Strong large-scale climate response to North American sulphate aerosols in CESM

The climate of Mexico and the United States (U.S.) has undergone substantial temperature and precipitation changes in the recent decades. For instance, a rapid warming has been identified (Pachauri et al 2014, Wuebbles et al 2017), in line with the current global trend. Annual precipitation has decreased over central and southern Mexico, while a positive trend has been observed in northern Mexico and most of the U.S. (Pachauri et al 2014, Wuebbles et al 2017). Besides, the intensity and severity of droughts in some regions of the U.S. and Mexico have increased (Stahle et al 2009, Wuebbles et al 2017, Vega-Camarena et al 2018). This has had profound impacts on society, water resources, and the local economy (Stocker et al 2013 and references therein). Furthermore, CMIP5 models project strong additional warming and a large precipitation reduction over Mexico throughout the 21st century, particularly during the summer, although with considerable uncertainty (Karmalkar et al 2011, Taylor et al 2012, Stocker et al 2013, Colorado-Ruiz et al 2018). Over the U.S., temperature is also projected to rise in the coming decades, while precipitation changes are dependent on the location and the season, with a drying trend in most of the U.S. during summer and more precipitation during winter in the northern states (Wuebbles et al 2017).

Aside greenhouse gases (GHGs), anthropogenic aerosols currently exert a considerable forcing on the Earth's radiative balance (Stocker et al 2013). In particular, the global radiative forcing from sulphate aerosols during the 20th century is estimated to be of the same order of magnitude, but of opposite sign, to that of GHGs (Pachauri et al 2014). Sulphate aerosol emissions have been declining worldwide since the early 1980s and are projected to decrease by up to 80% by the end of the 21st century, leading to an amplification of the GHG-related warming of up to 1°C globally and even more at regional scale (Westervelt et al 2015). Yet, aerosols represent the largest uncertainty in current estimates of human-driven climate change (Myhre et al 2014) due to compounding uncertainties associated with model representations of poorly-known aerosol processes, and with the estimation of aerosol emissions.

Anthropogenic aerosols can modify the climate by scattering or absorbing solar radiation, or by changing cloud properties and precipitation processes (e.g. Twomey 1977, Albrecht 1989, Charlson et al 1992, Ming and Ramaswamy 2009, Boucher et al 2013 and references therein). Worldwide, aerosols have been found to play a major role in driving the late 20th century weakening of the monsoon over South Asia (Bollasina et al 2011, Undorf et al 2018b), East Asia (Song et al 2014), and West Africa (Undorf et al 2018b), as well as in modulating multidecadal variability in sea surface temperature over the North Atlantic (Booth et al 2012, Undorf et al 2018a). Even though North America was one of the largest contributors (along with Europe) to global aerosol emissions, particularly of sulphur dioxide (SO₂, precursor of sulphate aerosols), up to the 1980s (Hoesly et al 2018), only a few studies have examined the climate response to anthropogenic aerosol variations over this region. Leibensperger et al (2012) found a cooling of 0.5 to 1 °C over the central and eastern U.S. in response to increased U.S. anthropogenic aerosols during 1970–1990. Westervelt et al (2017) reported a considerable rainfall increase over the central and eastern U.S. and over the North Atlantic associated with the recent SO₂ emissions decline. Yet, the physical mechanisms underlying these changes remain unclear.

The case for North America is particularly relevant as while surface temperature increased worldwide, a cooling trend (the so-called 'warming hole'), was observed over the southern U.S. from the early 1950s to the mid 1970s (e.g. Robinson et al 2002, Wang et al 2009, Leibensperger et al 2012). Yet, there is no consensus on the factors driving this muted warming, with some works emphasising the impact of aerosols (Leibensperger et al 2012, Yu et al 2014, Mascioli et al 2017), others the role of internal climate variability (mainly through teleconnections with Pacific sea surface temperatures; Robinson et al 2002, Wang et al 2009, Banerjee et al 2017) or possibly the combined effect of both (Kunkel et al 2006, Portmann et al 2009).

A better understanding of the regional as well as large-scale climate response to the 20th century changes in North American aerosol emissions is key to achieve more robust near-future projections in this highly vulnerable region (Karmalkar et al 2011). In this study, we assess the summertime climate impact of North American anthropogenic sulphate emissions using a state-of-the-art climate model and identify the underpinning mechanisms.

This study makes use of 8-member ensembles of transient coupled experiments with the U.S. National Center for Atmospheric Research (NCAR) Community Earth System Model (CESM) version 1.2.2 (Hurrell et al 2013). Model setup and experiments are thoroughly described in Undorf et al (2018a). The atmospheric component is the Community Atmosphere Model version 5.3 (Neale et al 2012), which uses a 3-mode aerosol scheme (MAM3, Ghan et al 2012) and includes a full prognostic representation of aerosol-cloud interactions (Ghan et al 2012, Meehl et al 2013). We analyse an all-forcing ensemble (ALL) driven by time-varying historical emissions from both natural and anthropogenic sources, and a perturbed ensemble identical to ALL but with anthropogenic emissions of sulphate aerosols and sulphur dioxide from North America (continental U.S. and Canada) fixed at their pre-industrial levels (NoNA). Assuming linearity in the combined responses (which has been shown to be a reasonable approach, e.g. Polson et al 2014), the difference ALL minus NoNA indicates, to a first order approximation, the impact of North American aerosols. Note that non-linear interactions between aerosols and other forcings (such as GHGs and remote aerosols) are removed in the NoNA ensemble, as these only arise when North American aerosols are present. Such an assumption is routinely made in studies investigating the impact of global forcing factors (e.g. Gillett et al 2016), and more specifically, that of regional aerosol emissions (e.g. Bollasina et al 2014, Persad and Caldeira 2018, Undorf et al 2018b, Westervelt et al 2018, Wilcox et al 2019). We also use several observational datasets to evaluate the present-day model performance: surface temperature from the climatic research unit (CRU) of the University of East Anglia (CRU TS 4.01, at 0.5° resolution, CRU et al 2017), the Berkeley Global surface temperatures (at 1° resolution, Rohde et al 2013), and the GISS Surface Temperature Analysis (GISTEMP, at 2° resolution, Hansen et al 201, 0); precipitation from CRU (CRU TS 3.26, at 0.5° resolution, CRU et al 2019) and from the Global Precipitation Climatology Centre (GPCC v7, at 1° resolution, Becker et al 2013); and wind fields from the National Centers for Environmental Prediction (NCEP)-NCAR reanalysis (NCEP-NCAR, at 2.5° resolution, Kalnay et al 1996). A comparison between relevant present-day observed and simulated summer climatology (Text S1 and figure S1 (stacks.iop.org/ERL/15/114051/mmedia)) shows that the model can well reproduce magnitude and location of the prominent regional circulation features.

The analysis focuses on summer (June–August), when a large percentage of annual precipitation falls over most of the region (over 60% for southern Mexico and up to 40% for southeast U.S.). The emphasis is on the period 1950–1975, when the cooling trend over the southern U.S.-northern Mexico was the largest (see discussion in section 3). This period encompasses the most pronounced near-linear increase in North American aerosol emissions since pre-industrial times to their peak in the 1970s (figure 1(a); Smith 2011, Hoesly et al 2018), resulting in a corresponding increase in SO₄ burden (figure 1(b)) and aerosol optical depth (not shown). Temporal changes are identified by least-square linear trends. To detect changes externally forced by anthropogenic aerosols, trends are computed for ensemble mean quantities which allows to largely filter out internal variability. A two-tailed Student's t test is used to assess the significance (at the 95% confidence level) of the difference in the ensemble-mean response between the ALL and NoNA experiments. The extent to which the robustness of the results is affected by internal variability of the climate system is qualitatively estimated by the agreement on the sign of the trends across individual ensemble members.

Zoom In Zoom Out Reset image size

At global scale, the climate response to increased sulphate aerosols features, in accordance with induced changes in the global energy balance, an overall cooling (−0.02°C per decade), particularly strong in the Northern Hemisphere (where aerosol emissions are located; −0.03°C per decade), as well as a global-mean precipitation reduction (−0.003 mm d⁻¹ per decade, or about 1% of the model summer climatology) accompanied by a southern shift of the Intertropical Convergence Zone towards the warmer hemisphere (not shown). This is consistent with previous studies (e.g. Ridley et al 2015, Allen et al 2015, Westervelt et al 2018). At regional scale, the climate response patterns display substantial spatial variability and result from the interplay between thermodynamical and dynamical adjustments to aerosol forcing, as well as local feedback mechanisms.

To place the analysis into context, figures 1(c) and (d) compares observed and simulated near-surface temperature trends over the southeastern U.S. (black box in figure 2(a)) as a function of the start and end years during the 20th century. A striking feature is the marked 1950–1975 observed cooling trend (figure 1(c), ∼−0.4 °C, statistically significant at the 95% confidence level), and the successive warming to present-day. While the largest temperature anomalies are located over the south-central U.S., they are part of a coherent large-scale pattern: the cooling is spatially extensive and spread over the eastern U.S. and northern Mexico, accompanied by a weak warming over the western U.S. (figure 2(a)). This spatial structure is consistent among various observational datasets (figures 2(b) and (d)) and is in agreement with previous studies (e.g. Leibensperger et al 2012, Yu et al 2014, Mascioli et al 201, 7). Albeit of weaker magnitude, the CESM_ALL ensemble is able to capture the spatial pattern of the observed 1950–1975 temperature trend reasonably well, in particular the core cooling over the southern U.S. (figures 1(d) and 2(a)–(d)). Notably, the cooling is robust across the eight ensemble members (figure S3), suggesting it to be primarily due to external forcing. The underestimated magnitude of the cooling in CESM_ALL, however, suggests a potential role of natural variability, or could also be the result of model biases and/or a compensation among different internal coupled processes (e.g. Stevens and Feingold 2009).

Zoom In Zoom Out Reset image size

Sulphate aerosols are found to be a key driver of the temperature anomalies described above. The large sulphate burden over the eastern U.S. (figure 1(b)) results in a significant regional surface cooling (up to −0.5°C per decade, figure 3(a)), enhancing the all-forcing trend and, although weaker, showing a similar spatial pattern to observations (figures 2(a)–(d)). Note that the largest negative temperature trends are located to the west of the region of maximum SO₄ burden and the cooling extends to the Gulf of Mexico and the North Atlantic Ocean (figure 3(a)). Furthermore, aerosols appear responsible for the weak warming along the western U.S. and southern Mexico.

Zoom In Zoom Out Reset image size

The land precipitation response to increased sulphate aerosols (figure 3(b)), while modest over the emission region, features a large-scale wettening of up to 0.15 mm d⁻¹ per decade over the Great Plains, the southern U.S. and northern Mexico, accompanied by a significant drying (−0.25 mm d⁻¹ per decade) over western Mexico. The aerosol imprint is recognisable in the all-forcing pattern, and the latter is broadly consistent with the observed precipitation trends, although with some regional differences and of weaker magnitude (figures 2(e)–(g)). Over the ocean, widespread drying is found over the western North Atlantic and the Gulf of Mexico, while a dipole of zonally-elongated anomalies forms over the north-equatorial eastern Pacific (figure 3(b)), suggestive of an aerosol-driven anomalous southwestward shift of the climatological rainfall.

The surface temperature and precipitation changes discussed above are associated with pronounced regional atmospheric circulation anomalies. Changes in the lower-tropospheric atmospheric circulation modulate heat and moisture transport and its convergence over land, an important component of the regional atmospheric water balance (Mo et al 2005, Nigam and Ruiz-Barradas 2006, Durán-Quesada et al 2010, Amador et al 2016). Additionally, possible variations in the relative contribution of moisture convergence and evaporation may have crucial implications for land water resources and storage under climate change (e.g. Ruiz-Barradas and Nigam 2006).

The 700 hPa streamfunction (figure 3(c)) shows the development of a low-tropospheric high pressure anomaly over the western North Atlantic with a corresponding pressure decrease towards the subtropical and equatorial Pacific, consistent with a thermodynamical response to the anomalous surface cooling from increased sulphate aerosols and subsequent mass redistribution. However, the centre of the anticyclonic anomaly is not geographically collocated with the largest increase in aerosols over the northeastern U.S. but is displaced northeastward over the Atlantic. This is suggestive of an atmospheric adjustment to aerosol changes, resulting in a large-scale dynamical response pattern extending beyond the source region.

As a result, the North Atlantic Subtropical High (NASH), a key dynamical feature modulating moisture transport towards Mexico and the central-eastern U.S., intensifies, especially on its northern flank, and extends southwestward across central Mexico (figure 3(c)). Anomalous low-tropospheric easterlies blow over the subtropical western Atlantic (figure 3(c)) obstructing the climatological southerlies over the southern U.S. and deflecting the climatological easterlies over the Caribbean southward, which leads to anomalous moisture flux divergence over the eastern seaboard of the U.S. and the northern Gulf of Mexico (figure 3(d)). The interaction between the enhanced easterly moisture transport and the elevated topography of central Mexico favours positive rainfall trends on the Atlantic side, while suppressing rainfall on the Pacific side. A stronger northeastward pressure gradient over the eastern tropical Pacific reinforces the climatological easterlies but also induces anomalous divergence, leading to anomalous drying there and the precipitation shift mentioned above.

The 850 hPa flow associated with the anomalous Atlantic high displays a secondary branch with cyclonic rotation over the eastern U.S. that later joins the flow over the Gulf of Mexico (figure 3(c)). Correspondingly, the anomalous northerly stationary moisture fluxes across the continental U.S. oppose their climatology (e.g. Ruiz-Barradas and Nigam 2006), and feature divergence over the Great Plains (figure 3(d)). The contribution of this drier northerly flow to the positive precipitation anomaly of the region turns out to be negligible, which suggests a key role of dynamically-induced convergence, rather than transport. Further evidence for this is provided by changes in the 700 hPa circulation (figure 3(c)), the lowest available level above regional topography, which hints to a plausible dynamical link with the moisture convergence pattern via Sverdrup balance and induced vertical motion. We will discuss this link below. The leading role of evaporation anomalies (figure S5) is striking over the southwest U.S. and northern Mexico, a dynamically active region enclosing the northern edge of the North American monsoon, where water recycling is particularly large (evaporation largely exceeds precipitation), and compensates for the regional anomalous vertically-integrated moisture divergence.

An examination of the changes in radiation and clouds sheds light on the realisation of the regional aerosol impact. Increased sulphate loading (figure 1(b)) leads to a marked reduction in all-sky and clear-sky downwelling shortwave radiation at the surface (figures 4(a) and (b)), with decreases of up to −7 and 1.25 W m⁻² per decade, respectively, over the southeastern U.S. There is also a decrease of −4 W m⁻² per decade at the top of the atmosphere (TOA, not shown). Shortwave cloud forcing (i.e. the difference between all-sky and clear-sky shortwave radiation) changes are predominant over the aerosol emission region (up to 80% of the all-sky changes), suggesting aerosol-cloud interactions to play a critical role there. Clear-sky shortwave radiation anomalies at the surface and TOA display considerable similarity, reflecting the scattering properties of sulphate aerosols. Over the central and western U.S., both surface and TOA all-sky radiation changes display a decrease while the corresponding clear-sky anomalies are negligible, indicative of increased radiation scattering by more abundant clouds and associated precipitation (figure 3(b)). Similarly, the positive all-sky radiation flux anomalies over western Mexico and further west over the subtropical Pacific are related to drier conditions.

Zoom In Zoom Out Reset image size

Changes in various cloud characteristics (figures 4(c)–(f)) show similar large-scale response patterns consistent with radiation and precipitation anomalies. Low-level cloud cover (figure 4(c)) features a widespread and significant positive trend over central and eastern U.S. These changes are accompanied by a significant increase in cloud droplet concentration (figure 4(e)) and, although more confined to the east, by a decrease in the droplet effective radius (figure 4(f)), a manifestation of the cloud-albedo effect in the presence of more abundant cloud condensation nuclei and assuming negligible changes in liquid water (Twomey 1977). However, liquid water path shows a pronounced increase over the eastern U.S. (figure 4(d)), possibly resulting from more abundant droplets held in clouds rather than precipitating out (the cloud-lifetime aerosol effect; Albrecht 1989) and from an enhanced moistened flux from the Atlantic Ocean. We note that in a large domain of the eastern U.S. there is an increase in the droplet effective radius (excepting the easternmost region mentioned above). A plausible explanation for this is the circulation-driven increase in liquid water path overcompensating for any microphysical-driven decrease in the droplet size. This highlights the complexity of the interplay between cloud microphysics and dynamics in addition to the important role of aerosol-cloud interactions found here. Decreased liquid water path and cloud fraction occur over the north-equatorial Pacific, consistent with the diminished rainfall and enhanced net shortwave radiation at the surface. Anomalies of the opposite sign stretch in a nearly zonal fashion, from the western coast of Central America towards the tropical Pacific, contributing, together with enhanced evaporation (figure S5) by stronger easterlies, to cooling sea surface temperatures (SSTs).The negative SST anomaly is then further spread westward by wind-driven advection. Over the northern Gulf of Mexico, dimming by widespread aerosols leads to cooler SSTs, which locally act to strengthen the anticyclone, as well as, regionally, to enhance the thermal contrast between the Atlantic and Pacific basins and thus the pressure gradient and associated flow.

Given the link between continental surface anomalies and those in the regional circulation, as well as with anomalies over the adjoining oceanic basins (which are known to modulate North American hydroclimate; Burgman and Jang 2015, Kushnir et al 201, 0), it is important to examine the large-scale dynamical context, aiding to an improved mechanistic understanding of how these interactions occur. The pattern of anomalous 500 hPa vertical velocity (not shown) bears a strong resemblance to that of rainfall: wetter (drier) areas generally correspond to ascent (descent), as expected from the approximate balance between diabatic heating and mid-tropospheric vertical motion in the tropics and subtropics. This is clearly discernible over the Pacific and Atlantic oceans, far from land and orographic effects (e.g. across Mexico and the southwestern U.S.). The precipitation excess over the central U.S. and northeastern Mexico is also accompanied by widespread ascent, while an area of strong subsidence is located over the drier northern central U.S.

The 200 hPa geopotential height and meridional wind anomalies (figure 5(a)) further reveal a coherent wave pattern across the Pacific-North American region towards the extratropical Atlantic, indicating that the surface anomalies over the eastern U.S. and Mexico are embedded in a hemispheric-wide upper-tropospheric teleconnection pattern. The equivalent-barotropic nature of the anomalies over North America (with a slight westward tilt with height) is suggestive of a remotely-forced stationary wave response (e.g. Qin and Robinson 1993). Interestingly, the anomalous height pattern resembles the stationary wave forced by diabatic heating in the central Pacific (Ting 1994); also the tri-polar pattern across North America bears striking resemblance to the summertime wave pattern associated with variability of the Great Plains low level jet (Weaver and Nigam 2011) and is further reminiscent of the summer hemispheric-wide wave train identified by Ding and Wang (2005). Insights into the nature of this remote forcing are provided by figures 3(b) and 5; the precipitation anomaly (dipole) over the central Pacific results in vertically-integrated diabatic heating anomalies (up to +0.16 K d⁻¹ per decade in the positive core, assuming all the heating is due to condensation) and an upper-tropospheric outflow.

Zoom In Zoom Out Reset image size

Yet, teleconnection patterns are not necessarily generated over the region of the forcing but can be displaced far downstream as determined by the Rossby wave source (RWS; Sardeshmukh and Hoskins 1988). A wave source dipole coincident with convection anomalies in the central north-equatorial Pacific is clearly recognisable (figures 5(b) and (c)); strong meridional divergent outflow associated with the rainfall anomalies coexist with large meridional vorticity gradients due to the Asian-Pacific jet (e.g. Sardeshmukh and Hoskins 1988, Qin and Robinson 1993, Weaver and Nigam 2008). The RWS distribution features other centres in the extratropics, which can be interpreted as secondary sources generated by the quasi-geostrophic adjustment of the circulation to the wave generated in the primary source region.

The anomalous hemispheric-wide wave pattern identified above, initially instigated by circulation anomalies over the eastern U.S. and of remote central Pacific origin, in turn has an important imprint downstream in modulating the continental aerosol-related signal. In agreement with the Sverdrup vorticity balance (e.g. Rodwell and Hoskins 200, 1), strong descent and convection suppression occurs to the east of the upper-tropospheric anomalous ridge and southward flow over the northern Great Plains. Conversely, off the coast of the northeastern U.S., ascent is associated with northward flow on the eastern flank of an anomalous trough. Correspondingly, a dry (wet) anomaly is seen in the precipitation distribution (figure 3(b)). A full mechanistic explanation of the dynamics underlying the formation of the continental precipitation pattern shown in figure 3(b) requires, however, to account also for the interaction between upper-tropospheric wave dynamics and the Rockies: anomalous northeasterlies, part of the anomalous upstream anticyclonic circulation, impinge on the eastern slope of the Rockies (around 30–40°N, 105°W), generating low-tropospheric convergence and ascent, and thus positive precipitation anomalies. It is also noteworthy that the wave pattern across the eastern U.S., notably the location of the anticyclone over the northern Atlantic, is largely coherent with the near-surface circulation anomalies. Particularly, the low-tropospheric anticyclone over the eastern U.S. is displaced northeastward over the ocean, despite the strong land negative radiative forcing. The surface extension of the upper-level anomalies is thus indicative of an interesting modulation of the aerosol sub-regional imprint by the subsequent large-scale circulation response instigated by induced tropical anomalies.

This work sought to characterise the summertime climate response to increased North American sulphate aerosols and to understand the underlying mechanisms, particularly the role of atmospheric circulation adjustments—a key factor modulating the conspicuous moisture transport and related hydroclimate over Mexico and the U.S. The focus is on the period 1950–1975, which encompasses the largest increase and subsequent peak in aerosol emissions, and features an anomalous cooling over the eastern U.S. amidst the general continental warming—the 'warming hole'—, whose drivers are still the subject of a controversial debate. We used two sets of historical experiments conducted with the CESM model to isolate the impact of regional aerosol changes: a set of all-forcing experiments and an identical one but with North American aerosol emissions kept at their pre-industrial levels.

Regionally, increased aerosols result in widespread large cooling over the central and eastern U.S. and northern Mexico and weak warming over the western U.S. and southern Mexico. Precipitation reduces along the eastern coast of the U.S., opposed to the wetter U.S. continental interior. This is accompanied by a strengthening and westward expansion of the NASH and subsequent intensification of the low-level easterlies and associated moisture transport across the Gulf of Mexico and the eastern north-equatorial Pacific. Both aerosol-radiation and aerosol-cloud interactions contribute to generating these anomalies. At larger scale, a zonal precipitation dipole appears over the eastern tropical Pacific, in contrast with the more meridional and weaker response in the Atlantic sector. The induced anomalous diabatic heating generates a coherent upper-tropospheric signal in the mid-latitudes from the Pacific to the Atlantic basin, which in turn modulates the local aerosol imprint over North America. This emphasises the prominent role of adjustments in the atmospheric circulation and the interplay between local and remote influences in realising the impact of North American aerosols.

One may wonder whether European aerosols, which also increased during the 1950–1975 period by a similar amount, had any influence. Analysis of an additional 8-member all-forcing ensemble with fixed European sulphate aerosol emissions at pre-industrial levels shows that the temperature and precipitation response patterns over Mexico and the U.S. are of smaller magnitude than those driven by North American aerosols (not shown). Aerosols are transported over the subtropical Atlantic basin by the climatological circulation. However, cooling of the underlying SST is minor (−0.03 °C per decade), with negligible changes in lower-tropospheric winds. Instead, regional aerosol dimming induces a large anomalous anticyclone over northeastern Europe extending throughout the troposphere, which in turn leads to an upper-tropospheric wave-train propagating across Eurasia (similarly to Undorf 2019). This reaches the maximum amplitude over Eastern Asia when interacting with the Asian Jet, and then progressively weakens while crossing the eastern Pacific and the U.S.

It is further reasonable to ask whether an aerosol signature is discernible also after the late 1970s, when stringent regulations aimed at improving air quality led to a rapid aerosol decline over the U.S. (halved in the following 30 years, Smith 2011). To ascertain this, we analyse the period 1976–2006. Observations show a warming trend over the whole domain, particularly large over the western U.S., western Mexico, and the northeastern U.S. and southeastern Canada (figures S2(b)–(d)), while the temperature increase is relatively modest over the central U.S. Also, an overall wettening, with the largest increase over the southeastern U.S. and the Great Plains, is observed (figures S2(f)–(g)). The ALL experiments reproduce the above features well S2(a) and (e). Analysis of the NoNA experiments indicate that aerosols, although decreasing during this period, produce cooling over the central and western U.S., and enhanced precipitation over the southern U.S. and part of the Great Plains. Despite opposite aerosol variations, these anomalous response patterns bear strong resemblance to those during the earlier period, hinting to a common driving mechanism. Support to their dynamically-rooted origin is found in the anomalous rainfall pattern over the Pacific (figure S4): decreased aerosols result in a nearly-uniform northward precipitation shift, with a core over the north-equatorial basin west of 135°W, which generates a wave-like upper-tropospheric response downstream (not shown) similarly to that of the 1950–1975 period. This further emphasises the role of Pacific anomalies as a key factor modulating the aerosol-driven continental anomalies as well as the fundamental contribution of large-scale circulation adjustments.

Although our findings are based on ensemble experiments with eight members each, the potential role of the internal variability in modulating multi-decadal climate variations over North America cannot be conclusively assessed. In this respect, the use of large ensembles, such as the CESM-LENS (Kay et al 2015), would help to more robustly isolate the external component in the presence of internally-driven fluctuations.

Furthermore, the results presented here are based on one model only and so they rely on the model's representation of aerosol, cloud, and circulation interactions, which could differ from those in other climate models and/or the real world given the large uncertainties associated with anthropogenic aerosols and their climate interactions. For instance, the aerosol effective radiative forcing (ERF) in CESM1 is known to be large (Zelinka et al 2014), which could give a stronger climate response to aerosols than that in other climate models. This will depend on how much of the total contribution to the ERF is coming from the processes that drive the regional climate response identified. Despite these limitations, the important role of regional aerosols and their large-scale footprint found here can translate into implications for near-future projections of climate variability over Mexico and the U.S., which affects not just seasonal mean quantities but also climate extremes (see Text S2 and figure S6). With SO₂ emissions considerably reduced in the U.S., and the expectation of a continued global decline throughout the 21st century, this study sheds light upon possible ongoing and future regional climate responses to changes in anthropogenic forcing.

IMGM was jointly funded by CONACyT-Mexico and MAB acknowledges support from the Natural Environment Research Council (NE/N006038/1). SU was supported by the European Research Council project (EC-320691) TITAN and the University of Edinburgh. Thank you to Luis Zavala Sansón for thoughtful observations. We thank two anonymous reviewers whose comments helped to improve this work. We acknowledge the use of ARCHER, the UK National Supercomputing Service. We thank the CESM project at NCAR. We acknowledge the use of the BEST, CRU, GISTEMP, GPCC, HADEX2 datasets and NCEP-NCAR reanalysis. The data used in this study is available from MAB upon request.