1. Introduction
The extratropical storm tracks form an important part of the global circulation (e.g., Blackmon et al. 1977; Chang et al. 2002; Hoskins and Hodges 2002). Cyclones developing and propagating over the storm tracks provide much of the high impact weather experienced in the midlatitudes, including heavy snow (Novak et al. 2008), coastal storm surge (Colle et al. 2008), and a significant fraction of extreme precipitation events (Pfahl and Wernli 2012; Kunkel et al. 2012). Thus, changes in the storm tracks under global warming can have significant impacts on the midlatitude climate.
Over the years, there have been two popular definitions of storm tracks. Traditionally, the storm tracks are defined by an aggregate of cyclone tracks (e.g., Klein 1958). This definition offers direct links to the relationship between cyclones and surface weather. Historically, progress was hampered by the tedious task of subjective cyclone tracking. With the advent of automatic objective tracking algorithms (e.g., Murray and Simmonds 1991; Hodges 1999; and many others), an increasing number of studies have made use of this approach (e.g., Hoskins and Hodges 2002). Nevertheless, cyclone statistics can depend on the details of the tracking algorithm (Raible et al. 2008; Ulbrich et al. 2009), but efforts are being made to intercompare the different algorithms in order to quantify such dependencies (Neu et al. 2013).
The alternative definition making use of bandpass-filtered transient eddy statistics was pioneered by Blackmon (1976). This alternative Eulerian definition has the advantage of highlighting eddy energy and fluxes that are closely related to the interactions between storm track and the low-frequency flow (e.g., Cai and Mak 1990). Most previous studies have made use of either one or the other definitions, but several studies that have examined multiple storm-track quantities (e.g., Paciorek et al. 2002; Pinto et al. 2007; Chang 2009) have found similarities and differences displayed by the different quantities, suggesting that more insight can be gained by examining the different storm-track quantities together. In this study, we will employ both approaches and examine cyclone track and transient eddy statistics together.
Making use of the multimodel projection provided by models participating in phase 3 of the Coupled Model Intercomparison Project (CMIP3; Meehl et al. 2007a), Yin (2005) found a robust poleward shift and upward expansion of the midlatitude storm tracks (defined in terms of bandpass-filtered eddy kinetic energy) by the end of the twenty-first century under the high-emission Special Report on Emissions Scenarios (SRES) A2 scenario. These changes have been linked to the projected warming in the upper troposphere and cooling in the lower stratosphere resulting from increased greenhouse gases, which will give rise to an enhanced temperature gradient in the upper troposphere/lower stratosphere (e.g., Butler et al. 2010, 2011; Chen et al. 2008), as well as an increase in the height of the tropopause (Lorenz and DeWeaver 2007).
In the Northern Hemisphere lower troposphere, polar amplification is expected to give rise to a significant reduction in the surface temperature gradient (Manabe and Stouffer 1980; Holland and Bitz 2003). This, together with increased static stability in the subtropics due to enhanced upper-tropospheric warming, likely accounts for the projected decrease in the frequency of midlatitude extratropical cyclones found in many studies (e.g., Geng and Sugi 2003; Lambert and Fyfe 2006; Bengtsson et al. 2006; Lim and Simmonds 2009). On the other hand, warming leads to increased water vapor in the atmosphere, which could lead to enhanced latent heat release, possibly giving rise to more intense cyclones (Lambert and Fyfe 2006). However, Bengtsson et al. (2009) found no significant increase in cyclone intensity measured in terms of extreme winds or vorticity in a relatively high-resolution modeling study. These contrasting impacts from changes in the upper and lower troposphere renders quantitative understanding of storm-track changes difficult and may account for the large spread in model projections of storm-track changes shown later in this study.
Around the United States, Teng et al. (2008) examined the projected changes in winter cyclone activity based on an experiment made with the Community Climate System Model, version 3 (CCSM3), with forcing based on the SRES A1B scenario and found some indications of decreasing frequency along the East Coast and an increase near the West Coast by the end of the twenty-first century. These results are consistent with a projected decrease in eddy kinetic energy (EKE) at 850 hPa over the East Coast and an increase over the western part of the United States found in the same model experiment. Long et al. (2009) examined storm-track projection near the east coast of North America using the Canadian Regional Climate Model, version 3.5, and found a significant decrease in cyclone track density along the Canadian east coast. Ulbrich et al. (2008) examined bandpass-filtered sea level pressure (SLP) variance statistics from 16 CMIP3 models, again based on the SRES A1B scenario, and found a projected decrease in the variance near the Great Lakes and the U.S.–Canadian east coast and indications of increased variance over the western United States, consistent with the results of Teng et al. (2008) and Long et al. (2009).
With the availability of data from phase 5 of CMIP (CMIP5; Taylor et al. 2012), several recent studies have compared CMIP3 and CMIP5 storm-track projections. Comparing CMIP5 to CMIP3, CMIP5 models generally have higher resolution, especially vertical resolution, and more sophisticated model physics. In addition, a greater number of modeling centers and models participated in CMIP5 compared to CMIP3 (see Table 1 for a list of CMIP3 and CMIP5 models used here and their expanded names). An overview of how well CMIP5 models simulate North American climate can be found in Sheffield et al. (2013a,b).
Names and expanded names of the models from CMIP3 and CMIP5 used here.
Chang et al. (2012) compared meridional velocity (at 300 and 700 hPa) as well as SLP variance statistics from CMIP3 based on the high-emission SRES A2 scenario (radiative forcing of ~8 W m−2 at 2100; see Solomon et al. 2007) and CMIP5 based on the representative concentration pathway 8.5 (RCP8.5; radiative forcing of ~8.5 W m−2 by 2100; Meinshausen et al. 2011; Moss et al. 2010). The authors found that projected changes in the Southern Hemisphere are largely consistent between CMIP3 and CMIP5 simulations, but there are significant differences in the Northern Hemisphere. In particular, CMIP5 models project much larger decrease in storm-track activity over North America than CMIP3 models. These results are consistent with those found by Harvey et al. (2012), who compared SLP variance statistics from CMIP3 simulations based on the SRES A1B scenario to those from CMIP5 models made with RCP4.5 radiative forcing.
While the two studies mentioned in the preceding paragraph displayed significant differences between CMIP3 and CMIP5 storm-track projections over North America, both studies focused on global comparisons and showed little details regarding regional differences. In this study, we will focus on projected changes over North America and show these projections in more details. We will also explore related mean flow changes to gain more insight on what might have led to these changes. Finally, we will explore possible impacts of the projected significant decrease in storm-track activity over North America.
2. Data and methodology
Storm-track activity is examined based on both cyclone and eddy variance statistics. For the eddy variance, daily or higher-frequency data are required. CMIP5 models provide 6-hourly data at three pressure levels (850, 500, and 250 hPa) as well as SLP. Unfortunately, the upper-tropospheric level provided (250 hPa) is different from the levels provided by CMIP3 (300 and 200 hPa). However, many models provide model level data at 6-h frequency. As in Chang et al. (2012), 6-hourly model level data (mostly on hybrid sigma-pressure coordinates) provided by 15 CMIP5 models [first 15 models listed in Table 2; more details about these models, including horizontal and vertical resolutions, can be found in Chang et al. (2012)] are used. Monthly variance statistics are first accumulated on the model levels, and then these statistics are interpolated onto 12 standard pressure levels from 1000 to 100 hPa. Chang et al. (2002) compared these statistics at 250 hPa to those computed directly using the available pressure level data provided by CMIP5 for several models and found that the results for projected changes are nearly identical. In this study, meridional velocity variance at 300 and 700 hPa, as well as SLP variance statistics (from all 23 models listed in Table 2), will be examined. Since CMIP3 only provides daily averaged data, the 6-hourly CMIP5 data are first averaged into daily means before the variance is computed. All monthly statistics are then interpolated from the model horizontal grid onto a common 2.5° × 2.5° latitude–longitude grid using bilinear interpolation so that multimodel mean can be computed.
CMIP5 projected percentage (%) change in SLP variance (pp) and cyclone statistics between 1980–99 (historical) and 2081–2100 (RCP8.5), over the region 30°–55°N, 130°–70°W, for all four seasons. Changes that are statistically significant at 95% are indicated by numbers in boldface font. Significant increases are further highlighted by a plus sign before the number.
For CMIP5, data from 1980 to 1999 from the “historical” experiments are used to compare with data from 2081 to 2100 made under the high-emission RCP8.5. We will also show some results from mid-twenty-first century (2041–60), as well as results from experiments made under the more-moderate-emission RCP4.5 (total radiative forcing leveled out at ~4.5 W m−2 by 2100) for comparison. For CMIP3, results from 11 models (see Table 3) based on the twentieth-century simulations (20C3M; for 1981–2000) will be compared to those made under the SRES A2 high-emission scenario for the years 2081–2100. For models that provide ensembles of more than one run, only one run (usually the r1i1p1 run) has been analyzed.
As in Table 2, but for CMIP3 projected percentage changes in SLP variance statistics (SRES A2 2081–2100 minus 20C3M 1981–2000).
For cyclone statistics, the automatic objective tracking algorithm developed by Hodges (1999) is employed. In this study, all negative SLP anomaly centers are tracked. The 6-hourly model data are first interpolated onto a common 2.5° × 2.5° latitude–longitude grid using a computer package for manipulating geophysical data on a global sphere (SPHEREPACK) if the model grid is either Gaussian or a regular latitude–longitude grid. If the model latitudinal grid is irregular, the cubic spline is first used to interpolate the data onto a regular latitude–longitude grid with the same number of grid points, before SPHEREPACK is again used for the interpolation onto the common grid. Before tracking, the monthly mean is first removed for each month to get rid of stationary disturbances (Donohoe and Battisti 2009), and spatial filtering is performed to remove waves with total wavenumber less than 5 to remove the large-scale, low-frequency background flow (Anderson et al. 2003). Negative SLP minima are linked together across time steps to form cyclone tracks using constraints on the track smoothness and maximum displacement distance [note that positive minima are not tracked; for details, see Hodges (1999)]. Only tracks lasting 2 days or more and moving over 1000 km are retained. Colle et al. (2013) evaluated tracking using a similar tracker for cyclones over the U.S. East Coast and western Atlantic and found an uncertainty of around 5%–10% in the tracking results. Cyclone-track statistics have been computed for all 23 CMIP5 models listed in Table 2 for the historical period of 1980–99 and under RCP8.5 forcing for 2081–2100. Previous studies have suggested that regional cyclone statistics can be rather noisy (e.g., Teng et al. 2008; Colle et al. 2013) and that results based on different trackers are more robust when averaged over larger regions (Neu et al. 2013). Hence, we will only display cyclone statistics averaged over the continental United States and southern Canada. In addition, Neu et al. (2013) suggested that statistics for strong cyclones are more robust; thus, we will highlight results for strong cyclones. Since CMIP3 only provides daily averaged data, which is too coarse for objective cyclone tracking, no cyclone statistics have been computed for CMIP3 model output.
Since variance statistics represent the passage of both cyclones and anticyclones and we have only tracked cyclones to produce the cyclone track statistics, we do not expect these two statistics to be equivalent because of the asymmetry between cyclones and anticyclones (e.g., Wallace et al. 1988; Donohoe and Battisti 2009). This is one reason why these two kinds of statistics are examined together in this study. Nevertheless, as discussed below, our results show that these two kinds of statistics are closely related to each other, at least over North America.
3. Projected storm-track changes
a. Variance statistics
The projected changes in meridional velocity (at 300- and 700-hPa levels) and SLP variance statistics between 1980–99 and 2081–2100, based on 15 CMIP5 models forced with RCP8.5, are shown in Fig. 1. Corresponding changes based on 11 CMIP3 models forced with the SRES A2 scenario are shown in Fig. 2. The shades shown in the figures indicate regions over which at least 80% of the models agree on the sign of the projected change.
Projected changes by CMIP5 models in variance statistics from 1980–99 to 2081–2100 under RCP8.5 for (a)–(d) 300-hPa meridional velocity variance, (e)–(h) 700-hPa meridional velocity variance, and (i)–(l) sea level pressure variance. Black contours represent model climatology (contour intervals of 50 m2 s−2 and 10 hPa2); red contours represent projected positive changes (contour intervals of 10 m2 s−2 and 2 hPa2); blue contours are the same as red contours, but for projected negative changes; shadings represent regions over which ≥80% (light) or 100% (dark) of the models agree on the sign of the change; pink shadings indicate projected increase; and light blue shadings denote decrease.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
As in Fig. 1, but for CMIP3 models under SRES A2.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
Near the tropopause (300 hPa; Figs. 1a–d), CMIP5 models project an increase in eddy variance over northern Canada during winter [December–February (DJF)], spring [March–May (MAM)], and fall [September–November (SON)], In winter, a decrease in storm-track activity is projected to the south of the storm-track maxima over Atlantic and central Pacific, while a significant decrease south of the storm-track peak is projected to spread from the central Pacific across North America into the Atlantic in the other three seasons. All these suggest a poleward shift of storm-track activity in the upper troposphere. In the lower troposphere (700 hPa; Figs. 1e–h), the projection is mainly a decrease in storm-track activity with maximum decrease again south of the climatological storm-track maximum for all four seasons. At sea level (Figs. 1i–l), CMIP5 models project a significant decrease in SLP variance stretching from Alaska across southern Canada and the United States into the western Pacific in winter. During the other seasons, projected decrease mainly occurs along a band south of the climatology maximum.
Comparing with CMIP3 projections (Fig. 2), in the upper troposphere (Figs. 2a–d), CMIP5 and CMIP3 projected changes are quite similar in winter and spring. However, in summer [June–August (JJA)] and fall, CMIP3 models project much smaller decrease across North America. In the lower troposphere (Figs. 2e–h), CMIP3 models project increase in υ variance in winter over much of the United States and Canada, while projected decrease for the other three seasons is much smaller than that projected by CMIP5. At the surface, CMIP3 models project a decrease in SLP variance centered near the Great Lakes and an increase over western North America in winter (Fig. 2i), consistent with the results of Ulbrich et al. (2008). Unlike CMIP5 projections, CMIP3 models project very little decrease in spring (Fig. 2j). In summer and fall (Figs. 2k,l), while CMIP3 models also project a decrease in SLP variance over a band stretching across the United States, the projected decrease is much less than that projected by CMIP5 models.
To quantify the projected changes over the United States and southern Canada and to show model-to-model variations, the projected percentage change in pp, averaged over the area 30°–55°N, 130°–70°W, is shown in Tables 2 and 3 for each CMIP5 and CMIP3 model, together with the multimodel mean. The climatological value for the average of pp over the region, normalized by the value based on the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Re-Analysis (ERA-Interim) for DJF, is given in Table 4 (first row for each season). Table 4 shows that pp has a maximum in DJF and minimum in JJA. The CMIP5 multimodel mean shows that CMIP5 models generally simulate a storm track that is too weak over North America, with the most severe bias in JJA and SON (~20% too low). In addition, individual models can have significant biases, as indicated by the model ranges (minimum and maximum values) and standard deviations shown in Table 4.
Climatological pp and cyclone statistics over North America from ERA-Interim and CMIP5 models for 1980–99. Note that pp is normalized by the value of that from ERA-Interim in DJF.
In Tables 2 and 3, changes that are statistically significant at the 95% level are highlighted by bold numbers and positive significant changes are further highlighted by the plus sign before the numbers. For each model, statistical significance is assessed based on month-to-month variability and for the ensemble mean, statistical significance is assessed based on model-to-model variations. Results are qualitatively similar (but more noisy) if we average over a smaller area that does not include the high terrain over the western United States and Canada west of 100°W (not shown).
For winter, 22 of the 23 CMIP5 models (Table 2) project a decrease in SLP variance over the United States and southern Canada, with 15 models projecting a decrease that is statistically significant at the 95% level. GFDL CM3 is the only model projecting an increase which is not statistically significant. While there is strong model consensus in the sign of the projected change, the magnitude ranges from +5.3% to −21.1%, with a multimodel mean of −9.1%. For CMIP3 (Table 3), only 4 of the 11 models project a decrease over the same region, with only 1 model projecting a decrease that is statistically significant and 2 models projecting an increase that is significant. The multimodel mean is +0.6%, which is not statistically significant. If we limit the area to 100°–70°W (not shown), CMIP5 models project a mean change of −9.3%, while CMIP3 models project a small decrease of −2.3%, which is not statistically significant.
For spring, 19 out of the 23 CMIP5 models project a decrease, with the projected change ranging between +9.6% and −19.2%, with a multimodel mean of −6.6%, which is still statistically significant. One of the projected increase and 12 of the decreases are statistically significant. For CMIP3, the multimodel mean comes out to be +2.0%. Projected changes in summer by CMIP5 models range from +0.2% to −46.9%, with a multimodel mean of −17.6%. Of the 23 models, 22 project a decrease with 19 projecting significant decrease. CMIP3 models also project a mean decrease (−8.2%), which is significant, with 8 out of the 11 models projecting a decrease and 6 of the projections being significant. For fall, 21 of 23 CMIP5 models project a multimodel mean decrease of −10.2%, while only 7 of the 11 CMIP3 models project a mean decrease of −3.1%. These results show that, in all four seasons, CMIP5 models generally project much more significant decreases in storm-track activity than CMIP3 models. The difference between the CMIP5 and CMIP3 multimodel-mean projections is statistically significant at the 95% level based on model-to-model variability for winter, spring, and summer and at the 90% level for fall. In terms of percentage change, the largest decrease is projected for summer when even CMIP3 projected decrease is significant (but note that the storm track is weakest in summer), while the smallest decrease is projected in spring.
Changes in SLP variance statistics projected by CMIP5 models between 1980–99 and 2041–60 for RCP8.5 are shown in Figs. 3a–d. Quantitatively, the changes over the United States and southern Canada come out to be approximately half of that projected for the end of the twenty-first century (Table 5). Changes for RCP4.5 at the end of twenty-first century (Figs. 3e–h) are roughly similar in pattern and amplitude to those projected for midcentury under RCP8.5, while for mid-twenty-first-century RCP4.5 (Figs. 3i–l) projected changes range between 65% and 90% of that projected for end of twenty-first century, consistent with a leveling off of the increase in radiative forcing under this pathway. From the first five rows of Table 5, it is clear that storm-track decreases projected even for mid-twenty-first century under RCP4.5 are still larger in magnitude than those projected for end of twenty-first century by CMIP3 models under the higher-emission SRES A2 for all seasons except summer.
As in Figs. 1i–l, but for (a)–(d) 2041–60 under RCP8.5, (e)–(h) 2081–2100 under RCP4.5, and (i)–(l) 2041–60 under RCP4.5.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
As in Table 3, but for comparisons between CMIP3 and CMIP5 projections. The last five rows are for projected percentage changes normalized to 5°C warming in the Northern Hemisphere.
b. Cyclone statistics
As discussed in section 2, projected changes in cyclone statistics (based on SLP cyclone anomalies as defined in section 2) over the continental United States and southern Canada (30°–55°N, 130°–70°W) between 1980–99 and 2081–2100 have been computed for 23 CMIP5 models under RCP8.5. Based on ERA-Interim data (Dee et al. 2011), during 1980–99, on average about 20.3, 18.4, 16.2, and 17.7 cyclones pass through the region each month in winter, spring, summer, and fall, respectively (see Table 4: note that each cyclone track is only counted once; unless otherwise stated, quantitative statistics for all four seasons will be listed in this order below). The CMIP5 multimodel means slightly underestimate these frequencies by between 2% and 6% (see Table 4). Average amplitude (defined as the magnitude of the most negative value on each track when the cyclone is within the region) ranges between 16.0 hPa in winter and 9.4 hPa in summer based on ERA-Interim data, with the CMIP5 multimodel mean overestimating the amplitude by about 4% in winter and underestimating it by 3% in fall.
As mentioned above, Neu et al. (2013) suggested that cyclone statistics derived using different tracking algorithms are more consistent for strong cyclones. Thus, the frequency of significant cyclones, defined here as cyclones reaching an amplitude of 20 hPa within the region during winter, spring, and fall and 15 hPa during summer, have also been compiled. Based on ERA-Interim data, on average, there are 5.3, 3.3, 1.9, and 3.5 such cyclones per month during those four seasons. Thus, these cyclones represent the strongest quartile in winter, and the top 12% in summer. In CMIP5 simulations, the corresponding multimodel-mean frequencies for such cyclones come out to be 6.0, 3.1, 1.8, and 3.1 events per month; thus, the bias of the multimodel mean is relatively moderate. However, while the bias of the multimodel mean is not large, individual models can have significant biases, especially in the frequency of significant cyclones.
Projected percentage change in the frequency and amplitude of all cyclones, as well as the frequency of significant cyclones, are shown in Table 2. Consistent with the projection of changes in SLP variance over the same area, all cyclone statistics are projected to decrease under global warming, with the largest decrease projected for summer (but recall that summer cyclones are weakest among the four seasons and there are fewest significant cyclones in summer) and the smallest decrease in spring. In fact, model-to-model variations in the projected percentage change in SLP variance is significantly correlated with the projected percentage change in the frequency of significant cyclones in all seasons (correlations of 0.80, 0.57, 0.70, and 0.55, respectively) and with the projected change in total cyclone frequency in spring and summer (r = 0.68 and 0.77, respectively), as well as with the projected change in average cyclone amplitude in all seasons (r = 0.69, 0.77, 0.92, and 0.70, respectively). In addition, the projected change in cyclone amplitude and frequency of significant cyclones are highly correlated in all four seasons, with values equal to 0.85, 0.79, 0.83, and 0.73, respectively.
Table 2 shows that under RCP8.5, by the end of the twenty-first century, CMIP5 models project a significant decrease in cyclone frequency over the United States and southern Canada ranging between −5.2% in spring and −12.2% in summer, with even larger decrease in the frequency of significant cyclones, ranging between −6.6% in spring and −32.6% in summer. While we have not computed track statistics for mid-twenty-first century, based on the strong correlation between changes in SLP variance and cyclone statistics, with SLP variance projected to change by roughly half that projected for the end of the twenty-first century (Table 5), we expect that changes in cyclone statistics by midcentury should also be roughly half of that shown in Table 2. Even for the more moderate RCP4.5, we should still expect significant reduction in cyclone frequency and amplitude by mid-twenty-first century.
4. Discussion
In the preceding section, we have shown that CMIP5 models project much larger decrease in storm-track activity over the United States and southern Canada than CMIP3 models and that there is a large model-to-model spread in the projected changes. In this section, we will examine some factors that may have contributed to these differences, as well as some possible consequences of the projected significant decrease in storm-track activity.
a. Impacts of model biases and resolution
Previous studies have suggested that in some cases, biases in a model’s climatology are correlated with the model’s projected change under global warming. For example, Chang et al. (2012, 2013) suggested that in DJF, for both CMIP3 and CMIP5, in the Northern Hemisphere the strength of a model’s climatological storm track (in terms of meridional velocity variance statistics) is negatively correlated with its projected percentage change in amplitude, while in the Southern Hemisphere models with larger equatorward biases in storm-track latitude tend to project larger poleward shifts [similar to the relationship found by Kidson and Gerber (2010) for the Southern Hemisphere eddy-driven jet]. In those cases, one may argue that bias corrections to model projections can be developed based on the relationship between model biases and future projections (Chang et al. 2012; Kidson and Gerber 2010).
We have examined whether the strength of a model’s climatological storm track (as indicated by pp) is correlated with the projected percentage change over North America. The correlations between these two quantities are −0.26, −0.01, 0.04, and 0.45 for the four seasons, respectively, and the correlation is statistically significant only in SON. However, the relationship during that season is positive rather than negative, contrary to the results of Chang et al. (2012) for DJF meridional velocity variance averaged over the entire Northern Hemisphere midlatitudes. The physical mechanism behind the relationship in fall (as well as whether it arises merely because of chance) should be further investigated.
Chang et al. (2013) also suggested that storm-track amplitude biases found in CMIP3 models are correlated with model horizontal resolution. However, we are unable to find similar relationships for CMIP5 models between the climatological storm-track amplitudes (in terms of pp) and horizontal grid spacings [see Chang et al. (2012) for the values of model grid spacing] over North America, with the correlation between these two quantities being less than 0.1 for all four seasons. There are some indications that the 11 higher-resolution models that have average grid spacing of less than 2° tend to project slightly less decreases in storm-track activity than the remaining 12 models (see the last two rows of Table 2). However, the correlations between model grid spacing and projected percentage change in pp for the four seasons are −0.35, −0.22, −0.32, and −0.13, respectively: none of which is significant even at the 90% level.
While our results indicate that, among CMIP5 models, horizontal resolution may not have significant impacts on either storm-track amplitude or future projection over North America, Willison et al. (2013) suggested that simulations run at high resolution (~20-km grid spacing) may be able to resolve impacts of latent heating on storm-track dynamics that cannot be resolved in these lower-resolution climate models. Whether high-resolution models may project different storm-track changes from those projected by low-resolution models remains to be investigated.
b. Surface temperature
As discussed in the introduction, one reason why storm-track activity may decrease in the Northern Hemisphere under global warming is the projected decrease in surface temperature gradient as a result of high-latitude amplification of the warming. Thus, in Fig. 4 the multimodel ensemble-mean projections of surface temperature change between the late-twentieth and late-twenty-first centuries are shown with the solid lines for CMIP5 and CMIP3 models. In Fig. 4, the projected surface temperature changes have been averaged between 160° and 60°W, over a longitude band covering most of North America and its upstream region, but the zonal mean over all longitudes show a similar pattern.
Projected change in surface temperature (°C) averaged over 160°–60°W for the four seasons, for (a) CMIP5 models under RCP8.5 and (b) CMIP3 models under SRES A2. The solid lines are for projected changes from end of the twentieth century to the end of the twenty-first century. Dashed lines in (a) are for projected changes to the mid-twenty-first century.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
Both CMIP5 and CMIP3 model projections show enhanced warming in the Northern Hemisphere high latitudes in fall and winter, with less high-latitude enhancement in spring and less warming near the pole in summer. This is consistent with previous studies that have shown that polar amplification of surface warming maximizes in early winter (e.g., Holland and Bitz 2003; Lu and Cai 2009).
Comparing CMIP5 and CMIP3 projections, it is clear that CMIP5 models are projecting much stronger polar amplification under RCP8.5 than those projected by CMIP3 models under SRES A2 during the cool seasons. With CMIP5 models also projecting a larger decrease in storm-track activity during fall and winter than in spring, it is tempting to hypothesize that the differences in projected storm-track changes could be partly related to less polar amplification in CMIP3 models during the entire cool season and CMIP5 models in spring, leading to less decrease in storm-track activity. However, CMIP5 model-to-model variations in the projected change in the magnitude of the temperature gradient (defined as the difference in temperature change between 30° and 75°N averaged over 160°–60°W), while positively correlated with the percentage change in SLP variance (i.e., larger decrease in the magnitude of the temperature gradient is correlated with larger decrease in SLP variance) for spring, summer, and fall, the correlation is only significant at the 95% level in summer (correlations equal 0.00, 0.33, 0.60, and 0.21 for DJF, MAM, JJA, and SON, respectively). Thus, while the differences between the degree of polar amplification might partially account for the differences between CMIP3 and CMIP5 projections as well as between the seasons, they do not appear to be a dominant factor in explaining model-to-model differences.
Figure 4 suggests that CMIP5 models project larger surface warming under RCP8.5 than CMIP3 models under A2; thus, part of the difference between the projected changes in storm-track activity could be a result of differences in projected temperature change. Hence, for each model, the projected change in SLP variance is normalized by its projected temperature change in the Northern Hemisphere. The average projected change normalized to a temperature increase of 5°C (the multimodel-mean projected warming for DJF under RCP8.5 is 5.2°C) is shown in the bottom five rows of Table 5. We can see that, for CMIP5 projections, when normalized to the same warming, differences between MAM and the other two cool seasons are reduced but projected decrease is still largest in JJA. The results also indicate a large degree of linearity between the projected storm-track and temperature changes and the similarities between RCP8.5 and RCP4.5 projections when normalized, consistent with previous results that suggested that many climate trends scale proportionally to global-scale temperature change (e.g., Murphy et al. 2007; Neelin et al. 2006). However, comparisons between CMIP5 and CMIP3 projections show that even when normalized to the same temperature change, CMIP5 models still project much larger decreases in storm-track activity compared to CMIP3 simulations in all seasons.
c. Mean available potential energy
Since the storm tracks are made up of baroclinic waves, one would expect storm-track activity to depend on the midlatitude baroclinicity. One measure of baroclinicity is the mean available potential energy (MAPE), which measures the effect of temperature gradient and static stability throughout the troposphere. O’Gorman (2010) showed that across 11 CMIP3 models, the projected percentage change in hemispheric-mean eddy kinetic energy scales linearly with the projected change in MAPE under global warming. O’Gorman (2010) also showed that the dry MAPE works about as well as the MAPE that includes the effects of moisture in scaling with the eddy kinetic energy; hence, in this study, the dry MAPE is computed based on the mean pressure–latitude temperature profile between 850 and 300 hPa from 30° to 60°N.
The MAPE has been computed for each CMIP5 model simulation for the historical and future (under RCP8.5) periods separately, and the percentage difference between the two periods is correlated with the projected percentage change in SLP variance over the United States and southern Canada. Traditionally, the MAPE is computed using the zonal-mean temperature profile. When that is used to correlate with change in SLP variance (values shown in Table 2) across CMIP5 models, the correlation is not significant at the 95% level for DJF (r = 0.35, significant at the 90% level), but is significant at >99% for MAM, JJA, and SON (r = 0.61, 0.66, and 0.53, respectively). The low correlation for DJF is not surprising, considering that the projected change in SLP variance is far from being zonally symmetric during winter (Fig. 1i). Thus, instead of using the zonal-mean profile, a local temperature profile, averaged across the longitude band of 130°–70°W, is used to compute the change in MAPE to indicate change in local baroclinicity. The correlation between this more local definition of MAPE and SLP variance is higher for all seasons except for spring (r = 0.46, 0.55, 0.71, and 0.82 for DJF, MAM, JJA, and SON, respectively) and is significant at the 95% level for all four seasons. Note that averaging across a broader longitude band (160°–60°W) that includes the upstream region gives slightly smaller correlations that are still significant at 95% for all four seasons. The relationship between these two quantities is shown in Fig. 5. Figure 5 suggests that model-to-model difference in the projected change in local baroclinicity (as measured by the local MAPE) can explain a significant portion of model-to-model difference in the projected change in storm-track activity over North America.
Projected percentage change in sea level pressure variance over North America (DPP), plotted against projected percentage change in local mean available potential energy (DAPE; computed based on the temperature profile averaged between 130° and 70°W). The gray triangle symbols are from CMIP5 models and the unfilled squares from CMIP3 models. The black triangle and square symbols correspond to the CMIP5 and CMIP3 ensemble means, respectively. The correlation listed in each panel is for CMIP5 models.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
To further illustrate this point, among the GFDL models, GFDL CM3 projects a much larger decrease in the SLP variance than GFDL-ESM2M in JJA (−46.9% versus −5.2%, see Table 2). At the same time, GFDL CM3 also projects a much larger change in the local MAPE than GFDL-ESM2M (−39.2% versus −2.1%) during the same season. The latitude–height cross sections of the projected change in temperature between the end of the twentieth and the twenty-first centuries for these two models, averaged over 130°–70°W, are shown in Fig. 6. In the midtroposphere between 30° and 60°N, GFDL CM3 projects a decrease in the meridional temperature gradient (larger warming at higher latitudes) and an increase in the static stability (larger warming at higher altitudes), with both acting to reduce the baroclinicity and MAPE. For GFDL-ESM2M, the projected temperature change is larger at lower latitudes, leading to a slight increase in the meridional temperature gradient. Again, with larger warming at higher altitudes, the static stability is projected to increase slightly. These two factors counteract each other, leading to a small change in the baroclinicity and MAPE. A similar picture also emerges when the zonal-mean temperature changes are examined. Why these two (and other) models project different structures of temperature change under global warming is clearly an issue that warrants further studies.
Latitude–height cross section of the projected temperature change in summer (contour interval 1°C), averaged across the longitude band of 130°–70°W, based on the difference between 1980–99 and 2081–2100 under RCP8.5, for (a) GFDL CM3 and (b) GFDL-ESM2M.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
In Fig. 5, changes in SLP variance as a function of changes in the local MAPE for the 11 CMIP3 models have also been plotted using unfilled square symbols. Taking the two groups together, it is apparent that by and large the two groups exhibit a consistent relationship between changes in the variance and change in the MAPE, except that CMIP3 results show a bit more scatter. In addition, CMIP3 models as a group project more positive (or less negative) changes in the MAPE than CMIP5 models, and this can partly account for the larger decrease in SLP variance projected by CMIP5 models. For reference, CMIP3 model-mean projections of change in the local MAPE are +4.5%, +7.7%, −9.0%, and +3.5% for the four seasons, respectively, while CMIP5 projections are −2.1%, +2.3%, −19.1%, and −3.0%. Taking the 34 CMIP5 and CMIP3 models together, the correlations for the four seasons are 0.50, 0.58, 0.59, and 0.64, respectively, with all being significant at the 99% level.
Note also that, while the CMIP5 models on average project a small and statistically not significant (based on model-to-model variability) increase in the local MAPE in spring, they project a significant decrease in SLP variance, suggesting that the decrease in storm-track activity is likely to be mainly a result of the projected decrease in the surface temperature gradient due to polar amplification in surface warming. In addition, CMIP5 models also only project a small and insignificant decrease in the local MAPE in fall and winter (in fact, the zonal-mean MAPE is projected to increase in both seasons because of an increase in the upper-tropospheric temperature gradient); thus, polar amplification in surface warming likely also plays a significant role in the projected near 10% decrease in the SLP variance during these two seasons.
d. Implications on precipitation changes
Extratropical cyclones strongly impact the weather of the regions they pass through, including bringing about a significant amount of precipitation (e.g., Chang and Song 2006; Field and Wood 2007). As mentioned above, Pfahl and Wernli (2012) and Kunkel et al. (2012) have suggested that a high percentage of precipitation extreme events in the middle and high latitudes, including events occurring over the United States and Canada, are associated with extratropical cyclones and their frontal systems. Thus, one might hypothesize that the projected significant decrease in storm-track activity by CMIP5 models over North America might have impacts on the projected change in precipitation over the region.
Under global warming, precipitation is projected to increase in the higher latitudes and decrease in the subtropics (Meehl et al. 2007b). These changes are related to increased water vapor in a warmer atmosphere, leading to increased moisture transport and hence enhanced moisture convergence in the high latitudes and the deep tropics and enhanced divergence in the subtropics (Wetherald and Manabe 2002; Kutzbach et al. 2005). The decrease in the subtropics can also be linked to the projected poleward expansion of the subtropical dry zone (Scheff and Frierson 2012). Here we examine how storm-track changes may modulate these projections.
The change in precipitation projected by CMIP5 models under RCP8.5, normalized to a Northern Hemisphere temperature increase of 5°C for each model, is shown in Fig. 7c for summer and Fig. 8c for winter. Consistent with previous results, across North America, CMIP5 models project an increase in precipitation in the high latitudes and a decrease in the subtropics and tropics. The main difference between the results shown here and those found in CMIP3 simulations (e.g., Christensen et al. 2007) is that the projected decrease in precipitation over central North America extends a bit farther northward toward southern Canada in summer.
CMIP5 ensemble-mean projection for changes in SLP variance and precipitation in summer, with projected change for each model normalized to a Northern Hemisphere temperature change of 5°C. (a) SLP variance for all 23 CMIP5 models. (b) Difference in SLP variance between the six models projecting the largest storm-track decrease over the United States and southern Canada and the six models projecting the smallest storm-track decrease. (c),(d) As in (a),(b), but for precipitation. (e) Projected precipitation change for the six models projecting the largest storm-track decrease. (f) As in (e), but for the six models projecting the smallest storm-track decrease. Contour intervals are 2 hPa2 (5°C)−1 in (a),(b) and 0.1 mm day−1 (5°C)−1 in (c)–(f). Shaded regions denote 90% significance based on model-to-model variability, with green shading denotes increase and brown shading denotes decrease.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
To see whether this difference might be related to the significant decrease in storm-track activity projected by CMIP5 models in summer, precipitation change projected by the six models that project the largest decrease in storm-track activity (as defined by the SLP variance over 30°–55°N, 130°–70°W) is shown in Fig. 7e, while that by the six models that project the smallest decrease is shown in Fig. 7f and the difference between the two is shown in Fig. 7d. The difference between the projected change in SLP variance between these two groups are shown in Fig. 7b, which can be compared to the 23-model mean shown in Fig. 7a. Consistent with the discussions above, there is large model-to-model difference in the projected change in storm-track activity and the difference between the two groups has similar amplitude as that found in the multimodel mean. Models projecting large decreases in storm-track activity project precipitation decrease covering the entire eastern part of the United States that extends to north of the Great Lakes into Canada to just south of the Hudson Bay (Fig. 7e), while those projecting small decreases in storm-track activity project significant increase in precipitation over the Great Lakes region (Fig. 7f). Differences between the two projections (Fig. 7d) show that models projecting large storm-track decreases also project more negative precipitation change over much of eastern North America, as well as over parts of the U.S. Southwest and northwestern Mexico.
The differences between the projected precipitation changes found in the CMIP5 models that project large and small decreases in storm-track activity in winter are shown in Fig. 8d. Negative but largely not statistically significant differences can be found over much of the western and southern United States, which is physically consistent with larger decrease in storm-track activity across the United States and southern Canada (Fig. 8b). Comparing the pattern shown in Fig. 8d to that shown in Fig. 8c, much of these differences occur over regions that lie close to the transition between projected decrease and projected increase in precipitation, suggesting that models projecting large decreases in storm-track activity also project precipitation decreases that extend slightly northward than those projecting small decreases in storm-track activity, which is indeed the case as shown by Figs. 8e,f. These results suggest that projected change in precipitation over Southern California is sensitive to how storm-track activity is projected to change (see also Neelin et al. 2013).
It is a bit surprising that the precipitation signal related to projected storm-track decrease appears to be only marginally significant in winter over the eastern United States, given the strong relationship between precipitation and cyclones mentioned above. Figure 8b suggests that the differences in the storm-track activity between the two groups of six models are dominated by the differences over the western part of the United States. Pfahl and Wernli (2012) suggested that the relationship between cyclones and extreme precipitation is stronger over the eastern United States than over the west. Hence, instead of selecting the group of model having the largest (and smallest) storm-track decrease over the entire United States and southern Canada, an alternative selection criterion is used, based on storm-track activity over the eastern United States (defined as the region 30°–50°N, 90°–70°W). The composite difference in storm-track activity between these two groups of models (Fig. 9a) show larger difference in SLP variance over the eastern United States than that shown in Fig. 8b. In terms of the precipitation projection, models projecting larger storm-track decrease over the eastern United States also project larger decrease in precipitation over much of the continental United States, with the difference between the two groups being statistically significant over the entire eastern United States (Fig. 9b). Comparing the precipitation projections based on the two groups separately, it is clear that the models projecting larger decrease in storm-track activity over the eastern United States (Fig. 9c) project a much more extensive area of decrease in precipitation that extends farther into the southern United States. If the scenario as depicted by Fig. 9c is realized, there will be huge consequences in terms of local hydrology over the southern United States.
As in Figs. 8b,d–f, but based on the projected storm-track changes over the eastern United States.
Citation: Journal of Climate 26, 24; 10.1175/JCLI-D-13-00209.1
While Figs. 7–9 suggest that the precipitation projection may be strongly tied to the storm-track projection in CMIP5 model simulations, it is not clear whether climate models can accurately simulate regional changes in precipitation related to large-scale circulation changes. For example, R. Fu et al. (2012, unpublished manuscript) suggest that most of the CMIP5 models they have examined cannot realistically capture the seasonal dependence of rainfall anomalies over the south-central United States on the El Niño–Southern Oscillation. Thus, further investigations on whether climate models can realistically simulate the relationship between the variations in cyclone statistics and precipitation are warranted.
5. Summary and conclusions
In this study, projections of storm-track changes over the continental United States and southern Canada made by 23 CMIP5 models under RCP8.5 and RCP4.5 have been examined and compared to changes projected by 11 CMIP3 models under SRES A2. Overall, under RCP8.5 forcing, CMIP5 models project a significant decrease in North American storm-track activity, with the largest decrease in summer and the smallest decrease in spring. The decrease is found both in sea level pressure variance and cyclone statistics. There is a strong consensus among the 23 CMIP5 models regarding the sign of the projected change, with less than 20% of the models (less than 10% in most cases) projecting changes in the opposite sign in any of the storm-track parameters examined. Nevertheless, there are also significant model-to-model differences in the magnitude of the projected changes.
In terms of sea level pressure variance, by the end of the twenty-first century CMIP5 models project a mean decrease ranging from −17.6% in summer to −6.6% in spring. CMIP3 models also project a significant decrease of −8.2% in summer but only a small and statistically not significant decrease in fall and small increases in winter and spring. Thus, CMIP5 models project much larger decreases in storm-track activity compared to CMIP3. This is true even if differences in projected Northern Hemisphere temperature change between CMIP5 and CMIP3 models are taken into account. Consistent differences between CMIP5 and CMIP3 results can also be found in the projected changes in meridional velocity variance at 700- and 300-hPa levels.
Cyclone statistics have also been derived from 6-hourly CMIP5 model data. Consistent with the results for sea level pressure variance statistics, CMIP5 models project an overall significant decrease in cyclone frequency and amplitude in all four seasons. The decrease is particularly acute for significant cyclones, defined as cyclones with perturbation amplitude of 20 hPa or larger in the cool seasons and 15 hPa or larger in summer, which represent the strongest 25% or fewer of all cyclones passing over the region in each season. CMIP5 models project an average decrease of −32.6% in the frequency of significant cyclones in summer. For winter, spring, and fall, the projected decreases amount to −15.9%, −6.6%, and −16.9%, respectively. While summer cyclones may be weaker (in terms of absolute amplitude) than those in the other seasons, recent studies (e.g., Kunkel et al. 2012) have suggested that extreme precipitation events over most regions of the United States (except for the Southeast in summer and fall) are predominantly caused by extratropical cyclones and their frontal systems, thus the possible consequences of a significant decrease in the frequency of summer cyclones should be further investigated.
Projected changes in the baroclinicity of the mean flow have also been examined. Model-to-model differences in the projected storm-track change as indicated by SLP variance statistics are found to correlate significantly with model-to-model differences in the projected change in a locally defined midtropospheric MAPE across the ensemble of 34 models made by combining the CMIP5 and CMIP3 projections, suggesting that the differences in the projected change in MAPE can partly account for not only the model-to-model differences but also the differences between CMIP5 and CMIP3 projections. Why the different models project different changes in MAPE and why CMIP5 models generally project more negative (or less positive) changes in MAPE compared to CMIP3 models are issues that still need to be addressed. In addition, the differences in MAPE clearly do not account for all the differences in the projections, and what other factors may also contribute still needs to be investigated.
On average, CMIP5 models project small and not statistically significant increase in the local MAPE in spring and only small and not significant decreases in the MAPE in winter and fall. Thus, decrease in the surface temperature gradient due to polar amplification of the surface warming during the cool seasons likely contributes to the significant decrease in storm-track activity projected in all three seasons. The more positive projected change in MAPE and less polar amplification of surface warming in spring, as compared to winter and fall, likely account for the smaller decrease in storm-track activity projected for spring.
Under global warming, precipitation is expected to increase in the higher latitudes and decrease in the subtropics. With cyclones being closely related to precipitation, a preliminary analysis on how the projected decrease in cyclone activity may modulate precipitation changes has been conducted. The results suggest that CMIP5 models projecting larger decrease in North American storm-track activity in summer also project a larger area of decrease in precipitation over the eastern United States and Canada. In winter, there are indications that models projecting larger storm-track decrease also tend to project a farther northward intrusion of the decrease in subtropical precipitation. Nevertheless, it is not clear whether CMIP5 models can accurately simulate the relationship between storm-track and precipitation changes but, with precipitation closely related to cyclones, the projected significant decrease in cyclone activity is expected to have significant impact on regional hydrology and this issue should be further examined.
While most of the results discussed in this study for CMIP5 models are for projected changes based on RCP8.5 experiments by the end of the twenty-first century, examination of results based on sea level pressure variance statistics at mid-twenty-first century, as well as those based on RCP4.5 forcings, suggests that, for CMIP5 projections, storm-track changes scale roughly linearly with projected Northern Hemisphere temperature change, implying that, even under the more moderate RCP4.5, significant decreases in storm-track and cyclone activity over North America should be expected by mid-twenty-first century.
This study has focused mainly on projected storm-track changes averaged over the entire continental United States and southern Canada, largely because CMIP5 models are not expected to provide accurate regional projections, with a majority of the models having horizontal grid spacing of about 200 km or coarser. In addition, as discussed above, recent studies suggested that diabatic impacts on cyclone dynamics may not be fully resolved at such coarse resolutions. These issues should be examined further, perhaps by examining time slice experiments conducted using high-resolution models having grid spacing much finer than 100 km.
Acknowledgments
The author acknowledges the support of NOAA Climate Program Office MAPP Program as part of the CMIP5 Task Force under Grant NA11OAR4310104. CMIP5 and CMIP3 data have been retrieved from the Earth System Grid and PCMDI data archives, and the author would also like to thank all the modeling centers for providing the model data and ECMWF for providing the ERA-Interim data. The author would also like to thank three anonymous reviewers for providing useful comments.
REFERENCES
Anderson, D., K. I. Hodges, and B. J. Hoskins, 2003: Sensitivity of feature-based analysis methods of storm tracks to the form of background field removal. Mon. Wea. Rev., 131, 565–573.
Bengtsson, L., K. I. Hodges, and E. Roeckner, 2006: Storm tracks and climate change. J. Climate, 19, 3518–3543.
Bengtsson, L., K. I. Hodges, and N. Keenlyside, 2009: Will extratropical storms intensify in a warmer climate? J. Climate, 22, 2276–2301.
Blackmon, M. L., 1976: A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere. J. Atmos. Sci., 33, 1607–1623.
Blackmon, M. L., J. M. Wallace, N.-C. Lau, and S. L. Mullen, 1977: An observational study of the Northern Hemisphere wintertime circulation. J. Atmos. Sci., 34, 1040–1053.
Butler, A. H., D. W. J. Thompson, and R. Heikes, 2010: The steady-state atmospheric circulation response to climate change–like thermal forcings in a simple general circulation model. J. Climate, 23, 3474–3496.
Butler, A. H., D. W. J. Thompson, and T. Birner, 2011: Isentropic slopes, downgradient eddy fluxes, and the extratropical atmospheric circulation response to tropical tropospheric heating. J. Atmos. Sci., 68, 2292–2305.
Cai, M., and M. Mak, 1990: Symbiotic relation between planetary and synoptic scale waves. J. Atmos. Sci., 47, 2953–2968.
Chang, E. K. M., 2009: Are band-pass variance statistics useful measures of storm track activity? Re-examining storm track variability associated with the NAO using multiple storm track measures. Climate Dyn., 33, 277–296, doi:10.1007/s00382-009-00532-9.
Chang, E. K. M., and S. Song, 2006: The seasonal cycles in the distribution of precipitation around cyclones in the western North Pacific and Atlantic. J. Atmos. Sci., 63, 815–839.
Chang, E. K. M., S. Lee, and K. L. Swanson, 2002: Storm track dynamics. J. Climate, 15, 2163–2183.
Chang, E. K. M., Y. Guo, and X. Xia, 2012: CMIP5 multimodel ensemble projection of storm track change under global warming. J. Geophys. Res.,113, D23118, doi:10.1029/2012JD018578.
Chang, E. K. M., Y. Guo, X. Xia, and M. Zheng, 2013: Storm-track activity in IPCC AR4/CMIP5 model simulations. J. Climate, 26, 246–260.
Chen, G., J. Lu, and D. M. W. Frierson, 2008: Phase speed spectra and the latitude of surface westerlies: Interannual variability and global warming trend. J. Climate, 21, 5942–5959.
Christensen, J. H., and Coauthors, 2007: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.
Colle, B. A., F. Buonaiuto, M. J. Bowman, R. E. Wilson, R. Flood, R. Hunter, A. Mintz, and D. Hill, 2008: New York City’s vulnerability to coastal flooding. Bull. Amer. Meteor. Soc., 89, 829–841.
Colle, B. A., Z. Zhang, K. A. Lombardo, E. K. M. Chang, P. Liu, and M. Zhang, 2013: Historical evaluation and future prediction of eastern North America and western Atlantic extratropical cyclones in the CMIP5 models during the cool season. J. Climate, 26, 6882–6903.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597.
Donohoe, A., and D. S. Battisti, 2009: The amplitude asymmetry between synoptic cyclones and anticyclones: Implications for filtering methods in feature tracking. Mon. Wea. Rev., 137, 3874–3887.
Field, P. R., and R. Wood, 2007: Precipitation and cloud structure in midlatitude cyclones. J. Climate, 20, 233–254.
Geng, Q., and M. Sugi, 2003: Possible change of extratropical cyclone activity due to enhanced greenhouse gases and sulfate aerosols—Study with a high-resolution AGCM. J. Climate, 16, 2262–2274.
Harvey, B. J., L. C. Shaffrey, T. J. Woollings, G. Zappa, and K. I. Hodges, 2012: How large are projected 21st century storm track changes? Geophys. Res. Lett., 39, L18707, doi:10.1029/2012GL052873.
Hodges, K. I., 1999: Adaptive constraints for feature tracking. Mon. Wea. Rev., 127, 1362–1373.
Holland, M. M., and C. M. Bitz, 2003: Polar amplification of climate change in coupled models. Climate Dyn., 21, 221–232.
Hoskins, B. J., and K. I. Hodges, 2002: New perspectives on the Northern Hemisphere winter storm tracks. J. Atmos. Sci., 59, 1041–1061.
Kidson, J., and E. P. Gerber, 2010: Intermodel variability of the poleward shift of the austral jet stream in the CMIP3 integrations linked to biases in 20th century climatology. Geophys. Res. Lett., 37, L09708, doi:10.1029/2010GL042873.
Klein, W. H., 1958: The frequency of cyclones and anticyclones in relation to the mean circulation. J. Meteor., 15, 98–102.
Kunkel, K. E., D. R. Easterling, D. A. R. Kristovich, B. Gleason, L. Stoecker, and R. Smith, 2012: Meteorological causes of the secular variations in observed extreme precipitation events for the conterminous United States. J. Hydrometeor., 13, 1131–1141.
Kutzbach, J. E., J. W. Williams, and S. J. Vavrus, 2005: Simulated 21st century changes in regional water balance of the Great Lakes region and links to changes in global temperature and poleward moisture transport. Geophys. Res. Lett., 32, L17707, doi:10.1029/2005GL023506.
Lambert, S. J., and J. C. Fyfe, 2006: Changes in winter cyclone frequencies and strengths simulated in enhanced greenhouse warming experiments: Results from the models participating in the IPCC diagnostic exercise. Climate Dyn., 26, 713–728.
Lim, E.-P., and I. Simmonds, 2009: Effects of tropospheric temperature change on the zonal mean circulation and SH winter extratropical cyclones. Climate Dyn., 33, 19–32, doi:10.1007/s00382-008-0444-0.
Long, Z., W. Perrie, J. Gyakum, R. Laprise, and D. Caya, 2009: Scenario changes in the climatology of winter midlatitude cyclone activity over eastern North America and the northwest Atlantic. J. Geophys. Res., 114, D12111, doi:10.1029/2008JD010869.
Lorenz, D. J., and E. T. DeWeaver, 2007: Tropopause height and zonal wind response to global warming in the IPCC scenario integrations. J. Geophys. Res., 112, D10119, doi:10.1029/2006JD008087.
Lu, J., and M. Cai, 2009: Seasonality of polar surface warming amplification in climate simulations. Geophys. Res. Lett., 36, L16704, doi:10.1029/2009GL040133.
Manabe, S., and R. J. Stouffer, 1980: Sensitivity of a global climate model to an increase of CO2 concentration in the atmosphere. J. Geophys. Res., 85, 5529–5554.
Meehl, G. A., C. Covey, K. E. Taylor, T. Delworth, R. J. Stouffer, M. Latif, B. McAvaney, and J. F. B. Mitchell, 2007a: The WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 1383–1394.
Meehl, G. A., and Coauthors, 2007b: Global climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–845.
Meinshausen, M., and Coauthors, 2011: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic Change, 109, 213–241, doi:10.1007/s10584-011-0156-z.
Moss, R. H., and Coauthors, 2010: The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756.
Murphy, J. M., B. B. B. Booth, M. Collins, G. R. Harris, D. M. H. Sexton, and M. J. Webb, 2007: A methodology for probabilistic predictions of regional climate change from perturbed physics ensembles. Philos. Trans. Roy. Soc., 365A, 1993–2028.
Murray, R. J., and I. Simmonds, 1991: A numerical scheme for tracking cyclone centres from digital data. Part I: Development and operation of the scheme. Aust. Meteor. Mag., 39, 155–166.
Neelin, J. D., M. Münnich, H. Su, J. E. Meyerson, and C. E. Holloway, 2006: Tropical drying trends in global warming models and observations. Proc. Natl. Acad. Sci. USA, 103, 6110–6115.
Neelin, J. D., B. Langenbrunner, J. E. Meyerson, A. Hall, and N. Berg, 2013: California winter precipitation change under global warming in the Coupled Model Intercomparison Project 5 ensemble. J. Climate, 26, 6238–6256.
Neu, U., and Coauthors, 2013: IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull. Amer. Meteor. Soc., 94, 529–547.
Novak, D., B. A. Colle, and S. Yuter, 2008: High-resolution observations and model simulations of an intense mesoscale snowband. Mon. Wea. Rev., 136, 1433–1456.
O’Gorman, P. A., 2010: Understanding the varied response of the extratropical storm tracks to climate change. Proc. Natl. Acad. Sci. USA, 107, 19 176–19 180.
Paciorek, C., J. Risbey, V. Ventura, and R. Rosen, 2002: Multiple indices of Northern Hemisphere cyclone activity, winters 1949–99. J. Climate, 15, 1573–1590.
Pfahl, S., and H. Wernli, 2012: Quantifying the relevance of cyclones for precipitation extremes. J. Climate, 25, 6770–6780.
Pinto, J. G., U. Ulbrich, G. C. Leckebusch, T. Spangehl, M. Reyers, and S. Zacharias, 2007: Changes in storm track and cyclone activity in three SRES ensemble experiments with the ECHAM5/MPI-OM1 GCM. Climate Dyn., 29, 195–210.
Raible, C. C., P. M. Della-Maria, C. Schwierz, H. Wernli, and R. Blender, 2008: Northern Hemisphere extratropical cyclones: A comparison of detection and tracking methods and different reanalyses. Mon. Wea. Rev., 136, 880–897.
Scheff, J., and D. M. W. Frierson, 2012: Robust future precipitation declines in CMIP5 largely reflect the poleward expansion of model subtropical dry zones. Geophys. Res. Lett., 39, L18704, doi:10.1029/2012GL052910.
Sheffield, J., and Coauthors, 2013a: North American climate in CMIP5 experiments. Part I: Evaluation of historical simulations of continental and regional climatology. J. Climate, in press.
Sheffield, J., and Coauthors, 2013b: North American climate in CMIP5 experiments. Part II: Evaluation of historical simulations of intraseasonal to decadal variability. J. Climate, in press.
Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, and H. L. Miller, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498.
Teng, H., W. M. Washington, and G. A. Meehl, 2008: Interannual variations and future change of wintertime extratropical cyclone activity over North America in CCSM3. Climate Dyn., 30, 673–686, doi:10.1007/s00382-007-0314-1.
Ulbrich, U., J. G. Pinto, H. Kupfer, G. C. Leckebusch, T. Spangehl, and M. Reyers, 2008: Changing Northern Hemisphere storm tracks in an ensemble of IPCC climate change simulations. J. Climate, 21, 1669–1679.
Ulbrich, U., G. C. Leckebusch, and J. G. Pinto, 2009: Extratropical cyclones in the present and future climate: A review. Theor. Appl. Climatol., 96, 437–459.
Wallace, J., G. Lim, and M. Blackmon, 1988: Relationship between cyclone tracks, anticyclone tracks and baroclinic waveguides. J. Atmos. Sci., 45, 439–462.
Wetherald, R. T., and S. Manabe, 2002: Simulation of hydrologic changes associated with global warming. J. Geophys. Res., 107, 4379, doi:10.1029/2001JD001195.
Willison, J. A., W. R. Robinson, and G. M. Lackmann, 2013: The role of storm-scale latent heating in the North Atlantic storm track and its dependence on model resolution. J. Atmos. Sci., 70, 2234–2250.
Yin, J. H., 2005: A consistent poleward shift of the storm tracks in simulations of the 21st century climate. Geophys. Res. Lett., 32, L18701, doi:10.1029/2005GL023684.