of western North America 2,3,34-36 during the entirety of the medieval era and the centuries immedi ately following 2,34,36 (800-1600 CE). These events include notable megadroughts across the south western USA during the 1100s and 1200s 36 ; in the mid 1100s over the Colorado River Basin 4,66 ; and across the south western USA, Mexico, and Central Plains during the late sixteenth century 8,31,67-69 , a drought that includes the driest year (1580 CE) of the last 1,200 years 61 .
The effects of these megadroughts are well docu mented in archaeological, historical, and palaeoecolog ical records. The late 1200s megadrought, for example, likely contributed to the depopulation of the Mesa Verde cliff dwellings in south western North America (built and occupied by the Ancestral Puebloans for more than 100 years) 64,65,70,71 , and possibly the collapse of the Cahokia settlements in the Mississippi River Valley 72 . The late sixteenth century megadrought, recorded in English and Spanish colonial records 67,73 , similarly caused the abandonment of Native American settle ments in the south western USA 74 . More broadly, the megadroughts caused increased wildfire activity 75,76 , mass forest mortality events [6][7][8] , and large scale increases in dune mobilization and dust storm activity [77][78][79][80] .

Mexico and Central America.
In Central America, the most recognized megadrought occurred during the Terminal Classic period, approximately 800-1000 CE. Palaeo climate records from lakes 37,81-84 , speleothems 85 , and coastal sediments 86 offer strong evidence for a major extended drought, or multiple drought events, centred over the Yucatán Peninsula and modern day Guatemala and Belize. Average annual precipitation deficits during the event are estimated to have ranged from 25 to 40% 87 or even from 41 to 54% 88 below average, with peak defi cits reaching as high as 52-70% 85,88 . Although debated 89 , these hydrological changes are believed to have con tributed to the putative 'collapse' of the southern Maya kingdoms.

www.nature.com/natrevearthenviron
Similarly, there is evidence of megadrought activ ity in other parts of Mexico, often overlapping with events in the south western USA and Central America. Major multi decadal megadroughts affected the Toltec (1149-1167 CE) and Aztec (1378-1404 CE) civilizations 21,44 , the former coinciding with the decline of the Toltec cap ital at Tula. The Terminal Classic megadrought is also believed to have extended into Central Mexico during 897-922 CE. Megadroughts were additionally observed during the Spanish Conquest (1514-1539 CE) 21 , followed by a mid 1500s event in central Mexico that predated the late sixteenth century megadrought in the south western USA and likely contributed to an outbreak of haemorrhagic fever in 1576 that caused ~2 million deaths 10,67 .

Box 1 | Defining droughts and megadroughts
Drought and megadrought characteristics are highly sensitive to the calculation of their start and termination dates. This point is demonstrated by applying three different methods of drought-event detection to a single reconstructed soil moisture time series (800-2021 Common era (Ce)) from south-western North America 61 : Williams -defining extended droughts as having at least 10 consecutive years with negative 10-year trailing mean values, trimming the start and end dates to avoid starting or ending the events with consecutive annual (non-smoothed) positive anomalies, and dismissing events shorter than 5 years 61 ; meehl -defining on consecutive, negative 11-year centred mean values 62 ; and Coats -beginning droughts with two consecutive negative annual values, continuing the event until two consecutive positive values occur 57 .
Focusing on extended (5-year or longer) droughts in the 1,200-year soil moisture reconstruction, the Williams, meehl and Coats methods identify 30, 41 and 56 events, respectively. many of the worst megadroughts are detected by all three methods, but often with differences (both minor and major) in their start and end dates. For example, although all three methods find the same start date for the late thirteenth-century megadrought, the event is terminated over a decade earlier for Coats compared with Williams or meehl (see the figure, panel a), and, consequently, the megadrought is much shorter (see the figure, panel c) and more moderate (see the figure, panel d) when using that methodology. even larger differences are evident for the late sixteenth-century megadrought (see the figure, panel b).
more generally (see the figure, panels c,d), comparing the different methods reveals that droughts based on the Williams method have longer durations and higher overall severity, those based on the Coats method are shorter because drought termination is easier, and those based on the meehl method have lower overall severity owing to inclusion of wet years during the beginning or end of the drought. In addition, each of the three methods could lead to varying results depending on the time period used to define baseline conditions when the long-term mean drought anomaly is set to zero. Thus, although there is no clear argument that any of the approaches are superior, a consistent drought definition method is required when comparing drought events, as is an awareness of the underlying assumptions and limitations associated with the chosen methodology. PDSI, Palmer Drought Severity Index. South America. Megadroughts were also commonplace in South America. Patagonia experienced a multi decadal megadrought at the same time as the first major California megadrought of the CE 5 , and seven other distinct medieval era megadroughts in Central Chile were concurrent in time with megadroughts over south western North America 17 . Subsequent events, however, were not temporally coincident. These include a dec adal drought in Patagonia at the end of the fifteenth century 45 ; extended periods of enhanced aridity in cen tral Chile and central western Argentina in the middle of the sixteenth, the eighteenth, and the beginning of the twenty first centuries 45 ; two decadal scale mega droughts in 1615-1637 CE and 1684-1696 CE; and an extreme 5 year drought in 1800-1804 in the Altiplano region. This latter event is referred to as the 'Silver Mine' drought 90 . Although relatively short in duration, this event was one of the driest and most spatially extensive periods of drought during the ~200 year period of inten sive Spanish mining in the region, extending from the Altiplano in Bolivia and into central Argentina 45 .
For tropical South America, less detailed informa tion is available on CE hydroclimate. Nevertheless, a 2,300 year lake sediment record from the Central Andes 91 indicates extended periods of reduced South American summer monsoon precipitation during the medieval era and over the last century 91 . Moreover, multi centennial tree ring based precipitation reconstructions 92 also highlight prolonged drought periods in the mid nineteenth century and late eighteenth century in the eastern Amazon.
Africa. In Africa, high resolution CE palaeoclimate records are only sparsely available, challenging efforts to resolve droughts on timescales shorter than several decades or even a century. The exception is in the Mediterranean region of North Africa, where tree rings have been successfully employed to reconstruct frequent and severe multi decadal droughts during the thirteenth and six teenth centuries 93 . Outside this area, lake records con stitute the primary source for drought reconstructions. However, such records often have limited precision age models and act as low pass filters on climate signals, making it challenging to identify the timing and severity of major hydroclimate events, particularly before the late eighteenth century. Following this period, information from higher resolution proxies, historical records, and documentary evidence is more widely available.
Despite their issues, lake records highlight several multi decadal to centennial scale periods of enhanced aridity over East Africa during 1-1000 CE that could cautiously be interpreted as megadroughts [94][95][96] . These include persistent periods of low flows into Lake Turkana, and reduced precipitation over the Ethiopian highlands from 200 BCE to 300 CE 97 ; reduced rainfall during the first two centuries of the CE at Lake Challa 98 ; and sev eral multi decadal droughts at Lake Edward between 400 and 890 CE 99 . East Africa was especially dry during the medieval era, with substantial megadroughts recorded in declining lake levels between 1000 and 1250 CE 98,100,101 .
From the late eighteenth century through to the first half of the nineteenth century 102,103 , nearly the entire con tinent transitioned into a major episode of aridity. These decades were some of the most arid in East Africa of the last 1,000 years, with exceptional declines recorded at multiple lakes 98,102,104,105 . Another major drought followed in the 1880s and 1890s over Ethopia 102,106,107 , causing severe famine from 1888 to 1892 (reF. 108 ).
In other parts of Africa, an approximately 300year megadrought is evident in West Africa from 1450 to 1750 CE, likely the driest in the region during the late Holocene [109][110][111][112][113]  A series of multi decadal megadroughts are evident in India during the fourteenth and fifteenth centuries, associated with 20-30% declines in monsoon rainfall, including at least one event that was 30 years long 120,121 . Megadroughts occurred in Southeast Asia around the same time, most notably two multi decadal events in the mid 1300s and early 1400s that might have con tributed to the collapse of the Angkor civilization in modern day Cambodia 122 (Box 2). Decadal scale mega droughts occurred regularly in northern China 123  Within tree-ring based regional average soil moisture reconstructions 43 , the Angkor drought stands out as an especially severe, persistent and spatially extensive event in Southeast Asia (see the figure, panels a,b). using a simple drought definition requiring at least 5 consecutive dry years, 21 total extended drought periods are identified -of which the Angkor drought is the longest, spanning 24 years from approximately 1344 to 1367 Common era (Ce). moreover, the event was the third most spatially extensive extended drought, with negative 24-year mean soil moisture anomalies affecting 89% of the regional land area. The single most severe soil moisture deficit of any extended drought also occurred during the Angkor drought in 1363 Ce, although the time-average soil moisture anomaly ranks as the fourth driest overall (see the figure, panel c). Thus, based on these analyses and megadrought-defining characteristics, this event qualifies as a megadrought, exhibiting exceptional severity, duration, and spatial extent compared with other droughts in the region over the last ~800 years.
The Angkor megadrought triggered a series of events that contributed to the demise of the Angkor civilization 122 . During this megadrought, streamflow in the mekong was approximately two or three standard deviations below normal 258 , causing Angkor's rulers to seal off reservoirs to minimize water losses. The modified infrastructure had much less hydraulic capacity and was subsequently heavily damaged by a severe flood in 1375 Ce 122,258 . With this loss of water storage capacity, Angkor could not sustain itself during the next major drought (approximately 1399-1404 Ce) and flood sequence. Consequently, much of the Greater Angkor region was eventually abandoned. PDSI, Palmer Drought Severity Index.   15 .
Ice core records suggest that south western Australia was especially drought prone in the mid 1300s 128 and during the eighteenth and nineteenth centuries. Tree ring records corroborate some of these events, indicating major multi decadal megadroughts in 1755-1785 CE, 1828-1859 CE, and 1889-1908 CE 15 . The coincidence of the tree ring derived 1828-1859 CE megadrought in western Australia with coral derived evidence of low streamflow in Queensland (the lowest in the region since the 1600s 38 ) suggests this event might have been especially widespread 15 .
In southeast Australia there is further evidence of multiple extended periods of drought, including multicentennial periods of aridity during the twelfth and thirteenth centuries 129,130 and major decadalscale megadroughts in the early 1500s, late 1700s and 1820s-1840s 46,[131][132][133][134] . The region also experienced mod erately dry summer conditions from 1550 to 1600 CE, and a more severe but shorter dry period from 1670 to 1704 CE 135 . The central coast in eastern Australia (South East Queensland) also experienced eight major megadroughts over the last millennium 16,136 , including a 39 year event from 1174 to 1212 CE 16 .
CE palaeoclimate research has produced an ever expanding body of literature documenting mega droughts on every continent outside Antarctica. It is, therefore, clear that megadroughts are not a phenomenon unique to the regional climate dynamics of western North America but are, instead, an intrinsic part of CE climate variability. Consequently, their global ubiquity strongly suggests that megadroughts are connected to fundamental, and globally important, climate system processes.

Drivers of Common Era megadroughts
Although instrumental and palaeoclimate records are useful for documenting global megadrought activity, deriving information on the driving processes from these data is more difficult. When analysed in conjunc tion with drought reconstructions, however, climate models can provide complementary insights into the three key hypotheses regarding megadrought causes: external forcing, ocean-atmosphere dynamics, and land-atmosphere interactions. Each of these aspects is now discussed.
External forcing. Many CE megadroughts occurred during periods of anomalous solar and volcanic forcing, motivating hypotheses that external forcing changes are a key driver of global megadrought activity 17,24,82,104,137 . For example, megadroughts in western North America are temporally clustered during the medieval era 2,3 , a period of enhanced solar output and reduced volcanic activity 138 . Solar and volcanic forcing are proposed to affect megadrought occurrence through several mecha nisms, including increased local temperature responses that enhance evaporative demand 40 ; reduced land-sea temperature gradients that weaken monsoons 123,139 ; and forced sea surface temperature (SST) changes in regions with teleconnections that cause regional drought 18,48,60,140,141 .
However, evidence for the impact of these forcings on megadroughts is mixed. For example, whereas the temporal clustering of North American megadroughts is higher than predicted from random noise alone, the evidence that this clustering was specifically caused by external forcing is much weaker 142,143 . Moreover, hydro climate responses to forcings vary spatially, as demon strated by volcanic eruptions being linked to drought in some regions 144 (tropical Africa) and pluvial periods in others [145][146][147] (Mediterranean, Southeast Asia). Finally, although some climate models can generate mega droughts in response to external forcing 17,24,40,60 , they can also generate megadroughts analogous to those in the palaeoclimate record through internal variability alone 57,58,[148][149][150] .
Atmosphere-ocean dynamics. Regional hydroclimate variability often results from complex interactions between the ocean and atmosphere. Persistent SSTs, arising from both large thermal inertia and slowly varying ocean circulations, create diabatic heating anomalies in the atmosphere which, in turn, generate atmospheric wave trains, shift storm tracks and modify mean circulation features. These changes subsequently cause droughts, pluvials, or floods in distant regions 151 which, owing to the persistence and slow evolution of SST anomalies, can also be long lived.
The tropical Pacific and Atlantic are especially impor tant given their strong inter annual to multi decadal SST variability and ability to drive teleconnections that influ ence climate worldwide 151 . Indeed, the influence of the tropical ocean can cause persistent hydroclimate states that would be unlikely to be sustained by atmospheric dynamics alone 17,40,152,153 . The temporal concurrence of spatially separate hydroclimatic events -a pattern www.nature.com/natrevearthenviron also unlikely to occur only through stochastic atmos pheric variability -further implicates SSTs as a com mon driver of CE megadroughts. For instance, Pacific SST variability is linked to the co occurrence of the late sixteenth century megadrought in south western North America and a major pluvial in eastern Australia 68 , as well as concurrent megadroughts in North and South America throughout the palaeo record 5,17,154 . Climate models can also generate megadroughts in response to low frequency variations in Pacific and Atlantic SSTs that have similar characteristics to events in the palaeo climate record 17,40 , even if the timing is inconsistent with observations 57,154 .
In the Pacific, decadal scale cold SST anomalies in the eastern tropical Pacific have been identified as the primary cause of many medieval era mega droughts in North America 40,153 and South America 17 . Cold tropical Pacific SSTs cool the tropical atmos phere and displace the jet streams and storm tracks in each hemisphere poleward 155 . Additionally, Rossby wave teleconnections create anomalous high pressure in the extratropical Pacific, west of North and South America 154 . This pattern resembles the atmospheric response to inter annually varying La Niña conditions, diverting precipitation bearing storms poleward of the drought regions 154 . Conversely, warm central Pacific SSTs, often related to positive phases of the Interdecadal Pacific Oscillation, have been strongly linked to mega droughts in eastern Australia 16,46,130 and Asia 43 as the locus of convection in the Indo Pacific sector shifts east. Megadroughts in eastern China have also been linked to western Pacific SST variability driving a weakening of the East Asia summer monsoon 156 .
SST variability outside the Pacific has also been linked to megadrought occurrence. Warm SSTs across the Atlantic sector likely contributed to megadroughts in North America 40 , in West Africa in approximately 1000-1200 CE and 1500-1700 CE 112 , and in Central America during the Terminal Classic period 157 . Cold SSTs in the North Atlantic, by contrast, are related to megadroughts in Central Europe 18 . The Indian Ocean can also contribute to multi decadal drought, especially when acting synergistically with west Pacific SSTs 93,158 .
There are uncertainties regarding the extent to which these megadrought related SST and corre sponding large scale circulation states were externally forced 17,18,40,48 or are a consequence of exceptional peri ods of internal climate variability 57,149 . For example, the dynamical thermostat mechanism has been invoked to explain how enhanced radiative forcing (increased solar output and weak volcanism) during the medieval era could have shifted the eastern tropical Pacific into a persistent cold state 140,141 , increasing megadrought risk in western North America. This mechanism, induced by increased radiative forcing from rising anthropogenic greenhouse gas concentrations, might be responsible for the observed shift in tropical Pacific SST gradients since the mid twentieth century that have contributed to the ongoing megadrought in south western North America 61,159,160 . Notably, this trend is counter to the weakened SST gradient projected by most global cli mate models 159 , and it is uncertain what proportion of these SST trends is forced or a consequence of internal ocean-atmospheric variability 57,149,160 .

Land-atmosphere interactions.
Land-atmosphere inter actions can also substantially affect drought severity and persistence. Declining soil moisture during drought con ditions increases sensible heat fluxes and decreases latent heating and evapotranspiration 161 . Warmer temperatures and reduced atmospheric moisture from these changes at the land surface can then amplify surface drying by increasing atmospheric aridity and evapotranspiration 162 or suppressing precipitation 161 , although not in all cases 163 . Additional feedbacks occur through declines in vegetation health and coverage and increases in wind erosion or dust aerosols 164 , which also reduce energy availability and evapotranspiration through similar mechanisms. Land-atmosphere interactions are strong in many megadrought prone regions 162,165 , and such processes likely contributed to many historical droughts, including the Dust Bowl of the 1930s 166 and the multi decadal Sahel drought from the 1970s to the early 1990s 167 .
There are at least two megadroughts in which landatmosphere interactions are thought to have had a role. During the medieval era megadroughts over the Central Plains of North America, geomorphological evidence indicates that declines in vegetation coverage caused high levels of wind erosion and dust storm activity 80 . Climate simulations suggest these changes would have significantly increased summer temperatures and sup pressed early summer precipitation 168 , enhancing the severity and persistence of the simulated megadroughts compared with simulations using SST forcing alone 168 . In the case of the Terminal Classic megadrought in Central America from 800 to 1000 CE, widespread deforestation (forest conversion to cropland) might account for a sub stantial fraction of the total precipitation deficit 87 . Using an empirically constrained estimate of Pre Columbian land use 169 , models suggest that deforestation could have caused a 10-20% reduction in late summer precipitation, contributing to an overall 5-15% decline in annual precip itation across southern Mexico and the Yucatán 169 . Other, more idealized experiments simulated a 15-30% reduc tion in July precipitation with complete deforestation of Mesoamerica 170 .
Although some uncertainties remain, the strongest available evidence strongly implicates tropical SSTs as the primary natural driver of CE megadrought activity in many regions, with possible secondary contributions from external climate forcings and land-atmosphere interactions. Given the robustness of these natural processes.

Climate change and megadroughts
Anthropogenic climate change affects drought risk and severity through forced changes in precipita tion [171][172][173] , snow 174,175 , and evaporative demand and evapotranspiration 176,177 . Increases in evaporative demand are predominantly caused by warmer temper atures and decreases in relative humidity 31,61,178,179 , with these effects at least partially counteracted by reductions in near surface wind speeds 180 . Plants also respond to changes in climate and atmospheric carbon dioxide concentrations in ways that could either diminish 181 or increase 182,183 evapotranspiration and drought impacts, although the net effect of these competing mechanisms is uncertain.
Through these processes, ACC has intensified soil moisture droughts in California 184,185 and south western North America 31 ; snow and streamflow droughts across the western United States [186][187][188][189][190][191][192] ; and precipitation droughts in the Mediterranean [193][194][195] , Central America 196 , the Caribbean 197 , Chile 26 , southern Africa 198,199 , and south western Australia 193,200 . Included among these are ongoing events that can be considered megadroughts in south western North America 61 and central Chile and central western Argentina 26 . These two events are now discussed, along with consideration of how megadrought risk might change in the future.

The south-western North America megadrought.
Beginning in 2000 CE 61,178 and extending at least through the summer of 2022 (the time of this writ ing), southwestern North America has experienced drought conditions that are unprecedented back to 800 CE 31,61 . A regional soil moisture reconstruction ranks 2000-2021 CE as the driest 22year soil moisture anom aly (−0.87σ) of the last 1,200 years 61 . This event is only comparable with the multidecadal megadroughts that afflicted the region before 1600 CE, slightly edging out the second driest 22year soil moisture anomaly (−0.83σ) from 1571 to 1592 CE during the late sixteenthcentury megadrought. This extended drought has caused severe declines in water resources 61,201,202 , major economic and agricultural losses 203 , and widespread wildfire activity 204,205 .
Precipitation deficits during this event are strongly influenced by natural variability 160,206 and there is no obvious evidence of a downward trend in precipita tion in the region over the last century. One important driver of the precipitation reductions contributing to the twenty first century megadrought is a change towards the cool (or negative) phase of Pacific decadal variability at the turn of the century and its persistence thereafter 160,206,207 . This shift is analogous to prior decadal changes in tropical Pacific SSTs that caused large scale precipitation deficits and drought over the south west, such as occurred during the 1950s and past megadrought periods.
However, although the natural precipitation defi cits alone would have established twenty first century drought conditions in the south west, the region has also experienced significant increases in vapour pres sure deficit owing to warmer temperatures and increas ing vapour pressure deficits [208][209][210] . The warming is largely attributable to anthropogenic forcing 179,209 , resulting in increased surface water losses that have amplified soil moisture deficits during this drought. Indeed, ACC is believed to account for nearly half (~42%) of the soil moisture deficit during 2000-2021 CE 61 .
The large contribution of ACC is also evident in the spatio temporal evolution of the observed mega drought (Fig. 3a,b). In the absence of anthropogenic forcing, 2000-2021 would have likely been composed of two distinct droughts with more moderate cumulative soil moisture deficits up through 2021 (reF. 61 ). Moreover, with anthropogenic effects removed, the most extreme soil moisture deficits would have been localized in southern California and Arizona, consistent with the drought pattern expected of predominantly cool phase conditions in the tropical Pacific, rather than the more spatially extensive observed drying. Anthropogenic cli mate change therefore likely turned what would have been a serious, but spatially and temporally constrained, drought with characteristics typical of historical vari ability in the region into the most severe and widespread megadrought of the last 1,200 years 61 .
The Chile-Argentina megadrought. Since 2008, Chile and Argentina have experienced extreme decadal scale drought conditions. In this region, such persis tent droughts are rare, motivating its labelling as a megadrought 26,27,211 . Indeed, the megadrought stands out as exceptional in both the historical record and palaeoclimate reconstructions of the last millennium (Fig. 3c,d). It is the longest consecutive run of drier than normal years during the twentieth century 26 , and also the driest, or near driest, decadal scale drought over the past 1,000 years 26,27,211,212 . Precipitation deficits during the megadrought reached 20-40% 26,27,213 below normal in some locations and years, causing 7-25% reductions in lake areal extent 214 , up to 90% declines in streamflow 27,215 , and major impacts on snow and glaciers in high alpine areas 27,216 . As a result, the megadrought has severely affected water resources 27,215,217 , wildfire 218 , and vegetation health 27 across the region.
As with south western North America, decadal var iations in tropical Pacific SSTs are an important natural driver of the current megadrought. However, there is some evidence that ACC is intensifying the precipita tion deficits 193,219 . For instance, anthropogenic green house gas emissions and stratospheric ozone depletion have contributed to positive trends in the Southern Annular Mode and poleward expansions in the south ern hemisphere Hadley Cell, shifting storm tracks and the jet stream and reducing precipitation over extra tropical South America. Anthropogenic warming has also contributed to warmer ocean temperatures in the subtropical south west Pacific Ocean, which might have amplified ridging and drying over the region 28 , further reducing regional precipitation. Collectively, these anthropogenically modified processes are estimated to explain 20-50% of the total precipitation reductions during the Chile-Argentina megadrought 26,27,213,220 .
The future of megadrought risk. Many of the most megadrought prone regions during the CE are loca tions expected to experience increased drought sever ity and risk with at least 'medium confidence' 221 . These include western North America, Central America, the Caribbean Islands, Europe and the Mediterranean, Chile, and southern Australia. Such changes are expected to occur in response to season specific precipitation declines 222,223 , decreases in snowpack storage 174,192,224,225 , and increases in evaporative demand 226 and plant water use 182,183 . In part because of relatively uncertain www.nature.com/natrevearthenviron regional precipitation responses in climate models, in most regions increased drought risk occurs as a direct response to warming induced declines in snow and increases in evapotranspiration 29,222 . A notable excep tion is in the Mediterranean climate regions of South America, the Mediterranean Sea, southern Africa, and south western Australia, where increased drought risk is caused by robust reductions in cool season precipitation related to anomalous high pressure occurring within an adjustment of planetary scale stationary waves 193,227 .
In line with increases in overall drought risk, CMIP6 simulations forced with a moderate warming sce nario (SSP2-4.5) indicate that many of the CE mega drought regions will experience substantial increases in multi decadal megadrought risk in the latter half of the twenty first century (Fig. 4). This finding is supported   59 . If realized, these changes would likely mean an unprecedented level of megadrought activity compared with even the most arid centuries of the Common Era. However, there is evidence that lower forcing scenarios would partially mitigate increases in megadrought risk and severity 32,59 , as increases in risk are attributable to background trends and shifts in the mean state which scale strongly with the magnitude of forcing, rather than changes in decadal or multi decadal hydroclimate variability 228 . Moreover, projected megadrought risks for regions such as the south western USA remain sensitive to internal climate variability, even under high forcing. Large initial condition ensembles of climate simulations are therefore necessary to better constrain uncertainties in future regional megadrought risk estimates 229,230 . Climate mitigation is thus likely to offer critical benefits for moderating megadrought severity and risk, even if future increases cannot be entirely avoided.

Summary and future perspectives
Megadroughts are distinguished by their exceptional nature compared with more typical droughts in the instrumental or palaeoclimate records. By this defini tion, megadroughts have occurred on every continent outside Antarctica during the CE, and have been most strongly linked to SST variability, especially in the  it is projected to further increase future megadrought risk to largely unprecedented levels by the end of the twenty first century. Such events would prob ably present substantial challenges to water resources 201 and ecosystem resilience 231 . Even so, climate mitigation is likely to reduce event risk and severity compared with the most extreme warming scenarios 32 . Despite progress in understanding natural and anthropogenic megadrought drivers, confidence in pro jections of megadrought risk is partially undermined by uncertainties in estimates of natural drought vari ability. This uncertainty is especially true for regions where hydroclimate proxy coverage during the CE is sparse (for example, the Amazon), where there is heavy reliance on remote proxies (for example, mainland Australia), or where most information comes from archives (for example, lake records) with low resolu tion and substantial time uncertainties (for example, sub Saharan Africa before the late 1700s). These limita tions make it difficult to fully constrain natural drought variability, the past occurrence of megadroughts, and the characteristics of these events. Addressing these lim itations will require prioritizing new efforts to collect and develop seasonally and annually resolved proxies in historically poorly sampled regions, and then syn thesizing these new records into spatially resolved reconstructions. Climate models introduce additional uncertainties. Climate models generally underestimate natural hydroclimate variability, and might therefore underestimate future megadrought risk 52,53,232 , and often disagree on projected changes in atmospheric circula tion and precipitation 233 . Although it is difficult to know what changes are needed to improve these models, more comprehensive evaluations against the much longer pal aeoclimatic record would probably offer insights into where and why the models are failing.
With temperature having an increasingly important role in modern and future droughts 18,[234][235][236][237][238][239][240] , past mega droughts are likely to be imperfect analogues of future events. Whereas historical megadroughts were caused by precipitation deficits, higher temperatures are an increasingly important contributor to soil moisture, streamflow, and snow drought risk and severity. For example, anthropogenic warming allowed the mega drought at the turn of the twenty first century to emerge across a broad swath of south western North America, even though drought promoting circulation anomalies were likely more severe and persistent during the mega droughts that occurred before 1600 CE, when anthropo genic forcing of the global climate was minimal 61 . These warmer temperatures are likely to amplify drought impacts on ecosystems 6,241 , although by how much is unclear owing to the complex and uncertain responses of vegetation to climate and atmospheric carbon diox ide concentrations [181][182][183] . Much of this uncertainty is centred on the representation of land surface and vege tation processes within climate models, which are often highly parameterized and simplified. Moving forward, it will be critical to understand and improve how these process representations affect model sensitivities, mean states, fluxes, and the underlying manifestation of drought events 242 .
Water management further complicates the picture of how future droughts and megadroughts are likely to affect human water resources, demand, and con sumption. Activities such as surface reservoir storage and management, irrigation, and sustainable groundwa ter management are all invaluable interventions that can increase drought resilience 202,[243][244][245][246] . However, it is unclear whether current implementations of these policies will be sufficient to adapt to a warmer and drier future, espe cially because adaptive capacities have generally not been tested in the context of megadroughts. In the case of the ongoing megadrought in south western North America, unsustainable groundwater withdrawals 202,247 and record breaking lows in reservoir storage at Lake Powell and Lake Mead 61 suggest that current policies are inadequate 248,249 . In some cases, management activities could even be maladaptive or have unintended conse quences. For example, increases in irrigation efficiency can reduce streamflow and groundwater recharge because run off is reduced as water is increasingly allo cated to evapotranspiration, often eliminating expected increases in total water availability 250 . Similarly, reliance on reservoir storage can create situations of higher demand and over reliance, increasing vulnerability to droughts 251 .
To address these future challenges, more explicitly considering megadroughts in drought planning exer cises will be required. Many water resource manage ment plans centre on what is often termed the 'drought of record' , usually the single worst drought event in the historical record designed to represent a potential worst case scenario for drought resiliency planning. Texas, for example, uses the 1950s drought as their drought of record 252,253 , an event that led to the crea tion of the Texas Water Development Board and new reservoir construction across the state 252,254 . Given the much more persistent and severe megadroughts in the palaeoclimate record and model projections, how ever, a more informed approach would be to use palaeo climatic records to define new droughts of record 39,130 or, at least, develop model exercises to assess how current plans would fare under the much more extreme con ditions associated with the megadroughts 255,256 . These approaches are already being considered or applied to water resource management in California, the Colorado River Basin, and the Missouri River Basin. Such exer cises should also consider that a megadrought prone future could mean shorter or less frequent wet intervals that can limit recovery between droughts, especially for groundwater and larger reservoirs. These issues are already beginning to manifest in the western USA, where persistent and frequent drought periods have impeded groundwater and reservoir recovery during the short intervening wet intervals 61,202,247 .
Megadroughts are already stressing adaptive capac ities in south western North America and Chile-Argentina, and it is plausible or even likely that other regions will experience a major megadrought in the coming decades, with similar impacts on water resources. One challenging aspect of these future events will be that much of the increased megadrought risk will occur because of a shift in regional climates to more arid mean states 228 characterized by higher temperatures and, in some cases, reduced precipitation. In some of these regions, megadrought might effectively become the new climate normal, a permanent shift towards drier conditions 201,228 that requires rethinking how droughts are defined given that they are, by definition, temporary events. Regardless, the palaeoclimate record and model projections can be used to better constrain and contex tualize these future changes in hydroclimate, droughts, and megadroughts, informing adaptation and allowing for more proactive development of resiliency plans to reduce the impact of these events.