Received: 17 January 2019 Revised: 23 May 2019 Accepted: 27 June 2019

DOI: 10.1002/joc.6219

RESEARCH ARTICLE

The Central Chile Mega Drought (2010–2018): A climate

dynamics perspective

René D. Garreaud1,2 | Juan P. Boisier2 | Roberto Rondanelli1,2 | Aldo Montecinos3,4 |

Hector H. Sepúlveda3 | Daniel Veloso-Aguila3

1
Geophysics Department, Universidad de
Chile, Santiago, Chile Abstract
2
Center for Climate and Resilience Central Chile, home to more than 10 million inhabitants, has experienced an
Research, Universidad de Chile, Santiago, uninterrupted sequence of dry years since 2010 with mean rainfall deficits of
Chile
20–40%. The so-called Mega Drought (MD) is the longest event on record
3
Department of Geophysics, Universidad de
Concepción, Concepción, Chile
and with few analogues in the last millennia. It encompasses a broad area,
4
Centro de Recursos Hídricos para la with detrimental effects on water availability, vegetation and forest fires that
Agricultura y Minería, Universidad de have scaled into social and economical impacts. Observations and reanalysis
Concepción, Concepción, Chile
data reveal that the exceptional length of the MD results from the prevalence
Correspondence of a circulation dipole-hindering the passage of extratropical storms over cen-
René D. Garreaud, Blanco Encalada 2002, tral Chile—characterized by deep tropospheric anticyclonic anomalies over
Santiago, Chile.
the subtropical Pacific and cyclonic anomalies over the Amundsen–
Email: rgarreau@dgf.uchile.cl
Bellingshausen Sea. El Niño Southern Oscillation (ENSO) is a major modula-
Funding information tor of such dipole, but the MD has occurred mostly under ENSO-neutral con-
CONICYT Chile, Grant/Award Numbers:
FONDAP 15110009, FONDECYT 1170286
ditions, except for the winters of 2010 (La Niña) and 2015 (strong El Niño).
Climate model simulations driven both with historical forcing (natural and
anthropogenic) and observed global SST replicate the south Pacific dipole
and capture part of the rainfall anomalies. Idealized numerical experiments
suggest that most of the atmospheric anomalies emanate from the subtropical
southwest Pacific, a region that has experienced a marked surface warming
over the last decade. Such warming may excite atmospheric Rossby waves
whose propagation intensifies the circulation pattern leading to dry conditions
in central Chile. On the other hand, anthropogenic forcing (greenhouse gases
concentration increase and stratospheric ozone depletion) and the associated
positive trend of the Southern Annular Mode also contribute to the strength
of the south Pacific dipole and hence to the intensity and longevity of the
MD. Given the concomitance of the seemingly natural (ocean sourced) and
anthropogenic forcing, we anticipate only a partial recovery of central Chile
precipitation in the decades to come.

KEYWORDS
Chile, climate change, drought, ENSO, PDO, SAM, South America

Int J Climatol. 2019;1–19. wileyonlinelibrary.com/journal/joc © 2019 Royal Meteorological Society 1

2 GARREAUD ET AL.

1 | INTRODUCTION (e.g., Miller, 1976) with annual mean precipitation ranging

between 100 and 2000 mm (Viale and Garreaud, 2015), mostly
When will a drought break? This is a pressing question posed concentrated in austral winter (May–September) and produced by
by stakeholders—from local farmers to national water cold fronts (Falvey and Garreaud, 2007; Viale and Nuñez, 2011).
authorities—during prolonged dry periods. Progress towards its The interannual rainfall variability is large (coefficient of variation
answer requires disentangling the role of natural climate vari- about 0.3) and partially driven by ENSO (e.g., Aceituno, 1988;
ability and anthropogenic climate change in sustaining precipita- Montecinos and Aceituno, 2003; see also Section 4.2). A regional
tion scarcity at a regional scale. On interannual time-scales, El drying trend, of particularly large magnitude since the late 1970s,
Niño Southern Oscillation (ENSO) is the leading driver of but observed since the mid-20th century (Quintana and Aceituno,
droughts in many regions worldwide (e.g., Vicente-Serrano 2012; Masiokas et al., 2016; Boisier et al., 2018), has been par-
et al., 2011; Schubert et al., 2016) followed by other ocean- tially attributed to anthropogenic climate change (Vera and Díaz,
forced modes of variability or by land surface processes 2015; Boisier et al., 2016; 2018). Such trend is expected to con-
(Seneviratne et al., 2012). On the other hand, there is mounting tinue in the future, as model-based regional climate projections
evidence that climate change is already increasing the fre- consistently indicate a reduction in mean annual precipitation
quency, duration and intensity of regional droughts (e.g., Dai, (up to 40% relative to current values) for the second half of this
2011; 2013), by either inducing circulation-mediated precipita- century under high emission scenarios (Fuenzalida et al., 2007;
tion deficit or enhancing evapotranspiration. If one concludes Bozkurt et al., 2018).
that a given drought is mostly of natural origin, reversal towards The observed decline of precipitation over central Chile
normal or even wetter conditions should occur in the foresee- has been greatly accentuated by an uninterrupted sequence
able future even though one cannot predict the exact time of of dry years since 2010 to the present, with annual rainfall
such shift. If, by the contrary, global warming is the main driver deficits ranging between 25 and 45%. This ongoing, multi-
of the rainfall deficit, one should expect (and adapt to) a year dry spell has been referred to as the Central Chile Mega
sustained trend towards drier conditions. Drought (MD; CR2, 2015) owing to its unprecedented lon-
Intense, short-lived (1–2 years) droughts are a rather com- gevity and large spatial extent in the historical record. The
mon feature in Mediterranean-like climates (where most of the MD impacts on hydroclimate and vegetation are described
annual rainfall is accounted for in a few winter storms) but dur- in Garreaud et al. (2017). The precipitation deficit resulted
ing the last decades, these regions have also experienced much in the diminished Andean snowpack, reservoir volumes and
longer dry spells. The multi-year (2012–2014) drought in Cali- groundwater levels across central Chile. Mean river dis-
fornia attracted considerable interest (e.g., Griffin and charge decreased up to 90% as well as nutrient exportation
Anchukaitis, 2014; Swain, 2015; Williams et al., 2015) given into the sea with potential impacts on coastal ecology
its unprecedented effects in hydrology, forest fires and agricul- (Masotti et al., 2018). The hydrological drought also
ture (e.g., AghaKouchak et al., 2014; Mao et al., 2015). The extended into westernmost Argentina (Rivera et al., 2017).
lack of winter rainfall over that region was directly associated A substantial decrease in vegetation productivity was
with an anomalously persistent high-pressure ridge over the observed in the shrubland-dominated, northern sector of cen-
northeast Pacific that in turn appears as a response to tropical tral Chile, but a mix of greening and browning patches
sea-surface temperature (SST) anomalies of mostly natural ori- occurred farther south where irrigated croplands and exotic
gin (Seager et al., 2015). A decade-long drought also afflicted forest plantations dominate. The MD coincided with the
southeastern Australia in the recent past (ca. 1997–2009, Saft warmest decade on record, leading to a substantial increase
et al., 2015) disrupting river ecosystems and agricultural pro- (~60%) of the burned area by forest fires (González et al.,
duction as reviewed in van Dijk et al. (2013). The rainfall defi- 2018). The biophysical impacts of the MD ultimately
cit during the so-called Millennium Drought has been partially resulted in detrimental consequences for the Chilean popula-
attributed to climate change -specifically to anthropogenic tion (~70% of which reside in Central Chile), especially in
greenhouse warming—throughout a trend in the Southern rural areas (Aldunce et al., 2017). Such social distress was
Annular Mode (SAM) towards its positive polarity and a pole- recognized by county, regional and national authorities, that
ward shift of the Hadley Cell (Cai et al., 2014). Multi-year rain- took several but weakly coordinated measures during the
fall deficits have also prevailed recently in-land areas MD (Verbist et al., 2016).
surrounding the Mediterranean Sea (Garcia-Herrera et al., The seemingly unprecedented character of the ongoing
2007; Hoerling et al., 2010), South Africa (Rouault and Rich- MD and the prospect of a drier climate emphasize our initial
ard, 2003) and the Middle East (Trigo et al., 2010; Kelley query. To this end, we investigate the large-scale circulation
et al., 2015). patterns and physical mechanisms sustaining the protracted
Central Chile, along the west coast of South America (30 – dry conditions in central Chile during 2010–2018. Observa-
38 S, Figure 1), exhibits an archetypical Mediterranean climate tions and model simulations are described in Section 2. We
GARREAUD ET AL. 3

FIGURE 1 Geographic and

climate features of Central Chile.
(a) Topographic map (dark grey:
Terrain elevation <500 m asl; light
grey: 500–3,000 m asl; white: >
3,000 m asl). Blue dots are rain gauges
stations operated by DMC/DGA. Red
circles are the location of the six
stations used to define the regional
precipitation index. Stations that
provide records for Figure 4 are also
indicated. (b) Station-based annual
mean rainfall (1980–2010) according
to latitude [Colour figure can be
viewed at wileyonlinelibrary.com]

then provide a brief description of the observed precipitation regional scale, the Chilean Directorate of water resources
anomalies in central Chile (Section 3), updating a more com- (DGA) and the National Weather Service (DMC) maintain
prehensive analysis in Garreaud et al. (2017). Atmospheric more than 500 rain gauges along central Chile (Figure 1).
reanalysis are used in Section 4 to describe the large-scale From the original daily observations, we computed monthly
circulation anomalies during the current MD and past accumulations when less than 5% of the days are missing,
droughts during the 20th century. We also evaluated the his- thus retaining 220 stations with data from 1960 onwards.
torical and present-day association between central Chile We also use lake and groundwater levels in a few stations
rainfall and relevant planetary-scale modes (ENSO, SAM operated by DGA. All the station data is available from the
and the Pacific Decadal Oscillation, PDO) to infer their role Center for Climate and Resilience Research Climate
during the MD. To assess natural and anthropogenic contri- Explorer (http://explorador.cr2.cl).
butions to the current MD, in Section 5, we use a large The large-scale circulation was characterized using the
ensemble of Global Climate Model (GCM) data, including National Centers for Environmental Prediction—National
both fully-coupled runs and simulations with prescribed Center for Atmospheric Research (NCEP-NCAR) reanalysis
SST. Additional numerical experiments were conducted to (NNR; Kalnay et al., 1996) available since 1948, including
isolate the ocean-sourced forcing of the dry conditions in gridded (2.5 × 2.5 latitude–longitude [lat–lon]) monthly
central Chile and propose a physical mechanism for such means of geopotential height and air temperature at several
connection. Conclusions are summarized in Section 6. levels. NNR-based results were contrasted against the
European Centre for Medium-Range Weather Forecasts
ERA-Interim Reanalysis (Dee et al., 2011). Our study also
2 | DAT A AN D M OD E L S employs monthly mean SST fields from the NOAA
extended-reconstructed sea surface temperature (ERSST)
2.1 | Observational dataset available from 1860 onwards on a 2 × 2 lat–lon grid
Several sources of information are used to characterize cen- (Smith et al., 2008) and the NOAA Optimum Interpolation
tral Chile historical droughts and the current event. At the (OI) SST - V2 available from 1981 onwards on a 1 × 1
4 GARREAUD ET AL.

lat–lon grid (Reynolds et al., 2002). Global precipitation is including three runs from the Community Earth System
depicted using monthly mean fields from the Global Precipi- Model CESM1, that used time-varying GHG and O3 con-
tation Climatology Project (GPCP) available from 1979 to centrations (Cionni et al., 2011). Table 1 presents the basic
present on a 2.5 × 2.5 lat–lon grid (Adler et al., 2003). features of these simulations. For each CMIP5 model, we
combine their historical runs (1950–2005) and RCP8.5 sce-
nario runs (2006–2040). Since averaging across the
2.2 | Atmospheric only simulations
models/runs effectively removes internal variability, the
In Section 5, we use several families of GCM simulations to multi-model mean isolates the signal due to anthropogenic
gauge the role of natural variability and anthropogenic forcing climate forcing. We refer to these simulations as CMIP5.
in sustaining the ongoing MD. First, we use a 40-member
ensemble carried out with the Community Atmosphere Model
2.4 | SPEEDY experiments
(CAM5, the atmospheric component of the Community Earth
System Model, Kay et al., 2015) by the NOAA Earth System SPEEDY (Simplified Parameterizations, primitivE-Equation
Research Laboratory. CAM5 was integrated from January 1900 DYnamics) is an atmospheric global circulation model
to February 2018 at approximately 1 × 1 lat–lon resolution, (AGCM) of intermediate complexity. It is based on the primi-
forced by observed SST and Sea Ice (Hurrell et al., 2008). tive equations with a spectral dynamical core and simplified
These kind of simulations, with prescribed ocean boundary con- physical parameterizations (Kucharski et al., 2013). It is a
ditions, are referred to as AMIP as they were employed in the hydrostatic model with a spectral transformation in the
Atmospheric Model Intercomparison Project (Gates et al., vorticity-divergence form described by Bourke (1974). The
1998). They also employ time-varying greenhouse gases model resolution used is T30 L8, which corresponds to a trian-
(GHG) and stratospheric ozone (O3) concentrations. The GHG gular spectral truncation with 30 wave numbers (96 × 48
evolution is based on observed estimates until 2005 and the Gaussian grid points), about 3.75 × 3.75 , and eight vertical
RCP6.0 scenario thereafter (Moss et al., 2010). The time- levels. The top two upper levels represent the stratosphere, the
varying O3 is from the SPARC observed database (Cionni lowest level the planetary boundary layer and three levels of
et al., 2011). Consistently, this simulation is referred to as free troposphere (Molteni, 2003; Kucharski et al., 2006). Physi-
AMIP-ORF (observed radiative forcing). Note that these simu- cal parameterizations include large-scale condensation, short-
lations include 8 out of 9 winters of the ongoing MD. Since the and longwave radiation, shallow and deep convection, surface
ensemble members are forced by identical boundary conditions fluxes of momentum and energy, and vertical diffusion.
and only differ in slightly different initial conditions, the ensem- In addition to topography and land–sea mask, SPEEDY
ble mean attempts to isolate the SST-forced response of the requires surface boundary conditions at the ocean (SST and sea
atmospheric circulation under observed levels of GHG and O3 ice fraction coverage [SIC]), soil temperature in the deep soil
while the ensemble spread informs about the SST-forced signal layer (1 m), moisture in the top soil layer and the root-zone
to internal atmospheric noise ratio. layer, snow depth, bare surface albedo, and fraction of land-
A second set of AMIP simulations, also performed with surface covered by vegetation. The last two levels are defined
CAM5, consists of a 30-member ensemble spanning January with an annual mean value, while the other fields are defined
1979 to December 2016, but with GHG and O3 concentra- as monthly means, which are linearly interpolated to get daily
tions kept fixed to their 1880s values (Cionni et al., 2011; values during the calculation. The climatological fields were
Meinshausen et al., 2011). In this case, SST has been originally calculated averaging the 1979–2008 period from
detrended and adjusted to 1880 equivalent mean conditions, ERA-Interim results (Dee et al., 2011). For the SPEEDY
AMIP run, we prescribed global monthly observed SST and
but retain the observed interannual and decadal variability as
SIC from 1870 to 2017 (Rayner et al., 2003). A total of
in the other experiments. Sea ice is set to a repeating seasonal
50 ensemble members for each run were created by adding ran-
cycle of roughly 1979–1990. This set of simulations is
dom diabatic forcing. Ensemble member 1 was perturbed
referred to as AMIP-PRF (past radiative forcing) and attempts
1 day (72 time steps), ensemble member 2 was perturbed for
to reproduce the atmospheric circulation in response to natural
2 days, and so on. We also employed SPEEDY to conduct a
SST variability previous to major human interference in the
numerical experiment described at the end of Section 5.1.
climate system. Thus, comparing AMIP-PRF against AMIP-
ORF is useful to precisely gauge that interference.
3 | REGIONAL FEATURES O F T HE
MEGA DROUGHT
2.3 | Fully-coupled GCM simulations
We used a total of 30 fully-coupled simulations from Despite of marked meridional rainfall gradients, year-to-year
26 GCMs participating in CMIP5 (Taylor et al., 2012), precipitation and streamflow variability exhibit a notable
GARREAUD ET AL. 5

TABLE 1 Main features of the fully coupled models used for the historical and RCP8.5 scenario (Taylor et al., 2012)

Model Nlon Nlat Institution

ACCESS1-0 192 144 CSIRO (Commonwealth Scientific and Industrial
Research Organization, Australia), and BOM
(Bureau of Meteorology, Australia)
ACCESS1-3 192 144
Bcc-csm1-1 128 64 Beijing Climate Center(BCC),China Meteorological
Administration, China
CanESM2 128 64 CCCma (Canadian Centre for Climate Modelling
and Analysis, Victoria, BC, Canada)
CCSM4 288 192 NCAR (National Center for Atmospheric Research)
Boulder, CO, USA
CESM1-BGC 288 192 NSF/DOE NCAR (National Center for Atmospheric
Research) Boulder, CO, USA
CESM1-CAM5 288 192
CMCC-CESM 96 48 CMCC—Centro Euro-Mediterraneo per i
Cambiamenti Climatici, Bologna, Italy
CNRM-CM5 256 128 CNRM (Centre National de Recherches
Meteorologiques, Meteo-France, Toulouse,
France)
CSIRO-Mk3-6-0 192 96 CSIRO (Commonwealth Scientific and Industrial
Research Organization, Australia), and University
of Queensland, Australia
FGOALS-s2 128 108 IAP (Institute of Atmospheric Physics), CAS
(Chinese Academy of Sciences), Beijing, China
FIO-ESM 128 64 FIO (The First Institution of Oceanography, SOA,
Qingdao, China)
GFDL-CM3 144 90 NOAA GFDL (Geophysical Fluid Dynamics
Laboratory, USA)
GFDL-ESM2G 144 90
GISS-E2-H 144 89 NASA/GISS (Goddard Institute for Space Studies)
New York, NY
HadGEM2-CC 192 144 Met Office Hadley Centre, Fitzroy road, Exeter,
Devon, EX1 3 PB, UK, (http://www.metoffice.
gov.uk)
HadGEM2-ES 192 144
inmcm4 192 144 INM (Institute for Numerical Mathematics, Moscow,
Russia)
IPSL-CM5A-LR 96 96 IPSL (Institut Pierre Simon Laplace, Paris, France)
IPSL-CM5A-MR 144 143
MIROC5 256 128 AORI (Atmosphere and Ocean Research Institute,
the University of Tokyo, Chiba, Japan)
MPI-ESM-LR 192 96 Max Planck Institute for Meteorology
MRI-CGCM3 320 160 MRI (Meteorological Research Institute, Tsukuba,
Japan)
MRI-ESM1 320 160
NorESM1-M 144 96 Norwegian Climate Centre

Nlon and Nlat indicates the number of points in longitude and latitude, respectively.
6 GARREAUD ET AL.

degree of spatial homogeneity in central Chile (Montecinos two analogues of the ongoing MD during the last millen-
et al., 2000). Following Garreaud et al. (2017), these com- nium (Garreaud et al., 2017). Precipitation deficits over 25%
mon variations were tracked by considering annual rainfall were observed in more than three-quarters of the stations
accumulations in six stations between 32 and 37 S with along central Chile in every year conforming the MD,
nearly complete records from 1915 onwards (highlighted in although the rainfall anomalies have some spatial variability
Figure 1). For each station, the annual series of observed as illustrated by the station-based maps for individual years
accumulation was divided by its climatological value (Figure 3). These data support the findings in Garreaud et al.
(1980–2010). The regional precipitation index (RPI, (2017) regarding the more persistent and extraordinary char-
Figure 2a) was then calculated every year as the median of acter of the MD in the southern half of central Chile.
the seven station values. RPI has a high correlation (r ~ 0.7)
with individual precipitation time series almost everywhere
in central Chile. 4 | O B S E R V E D LA R G E - S C A L E
Given our focus on regional dry spells in central Chile, ANOMALIES
drought events are identified as those years in which
RPI ≤ 80% (≥20% deficit in rainfall), a threshold often used In this section, we use reanalysis and observed data to
for hydrological and agricultural applications. This simple describe the large-scale circulation and attending SST pat-
identification agrees well with those based on more sophisti- tern during the MD. Most analyses are performed using aus-
cated indices (e.g., SPI or the Palmer Drought Severity tral winter (MJJAS) mean anomaly fields from 2010 to
Index; see Garreaud et al., 2017 for details). Between 1915 2018, calculated as departures from the 1980–2010 climatol-
and 2009, 24 years were classified as regional droughts ogy. We also consider dry years occurring before 2010 (see
(about a fourth of the time) mostly composed by one to Figure 2a) to provide a historical background. For both the
3-year long events (Figure 2a). MD and past droughts, the precipitation anomaly maps
The Central Chile MD stands out in the RPI time series (Figures 4a,d) reveals that dry conditions in central Chile are
as the uninterrupted sequence of dry years since 2010, with the easternmost expression of a broad band of rainfall deficit
RPI ranging between 55 and 80% (Figure 2a). This 9-year across much of the subtropical southeast Pacific. Dry condi-
drought is substantially longer than any other event in the tions extend across the subtropical Andes into central Argen-
20th century, although it does not include extremely dry tina but with weaker amplitude. To the south of 45 S, there
(RPI < 50%) years like 1924, 1968 or 1998. The histogram is a tendency for wet anomalies from the central Pacific to
of mean RPI considering 9-year blocks further illustrates the the south Atlantic. A narrow dry band over the equatorial
extraordinary character of the MD in the observational Pacific—all the way from the maritime continent to the coast
period (Figure 2b). In a longer-term context, tree-ring-based of South America—is very prominent in the historical
precipitation reconstruction for central Chile reveals only drought composite but mostly absent in the MD mean. As

FIGURE 2 (a) Annual series of Central Chile regional precipitation index (RPI). Droughts, defined as years with RPI < 80%, are identified by
the red circles. (b) Histogram of 9-year average of RPI for the period 1915–2009. The light blue bars show the observed frequency, considering a
9-year sliding window throughout the 1915–2009 record. The blue thick line is the distribution obtained from 5,000 randomly selected 9 years from
the historical period. The orange arrow indicates the RPI averaged during the MD (2010–2018) [Colour figure can be viewed at
wileyonlinelibrary.com]
GARREAUD ET AL. 7

FIGURE 3 Station-based rainfall anomalies during the MD. In each station, the anomalies are calculated as the annual deficit (observed minus
climatology values) divided by the climatology value (precipitation average 1980–2010). The map in the left shows the rainfall anomalies for the
full period (2010–2018). The graphs show the anomalies for each year and station, sorted according to their latitude. The thick box encompasses
Central Chile [Colour figure can be viewed at wileyonlinelibrary.com]

we describe later, this difference is connected with the vary- and less frequent weather fronts reaching this region
ing impact of ENSO on central Chile hydroclimate. (Garreaud, 2007; Solman and Menéndez, 2002; see also
Figure S2a). Reduced westerlies also weaken moisture trans-
port from the Pacific (Campos et al., 2018) and orographic
4.1 | Circulation pattern and SAM impact
precipitation forced by the Andes range (Falvey and
The composite map of 500 hPa geopotential height (Z500) Garreaud, 2007). By the contrary, enhanced transient
anomalies during dry winters (Figure 4b) provides a dynami- activity—and hence precipitation—occurs over the tip of the
cal perspective on central Chile historical droughts. Of par- continent and the Drake's passage, immediately downstream
ticular relevance is a dipole of positive anomalies across the of the anomalous low over the ABS (Figure S2a).
subtropical Pacific and negative anomalies at mid-latitudes. The strength of the geopotential height dipole over the
Such dipole, strongly related to ENSO variability, has southeast Pacific can be gauged by the difference in Z500
been referred to as South Pacific Oscillation (SPO) and between a subtropical box (centred at 30 S–110 W) and a
emerge as the leading mode of a Principal Component Anal- mid-latitude box (centred at 65 S–90 W). Figure 5 shows
ysis of the pressure field over this ocean basin (You and the scatter plot between the winter mean height difference
Furtado, 2017). and central Chile rainfall anomalies (RPI) for the period
The South Pacific dipole is very prominent in the MD 1948–2018. Consistent with the previous discussion, there
composite (Figure 4e) and evident every winter during this is a negative relationship between both variables with a
period, although the centres of the pressure anomalies correlation coefficient of −0.6. The strength of the dipole
exhibit some variability in their position and intensity has been consistently high from 2010 onwards resulting
(Figure S1). In average during the MD, the positive anoma- in a MD-average about 20% larger than the historical
lies further extend over the tropical Pacific and the negative mean (Figure 5). The winter mean series of Z500 used
anomalies are very deep over the Amundsen–Bellinhausen to calculate the strength of the dipole over the South
Sea (ABS). The positive anomalies in the subtropics drive Pacific are shown in Figure 6. Both series exhibit sub-
tropospheric-deep easterly wind anomalies over much of stantial year-to-year variability, caused by either internal
central Chile, a key factor behind regional dry conditions dynamics or remotely forced by SST anomalies in the
(Montecinos et al., 2011; Garreaud et al., 2013). Weaker equatorial Pacific. Superimposed on that variability there
westerlies associate with suppressed baroclinic instability are negative/positive trends in Z500 in the mid-
8 GARREAUD ET AL.

F I G U R E 4 Composite anomalies (departures from de 1980–2010 mean) of austral winter (May–September) precipitation (GPCP, 1979
onwards), 500 hPa geopotential height (NNR; 1948 onwards) and sea surface temperature (ER-SST until 1980 and OI-SST from 1981 onwards).
Upper panels: Historical droughts (identified in Figure 2a). Lower panels: MD period (2010–2018) [Colour figure can be viewed at
wileyonlinelibrary.com]

latitude/subtropical box, evident since the 1970s and sta- stratospheric O3 depletion and increased GHG concentra-
tistically significant at the 1% level. These opposing tion (Arblaster and Meehl, 2006; Eyring et al., 2013; Gillett
trends result in a southeast Pacific Z500 dipole strength- et al., 2013), thus, supporting the hypothesis that anthropo-
ening of ~20 m/decade between 1980 and 2018. The genic forcing have played a relevant role on the drying ten-
magnitude of this trend is substantial and accounts for dencies and on the maintenance of MD in central Chile
one-third of the dipole anomaly observed in average (Boisier et al., 2016).
between 2010 and 2018.
The positive trend in mid-level geopotential height over
4.2 | Influence of ENSO
subtropical latitudes can be linked to the overall tropo-
spheric warming during the last decades (e.g., Sherwood ENSO impacts on the hydroclimate of central Chile in the
et al., 2008). The negative height trend over the ABS dur- form of a warm/wet - cold/dry relationship by exciting the
ing wintertime is more puzzling. Only weak trends were South Pacific dipole (e.g., Aceituno, 1988; Rutllant and
found in that season in the last part of the 20th century Fuenzalida, 1991; Montecinos and Aceituno, 2003). Ind-
(Fogt et al., 2012), but after year 2000 reanalysis data eed, the composite map of SST anomalies during historic-
shows a marked drop (R. Fogt 2018, personal communica- al droughts exhibits a marked cooling across much of
tion). The SAM pattern strongly projects upon the strength equatorial Pacific, weaker cool anomalies over the subtro-
of the height anomaly dipole (r ≈ 0.75; see also Fogt et al., pical southeast Pacific, and a horse-shoe pattern of warm-
2012), so part of the augmented meridional gradient over er anomalies rooted in the tropical western Pacific
the south Pacific arises from the SAM trend towards its (Figure 4c). All these features conform the well-known
positive polarity (Marshall, 2003; Jones et al., 2016). In pattern of SST anomalies during a La Niña event
turn, the SAM trend has been largely attributed to (e.g., Rasmusson and Carpenter, 1982; Capotondi et al.,
GARREAUD ET AL. 9

and that during ENSO neutral winters rainfall is distributed

over a wide range.
The SST anomaly field during the MD shares some fea-
tures with its historical counterpart (Figures 4c,f), namely,
the weak cold anomalies over most of the subtropical SE
Pacific and the horse-shoe pattern of warm anomalies rooted
in the maritime continent. A remarkable difference, however,
is the lack of significant cold—La Niña-like—anomalies
along the equatorial Pacific during the MD, even if the win-
ter 2015 (when a strong ENSO was developing) is excluded.
Considering a threshold of ±0.5 C of the winter mean
Nino3.4 index for El Niño/La Niña classification, only 2010
qualified as La Niña, ENSO-neutral conditions prevailed
from 2011 to 2014, 2015 qualified as a strong El Niño, and
ENSO-neutral winters return in 2016–2018 (Figure 7). This
suggests that ENSO had little influence in the maintenance
of the rainfall deficit in central Chile during the MD period.
FIGURE 5 Scatter plot between the Central Chile precipitation Moreover, although dry winters under ENSO-neutral
index (RPI) and the difference of winter mean 500 hPa geopotential conditions are not uncommon, a 6-year drought chain seems
height between a box over the subtropical Southeast Pacific (30 –40 S; unlikely. To assess the likelihood of such dry sequence we
140 –110 W) and a box over the Amundsen Bellinhausen Sea (60 – made 5,000 random extractions of 6 ENSO-neutral years
70 S; 110 –80 W). Data from 1948–2018. The years forming the MD (without replacement) from the historical RPI time series
are indicated in red [Colour figure can be viewed at
(1920–2009). The probability of having a 6-year sample
wileyonlinelibrary.com]
with a mean rainfall deficit ~25% is less than 2.5%, and the
probability of having such deficit in each individual year
2015). The scatter plot between winter values of Niño3.4 (as the current MD) is 0.5% or less, indicative of the influ-
and central Chile rainfall anomalies (Figure 7), however, ence of other modes in sustaining the long drought in central
shows that the statistical relation is moderate (r ~ +0.6) Chile.

FIGURE 6 The lines with circles

are the time series of the observed
(NNR) winter mean (May–September)
500 geopotential height at a box over
the subtropical Southeast Pacific (30 –
40 S; 140 –110 W) and a box over the
Amundsen Bellinhausen Sea (60 –
70 S; 110 –80 W). The MD period is
highlighted. The orange (light blue)
thick line is the CMIP5 multi model
mean 500 geopotential height in the
subtropical (mid-latitudes) box
considering the historical runs
(1940–2005) and the RCP8.5 runs
(2006 onwards). Each simulated series
(from 43 fully coupled models) were
previously low-pass filtered (5-year
moving average) and adjusted so their
1960–1990 mean coincided with the
observed (NNR) mean for that period
[Colour figure can be viewed at
wileyonlinelibrary.com]
10 GARREAUD ET AL.

4.3 | Influence of ocean-sourced decadal

variability
Rainfall variability in central Chile has also been related to
low frequency phenomena, particularly to the Pacific
Decadal Oscillation. Consistent with the ENSO-like struc-
ture of the PDO and attending teleconnections in the SH
(Garreaud and Battisti, 1999), its cold (warm) phase tends
to produce extended periods that are in average drier (wet-
ter) than the long term mean in central Chile (Garreaud
et al., 2009; Masiokas et al., 2010; Muñoz et al., 2016;
González-Reyes et al., 2017). This relation is summarized
in Figure 8 by the scatter plot between multi-year average
of the PDO index and the corresponding RPI in central
Chile. The periods were selected as sequences of years in
which the PDO was predominantly in its warm or cold
phase. The statistical relationship is modest (r ≈ 0.4), but
we acknowledge that such analysis is based on only a few
PDO multi-year blocks.
After the regime shift in the mid 1970s (Jacques-Coper
and Garreaud, 2014), the PDO remained in its positive phase
during the next three decades but for brief negative excur- F I G U R E 7 Scatter plot between the winter (MJJAS) average of
sions in 1990 and 1999–2002. Around 2007, the PDO index the Niño3.4 index and the Central Chile precipitation index (RPI). Data
became negative until the end of 2014, when it flipped again from 1915 to 2018. The years forming the MD are indicated in red.
to its positive phase. The mostly positive PDO indices from The Niño3.4 index is the area average SST anomaly in the region 5 S–
1980s to 1990s and mostly negative indices after 2005 5 N and 170 –120 W [Colour figure can be viewed at
resulted in a negative trend year round during the last three wileyonlinelibrary.com]
decades (Clem and Fogt, 2015). During the MD period, the
PDO index has exhibited both polarities during, with a
Saurral et al., 2018) that also involves the deep ocean
multi-year median of −0.15 and a mean of +0.1. Consider-
(Volkov et al., 2017). In turn, part of the surface warming is
ing the PDO-rainfall relationship from the historical record
related to the shift in the PDO from its positive to negative
(Figure 8) one would expect near average rainfall conditions
polarity over the past decades (England et al., 2014; Saurral
for the 2010–2018 period. Even if we restrict the analysis to
2010–2014 (when the mean PDO was −0.45) the observed et al., 2018). The impact of the SSWP warming upon central
rainfall deficit (30%) is well below the PDO-congruent range Chile rainfall deficit is explored in Section 5.1.
(15–22%). Considering the uncertainty in the PDO-rainfall
relationship, this analysis suggests that the mostly negative
PDO phase during the last decade has contributed to the dry 5 | INS I G H T S FR O M GL O B A L
conditions in central Chile but it is insufficient to explain the CLIMATE M ODELS
full intensity and persistence of the current Mega Drought.
The observational evidence presented in the previous sections
Similarly, Boisier et al. (2016) found a substantial but partial
role of the PDO negative trend in the observed drying trend has revealed that the exceptional length of the ongoing drought
over the southeast Pacific (including central Chile) during in central Chile results from the uninterrupted reiteration of a
the last few decades. large-scale circulation pattern disfavouring the passage of frontal
In describing the composite SST anomalies during the systems over central Chile. This pattern is characterized by
MD, the warming of the subtropical southwest Pacific tropospheric-deep positive pressure anomalies over the central-
(SSWP) stands out (>1 C just northeast of New Zealand, eastern subtropical Pacific and negative anomalies over the
Figure 4f). Warmer than normal conditions in that region Amundsen-Bellinhausen Sea. Both natural (e.g., the cold PDO
also appears in the historical drought composite -in connec- phase) and anthropogenic forcing (through the positive phase of
tion with La Niña conditions- but with much lesser magni- SAM) can produce such pattern and appear to be acting during
tude (<0.3 C). The marked anomalies at the surface of the the central Chile MD. In this section, we use three families of
SSWP during the last decade are partially explained by a global simulations—described in Section 2.1 and 2.2—to gauge
warming trend since the late 1980s (Figure 9a,b; see also the importance of these factors.
GARREAUD ET AL. 11

5.1 | Ocean forcing an MD-average rainfall deficit that reach and even exceed
the observed value as summarized by the box plot in
The left panels of Figure 10 show the CAM5 AMIP-ORF
Figure 11. Similar findings emerge when using the AMIP
(i.e., observed SST and radiative forcing) ensemble mean
ensemble runs with SPEEDY.
anomalies of Z500 and precipitation averaged during the These AMIP results strongly suggest that the protracted
MD (austral winter of 2010–2017 given the data availabil- Central Chile MD was substantially driven by global SST
ity). The ensemble mean replicates the key dynamical pat- anomalies, while internal atmospheric variability would
tern sustaining dry conditions in central Chile: negative enhance its severity. Such inference is somewhat surprising
anomalies over the ABS and positive anomalies over the given the mostly ENSO-neutral conditions during the MD
subtropical Pacific (Figure 4e). Consistently, the simulated but consistent with the mostly negative PDO phase during
area of negative rainfall anomalies over the west coast of this period. The outstanding question is then, from where in
South America and adjacent ocean is in good agreement the SH extratropical oceans is emerging the atmospheric
with its observational counterpart (Figure 4d). The MD forcing of the dry conditions in central Chile? A plausible
mean pattern of Z500 and rainfall anomalies is also found candidate is the subtropical southwest Pacific (SSWP,
in individual winters. The amplitude of the CAM5 ensem- 30–40 S, 190–210 W), a region that has exhibited a sub-
ble mean Z500 anomalies at the mid-latitude and subtropi- stantial warm anomaly throughout the MD period
cal sectors of the south Pacific are about two-thirds of the (Figure 4f). Such surface warming may shift the tropical
observed values. Likewise, the MD-averaged ensemble convection or lead to anomalous deep convection locally
mean rainfall anomalies in Central Chile reaches about (Figure 2b), perturbing the tropospheric flow throughout
−17 ± 5%, below the observed value of −30% Rossby wave generation (e.g., Hoskins and Karoly, 1981;
(Figure 10). Nonetheless, five members (out of 40) show Hoskins and Ambrizzi, 1993; Mo and Higgins, 1998). This

FIGURE 8 Scatter plot between multi-year winter (MJJAS) average of the PDO index and the concurrent Central Chile precipitation index
(RPI) average. The periods were selected as sequences of years in which the PDO was predominantly in its cold or warm phase (see inset with the
PDO time series). The MD period in the PDO-RPI space is highlighted. The PDO index is defined as the leading principal component of North
Pacific monthly sea surface temperature variability (poleward of 20 N); monthly values obtained from http://research.jisao.washington.edu/pdo/
[Colour figure can be viewed at wileyonlinelibrary.com]
12 GARREAUD ET AL.

F I G U R E 9 (a) Time series of

winter (MJJAS) mean of sea surface
temperature (NOAA ER-SST) over the
subtropical Southwest Pacific (SSWP,
30 –40 S, 190 –210 W). (b) Trends in
winter mean SST (from NOAA OI-
SST) over the Pacific from 1981
to 2018 [Colour figure can be viewed at
wileyonlinelibrary.com]

perturbation eventually reaches South America in a way that Sensitivity simulations with intermediate SST perturbations
is favourable to dry conditions in Chile. indicate a mostly linear response of central Chile precipita-
To test this hypothesis, we conducted an ad-hoc numeri- tion deficit to the warming over subtropical southwest
cal experiment using SPEEDY (Section 2.3). We first pro- Pacific.
duced a 50-member ensemble of 30-year SPEEDY Specific details on the mechanism linking the SSWP
simulations using climatological mean SST (CTR simula- warming and central Chile drought warrant further investiga-
tions). The ensemble mean of wintertime precipitation, SLP tion. As a first approximation, Figure 13 show the
and Z500 are shown in Figure 12a and reproduce well the streamfunction anomalies for the 0.2101 sigma-level from
key elements of the large scale circulation. We then repeated NCEP NCAR reanalysis. The streamfunction anomalies are
these simulations (50 runs in each case) but adding an SST calculated as the average of the anomalies for individual
perturbation of {+0.5, +1.0, +1.5, +2.0, +2.5} C in the months from May to September in the period from 2010 to
SSWP domain. The difference between SSWP+2.5 and 2017. The period between 1980 and 2010 is used as a base
CTR ensembles in precipitation, SLP and Z500 are shown in for climatologies. The W-vector shows the horizontal wave
Figures 12b,c. The warming of the SSWP produces a local, activity flux (Takaya and Nakamura, 2001) corresponding to
baroclinic response that features a drop in SLP, enhanced this stream function anomaly field. W-vector has been previ-
precipitation and increased Z500. There is also a barotropic ously used to diagnose Rossby wave propagation from the
response downstream of the region of warming that include tropical Pacific into South America in time scales from daily
negative anomalies at higher latitudes, centred over the to seasonal (Montecinos et al., 2011; Rondanelli et al.,
Antarctic Peninsula, and positive anomalies at subtropical 2019). Anticyclonic and cyclonic anomalies in the upper
latitudes, largest over the west coast of South America. atmosphere are located nearly in the same position as the
This pattern strongly projects upon the dipole that favour a 500 hPa geopotential high and low anomalies previously
reduction of precipitation in central Chile (cf. Figures 4 and identified as a dipole. The anomalies have a barotropic
12). Indeed, the SSWP+2.5 wintertime precipitation in cen- equivalent vertical structure. Wave activity flows from the
tral Chile is about 25% lower than its CTR counterpart, a maximum heating region (indicated with a rectangle) and
significant difference considering the role of internal vari- into the polar upper level cyclonic anomaly. The maximum
ability within the various ensemble members (Figure 12d). mid-latitude upper level anticyclonic anomaly also seems to
GARREAUD ET AL. 13

F I G U R E 1 0 Mean anomalies during the MD of 500 hPa geopotential height (top panels) and precipitation (lower panels) from the (a,b)
CAM5 simulations using observed SST, SIC and radiative forcing (AMIP-ORF) and (c,d) CAM5 simulations using observed SST, SIC and past
radiative forcing (AMIP-PRF). The ensemble-mean (40 members) was first calculated for each winter (MJJAS). We then calculated the anomaly by
subtracting the ensemble-mean climatology (1970–2010) and finally we calculated the average over the period 2010–2017. Panels (e,f): Difference
between present-day (2010–2020) minus recent past (1970–2000) in winter (MJJAS) fields on the basis of the multi-model mean from 43 CMIP5
fully-coupled simulations [Colour figure can be viewed at wileyonlinelibrary.com]

contribute with wave activity towards the polar low. Since relevance, the ensemble mean rainfall anomalies in central
W-vectors are nearly parallel to a quasi-stationary Rossby Chile is −8 ± 4% and none of the members produce a def-
wave at each point, poleward wave activity flowing from the icit as large as observed (Figure 11). Thus, while both
surface heating region seems to be consistent with a low CAM5 AMIP-ORF and AMIP-PRF simulations reveal an
wave number quasistationary Rossby wave. A low important ocean-forced component of this drought, this
wavenumber propagation is also consistent with the horizon- forcing seems to produce an event with a severity close to
tal scale of the anomalies, cyclonic and anticyclonic anoma- the observations only when the model incorporates the
lies covering about 90 longitude, translating into horizontal current—anthropogenic—atmospheric levels of GHG, O3
wavenumber 2. At around 75 S and near the Antarctic Pen- and aerosols.
insula, wave activity shifts equatorward and into the The differences between CAM5 AMIP-PRF and AMIP-
Atlantic. ORF may be interpreted as an indication of the anthropo-
genic interference in sustaining the Central Chile MD. A
more direct evidence is provided by the fully coupled simu-
5.2 | Anthropogenic forcing
lations (CMIP5) including all climate forcing (Section 2.3).
We now return to the CAM5 AMIP simulations but now The rightmost panels of Figure 10 shows the multi-model
considering those integration using 1880s radiative forcing mean difference between present (2010–2020) and recent
(CAM5 AMIP-PRF). The overall structure of the ensemble past (1970–2000) global Z500 and precipitation. Note that to
mean Z500 and precipitation anomalies during the MD increase the signal to noise ratio, the averaging period
also agree well with the observations but the amplitude of (2010–2020) for these simulations is slightly larger than the
key features is reduced (Figure 10c,d). Of particular observed MD interval. As noted in previous studies
14 GARREAUD ET AL.

the dynamical pattern that favours dry conditions in cen-

tral Chile (Section 4.1). Indeed, the multi-model mean
precipitation anomalies exhibit a dry band over the sub-
tropical southeast Pacific reaching central Chile
(Figure 10d) resembling the observations. The CMIP5
simulations also predicts an overall increase in SST but
less marked along the west coast of South America than
in the rest of the Pacific. A more quantitative view of the
climate change impacts on central Chile in present time
can be obtained from the distribution of rainfall anoma-
lies from the 26 CMIP5 models for the period 2010–2020
(Figure 11). The multi-model mean deficit for central
Chile is −6 ± 3%, so anthropogenic-induced precipitation
changes appears to explain about a quarter of the
observed rainfall anomaly in this region. Note that the
mean rainfall deficit “predicted” by the CMIP5 simula-
tions is similar to the RPF value (although the last num-
ber comes from just one model). To test its significance
we employed the historical simulations (1850–2005) of
the same 26 models. To avoid a major influence of the
anthropogenic forcing we only consider the period
1900–1950 and from there we randomly selected ten
10-year periods. In each of those periods, we calculated
the multi-model mean central Chile rainfall anomalies,
F I G U R E 1 1 Summary of the Central Chile MD mean rainfall
and then we obtained the ensemble mean rainfall statis-
deficit from different simulations. The winter (MJJAS) rainfall anomaly
tics. The multi-model mean over this synthetic
over the region 32.5–37.5 S, 74–70 W was calculated for every year
and family of simulations. We then calculated the yearly anomalies by “unperturbed” period is close to zero (+0.3%) with a SD
subtracting the climatology (1970–2010), that were averaged during of 2.2%. On the basis of these values, we posit that the
the MD period. Each family of simulations include multiple runs (for predicted rainfall anomalies (−6 ± 3%) for the MD are
CAM5) or multiple models (form CMIP5), that allows to construct box significant at the 95% confidence level against the null
plots showing the median value (colour line), interquartile range (pink hypothesis of no anthropogenic driven drying and repre-
box) and the [5%,95%] range (black vertical line). The observed sents about a quarter of the total (observed) signal (~25%
(station based) MD rainfall deficit is also indicated. The red and green rainfall deficit). Such level of the human contribution was
lines correspond to the average during 2010–2017 and 2010–2016,
also reported by Boisier et al. (2016) and is much larger
respectively [Colour figure can be viewed at wileyonlinelibrary.com]
than its counterpart in the 2011–2014 California droughts
(e.g., Seager et al., 2015). Yet, the diagnosed rainfall
(e.g., Fyfe et al., 1999; Arblaster et al., 2011; Gillett and anomalies vary substantially among the models, probably
Fyfe, 2013; Boisier et al., 2016), the anthropogenic forcing because of the short difference in time between the pre-
leads to a marked Southern Annular Mode pattern in its pos- sent (2010–2020) and the past (1970–2000) periods. In
itive polarity, with negative height anomalies over the Ant- such case, internal variability within each model can be
arctic periphery and positive anomalies at lower latitudes. stronger than the forced signal (Deser et al., 2012). We
The Z500 anomalies found in the 2010–2020 period are part also verified that among the CMIP5 simulations those
of gradual trends over the subtropical Pacific and ABS of predicting the largest drying in central Chile (15–20%
the same sign of their observed counterparts, evident in the rainfall deficit) feature an SST anomaly pattern resem-
multi-model mean time series of 500 hPa geopotential height bling the observed in the last decade (warm in the west
in the boxes defined before (Figure 6). Notably, the positive Pacific, cool in the far east; not shown). This SST pattern
height trend since the late 1970s over the subtropics is quite emerged in those simulations given its own internal,
comparable with its observed counterpart, but the simulated coupled variability and stress the hypothesis that direct
height decline over the ABS is weaker than the reanalysis (atmospheric-circulation) anthropogenic forcing can cause
value. a drought as observed only if coupled with a La-Niña like
The positive SAM polarity shown by CMIP5 models SST pattern of enough longevity (i.e., during the cold
for the present decade (Figure 10e) strongly projects upon phase of PDO).
GARREAUD ET AL. 15

F I G U R E 1 2 (a) Ensemble mean

(50 members), multi-year mean of
winter precipitation (colour), sea level
pressure (SLP, grey contours every
3 hPa) and 500 hPa geopotential height
(Z500, yellow contours at 5250, 5400
and 5,600 m) from a 30-year long
SPEEDY control simulation (CTR)
forced by climatological monthly mean
SST. (b) Differences of ensemble mean
precipitation (colours) and SLP
(contours, every 0.15 hPa) between
SSWP+2.5 minus CTR simulations.
SSWP+2.5 is identical to CTR but for
a permanent warm anomaly of +2.5 C
in the subtropical Southwest Pacific
(30–40 S, 190–210 W). (c) Differences
of ensemble mean precipitation
(colours) and Z500 (contours, every
7.5 m) between SSWP+2.5 minus CTR
simulations. (d) Multi-year winter mean
precipitation over Central Chile
(32.5–37.5 S, 74–70 W) for the
climatological SPEEDY simulations.
The circle is the ensemble mean and the
vertical lines indicate the interquartile
range. The horizontal axis indicate the
SST anomaly imposed in the SSWP
region [Colour figure can be viewed at
wileyonlinelibrary.com]

F I G U R E 1 3 NNR
streamfunction anomalies (colours in
m2/s) and W vectors (m2/s2). Both
fields are calculated from monthly
anomalies in the period 2010–2017 for
the months from may to September.
W-vectors after Takaya and Nakamura
(2001). Climatologies are with respect
to the period 1980 to 2010. H and L
indicate the center of anticyclonic and
cyclonic anomalies, respectively. The
black rectangle shows the subtropical
Southwest Pacific (SSWP) region
[Colour figure can be viewed at
wileyonlinelibrary.com]

6 | CONCLUDING REMARKS observed in water availability, vegetation and forest fires

(Garreaud et al., 2017). In this work, we have used observa-
The central Chile MD(2010–2018) has been the longest tional datasets to document regional- and large-scale features
sequence of dry years during the observational period (1914 of the MD, emphasizing those aspects that differ from previ-
onwards) and has few (if any) analogues in the last ous dry events. We also took advantage of atmospheric-only
millennia, perhaps heralding the dry conditions projected for and fully coupled GCMs to gauge the role of natural vari-
this region during the rest of the 21st century (Boisier et al., ability and anthropogenic forcing in sustained such extended
2018; Bozkurt et al., 2018). Its detrimental effects have been drought. Our main findings are as follow:
16 GARREAUD ET AL.

• The exceptional length of the MD results from an SAM trend has been largely attributed to stratospheric O3
uninterrupted reiteration of a large-scale circulation pat- depletion and increased GHG concentration (Arblaster
tern during austral winter hindering the passage of and Meehl, 2006; Eyring et al., 2013; Gillett et al., 2013).
extratropical storms over central Chile. This pattern is Indeed, fully-coupled GCM forced by observed radiative
characterized by tropospheric-deep positive pressure forcing also produce a dry band over the subtropical
(anticyclonic) anomalies over the central-eastern subtropi- southeast Pacific reaching central Chile, with a multi-
cal Pacific and negative pressure (cyclonic) anomalies model mean deficit of about 5% for the present decade
over the Amundsen-Bellingshausen Sea. The strength of relative to the recent past.
this dipole has been substantially above the long-term • A subset of CMIP5 simulations produce a rainfall deficit
mean during the MD. closer to the observed values and feature a La-Niña like
• Historically, ENSO is a major modulator of the south SST anomaly pattern, stressing again the relevance of the
Pacific dipole, producing low (high) pressure anomalies ocean sourced forcing of the MD, that is currently
over the ABS (subtropical latitudes) during La Niña enhanced by changes in atmospheric circulation driven by
years. Unlike historical droughts, however, the ongoing anthropogenic forcing.
dry period has occurred mostly under ENSO-neutral con-
ditions, except for the winters of 2010 (La Niña) and In summary, both natural variability (PDO shift causing
2015 (strong El Niño). This has resulted in a substantial SSWP warming) and anthropogenic forcing (SAM trends
decrease in the ability to foresee central Chile hydro- towards its positive polarity) are at play in sustaining the
climate based on ENSO predictions (J. Vicencio, personal Central Chile Mega Drought. While it is not possible to
communication).
obtain an exact partitioning, the seemingly natural, ocean-
• Nonetheless, atmospheric-only GCM simulations forced
forced component appears as dominant, accounting for at
with observed global SST and present-day radiative forc-
least half of the dry signal in central Chile. Given its origin
ing (AMIP ensembles) replicates the south Pacific dipole
in ocean–atmosphere variability, a reversal of its sign can be
and the area of negative rainfall anomalies during the
expected for the next decades, partially relieving the dry
MD. The simulated precipitation deficit is 1/2 to 2/3 of
conditions afflicting central Chile. Nevertheless, the anthro-
the observed values, and much smaller if the radiative
pogenic forcing is also important—about a quarter of the
forcing is set to pre-industrial levels. These results con-
MD signal—and it will keep pushing Central Chile towards
firm that the protracted Central Chile MD is largely
a dry condition during the rest of 21st century with an inten-
driven by global SST anomalies of the last decades under
sity that depend upon the emission scenario that humanity
the current (anthropogenic) levels of greenhouse gases
will follow.
and stratospheric ozone.
• Further numerical experiments suggest that most of the
atmospheric forcing emanates from the subtropical south- A C KN O W L E D G E M E N T S
west Pacific (SSWP), although specific details of such
This research emerged from the collaboration with many
link deserve further investigation. This area has experi-
colleagues at the Center for Climate and Resilience Research
enced a marked sea surface warming over the last
(CR2, CONICYT/FONDAP/15110009) that also provided
20 years, with mean anomalies over +1 C during the MD
partial funding. R.G. also acknowledges for support by Fon-
period. Such warming is capable to excite a Rossby
decyt project 1170286. We thank two anonymous reviewers
waves response whose propagation lowers the pressure
for constructive criticism and comments on the manuscript.
over the ABS and increases the pressure over western
South America, projecting upon the pattern that lead to
dry conditions in central Chile. OR CI D
• The origin of the SSWP warming is unclear, but partially
René D. Garreaud https://orcid.org/0000-0002-7875-2443
connected with the shift in the PDO from its positive to
negative polarity over the past decades (e.g., Saurral
et al., 2018). Yet, climate change projections also reveals REF ER ENC ES
substantial west–east asymmetries in the SST change
Aceituno, P. (1988) On the functioning of the southern oscillation in
across the south Pacific at subtropical latitudes, thus, con-
the south American sector. Part I: surface climate. Monthly Weather
tributing to the anomalous warming of the SSWP. Review, 116, 505–524.
• The tendency of the Southern Annular Mode towards its Adler, R.F., Huffman, G.J., Chang, A., Ferraro, R., Xie, P.-P.,
positive polarity also contributes to strength of the dipole Janowiak, J., Rudolf, B., Schneider, U., Curtis, S. and Bolvin, D.
leading to rainfall deficit in central Chile. In turn, the (2003) The version-2 global precipitation climatology project
GARREAUD ET AL. 17

(GPCP) monthly precipitation analysis (1979–present). Journal of Dee, D., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P.,
Hydrometeorology, 4, 1147–1167. Kobayashi, S., Andrae, U., Balmaseda, M.A., Balsamo, G.,
AghaKouchak, A., Cheng, L., Mazdiyasni, O. and Farahmand, A. Bauer, P., Bechtold, P., Beljaars, A.C.M., van de Berg, L.,
(2014) Global warming and changes in risk of concurrent climate Bidlot, J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M.,
extremes: insights from the 2014 California drought. Geophysical Geer, A.J., Haimberger, L., Healy, S.B., Hersbach, H., Hólm, E.V.,
Research Letters, 41, 8847–8852. https://doi.org/10.1002/ Isaksen, L., Kållberg, P., Köhler, M., Matricardi, M., McNally, A.
2014GL062308. P., Monge-Sanz, B.M., Morcrette, J.-.J., Park, B.-.K., Peubey, C.,
Aldunce, P., Araya, D., Sapiains, R., Ramos, I., Lillo, G., Urquiza, A. de Rosnay, P., Tavolato, C., Thépaut, J.-.N. and Vitart, F. (2011)
and Garreaud, R. (2017) Local perception of drought impacts in a The ERA-interim reanalysis: configuration and performance of the
changing climate: the mega-drought in Central Chile. Sustainabil- data assimilation system. Quarterly Journal of the Royal Meteoro-
ity, 9, 2053. https://doi.org/10.3390/su9112053. logical Society, 137, 553–597. https://doi.org/10.1002/qj.828.
Arblaster, J.M., Meehl, G.A. and Karoly, D.J. (2011) Future climate Deser, C., Phillips, A., Bourdette, V. and Teng, H. (2012) Uncertainty
change in the southern hemisphere: competing effects of ozone and in climate change projections: the role of internal variability. Cli-
greenhouse gases. Geophysical Research Letters, 38(2), L02701. mate Dynamics, 38, 527–546.
https://doi.org/10.1029/2010GL045384. England, M.H., McGregor, S., Spence, P., Meehl, G.,
Arblaster, J.M. and Meehl, G.A. (2006) Contributions of external forc- Timmermann, A., Cai, W., Gupta, A.S., McPhaden, M., Purich, A.
ings to southern annular mode trends. Journal of Climate, 19, and Santoso, A. (2014) Recent intensification of wind-driven circu-
2896–2905. https://doi.org/10.1175/JCLI3774.1. lation in the Pacific and the ongoing warming hiatus. Nature Cli-
Boisier, J.P., Rondanelli, R., Garreaud, R.D. and Muñoz, F. (2016) mate Change, 4, 222–227.
Anthropogenic and natural contributions to the Southeast Pacific Eyring, V., Arblaster, J., Cionni, I., Sedlácek, J., Perlwitz, J.,
precipitation decline and recent megadrought in Central Chile. Geo- Young, P., Bekki, S., Bergmann, D., Cameron-Smith, P. and
physical Research Letters, 43, 413–421. Collins, J. (2013) Long-term ozone changes and associated climate
Boisier, J.P., Alvarez-Garreton, C., Cordero, R.R., Damian, A., impacts in CMIP5 simulations. Journal of Geophysical Research:
Gallardo, L., Garreaud, R.D., Lambert, F., Ramallo, C., Rojas, M. Atmospheres, 118, 5029–5060.
Falvey, M. and Garreaud, R. (2007) Wintertime precipitation epi-
and Rondanelli, R. (2018) Anthropogenic drying in Central-
sodes in Central Chile: associated meteorological conditions
Southern Chile evidenced by long term observations and climate
and orographic influences. Journal of Hydrometeorology, 8,
model simulations. Elementa: Science of the Anthropocene, 6, 74.
171–193.
https://doi.org/10.1525/elementa.328.
Fogt, R., Wovrosh, A., Langen, R. and Simmonds, I. (2012) The char-
Bozkurt, D., Rojas, M., Boisier, J.P. and Valdivieso, J. (2018) Projected
acteristic variability and connection to the underlying synoptic
hydroclimate changes over Andean basins in Central Chile from
activity of the Amundsen-Bellingshausen seas low. Journal of Geo-
downscaled CMIP5 models under the low and high emission sce-
physical Research, 117, D07111. https.//doi.org//10.1029/
narios. Climatic Change, 150, 131–147.
2011JD017337.
Cai, W., Purich, A., Cowan, T., van Rensch, P. and Weller, E. (2014)
Fuenzalida, H., Aceituno, P., Falvey, M., Garreaud, R., Rojas, M. and
Did climate change–induced rainfall trends contribute to the
Sanchez, R. (2007) Study on Climate Variability for Chile during
Australian millennium drought? Journal of Climate, 27,
the 21st Century. Technical Report. Santiago, Chile: National Envi-
3145–3168.
ronmental Committee. Available at: http://dgf.uchile.cl/PRECIS.
Capotondi, A., Wittenberg, A.T., Newman, M., Di Lorenzo, E., Yu, J.-
Accessed March 5, 2015.
Y., Braconnot, P., Cole, J., Dewitte, B., Giese, B. and Guilyardi, E. Fyfe, J., Boer, G. and Flato, G. (1999) The Arctic and Antarctic oscilla-
(2015) Understanding ENSO diversity. Bulletin of the American tions and their projected changes under global warming. Geophysi-
Meteorological Society, 96, 921–938. cal Research Letters, 26, 1601–1604.
Cionni, I., Eyring, V., Lamarque, J., Randel, W., Stevenson, D., Garcia-Herrera, R., Hernández, E., Barriopedro, D., Paredes, D.,
Wu, F., Bodeker, G., Shepherd, T., Shindell, D. and Waugh, D. Trigo, R.M., Trigo, I.F. and Mendes, M. (2007) The outstanding
(2011) Ozone database in support of CMIP5 simulations: results 2004/05 drought in the Iberian Peninsula: associated atmospheric
and corresponding radiative forcing. Atmospheric Chemistry and circulation. Journal of Hydrometeorology, 8, 483–498.
Physics, 11, 11267–11292. Garreaud, R., Lopez, P., Minvielle, M. and Rojas, M. (2013) Large
Clem, K.R. and Fogt, R.L. (2015) South Pacific circulation changes scale control on the Patagonia climate. Journal of Climate, 26,
and their connection to the tropics and regional Antarctic warming 215–230.
in austral spring, 1979–2012. Journal of Geophysical Research Garreaud, R., Vuille, M., Compagnucci, R. and Marengo, J. (2009) Pre-
Atmosphere, 120, 2773–2792. https://doi.org/10.1002/2014JD02 sent-day south american climate. Palaeogeography, Pal-
2940. aeoclimatology, Palaeoecology, 281, 180–195.
CR2: Report to the Nation: The Central Chile Mega-Drought. Techni- Garreaud, R. (2007) Precipitation and circulation covariability in the
cal report from the Center for Climate and Resilience Research. extratopics. Journal of Climate, 20, 4789–4797.
30 pp. Santiago-Chile. Available at: http://www.cr2.cl/megasequia, Garreaud, R.D., Alvarez-Garreton, C., Barichivich, J., Boisier, J.P.,
2015. Christie, D., Galleguillos, M., LeQuesne, C., McPhee, J. and
Dai, A. (2011) Drought under global warming: a review. Wiley Inter- Zambrano-Bigiarini, M. (2017) The 2010–2015 megadrought in
disciplinary Reviews: Climate Change, 2, 45–65. Central Chile: impacts on regional hydroclimate and vegetation.
Dai, A. (2013) Increasing drought under global warming in observa- Hydrol Earth Syst Sci, 21, 6307–6327. https://doi.org/10.5194/
tions and models. Nature Climate Change, 3, 52–58. hess-21-6307-2017.
18 GARREAUD ET AL.

Gates, W., Boyle, J., Covey, C., Dease, C., Doutriaux, C., Drach, R., Kucharski, F., Molteni, F., King, M., Farneti, R., Kang, I. and
Fiorino, M., Gleckler, P., Hnilo, J., Marlais, S., Phillips, T., Feudale, L. (2013) On the need of intermediate complexity general
Potter, G., Santer, B., Sperber, K., Taylor, K. and Williams, D. circulation models: a “SPEEDY” example. Bulletin of the American
(1998) An overview of the results of the atmospheric model Meteorological Society, 94, 25–30.
Intercomparison project (AMIP I ). Bulletin of the American Meteo- Mao, Y., Nijssen, B. and Lettenmaier, D. (2015) Is climate change
rological Society, 73, 1962–1970. implicated in the 2013–2014 California drought? A hydrologic per-
Gillett, N., Fyfe, J. and Parker, D. (2013) Attribution of observed sea spective. Geophysical Research Letters, 42, 2805–2813.
level pressure trends to greenhouse gas, aerosol, and ozone Marshall, G. (2003) Trends in the southern annular mode from observa-
changes. Geophysical Research Letters, 40(10), 2302–2306. tions and reanalyses. Journal of Climate, 16, 4134–4143.
Gillett, N. and Fyfe, J. (2013) Annular mode changes in the CMIP5 Masiokas, M., Christie, D., Luckman, B. and Nussbaumer, S. (2016)
simulations. Geophysical Research Letters, 40, 1189–1193. Reconstructing the annual mass balance of the Echaurren Norte gla-
González, M.E., Gómez-González, S., Lara, A., Garreaud, R. and Díaz- cier (Central Andes, 33.5 S) using local and regional hydroclimatic
Hormazábal, I. (2018) The 2010-2015 Megadrought and its influ- data. The Cryosphere, 10, 927.
ence on the fire regime in central and south Central Chile. Eco- Masiokas, M., Villalba, R., Luckman, B. and Mauget, S. (2010) Intra-
sphere, 9, 1–17. https://doi.org/10.1002/ecs2.2300. to multidecadal variations of snowpack and streamflow records in
González-Reyes, A.,  McPhee, J., Christie, D., Le Quesne, C., the Andes of Chile and Argentina between 30 and 37 S. Journal of
Szejner, P., Masiokas, M., Villalba, R., Muñoz, A. and Crespo, S. Hydrometeorology, 11(3), 822–831.
(2017) Spatiotemporal variations in Hydroclimate across the Medi- Masotti, I., Aparicio-Rizzo, P., Yevenes, M., Garreaud, R.,
terranean Andes (30 –37 S) since the early twentieth century. Belmar, L. and Farias, L. (2018) The influence of river discharge
Journal of Hydrometeorology, 18(7), 1929–1942. on nutrient export and phytoplankton biomass off the Central
Griffin, D. and Anchukaitis, K.J. (2014) How unusual is the Chile coast (33 -37 S). Seasonal cycle and interannual variability.
2012–2014 California drought? Geophysical Research Letters, 41, Frontiers in Marine Science, 5, 423. https://doi.org/10.3389/
9017–9023. https://doi.org/10.1002/2014GL062433. fmars.2018.00423.
Hoerling, M., Eischeid, J. and Perlwitz, J. (2010) Regional precipitation Meinshausen, M., Smith, S., Calvin, K., Daniel, J., Kainuma, M.,
trends: distinguishing natural variability from anthropogenic forc- Lamarque, J., Matsumoto, K., Montzka, S., Raper, S. and Riahi, K.
ing. Journal of Climate, 23, 2131–2145. (2011) The RCP greenhouse gas concentrations and their exten-
Hoskins, B. and Ambrizzi, T. (1993) Rossby wave propagation on a sions from 1765 to 2300. Climatic Change, 109(1–2), 213–241.
realistic longitudinally varying flow. Journal of the Atmospheric Miller, A. (1976) The climate of Chile. In: Schwrdtfeger, W. (Ed.) The
Sciences, 50, 1661–1671. Climates of Central and South America. Amsterdam: Elsevier,
Hoskins, B. and Karoly, D. (1981) The steady linear response of a pp. 113–145.
spherical atmosphere to thermal and orographic forcing. Journal of Mo, K. and Higgins, R. (1998) The Pacific–south American modes and
the Atmospheric Sciences, 38, 1179–1196. tropical convection during the southern hemisphere winter. Monthly
Hurrell, J.W., Hack, J., Shea, D., Caron, J. and Rosinski, J. (2008) A Weather Review, 126, 1581–1596.
new sea surface temperature and sea ice boundary dataset for the Molteni, F. (2003) Atmospheric simulations using a GCM with simpli-
community atmosphere model. Journal of Climate, 21, fied physical parametrizations. I: model climatology and variability
5145–5153. in multi-decadal experiments. Climate Dynamics, 20, 175–191.
Jacques-Coper, M. and Garreaud, R. (2014) Characterization of Montecinos, A., Diaz, A. and Aceituno, P. (2000) Seasonal diagnostic
the 1970s climate shift in South America. International Jour- and predictability of rainfall in subtropical South America based on
nal of Climatology, 35, 2164–2179. https://doi.org/10.1002/joc. tropical Pacific SST. Journal of Climate, 13, 746–758.
4120. Montecinos, A., Kurgansky, M.V., Muñoz, C. and Takahashi, K.
Jones, J.M., Gille, S., Goosse, H., Abram, N., Canziani, P., (2011) Non-ENSO interannual rainfall variability in Central Chile
Charman, D., Clem, K., Crosta, X., de Lavergne, C. and during austral winter. Theoretical and Applied Climatology, 106,
Eisenman, L. (2016) Assessing recent trends in high-latitude south- 557–568.
ern hemisphere surface climate. Nature Climate Change, 6, Montecinos, A. and Aceituno, P. (2003) Seasonality of the ENSO-
917–926. related rainfall variability in Central Chile and associated circula-
Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., tion anomalies. Journal of Climate, 16, 281–296.
Gandin, L., Iredell, M., Saha, S., White, G. and Woollen, J. (1996) Moss, R.H., Edmonds, J., Hibbard, K., Manning, M., Rose, S.,
The NCEP/NCAR 40-year reanalysis project. Bulletin of the Ameri- Vuuren, V., Carter, T., Emori, S., Kainuma, M. and Kram, T.
can Meteorological Society, 77, 437–471. (2010) The next generation of scenarios for climate change research
Kay, J., Deser, C., Phillips, A., Mai, A., Hannay, C., Strand, G., and assessment. Nature, 463, 747–756.
Arblaster, J., Bates, S., Danabasoglu, G. and Edwards, J. (2015) Muñoz, A., González-Reyes, A., Lara, A., Sauchyn, D., Christie, D.,
The community earth system model (CESM) large ensemble pro- Puchi, R., Urrutia-Jalabert, R., Toledo-Guerrero, I., Aguilera-
ject: a community resource for studying climate change in the pres- Betti, I. and Mundo, I. (2016) Streamflow variability in the Chilean
ence of internal climate variability. Bulletin of the American temperate-Mediterranean climate transition (35 S–42 S) during the
Meteorological Society, 96, 1333–1349. last 400 years inferred from tree-ring records. Climate Dynamics,
Kelley, C.P., Mohtadi, S., Cane, M., Seager, R. and Kushnir, Y. (2015) 47, 4051–4066.
Climate change in the Fertile Crescent and implications of the Quintana, J. and Aceituno, P. (2012) Changes in the rainfall regime
recent Syrian drought. Proceedings of the National Academy of Sci- along the extratropical west coast of South America (Chile): 30-43
ences, 112, 3241–3246. S. Atmosfera, 25, 1–22.
GARREAUD ET AL. 19

Rasmusson, E. and Carpenter, T. (1982) Variations in tropical sea sur- Taylor, K., Stouffer, R. and Meehl, G. (2012) An overview of CMIP5
face temperature and surface wind fields associated with the south- and the experiment design. Bulletin of the American Meteorological
ern oscillation/El Niño. Monthly Weather Review, 110, 354–384. Society, 93(4), 485–498.
Rayner, N., Parker, D., Horton, E., Folland, C., Alexander, L., Trigo, R., Gouveia, C. and Barriopedro, D. (2010) The intense
Rowell, D., Kent, E. and Kaplan, A. (2003) Global analyses of sea 2007–2009 drought in the Fertile Crescent: impacts and associated
surface temperature, sea ice, and night marine air temperature since atmospheric circulation. Agricultural and Forest Meteorology, 150,
the late nineteenth century. Journal of Geophysical Research, 108 1245–1257.
(D14), 4407. https.//doi.org.//10.1029/2002JD002670. Van Dijk, A., Beck, H., Crosbie, R., De Jeu, R., Liu, Y., Podger, G.,
Reynolds, R., Rayner, N., Smith, T., Stokes, D. and Wang, W. (2002) Timbal, B. and Viney, N. (2013) The millennium drought in
An improved in situ and satellite SST analysis for climate. Journal Southeast Australia (2001–2009): natural and human causes and
of Climate, 15(13), 1609–1625. implications for water resources, ecosystems, economy, and society.
Rivera, J., Penalba, O., Villalba, R. and Araneo, D. (2017) Spatio- Water Resource Research, 49, 1040–1057. https://doi.org/10.1002/
temporal patterns of the 2010–2015 extreme hydrological drought wrcr.20123.
across the Central Andes, Argentina. Water, 9, 652. Vera, C. and Díaz, L. (2015) Anthropogenic influence on summer pre-
Rondanelli, R., Hatchett, B., Rutllant, J., Bozkurt, D. and Garreaud, R. cipitation trends over South America in CMIP5 models. Interna-
(2019) Strongest MJO on record triggers extreme Atacama rainfall tional Journal of Climatology, 35(10), 3172–3177. https://doi.org/
and warmth in Antarctica. Geophysical Research Letters, 46, 10.1002/joc.4153.
3482–3491. https://doi.org/10.1029/2018GL081475. Verbist, K., Amani, A., Mishra, A. and Cisneros, B. (2016) Strengthen-
Rouault, M. and Richard, Y. (2003) Intensity and spatial extension of ing drought risk management and policy: UNESCO international
drought in South Africa at different time scales. Water SA, 29, 489–500. hydrological Programme's case studies from Africa and Latin
Rutllant, J. and Fuenzalida, H. (1991) Synoptic aspects of the Central America and the Caribbean. Water Policy, 18, 245–261.
Chile rainfall variability associated with the southern oscillation. Viale, M. and Garreaud, R. (2015) Orographic effects of the sub-
International Journal of Climatology, 11, 63–76. tropical and extratropical Andes on upwind precipitating clouds.
Saft, M., Western, A., Zhang, L., Peel, M. and Potter, N. (2015) The Journal of Geophysical Research: Atmospheres, 120, 4962–
influence of multiyear drought on the annual rainfall-runoff rela- 4974.
tionship: an Australian perspective. Water Resource Research, 51, Viale, M. and Nuñez, M. (2011) Climatology of winter orographic pre-
2444–2463. https://doi.org/10.1002/2014WR015348. cipitation over the subtropical Central Andes and associated synop-
Saurral, R., Doblas-Reyes, F. and García-Serrano, J. (2018) Observed tic and regional characteristics. Journal of Hydrometeorology, 12,
modes of sea surface temperature variability in the South Pacific 481–507.
region. Climate Dynamics, 50, 1129–1143. Vicente-Serrano, S., López-Moreno, J., Gimeno, L., Nieto, R., Morán-
Schubert, S., Stewart, R., Wang, H., Barlow, M., Berbery, E., Cai, W., Tejeda, E., Lorenzo-Lacruz, J., Beguería, S. and Azorin-Molina, C.
Hoerling, M., Kanikicharla, K., Koster, R. and Lyon, B. (2016) (2011) A multiscalar global evaluation of the impact of ENSO on
Global meteorological drought: a synthesis of current understand- droughts. Journal of Geophysical Research, 116, D20109. https.//
ing with a focus on SST drivers of precipitation deficits. Journal of doi.org//10.1029/2011JD016039.
Climate, 29, 3989–4019. Volkov, D., Lee, S., Landerer, F. and Lumpkin, R. (2017) Decade-long
Seager, R., M., Schubert, S., Wang, H., Lyon, B., Kumar, A., deep-ocean warming detected in the subtropical South Pacific. Geo-
Nakamura, J. and Henderson, N. (2015) Causes of the 2011–14 physical Research Letters, 44, 927–936.
California drought. Journal of Climate, 28, 6997–7024. Williams, A., Seager, R., Abatzoglou, J., Cook, B., Smerdon, J. and
Seneviratne, S., et al. (2012) In: Field, C.B. , et al. (Eds.) Managing the Cook, E. (2015) Contribution of anthropogenic warming to Califor-
Risks of Extreme Events and Disasters to Advance Climate Change nia drought during 2012–2014. Geophysical Research Letters, 42,
Adaptation. Cambridge, MA: IPCC, Cambridge University Press, 6819–6828.
pp. 109–230. You, Y. and Furtado, J. (2017) The role of South Pacific atmospheric
Sherwood, S., Meyer, C., Allen, R. and Titchner, H. (2008) Robust tro- variability in the development of different types of ENSO. Geo-
pospheric warming revealed by iteratively homogenized radiosonde physical Research Letters, 44, 7438–7446.
data. Journal of Climate, 21, 5336–5352.
Smith, T., Reynolds, R., Peterson, T. and Lawrimore, J. (2008)
Improvements to NOAA's historical merged land–ocean surface
temperature analysis (1880–2006). Journal of Climate, 21,
S UP P OR T I N G I NF OR M A T I ON
2283–2296. Additional supporting information may be found online in
Solman, S. and Menéndez, C. (2002) ENSO-related variability of the
the Supporting Information section at the end of this article.
southern hemisphere winter storm track over the eastern Pacific–
Atlantic sector. Journal of the Atmospheric Sciences, 59, 2128–2141.
Swain, D. (2015) A tale of two California droughts: lessons amidst How to cite this article: Garreaud RD, Boisier JP,
record warmth and dryness in a region of complex physical and
Rondanelli R, Montecinos A, Sepúlveda HH, Veloso-
human geography. Geophysical Research Letters, 42, 9999–10,003.
Takaya, K. and Nakamura, H. (2001) A formulation of a phase-
Aguila D. The Central Chile Mega Drought (2010–
independent wave-activity flux for stationary and migratory 2018): A climate dynamics perspective. Int J
quasigeostrophic eddies on a zonally varying basic flow. Journal of Climatol. 2019;1–19. https://doi.org/10.1002/joc.6219
the Atmospheric Sciences, 58, 608–627.