Paleo-proxy and modeling evidence suggest that a shutdown of the Atlantic meridional overturning circulation (AMOC) would decrease North Atlantic Ocean sea surface temperatures and have far-reaching climate impacts. The authors use a regional climate model to examine the warm season response over North America to a hypothetical late-twenty-first-century shutdown of the AMOC with increased atmospheric CO2. In the future simulation, precipitation decreases over the western and central United States by up to 40% and over eastern Mexico by up to 50%. Over the eastern United States rainfall generally increases except during July. Variations in the moisture convergence associated with large-scale circulation changes dominate the rainfall variations, while evaporation plays a critical role over the northeastern United States in spring and the north-central United States in summer. During April–June the westward extension of the North Atlantic subtropical high enhances southwesterly moisture fluxes from the Gulf of Mexico into the eastern and south-central United States. Increases in low-level moisture content reduce the stability of the atmosphere. Enhanced southerly winds promote convergence over the eastern United States through the Sverdrup vorticity balance and precipitation increases. In July–August anomalous anticyclonic moisture fluxes associated with an anomalous high over the Gulf of Mexico and eastern Pacific decrease the moisture supply into the United States and Mexico. Over the central United States decreases in evaporation support decreases in low-level moisture content and increases in atmospheric stability. Over the eastern United States the Sverdrup balance weakens in summer and anomalous moisture convergence is mainly located over the East Coast.
Cold, dense waters sink in the high-latitude North Atlantic Ocean forming deep water that flows southward to the Southern Hemisphere as part of the global thermohaline circulation. At the surface northward flow transports relatively warm, saline water to high latitudes. This circulation system is known as the Atlantic meridional overturning circulation (AMOC). The overall northward heat transport is large, with a magnitude of about 0.7 ± 0.2 PW (1 PW = 1015 W) in the tropical Atlantic and 1.3 ± 0.15 PW near 25°N (Ganachaud and Wunsch 2000, 2003; Stammer et al. 2003). When the formation of the deep water is inhibited by a large amount of freshwater input due to glacial melting, the AMOC may be slowed or even shutdown. It is believed that such reorganizations of the AMOC are associated with abrupt cooling events in paleoclimate records (Alley and Agustsdottir 2005; Clark et al. 2001; Broecker 2003; Clarke et al. 2003).
Coupled atmosphere–ocean general circulation model (AOGCM) simulations suggest that the strength of the AMOC is sensitive to increases in CO2 concentration, weakening as the sea surface buoyancy increases (Gregory et al. 2005). While a slowdown of the AMOC has not been observed definitively (Kerr 2005; Latif et al. 2006), most of the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) coupled models predict a slowdown of the AMOC during the twenty-first century (Meehl et al. 2007). Although none of the models predicts a complete shutdown of the AMOC by the end of the twenty-first century, the models themselves may not be sensitive enough to changes in freshwater forcing and may be underestimating the potential for a shutdown (Alley and Agustsdottir 2005). Even though the probability of a shutdown may not be great, understanding the climate’s response is important for risk management and strategic resource planning given the potentially large economic and societal consequences.
The purpose of this paper is to study the regional response to a shutdown of the AMOC in the context of the contemporary climate change problem, that is, in the presence of increased atmospheric CO2 levels. A review of previous modeling and observational (paleoclimate) studies of the consequences of an AMOC shutdown/slowdown is provided in section 2. Section 3 includes a description of the regional climate model and the simulation design. Validation of the twentieth-century control simulation is presented in section 4, and section 5 presents results. Conclusions are summarized in section 6.
Paleo-proxy evidence indicates that about 8200 years ago Europe and northeastern North America experienced abrupt cooling accompanied by drying over the Sahara, the western Asian monsoon, and the U.S. Great Plains with a southward shift of the intertropical convergence zone (ITCZ). This “8 kya” event, along with other cooling events such as the Younger Dryas (about 12.8–11.5 ka) and Heinrich events, is believed to be forced by changes in the AMOC (Alley and Agustsdottir 2005; Clark et al. 2001; Broecker 2003). The idea is that enhanced freshwater influxes to the North Atlantic freshen and lighten the surface waters of the North Atlantic Ocean and suppress North Atlantic deep-water formation (Clarke et al. 2003, 2001). Water-hosing simulations, in which freshwater influxes are prescribed in the North Atlantic, produce anomalous responses consistent with the paleorecords (Renssen et al. 2001a,b, 2002).
Global warming has the potential to change the buoyancy of surface water over the North Atlantic and affect the AMOC by modifying the hydrological cycle. Observational evidence indicates that freshwater input into the high-latitude oceans is increasing (Dickson et al. 2003; Curry and Mauritzen 2005) and the strength of the AMOC is decreasing at 25°N (Bryden et al. 2005). However, since direct observations of ocean circulations are limited, there are uncertainties about such observed change in the AMOC, and it is unclear if the observed variations fall within the natural variability or represent a trend (Kerr 2005; Latif et al. 2006). On the other hand, AOGCMs generally simulate a weakening of the AMOC when forced with increasing atmospheric CO2 concentrations (Dixon et al. 1999; Wood et al. 1999; Manabe and Stouffer 1999; Thorpe et al. 2001; Gregory et al. 2005; Schmittner et al. 2005). This decrease ranges from a minimal weakening to a 50% decrease, depending on the AOGCM.
Climate impacts of an AMOC shutdown are studied in coupled models by manually increasing freshwater fluxes and/or reducing the salinity at high latitudes over the North Atlantic to induce a shutdown (Vellinga and Wood 2002; Stouffer et al. 2006; Barreiro et al. 2008). In these studies, the CO2 concentration was either fixed at preindustrial or present day levels. In association with a shutdown, precipitation decreases over most of the Northern Hemisphere, and the Atlantic and eastern Pacific ITCZs shift southward. Northern Hemisphere annual-mean surface air temperatures decrease by 1°–3°C with a maximum cooling of 8°–10°C over the northwest Atlantic. A slight warming, up to 0.2°C on average, is simulated in the Southern Hemisphere, with maxima over the South Atlantic (up to 1°C, Vellinga and Wood 2002) and the Weddell Sea (Stouffer et al. 2006). Warming in the Southern Hemisphere lags the Northern Hemisphere cooling by anywhere from several decades to 300 years (Chang et al. 2008; Vellinga and Wood 2002; Roche et al. 2010).
The effects of an AMOC shutdown in the presence of increasing CO2 concentrations will differ from the effects of an AMOC shutdown alone as the cooling effects of the shutdown may offset or perturb the effects of greenhouse gas warming. Vellinga and Wood (2008) examine the response to an AMOC shutdown in 2050 in an AOGCM along with greenhouse gas warming. During the first decade of their simulation, the Northern Hemisphere cools by about 1.7°C on average, while the southern Atlantic warms by 0.5°–2°C. Over western Europe and most of the North Atlantic cooling associated with the AMOC shutdown dominates the global warming signal, and simulated air temperatures are close to their preindustrial values. Precipitation changes associated with an AMOC shutdown generally oppose the rainfall response associated with greenhouse gas warming. However, over Central America, southern Europe, and Southeast Asia, precipitation decreases associated with the greenhouse warming are enhanced by the shutdown of the AMOC.
AOGCM simulations provide a large-scale perspective on global climate changes and they include atmosphere–ocean interactions. Regional model simulations complement the global modeling approach by providing higher resolution to evaluate regional climate change. For example, Jacob et al. (2005) used a regional climate model at 0.5° resolution to dynamically downscale AOGCM output and found that it provided information that improves the risk assessment associated with an AMOC-induced climate change over Europe. The regional model predicted a much larger snow fraction in total precipitation than the GCM because it resolved orographic features more accurately. Here we use a regional climate model forced by sea surface temperature anomalies (SSTAs) derived from AOGCM simulations to investigate the climate response over North America to a future AMOC shutdown in combination with atmospheric CO2 increases.
3. Simulation design
The National Center for Atmospheric Research (NCAR)– National Oceanic and Atmospheric Administration (NOAA) Weather Research and Forecasting (WRF) regional model, version 3.1.1 (Skamarock et al. 2008), is used with a large domain covering the North Atlantic and adjacent continents ( 5°S–70°N, 125°W–30°E; Fig. 1). Here we focus on climate variations over North America, while the climate response over West Africa and Europe is addressed by K. H. Cook et al. (2012, manuscript in preparation).
The model is used at 90-km horizontal resolution with 30 vertical levels and the top of the atmosphere set at 30 hPa. The vertical coordinate in the WRF model is a terrain-following hydrostatic-pressure vertical coordinate that varies from 1 at the surface to 0 at the top of the model. Our previous studies indicate that at this spatial resolution the model is able to realistically simulate the warm season climate over northern Africa (e.g., Patricola and Cook 2010, 2011a) and North America (Patricola and Cook 2011b, manuscript submitted to Climate Dyn., 2012). The physical parameterizations chosen include the Mellor–Yamada–Janjic planetary boundary layer (Mellor and Yamada 1982; Janjic 1990, 1996, 2002), the Monin–Obukhov–Janjic surface layer (Monin and Obukhov 1954; Janjic 1994, 1996, 2002), new Kain–Fritsch cumulus convection (Kain and Fritsch 1990, 1993; Kain 2004), Purdue Lin microphysics (Lin et al. 1983; Rutledge and Hobbs 1984; Tao et al. 1989; Chen and Sun 2002), Rapid Radiative Transfer Model (RRTM) longwave radiation (Mlawer et al. 1997), Dudhia shortwave radiation (Dudhia 1989), and the unified Noah land surface model (LSM) (Chen and Dudhia 2001).
Two ensemble simulations are conducted. One represents the late twentieth century and is referred to as “20C.” The other represents a possible late-twenty-first century future in which atmospheric CO2 levels have increased significantly and the AMOC has shutdown, referred to here as AMOC/CO2. Each ensemble consists of 20 warm season simulations, and these are averaged together to form the 20C and AMOC/CO2 climatologies. Each ensemble member is initialized on 15 March of a different year (1981–2000) and run for 200 days to 30 September. The first 17 days are disregarded for spin up. The upper soil layers, which are most important for the land–atmosphere coupling, adjust to the atmospheric forcing within a few days, indicating that the spinup period is sufficiently long enough. Monthly varying vegetation fraction and static U.S. Geological Survey land use and soil categories are prescribed for both the 20C and AMOC/CO2 runs.
In the 20C ensemble members, initial surface and lateral boundary conditions are specified from the National Centers for Environmental Prediction–U.S. Department of Energy reanalysis 2 (NCEP2) (Kanamitsu et al. 2002) for the required 6-hourly values of winds, temperature, relative humidity, surface temperature, mean sea level pressure, SST, soil temperature, and soil moisture for each year. Soil moisture and soil temperature are initialized on 15 March of each year and updated by the land surface model. The CO2 concentration is held fixed at a representative late twentieth century value of 330 ppmv.
To generate the AMOC/CO2 ensemble, initial, lateral, and surface boundary conditions are calculated using monthly mean anomalies derived from IPCC AR4 AOGCM simulations that used the Special Report on Emissions Scenarios (SRES) A2 emissions scenario (Nakicenovic et al. 2000). The anomalies derived from the AOGCM simulations are added to the NCEP2 reanalysis 20C lateral and boundary conditions. The SRES A2 scenario is chosen because it represents an estimate of the upper limit of the projected CO2 changes in the twenty-first century and, presumably, would be most likely to induce an AMOC shutdown.
This simulation design differs from a conventional downscaling approach in which full fields from an AOGCM are used to constrain both present day and future conditions. This approach is not used here because it is important to have an accurate present day simulation, and using boundary conditions directly from an AOGCM has been shown to degrade the results in previous applications (e.g., Cook et al. 2008; Patricola and Cook 2010, 2011a). Additionally, AOGCM output for all of the needed fields is not available at the 6-hourly interval. AOGCM anomalies are formed by taking differences between late twenty-first century (2081–2100) simulations forced with the SRES A2 emission scenario and late twentieth-century simulations. Monthly mean anomalies are averaged for nine models: Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model, version 3.1 (CGCM3.1) (T47), Centre National de Recherches Météorologiques Coupled Global Climate Model, version 3 (CNRM-CM3), ECHAM/Max Planck Institute Ocean Model (MPI-OM), Geophysical Fluid Dynamics Laboratory Climate Model, version 2.0 (GFDL CM2.0), Model for Interdisciplinary Research on Climate 3.2, medium-resolution version [MIROC3.2 (medres)], Meteorological Research Institute Coupled General Circulation Model, version 2.3.2 (MRI CGCM2.3.2), NCAR Community Climate System Model, version 3 (CCSM3), NCAR Parallel Climate Model (PCM), and Met Office (UKMO) third climate configuration of the Met Office Unified Model (HadCM3). These anomalies are assumed to be midmonth values and linearly interpolated to produce 6-hourly anomalies.
AOGCMs do not provide output needed to determine future soil moisture and soil temperature anomalies to initialize the land surface model. Here, present day value for the initialization for each of the 20 future AMOC shutdown runs is used for the AMOC/CO2 integrations.
An idealized SSTA pattern (Fig. 1) is designed and added to late-twentieth-century-observed SSTs for each run to represent Atlantic SST anomalies after a future AMOC shutdown in the presence of increased atmospheric CO2. Guided by the results from water-hosing experiments (Vellinga and Wood 2002, 2008; Stouffer et al. 2006; Chang et al. 2008; Barreiro et al. 2008), a Gaussian-shaped region of negative SSTAs is placed in the North Atlantic, with maximum cooling of −7 K at 20°N, 55°W. (AOGCMs project that the North Atlantic would cool by 5–12 K during an AMOC shutdown.) To represent cool water advection by the eastern boundary current, cool SSTAs are prescribed along the west coast of Europe. In addition to these SSTAs representing a shutdown on the AMOC, uniform 2.5-K warming is applied everywhere in the Atlantic to represent the effect of greenhouse gas warming by the end of the twenty-first century. Here 2.5 K is the domain-averaged warming of the nine AOGCMs ensemble mean for 2081–2100. Over the Mediterranean Sea, the cooling associated with an AMOC shutdown (e.g., Vellinga and Wood 2008; Barreiro et al. 2008) is offset by the warming associated with future increased greenhouse gas forcing.
In the AMOC/CO2 simulations, the CO2 concentration is increased to 757 ppmv, the average from the SRES A2 scenario for 2081–2100.
4. Model validation
Figure 2 shows 850-hPa wind and geopotential heights from the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) (Uppala et al. 2005) 1981–2000 climatology and the present day control 1981–2000 climatology. In April–June (AMJ) (Figs. 2a,b), the location of the North Atlantic subtropical high (NASH) and the Icelandic low are realistically simulated in 20C and the strength of both is only 10–20 gpm greater than that in the ERA-40 reanalysis. The westerly zonal flow at 45°–55°Ν is stronger by 2 m s−1.
In July–September (JAS), the subtropical high is stronger and more zonally elongated than in AMJ in both the reanalysis (Fig. 2c) and 20C (Fig. 2d). The simulated NASH is slightly weaker (by about 10 gpm) over the eastern United States than in the reanalysis, while over the eastern Atlantic Ocean, western Europe, and northern Africa it is slightly stronger (10–20 gpm). Over the eastern tropical Pacific a weak unobserved low forms off the coast of Mexico in the simulation associated with relatively high (~0.6 K) SSTs along the west coast of Mexico in the NCEP2 reanalysis that do not appear in the ERA-40 reanalysis. A similar cyclonical structure is found in individual years (e.g., 1998) in the ERA-40 reanalysis (not shown). The low geopotential heights centered over Baffin Bay are captured in 20C, and the storm track is accurately simulated.
Figures 3a–f show AMJ and JAS precipitation from the North American Regional Reanalysis (NARR) (Mesinger et al. 2006) climatology (1981–2000), the Tropical Rainfall Measuring Mission (TRMM) 3B43, version 6 (Huffman et al. 2007) climatology (1998–2010), and 20C. NARR provides high-resolution (~32 km) assimilated rainfall values. However, while precipitation is fully assimilated south of 27.5°N over the oceans, there is no assimilation north of 42.5°N and over Mexico data is from the daily gauge-based 1° grid analysis (Mesinger et al. 2006). Thus, the high resolution (0.25° by 0.25°) TRMM observations are also used as a reference. In AMJ maximum precipitation in the NARR is located over the lower Mississippi River valley, with relatively high rainfall over the eastern United States and in the Northwest (Fig. 3a). In the TRMM data, the rainfall maximum is located a bit farther north, centered over Kansas and Missouri (Fig. 3b). In 20C (Fig. 3c), the zonal precipitation gradient is captured, but maximum values are too far east. Precipitation maxima are well placed over Montana and Idaho, but the magnitude is 1–2 mm day−1 larger than in the NARR. The regional model also has a 2–6 mm day−1 wet bias over southern Mexico. Over the northwestern Atlantic, the convergence zone shown in the TRMM is well captured.
In JAS the NARR climatological precipitation maxima are located over the upper Mississippi River valley and along the southeast U.S. coast (Fig. 3d). In the TRMM climatology (Fig. 3e), rainfall maxima are also located over the southeast coast and the central Mississippi River valley. Rainfall rates over southern Mexico are about 2 mm day−1 greater than in the NARR. In 20C (Fig. 3f), the precipitation maximum over the southeast coast is simulated, but the maximum over the upper Mississippi River valley is underestimated. The model also has a wet bias over Colorado and New Mexico. The observed precipitation pattern over Mexico is generally captured, but rainfall intensity is 2–10 mm day−1 greater than in the TRMM climatology. Over the northwestern Atlantic, the 20C simulation captures the convergence zone that is misrepresented in the NARR.
Figures 4a–d show AMJ and JAS surface temperatures from the NARR climatology and 20C. In AMJ temperature maxima up to 303 K in the NARR climatology (Fig. 4a) are located over southern Texas, the western Gulf of Mexico coast, and the Pacific coast of Mexico. Minimum temperatures of 279–285 K occur over the Rocky Mountains, while over the central and eastern United States there is a negative meridional temperature gradient. A similar springtime temperature distribution is simulated in 20C (Fig. 4b). During JAS (Figs. 4c,d) the warmest temperatures are located over Texas and southeastern California in the NARR climatology as well as in 20C, although warm temperatures over the south-central Great Plains extend farther north in the simulation.
a. Precipitation and circulation changes
Figure 5 shows monthly precipitation from the 20C and AMOC/CO2 climatologies for the six averaging regions shown in Fig. 3c. The regions are the north-central United States (35°–45°N, 95°–105°W), south-central United States (26°–35°N, 95°–105°W), northwestern United States (40°–49°N, 105°–120°W), southwestern United States (32°–40°N, 105°–120°W), eastern United States (31°–45°N, 75°–95°W), and eastern Mexico (15°–26°N, 86°–100°W). These regions are chosen because rainfall anomaly patterns and seasonal variations are relatively uniform within each region. The spread among the individual members for the 20C (open squares) and AMOC/CO2 (closed triangles) simulations are also shown.
Precipitation decreases by 10%–36% throughout the warm season over the north-central United States (Fig. 5a) in AMOC/CO2 compared with 20C. Results from the calculation of a two-tailed Student’s t test indicate that the monthly rainfall anomalies in April, July, and August are significant at the 90% confidence level, while anomalies in May, June, and September are not. (We select the 80% level in the two-tailed significance test, which provides confidence at 90% for positive anomalies and 90% for negative anomalies.)
Over the south-central United States (Fig. 5b), May precipitation in AMOC/CO2 is about 0.3 mm day−1 greater than in 20C, but smaller in other months with a minimal anomaly of −0.8 mm day−1 (−40%) in August. A two-tailed t test indicates that rainfall anomalies are significant at the 90% confidence level in July and August but not significant in other months.
Over the northwestern United States (Fig. 5c), there is a minimal difference in precipitation in April, but conditions become drier from May to September, with the largest decrease of about −0.4 mm day−1 (−30%) in July. Rainfall anomalies in May, July, and August are significant at the 90% confidence level.
Rainfall is reduced from April to September over the southwestern United States (Fig. 5d) in the AMOC/CO2 simulation, with the largest decrease of −0.4 mm day−1 in August (−20%). Rainfall anomalies in April, May, July, and August are significant at the 90% confidence level.
Over the eastern United States (Fig. 5e), precipitation increases in April–June and August–September with a maximum increase of about 0.4 mm day−1 (10%) in May. In July precipitation decreases up to 0.2 mm day−1 (−5%). Rainfall anomalies are significant at the 90% confidence level only in April.
Over eastern Mexico (Fig. 5f) rainfall decreases in every month, with a maximum decrease of 6 mm day−1 in July (about −51%). The rainfall anomalies are significant at the 90% confidence level from May to September.
In summary, rainfall decreases over most of the western and central United States and eastern Mexico in AMOC/CO2 while over the eastern United States precipitation increases except in July. The big spread among the 20C and AMOC/CO2 members indicates that the variability of rainfall is large. However, variations of the spread are generally consistent with the variations of the means. In most regions the precipitation anomalies have the same sign during AMJ or JAS. Since the circulation anomalies are also similar during these months in AMOC/CO2 (not shown), the analysis below is based on AMJ and JAS averages.
The vertically integrated atmospheric column moisture balance is examined to relate precipitation anomalies to circulation anomalies. As a first step, local precipitation P is decomposed into contributions from local evaporation E and moisture convergence in the atmospheric column according to
where g is the acceleration due to gravity, q is specific humidity, ps is surface pressure, ptop is the pressure at the top of the atmosphere, and is horizontal wind; u and υ are the zonal and meridional components, respectively.
Figure 6 shows anomalies of precipitation, evaporation, and vertically integrated moisture convergence (calculated as P − E) for AMOC/CO2–20C in AMJ and JAS. In AMOC/CO2, AMJ precipitation is greater over the eastern United States, central Mexico, and western Texas and smaller over the north-central and western United States (Fig. 6a). Evaporation plays an important role over the eastern United States, with anomalies up to 50% of the rainfall anomalies in the northeast (Fig. 6c). Overall, the patterns of anomalous moisture convergence are similar to the patterns of anomalous precipitation over the eastern and central United States (Fig. 6e). Over the western and north central United States, decreases in precipitation are mainly related to anomalous moisture divergence, while over the eastern United States between 75° and 90°W, anomalous moisture convergence contributes up to 60% (50%) of the precipitation increase over the southeast (northeast).
In JAS rainfall rates are reduced over most of the United States except the east coast (Fig. 6b). Over Mexico, precipitation decreases more in the south than in the northwest. Evaporation (Fig. 6d) decreases over the central and western United States and is the primary contributor to the future rainfall reduction. Evaporation increases over the eastern United States, between 75° and 90°W, and central Mexico. Anomalous moisture divergence occurs over the western United States centered over Colorado, while anomalous moisture convergence is located over the central Great Plains centered over Iowa (Fig. 6f). Over the eastern United States between 35° and 40°N, 78° and 85°W anomalous moisture divergence reduces moisture increases from enhanced evaporation. Along the east coast between 32° and 45°N, anomalous moisture convergence is about one-half of the precipitation anomaly. Over southern Mexico and the Gulf of Mexico, anomalous moisture divergence is the main contributor to future reductions in rainfall.
As revealed in Fig. 6, variations of large-scale moisture convergence contribute significantly to the precipitation anomalies over the northeastern, southeastern, and south-central United States in AMJ and over the western central United States and Mississippi Valley in JAS. Evaporation anomalies play an important role over the northeastern United States in AMJ and the north-central United States in JAS. The simulated JAS evaporation anomalies resemble the AMJ soil moisture anomalies (not shown) which are, in turn, associated with AMJ precipitation anomalies.
To investigate the association between moisture convergence anomalies [Eq. (1)] and circulation anomalies, the vertically integrated, mass-weighted moisture transport (flux) M is examined:
where is the sigma-level thickness. The summation is from the surface (σ = 1) to the top model level (σ = 0) and is calculated as a finite sum in sigma coordinates to avoid errors associated with topography.
Figures 7a–d show M from the 20C ensemble mean and for AMOC/CO2–20C for AMJ and JAS. In 20C during AMJ (Fig. 7a), the flux is anticyclonic with easterly moisture transport over the northern Caribbean Sea, southerly/southeasterly transport over the western Gulf of Mexico, and southwesterly/westerly transport over the central and eastern United States. This transport is consistent with circulation around the NASH during spring, including the Caribbean low-level jet (CLLJ) along its southern flank, the Great Plains low-level jet (GPLLJ) along its western flank, and the westerly midlatitude storm track along its northern flank. The largest simulated moisture flux transports (>210 kg m−1 s−1) occur over the Caribbean Sea, the Tennessee Valley, and due east of the Carolina coast.
In AMOC/CO2 in AMJ (Fig. 7b), anomalous southwesterly moisture fluxes are located between 70° and 95°W, enhancing moisture transport from the Gulf of Mexico onto the continent. This increased moisture transport is associated with increased moisture convergence (Fig. 6c) and rainfall (Fig. 6a) over the southeastern United States and the Great Lakes region. Over the Pacific Ocean an anomalous southerly flux is associated with an increase in moisture transported over the Baja California peninsula and the southwestern United States.
In JAS in the 20C simulation, the easterly moisture flux is strong, exceeding 210 kg m−1 s−1, over most of the Gulf of Mexico and Caribbean Sea and transporting Atlantic moisture over Mexico and southern Texas (Fig. 7c). Over the central United States the southerly moisture flux associated with the GPLLJ over the southern (northern) Great Plains is weaker (stronger) than in the springtime. The core of strong westerly moisture transport centered around 38°N over the Tennessee Valley in AMJ shifts northward to ~45°N in JAS, consistent with the summertime development and intensification of the NASH over the southeastern United States.
In the AMOC/CO2 climatology for JAS (Fig. 7d), the easterly moisture flux over the southern Gulf of Mexico/Caribbean Sea and the westerly flux north of 40°N over the continental United States are enhanced. Southerly moisture transport over the north-central United States associated with the GPLLJ is also enhanced, but this increase does not translate into increased rainfall over the north-central United States as changes in the evaporation field dominate (Fig. 6c) the increases in moisture convergence (Fig. 6f). Between 20° and 30°N, 80° and 110°W southeasterly moisture transport into Mexico and the south central United States decreases along with the rainfall (Fig. 6b). Over the Rocky Mountains from to 30° to 40°N, 100° to 115°W the southerly moisture flux weakens and rainfall decreases over Arizona, Utah, Colorado, and New Mexico (Fig. 6b). Anomalous southerly moisture transport is located over the western United States between 115° and 120°W, enhancing rainfall over western Nevada and southern Idaho (Fig. 6b).
Height–longitude cross sections of wind anomalies (shaded) and moisture transport anomalies (contoured) averaged between 30° and 35°N for the meridional and zonal components for AMJ and JAS are shown in Fig. 8. The spring southerly flow anomalies (Fig. 8a) indicate that the GPLLJ (e.g., between 90°–105°W centered at 950 hPa) is approximately 15% stronger in the future [consistent with the GCM results reported in Cook et al. (2008)], while southerly flow over the Gulf Stream (e.g., between 70°–80°W below 900 hPa) is about 25% stronger, indicating a strengthening of the western flank of the NASH. In the upper troposphere near 300 hPa, northerly anomalies are located over the Rocky Mountains between 102° and 110°W, while southerly anomalies are located over the eastern United States between 75° and 90°W. Low-level moisture flux anomalies are generally collocated with the wind anomalies. Anomalous moisture flux anomalies are northerly over the Rocky Mountains, but southerly farther west between 115° and 120°W.
In the zonal direction (Fig. 8b), the westerly wind anomalies dominate between 1000 and 500 hPa and extend up to 300 hPa at 110°–120°W. Anomalous westerly moisture flux extends from the surface to ~400 hPa, with two maxima collocated with the maximum wind anomaly at low levels.
The JAS meridional cross section (Fig. 8c) indicates that the GPLLJ is 25% stronger in AMOC/CO2, while southerly flow over the Gulf Stream is 25%–50% stronger. Over the Rocky Mountains, anomalous northerly winds are centered at 110°W, 750 hPa while farther west between 115° and 120°W southerly wind anomalies are relatively strong, extending from the surface to 450 hPa. Meridional moisture flux anomalies coincide in position with the wind anomalies at low levels with magnitudes greater than those in spring.
Zonal wind and moisture flux anomalies are westerly from the surface to approximately 700 hPa (Fig. 8d), which is more shallow than the spring anomaly, while between 650 and 450 hPa the anomalous moisture flux and wind are easterly.
According to Fig. 8, the moisture transport anomalies are dominated by changes in the flow and are greater in the lower troposphere than in the upper troposphere. In AMJ an enhanced GPLLJ and southwesterly winds over the coast increase the moisture transport onto the United States. During JAS northerly midlevel wind anomalies (between 700 and 300 hPa) over the Rocky Mountains and the east coast reduce the southerly moisture fluxes, while easterly wind anomalies above 650 hPa decrease the westerly moisture flux from the south-central to the eastern United States. Thus, the vertically integrated moisture transport over the south-central and eastern United States at 30°–35°N is weaker compared to that in 20C (Fig. 7d).
Figures 9a,b show 850-hPa geopotential height (shading) and wind (vectors) anomalies for AMJ and JAS. The black thick contours indicate the location of the NASH in the 20C climatology. Since geopotential heights increase domainwide in the AMOC/CO2 climatology (about 18–29 gpm), the height anomalies are normalized as in Cook et al. (2008) to clarify changes in gradients. Monthly domain-averaged geopotential height differences between AMOC/CO2 and 20C are subtracted from the monthly anomaly.
As shown in Fig. 9a, AMJ geopotential heights are higher over the western Atlantic basin including the Gulf of Mexico and northeastern Atlantic east of 25°W. Over the central North Atlantic and northern/central North America, normalized geopotential heights are lower by up to 15 gpm. This anomalous pattern over the Atlantic is similar to the anomalous “quadrupole” response identified by Cook et al. (2008) in the AR4 AOGCM simulations of the later twenty-first century. They found that the AOGCMs simulated an enhanced GPLLJ in the spring for 2079–99 owing to a westward extension of the NASH that increased zonal geopotential height gradients over the central United States, and rainfall increased in the upper Mississippi Valley. However, the westward extension of the subtropical high in the AMOC/CO2 climatology does not penetrate as deeply into the continental interior of the United States. Over the south-central and eastern United States, meridional height gradients are also enhanced. The GPLLJ is strengthened with a westerly component and anomalous southwesterly winds dominate over the eastern United States, consistent with the enhanced southwesterly moisture transport (Figs. 8a,b) during AMJ.
In JAS the NASH extends farther west and north over the continent (Fig. 9b) in AMOC/CO2. The GPLLJ is stronger, accompanied by enhanced geopotential height gradients and southerly moisture transport over the southern Great Plains (Fig. 8c). The anomalous high center located near the southeast coast of the United States shifts northward to 40°N, 60°W. Between 15° and 35°N anomalous highs are located over the eastern North Pacific and the east Gulf of Mexico. The anomalous northerly flow over the Rocky Mountains and the southerly flow over the west coast (Figs. 8a,c) are part of the anticyclonic flow about these anomalous highs.
A moist static energy (MSE) analysis is used to understand changes in atmospheric stability. MSE is the sum of the sensible, latent, and geopotential energy according to
where cp is the specific heat of air at constant pressure, T is air temperature, L is the latent heat of vaporization of water, q is specific humidity, and z is height. MSE increasing with altitude denotes a stable atmosphere, so increases (decreases) in low-level MSE destabilize (stabilize) the vertical column and promote (suppress) convection.
Figures 10a,b show MSE anomalies over the eastern United States in AMJ and north-central United States in JAS, respectively. Precipitation anomalies (not shown) are relatively large in these regions. Over the eastern United States, MSE anomalies (solid line) in AMJ have a negative slope between the surface and 600 hPa, and the stability of the vertical air column decreases (Fig. 10a). This change in stability is related to increases in low-level moisture (Lq, dashed line) associated with the enhanced low-level southwesterly moisture transport (Figs. 8a,b), while temperature anomalies (cpT, dot-dash line) contribute little to the slope.
Over the north-central United States in JAS (Fig. 10b), the MSE profile changes little between the surface and 600 hPa. While increases in low-level temperature tend to destabilize the atmosphere, moisture at 600 hPa counteracts the effect. Enhanced southwesterly moisture transport increases the moisture content around 600 hPa in AMOC/CO2 (Fig. 7d), while a lowered moisture content below 700 hPa is associated with evaporation decreases (Fig. 6d). Between 400 and 250 hPa, the atmosphere is stabilized in association with CO2-forced higher temperature anomalies at 200 hPa than at 500 hPa.
Figure 10 shows that enhanced moisture transport associated with the westward extension of the NASH decreases the stability of the lower atmosphere over the eastern United States during springtime and promotes convective precipitation, while over the north-central United States decreases in evaporation reduce low-level moisture content and enhance stability. Such an increase in stability related to decreases in evaporation is also seen over the south-central United States (not shown).
Nigam and Ruiz-Barradas (2006) suggested that the western flank of the NASH is related to summer precipitation over the eastern seaboard through the Sverdrup vorticity balance. Here we examine this relationship over the eastern United States below the western flank of the NASH. The climatological vorticity equation is
where is the relative vorticity, f is the Coriolis parameter in which is the vertical p velocity, is friction, and k is the unit vector in the vertical direction. Variables are averaged over time, such that . The five terms in Eq. (4) represent the vorticity tendency due to convergence, advection, titling, friction, and the transients ( in which the overbar denotes the climatological mean and the prime a perturbation), respectively. Effects of transients and the frictional generation of relative vorticity are calculated as a residual.
The Sverdrup balance states that for a large-scale steady flow the vorticity tendency due to the meridional advection of planetary vorticity is balanced by convergence (for midlatitude synoptic-scale flow ) and the vorticity equation simplifies to
Figure 11 shows anomalies (AMOC/CO2 minus 20C) for each term in Eqs. (4) and (5) averaged over the eastern and coastal United States (31°–45°N, 75°–95°W) at 850 hPa. Results are similar if calculated at 925 hPa. The convergence (black solid line) and advection (gray solid line) terms are the largest and nearly balance each other during AMJ and September. The residual term (dot-dot-dash line) is relatively large during July–August, while the tilting term (short-dashed line) is always small. Planetary vorticity advection (gray long dashed line) dominates advection variations while (black long-dashed line) dominates convergence variations. The Sverdrup balance holds during AMJ. As the NASH extends westward, the negative vorticity tendency due to enhanced northward transport of low planetary vorticity air requires anomalous convergence (or stretching) to maintain the balance and promote convective precipitation. During July–August the balance is broken because of a greater residual term, which is probably related to enhanced turbulence mixing in the summer. The Sverdrup balance is reestablished in September.
b. Surface temperature changes
Figures 12a,b show surface skin temperature anomalies for AMJ and JAS. In AMJ (Fig. 12a) temperatures increase over the United States and Mexico, with a maximum increase of about 4 K over the north-central United States and southern Mexico. Over the eastern United States skin temperatures increase about 3 K. The overall warming over land suggests that greenhouse warming outweighs the cooling effect of the AMOC shutdown. In JAS (Fig. 12b) over most of the United States and Mexico, skin temperatures increase more than 4 K. Over the north-central United States between 40° and 45°N, 90° and 100°W temperature anomalies reach 7 K.
Figure 12 shows that surface temperature increases are greater over land than over the ocean, and this is also true over the entire simulation domain (not shown). The resulting differences in land–sea temperature gradients are associated with large-scale circulation changes. Miyasaka and Nakamura (2005) mentioned that the land–sea temperature difference, especially the relatively cold SSTs over the northeastern Atlantic, is a major factor influencing the formation of the subtropical high. Cook et al. (2008) found that the westward extension of the NASH in GCM simulations is mainly due to the changes of land–sea temperature gradients. Similarly, the extension of the Atlantic subtropical high in AMOC/CO2 is associated with cooling over the northeastern Atlantic and warming over land and the low-latitude oceans. Also, as mentioned above, enhanced precipitation over the eastern United States increases diabatic heating, which could reinforce the westward extension of the NASH.
As shown in Fig. 12, strong warming occurs over the north-central United States. The net surface heating, , is calculated using
where , , , and are the net downward solar radiation, net upward longwave radiation, upward sensible heat flux, and upward latent heat flux, respectively. Horizontal heat exchanges are neglected. Figure 13 shows the differences between the AMOC/CO2 and 20C simulations for each term in Eq. (6) averaged over the north-central United States (35°–45°N, 95°–105°W). Positive (negative) values denote downward (upward) fluxes. Net heating is positive from April to September, consistent with surface warming (Fig. 12). Differences in the net solar radiation are always positive with maxima in May and August. This is mainly related to decreases in cloud amount (not shown), which is consistent with decreases in rainfall (Fig. 5a). Latent cooling differences are positive in April and May in association with increases in evaporation and negative from June to September when evaporation is smaller (Figs. 6c,d). The anomalous sensible heat flux is positive, enhancing surface heating, in April and negative during May–September as the land surface becomes warmer than the atmosphere. Figure 13 shows that the surface temperature increases over north-central United States in the AMOC/CO2 simulation are primarily associated with increases in the net solar flux in April and May and decreases in latent cooling during June–September.
Figure 14a displays histograms of daily maximum 2-m temperature over the north-central United States in AMJ. The distribution shifts to warmer temperatures, ranging from 272–315 K in 20C to 276–320 K in AMOC/CO2, and the peak of the distribution shifts from ~298 to ~303 K. Daily temperature maxima that were unprecedented in the 20C simulation, that is, daily maxima above 315 K, are relatively common in the AMOC/CO2 simulation.
In JAS (Fig. 14b), the shift to significantly warmer daily temperature maxima is even more pronounced than in the spring. The range changes from 280–315 K in 20C to 285–322 K in AMOC/CO2, and the peak changes from 306–307 to 315–316 K. Nearly half of the daily maxima in the AMOC/CO2 simulation are unprecedented in the 20C simulation. Similar changes are also found over the eastern United States (not shown).
Both paleoclimatic records and model simulations reveal that reorganization of the Atlantic meridional overturning circulation (AMOC) has far-reaching climate impacts over the Northern Hemisphere. Observations and model simulations also suggest that the strength of the AMOC is sensitive to greenhouse gas concentration and weakens by various amounts (model dependent) as CO2 increases. Although none of the IPCC AR4 models predicts a complete shutdown of the AMOC by the end of the twenty-first century, the possibility cannot be excluded (Meehl et al. 2007), so the consequences should be evaluated.
Model simulations suggest that, while the climate responses to an AMOC shutdown are generally opposite to those forced by greenhouse gas warming, the two forcings may add nonlinearly and the total effects will be regionally dependent (Vellinga and Wood 2008). Here we study the climate response over the United States and Mexico to a hypothetical AMOC shutdown during 2081–2100 using a regional climate model (WRF). An idealized SSTA (Fig. 1) and boundary perturbations predicted by the IPCC AR4 AOGCMs for 2081–2100 are applied to the NCEP2 reanalysis (1981–2000) to constrain the regional model, while the CO2 concentration is fixed to 757 ppmv, the mean value of the IPCC SRES A2 emissions scenario for 2081–2100.
In the simulation, an AMOC shutdown with CO2 increases causes precipitation decreases over most of the United States and Mexico from April to September. Rainfall reductions are most severe in July and August with negative anomalies up to −40% over the central United States and −50% over eastern Mexico. Over the eastern United States, however, rainfall increases in April–June and August–September, and over the south-central United States rainfall increases slightly in May. A moisture budget analysis shows that the precipitation differences over Mexico and the eastern and western United States are mainly associated with changes in moisture convergence. Decreases in evaporation contribute about one-half to the precipitation anomalies over the central United States during July–September while increases in evaporation support enhanced rainfall over the East Coast.
In spring, enhanced southwesterly moisture transport brings added moisture from the Gulf of Mexico into the eastern United States. A moist static energy analysis shows that enhanced low-level moisture decreases the stability of the atmosphere over the eastern United States and promotes anomalous convective precipitation.
In July–September, the moisture supply from the ocean into Mexico and the United States is decreased in association with anomalous anticyclonic moisture fluxes centered over the eastern Pacific and the Gulf of Mexico. Over the north-central United States, while enhanced southwesterly moisture transport increases the moisture content around 600 hPa, the moisture content below 750 hPa decreases associated with reduced evaporation, and the stability of low-level atmosphere increases.
These changes in moisture transport are related to differences in the large-scale circulation. A prominent feature of the shutdown simulation is the strengthening and westward extension of the NASH, which enhances southwesterly winds and moisture transport over the Gulf of Mexico and the eastern United States in spring and over the central and eastern coastal United States in summer. This strengthening of the NASH also provides a convective environment along its western flank through the Sverdrup vorticity balance. In the spring, enhanced southerly flow over the eastern United States requires anomalous convergence to balance the negative vorticity tendency due to the northward transport of low planetary vorticity. In July–September, the relationship is weaker and the anomalous moisture convergence is located over the east coast of the United States. The Sverdrup balance relationship also suggests a possible interaction between the rainfall and the NASH since enhanced diabatic heating associated with rainfall increases would increase convergence and, in turn, reinforce the westward extension of the NASH.
Greenhouse gas warming dominates land surface temperature changes. Variations in regional hydrological cycles and circulation further modify the temperatures through latent cooling and cloud distribution. In July–September, land surface temperatures increase up to +7 K over the north-central United States along with marked increases in extremely warm daily maximum temperatures.
Finally, additional simulations at finer resolutions are needed to test the sensitivity of the results to model resolution. The results presented here are based on simulations conducted at 90-km resolution. Increasing the model resolution may affect the results as the orography and small-scale processes would be better resolved.
This research was supported by award DE-FG02-08ER64610 from the U.S. Department of Energy Office of Science Biological and Environmental Research Abrupt Climate Change program. Insightful and constructive suggestions from Professors Peter Hess and Stephen J. Colucci are sincerely appreciated. The authors also acknowledge Dr. Christina M. Patricola, Naresh Neupane, Meredith Brown, and Jamie Favors for their helpful discussion and encouragement during the course of this work. Helpful comments from two anonymous reviewers improved the paper. The Texas Advanced Computing Center (TACC) at The University of Texas at Austin provided high performance computing resources for the project.
Current affiliation: Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, Austin, Texas.