1. Introduction
Damaging surface wind gusts, large hail, and tornadoes that comprise hazardous convective weather (HCW) pose a significant threat to life and property and remain a prominent source of weather-related losses in the United States. Indeed, HCW accounted for over one-third of the 151 weather-related disasters exceeding $1 billion in damages during the time period 1980–2013 (NOAA 2014). While heightened risk for weather disasters may result from enhanced exposure as population grows and expands into previously less populated areas (e.g., Hall and Ashley 2008; Ashley et al. 2014; Strader and Ashley 2015), it has been postulated that a warming global climate may also contribute. Should such a physical increase in frequency and/or severity be realized, future losses due to HCW may be expected to increase, and even more so when nonmeteorological factors that impact human vulnerability are considered (Gensini et al. 2014).
Recent efforts to address how HCW may change in the future have focused mainly upon the environmental parameters that characterize the potential for the development of severe thunderstorms, namely, convective available potential energy (CAPE) and deep-layer vertical wind shear, which are assumed to be adequately resolved by global climate model (GCM) projections. Results largely agree that favorable environmental conditions for HCW are likely to become more frequent through the twenty-first century (Del Genio et al. 2007; Trapp et al. 2007a, 2009; Van Klooster and Roebber 2009; Lee 2012; Diffenbaugh et al. 2013; Paquin et al. 2014; Seeley and Romps 2015), with climate model signals nearly unanimous in their projections of increased CAPE values. Despite the projected reductions in mean deep-layer vertical wind shear, the literature suggests that favorable daily covariation of these two parameters will result in an uptick in days with favorable severe environments. Similar conclusions have been reached elsewhere in the world, such as Europe and Australia (Marsh et al. 2009; Allen et al. 2014a,b).
This environmental, or implicit, modeling approach is undoubtedly limited by the assumption that the convective-scale phenomena will be realized within these environments. Thus, researchers have had to assume that the proportion of favorable environments that result in severe storms will remain unchanged in a globally warmed climate, despite acknowledgment of this drawback (e.g., Trapp et al. 2007a). While necessary with this methodological approach, the presumption of constant environmental efficiency in producing severe storms remains a significant source of uncertainty, as climate change may modify factors pertinent to the development of storms, such as extratropical cyclone frequency and storm track, fronts, and convective inhibition (CIN), to name a few. As a consequence, dynamical downscaling approaches at convective permitting resolution have been suggested as an alternative approach (Trapp et al. 2007a,b; Brooks 2013; Gensini and Mote 2014).
With the advancement of computing resources, in addition to the growing availability of GCM data on a subdaily temporal frequency, it has now become feasible to produce high-resolution simulations through dynamical downscaling of climate model projections. While computationally expensive, utilizing a high-resolution convection-permitting mesoscale model allows the convective events to develop intrinsically within the large-scale environment, thus accounting for convective initiation. Convection permitting refers to the lack of a cumulus parameterization in the numerical model; in general, a horizontal grid spacing of 4 km or less has been deemed sufficient to resolve the larger convective circulations without the need for parameterization (Weisman et al. 1997). Previous work has demonstrated the concept of dynamical downscaling using global reanalysis data to construct synthetic regional climatologies of severe storms (Trapp et al. 2011; Robinson et al. 2013). Trapp et al. (2011) used a sequence of daily reinitialized, 24-h convection-permitting forecasts (4.25-km horizontal grid spacing) using the Weather Research and Forecasting (WRF) Model. A model proxy was used to estimate the occurrence of a simulated severe weather occurrence (hail, wind, and/or tornado) based on the coupling of updraft helicity (UH) in the 2–5-km AGL layer, a measure of storm rotation, and simulated radar reflectivity factor Z. Results demonstrated that this approach had some degree of skill in emulating the climatological distribution of HCW. Robinson et al. (2013) conducted a similar study for the 20-yr period 1990–2009 but used an artificial neural network to estimate severe occurrences, finding that UH and CAPE were the most important factors in determining a model-simulated storm to be severe. Simulated severe events and environmental controls over the period exhibited little to no trend, which overall agrees with the findings of Gensini and Ashley (2011). More recently, Gensini et al. (2014) produced a high-resolution (4 km) dynamically downscaled climatology using WRF and driving data from a historical global climate simulation for the months of March–May for the period 1980–90 using the Community Climate System Model, version 3 (CCSM3). As in Trapp et al. (2011), the UH–Z proxy was used to estimate all severe hazards; downscaled simulations were found to reasonably capture the interannual variability and diurnal cycle of observed severe reports. However, as in Trapp et al. (2011), an underestimation of severe occurrences was noted in May, which may be attributed to the inability of the model proxy to account for convective mode or the scale of forcing for ascent. Overall, each of these studies have shown that the climatological distribution of severe convection can reasonably be replicated using coarse initial and boundary conditions, either furnished by a reanalysis or GCM dataset, to drive high-resolution convection allowing models.
In regard to future climate, a growing number of studies have produced high-resolution regional climate simulations to study future changes at localized scales (e.g., Kendon et al. 2012; Mahoney et al. 2013; Lauer et al. 2013; Harding et al. 2013; Prein et al. 2013; Rasmussen et al. 2011, 2014; Prein et al. 2015). Yet, few studies have used such an approach to assess how HCW may be impacted in the future. Mahoney et al. (2012) produced high-resolution WRF simulations (1.33 km) over Colorado and examined how hailstorm frequency and intensity may change in the future, and they found that for warm-season extreme precipitation events, hail reaching the surface was significantly decreased owing to the increased height of the melting level. More recently, Gensini and Mote (2015) generated high-resolution (4 km) WRF simulations for the months of March, April, and May during a limited future period (2080–90) climate. Results indicate that compared to the historical baseline, a statistically significant increase in HCW is noted in March with an overall increased tendency for severe weather during the March–May months. Additionally, variability of severe weather was also shown to increase under future climate in the spring months.
This work seeks to produce high-resolution dynamically downscaled simulations for the entire annual cycle over longer time periods (30 yr) for both a historical and future climate period. Descriptions of the GCM data used for downscaling, the experimental setup, and different analysis methods used for assessing projected changes are provided in section 2. In section 3, results are presented for both the changes in favorable convective environments projected by the chosen GCM and the ensuing downscaled estimates of HCW. A comparison of outcomes depicted by the environmental response of the parent GCM and the resulting synthetic events in the downscaled simulations are then examined in section 4 to assess the environment–event relationship and how this may or may not complicate the interpretation of the large-scale environment analysis from the GCM. A discussion of these results is provided that addresses the limitation of using only large-scale environmental changes alone to infer projected impacts upon the spatiotemporal frequency of HCW. Last, concluding remarks and suggestions for future research directions are given in section 5.
2. Data and methods
a. Global climate model
The Geophysical Fluid Dynamics Laboratory Climate Model, version 3 (GFDL CM3), serves as the parent model in this study. Results from both Diffenbaugh et al. (2013) and Seeley and Romps (2015) have demonstrated GFDL CM3 to be a high-performing GCM in terms of its ability to represent the historical climatology of convective parameters and severe weather environments as compared to both reanalysis and radiosonde observations. The atmospheric model component (AM3) of GFDL CM3 utilizes a finite-volume dynamical core, a hybrid sigma–pressure vertical coordinate system with 48 vertical levels and a model top of 1 hPa, and a 2.5° × 2° longitude–latitude grid spacing. Further details, including a description of the physical parameterizations used within GFDL CM3 simulations are outlined in Donner et al. (2011). Historical and future climate experiments available from phase 5 of the Coupled Model Intercomparison Project (CMIP5; Taylor et al. 2012) are utilized here, specifically the historical and the most aggressive representative concentration pathway 8.5 (RCP8.5) simulations. The CMIP5 historical model experiment is initialized in 1850 and run through 2005, incorporating observed changes in atmospheric composition (greenhouse gases, aerosols, and volcanic influences), solar output, and time-evolving land cover (Taylor et al. 2012). Meanwhile, the RCP8.5 experiment (2006–2100) accounts for an increase in radiative forcing to 8.5 W m−2 by the year 2100, which equates to a quadrupling of CO2 concentration (>1370 ppm) from preindustrial levels (Moss et al. 2010; Van Vuuren et al. 2011).
b. Regional climate model
The WRF-ARW version 3.6 acts as the regional climate model (RCM) to dynamically downscale GFDL CM3 under historical (1971–2000) and RCP8.5 (2071–2100) conditions (r1i1p1 realization member). Simulations encompass the entire annual cycle, which is a significant departure from previous downscaling efforts that focused solely on warm season months when severe weather frequency traditionally peaks. A single computational domain encompassing the entire contiguous United States (CONUS) serves as the region of study (Fig. 1). The horizontal grid spacing over this domain is 4 km, thus precluding the need for cumulus parameterization. Initial and boundary conditions are supplied by the 6-hourly GCM data for the large-scale atmospheric fields and monthly average soil parameters (temperature and moisture) and sea surface temperatures. The 3D atmospheric variables are interpolated from the native hybrid vertical coordinates to pressure levels. A 10-point buffer zone is applied at the lateral boundary (Davies and Turner 1977) to provide a smoother transition at the lateral boundaries between the coarser boundary conditions and the model interior (Giorgi and Mearns 1999; Gula and Peltier 2012). Specific information regarding model configuration, including choice of parameterizations, may be found in Table 1.
WRF domain for dynamically downscaled simulations.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
WRF configuration information.
The procedure for downscaling involves a succession of daily initialized (0600 UTC) 30-h WRF Model integrations with hourly data output. The first six forecast hours of integration are considered a model spinup period (Skamarock 2004) and thus are disregarded prior to computing climate statistics (e.g., Kotlarski et al. 2012; Lucas-Picher et al. 2013). In this sense, the method of downscaling treats the regional climate as a collection of 24-h simulations (valid between 1200 and 1200 UTC), as in the approach of Trapp et al. (2011) and Robinson et al. (2013). Previous work has been shown that frequent reinitialization can improve the sequence of weather events in simulations, better represent the spatial patterns and diurnal cycle of precipitation, prevent significant drift away from the large-scale conditions within the parent model, and limit error growth within the domain (e.g., Pan et al. 1999; Qian et al. 2003; Lo et al. 2008; Kotlarski et al. 2012; Lucas-Picher et al. 2013; Hong and Kanamitsu 2014). While additional computing is required to account for spinup, the method also allows for a large number of simulations to be executed simultaneously. A drawback of this approach regards the frequent updating of the land surface variables (e.g., soil moisture). Soil variables are slow to spin up and may require a season to several years to reach equilibrium (e.g., Giorgi and Mearns 1999). On the other hand, it has been suggested that the effects of long memory processes such as soil moisture may be of secondary importance to atmospheric forcing (Pan et al. 1999). It is also assumed that these processes are being handled adequately by the GCM land surface model, though feedbacks from the higher-resolution simulation are not taken into account. Furthermore, the one-way nesting approach does not account for any upscale feedback from the regional domain to the larger-scale environment used for initial and boundary conditions.
c. HCW environments from GFDL CM3
As in previous studies, the product of CAPE and 0–6-km vertical wind shear (S06) greater than a specified threshold is used here to identify a favorable environment for HCW (Craven and Brooks 2004; Marsh et al. 2007, 2009; Trapp et al. 2007a, 2009; Gensini and Ashley 2011; Robinson et al. 2013; Diffenbaugh et al. 2013). A threshold of 20 000 m3 s−3 is used to identify an environment supportive of HCW, which is within the range of values applied in prior studies (e.g., Trapp et al. 2007a; Gensini and Ashley 2011; Diffenbaugh et al. 2013; Seeley and Romps 2015). CAPE is calculated using a parcel representing the mixed layer within the lowest 100 hPa of the atmosphere and employs the virtual temperature correction. S06 is computed as the magnitude of the vector difference between winds at 6 km AGL and those near the surface. Here, we further require that CAPE ≥ 100 J kg−1, S06 ≥ 5 m s−1, and CIN > −100 J kg−1. The latter is introduced to filter environments that are strongly capped and unlikely to produce convection, as in the approach taken by Gensini and Ashley (2011). Variables are interpolated to a 1° grid to maintain consistency with previous studies (e.g., Diffenbaugh et al. 2013; Seeley and Romps 2015). A severe environment day (NDSEV) is said to occur if the minimum threshold is exceeded at any time over the 24-h period between 1200 and 1200 UTC.
d. Synthetic HCW from dynamically downscaled simulations
The severe storms community has developed diagnostic variables to serve as surrogates for simulated severe storms at convection-permitting grid lengths (e.g., Kain et al. 2008, 2010; Sobash et al. 2009, 2011; Carley et al. 2011). The use of UH exceeding a specified threshold has been a common proxy and has been shown to adequately capture the geographical progression of observed HCW climatology during the spring (Trapp et al. 2011; Robinson et al. 2013; Gensini et al. 2014; Gensini and Mote 2015). However, Trapp et al. (2011) and Robinson et al. (2013) found overall HCW occurrences were not as well captured during June, with a notable underestimation in the eastern United States. This result may be due to HCW occurrences (severe wind occurrences, in particular) that go undetected via the UH diagnostic owing to the association of such events with a nonsupercell mode of convection. Thus, UH may not be the most suitable proxy for use in all seasons and geographic regions. As an alternative, we explore the use of the maximum upward vertical velocity (UVV) in the lowest 400 hPa as a measure of storm intensity, with a greater magnitude indicative of a stronger storm capable of producing HCW. Here, we provide a brief evaluation of the HCW day climatology from WRF using both the run-time maximum values (e.g., Kain et al. 2010) of UH and UVV as compared to the observed HCW day climatology over the period 1971–2000. Observations of HCW are obtained from the National Centers for Environmental Information (NCEI) publication Storm Data and are available from the Storm Prediction Center (SPC) website (http://www.spc.noaa.gov/wcm/); an HCW day is considered to have occurred if a severe hail, wind, or tornado report had occurred at any time during the 24-h period between 1200 and 1200 UTC. We tested 11 different thresholds for UH and UVV, beginning with a minimum threshold based upon the 99.995th percentile of the frequency distribution (35 m2 s−2 and 20 m s−1 for UH and UVV, respectively) and increasing at regular intervals (5 m2 s−2 and 1 m s−1) thereafter. Summary measures gauging the mean error and overall spatial pattern agreement [here, the root-mean-square error (RMSE) and pattern correlation (PC)] between the observed and synthetic WRF seasonal climatology are shown in Fig. 2. Overall, UVV is shown to better replicate the spatial pattern of HCW for all seasons, JJA in particular as UH fails to capture HCW east of the Mississippi River (not shown), despite having a high bias. Increasing the minimum threshold to 22 m s−1 during JJA aids in reducing the error without significantly degrading the pattern correlation and RMSE for each season; as such, we opt for better replication. For this reason, a severe weather day in WRF, hereafter referred to as synthetic HCW, occurs when UVV exceeds 22 m s−1 at any time over the 24-h period (1200–1200 UTC) within the latitude–longitude bounds defined by the interpolated 1° GCM grid. While this approach does not take into account subdaily frequency of individual storms, it allows for a relatively straightforward comparison to the daily frequency of severe environments depicted by the GCM.
RMSE (solid lines) and PC (dashed lines) between mean seasonal number of days with (a) UH and (b) UVV greater than varying thresholds as compared to observed HCW day climatology for the period 1971–2100.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
3. Results
The response in mean annual NDSEV anomaly from GFDL CM3 over the period 1950–2100, expressed as percentage change relative to the 1971–2000 baseline mean, is shown in Fig. 3. Substantial increases in mean frequency are projected over the twenty-first century, with nearly a 150% increase by 2100. The annual anomalies for synthetic HCW days are strongly correlated to the environmental response (0.860 and 0.885 for the historical and future periods, respectively) and similarly depict sizable increases in the future, though the relative increase from the climatological baseline is less than that of NDSEV, between 50% and 80% (Fig. 4). Spatially, the mean seasonal response in NDSEV and synthetic HCW days are illustrated in Fig. 5. Statistically significant increases in daily frequency are found throughout the entire year, although this varies regionally and in magnitude. To measure changes on a regional basis, five distinct geographical regions are defined (Fig. 6). During the winter months of December–February (DJF), the most significant changes are confined to the U.S. Southeast region, with a regional mean increase of 4.4 and 1.5 days season−1 for NDSEV and synthetic HCW, respectively (Table 2). Nearly the entire United States east of the Continental Divide, from the Gulf of Mexico to the Canadian border, exhibits a significant rise in frequency for both NDSEV and synthetic HCW during MAM. On average, regionally averaged increases of approximately 6 and 3 days for NDSEV and synthetic HCW are noted over the eastern two-thirds of the CONUS, respectively, with larger regional increases in the southern Great Plains, Midwest, and the Southeast regions. The most robust increases in frequency materialize during the summer (JJA), accentuating the north-central portion of the CONUS, with some areas over Montana, Wyoming, Colorado, and the Dakotas experiencing mean seasonal increases in NDSEV by as much as 20–25 days by the 2071–2100 time period. Likewise, the maximum increases in synthetic HCW are concentrated in the lee of the Rocky Mountains, with maximum gains in the range of 15–20 days. These areas of marked frequency increase represent a northward and westward a shift in the historical maxima of NDSEV and synthetic HCW by approximately 1° in both latitude and longitude. Elsewhere, the upper Midwest and much of the United States east of the Mississippi River also see a statistically significant rise in frequency NDSEV and synthetic HCW. Meanwhile, relatively little change and even slight reductions in synthetic HCW are evident over the portions of the Midwest and southern Great Plains regions, despite NDSEV increases over much of these areas. These decreases in synthetic HCW are on the order of 1.5–2 days and are most pronounced during July (not shown). Finally, in the autumn (SON), a relatively large area of the United States has statistically significant rises in the mean NDSEV, from the Great Plains eastward through the Midwest and into the U.S. Northeast, and down along the eastern seaboard, with the largest change in the seasonal mean frequency concentrated in the north-central United States. The mean increases are on the order of 1–2 days season−1 across each region. Additionally, similar areas undergo changes in synthetic HCW frequency, with an average increase in the range of 0.7–1.7 days for each region, with the bulk of these seasonal changes originating from the month of September. Overall, the projected mean change in NDSEV tends to be 2–4 times greater than that of synthetic HCW on an annual basis, with the largest departures occurring during JJA and even more significantly in the southern Great Plains (Table 2).
Time series of annual anomaly in regional (land points in the United States east of 105°W) mean NDSEV from GFDL CM3. Anomalies are expressed as a percentage departure from the baseline (1971–2000) mean. The historical experiment (1950–2005) is represented by the blue line, and the RCP8.5 scenario (2006–2100) is represented by the red line. A Gaussian filter is applied with σ = 5 yr represented by the thick lines. Vertical dashed lines indicate the two 30-yr periods downscaled by WRF.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Time series of annual regional anomaly (CONUS land points east of 105°W) in both NDSEV from GFDL CM3 (solid line) synthetic HCW days from WRF (dashed). The historical (1971–2000) values are represented in blue and the future RCP8.5 (2071–2100) values in red. Anomalies are computed as in Fig. 3, and thick lines represent data smoothed with a Gaussian filter (σ = 5 yr).
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Mean seasonal response in NDSEV from GFDL CM3 and synthetic HCW days from WRF in the future (2071–2100) relative to the historical (1971–2000) period. Stippling indicates where the distribution of seasonal means between the two periods are statistically significant from one another at the 95% confidence level using the Mann–Whitney U test.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Regional boundaries used for localized geographic analysis. Because of the relative infrequency of severe weather activity in the western United States, the regional analysis will be focused on areas in the CONUS east of the Continental Divide. The number of grid boxes contained within each region are 152, 67, 110, 160, and 137 for the Midwest, Northeast, Southeast, southern Great Plains, and the northern Great Plains, respectively.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Annual and seasonal mean changes (in days) of NDSEV from GFDL CM3 and WRF synthetic HCW. Changes are representative of the mean differences between the historical (1971–2000) and future (RCP8.5; 2071–2100) periods for the CONUS east of 105°W and each subregion domain.
A finer temporal analysis based upon Julian date provides additional insight into these projected changes. To start, the probability of experiencing NDSEV and synthetic HCW at any grid point in the eastern United States (east of 105°W) during the historical and the future 30-yr periods are illustrated in Figs. 7a and 7b, respectively. The largest changes are generally seen in the tails of the annual probability distribution, though increases in probability occur throughout the annual cycle. Given previous results indicating increases in seasonal mean frequency, it is unsurprising that the total number of days with NDSEV and synthetic HCW occurring anywhere in the United States also increases, with corresponding mean increases of 32% and 19% (Figs. 7c,d). Furthermore, the widening of the future annual probability distribution relative to the historical period is noteworthy and suggestive of a lengthening of the HCW season in the future, a notion supported by examination of the mean empirical cumulative distribution function (CDF; Figs. 7e,f) of accumulated days. The CDF quantifies the proportion of the total number of days that are accumulated by day of the year. If we consider the season to be defined, albeit arbitrarily, as encompassing the amount of time where 80% of the total number of days are accumulated (i.e., between the 10th and 90th percentiles of the cumulative probability), both NDSEV and synthetic HCW indicate the season lengthens by 36 days, on average (Table 3).
(a),(b) The 30-yr mean probability estimates of NDSEV and WRF synthetic HCW day anywhere in the United States (east of 105°W; land points only) by Julian day. Raw probabilities are represented by the scatter points, and Gaussian smoothed (σ = 15 days) by the lines as in Brooks et al. (2003). Smoothed 95% confidence intervals are shaded. The historical period (1971–2000) is represented by the blue line and RCP8.5 scenarios (2071–2100) by the red lines, respectively. (c),(d) Accumulated GFDL CM3 NDSEV and WRF synthetic HCW days for both the historical and RCP8.5 periods. (e),(f) Mean empirical CDF of accumulated GFDL CM3 NDSEV and WRF synthetic HCW days for both the historical and RCP8.5 periods. Bootstrapped 95% confidence intervals are shaded.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Mean differences (in days) between the historical (1971–2000) and future (RCP8.5; 2071–2100) periods for several cumulative probability levels of the empirical CDFs for GFDL CM3 NDSEV and WRF synthetic HCW days.
To complement analyses focusing solely on daily occurrences, here we also examine the number of daily gridpoint occurrences, which essentially gauges the areal coverage of activity on a given day. In Fig. 8a, the smoothed daily mean gridpoint frequency for NDSEV and synthetic HCW are shown for each climatological period, and shown are a greater gridpoint frequency for both NDSEV and synthetic HCW in the future relative to the historical climate period, with the most significant increases occurring from May to September. Unmistakably, the increase in future NDSEV cumulative gridpoint frequency is considerable compared to synthetic HCW, with a mean annual increase of 130% for NDSEV yet only a 58% increase for synthetic HCW. Subsequently, this illustrates that the areal coverage of NDSEV far outpaces the change in synthetic HCW.
(a) Mean CONUS gridpoint frequency (east of 105°W, land points only) by Julian date (smoothed with Gaussian filter; σ = 15 days) for NDSEV (solid line) and WRF synthetic HCW (dashed). Smoothed 95% confidence intervals are shaded. The historical period (1971–2000) is represented in blue and the future RCP8.5 scenarios (2071–2100) by the red lines. (b) Cumulative gridpoint frequency, with bootstrapped 95% confidence intervals shaded.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Analysis thus far has focused on the frequency of synthetic HCW; however, of additional interest is the potential change in intensity of severe thunderstorms. Here, intensity is gauged by the magnitude of UVV, with a larger vertical velocity indicative of a stronger storm updraft. The storms with the strongest updraft velocities, not unexpectedly, occur during MAM and JJA. Using thresholds of 30, 35, and 40 m s−1, statistically significant increases in daily frequency occur only during these seasons (Fig. 9). Predominantly, the enhanced frequency of days with the strongest updrafts is confined to the south-central United States during the spring and the northern Great Plains into the Midwest during the summer. Of note are the statistically significant increases in frequency (at the 30 and 35 m s−1 thresholds) during JJA in some areas in the lower Midwest, namely Missouri, Illinois, and Indiana, which previously exhibited relatively little, if any, change at the lower threshold of 22 m s−1. This demonstrates that there may be little change in the frequency of overall HCW activity, but HCW events produced by stronger storms may become more common. Ultimately, stronger updrafts are supportive of, but do not necessarily equate to, more significant convective hazards (i.e., stronger tornadoes, convective wind gusts, and/or larger hail). Beyond updraft speed, related factors that affect the characteristics of HCW require additional study and will be the subject of future work.
MAM and JJA seasonal mean response in days with UVV exceeding (top)–(bottom) 30, 35, and 40 m s−1 for the future (2071–2100) period relative to the historical (1971–2100) period. Stippling indicates where the distributions of seasonal means between the two periods are statistically significant from one another at the 95% confidence level.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
4. Discussion
Because of the large computational expense of dynamical downscaling, it becomes relevant to question whether this undertaking lends greater insight into potential changes in HCW over the conventional environmental approach. A noted limitation of the environmental approach is the assumption that the proportion of favorable environments that yield HCW will remain unchanged in the future; this is a significant source of uncertainty as it does not account for factors that may promote or inhibit the initiation of convective storms. Based on previous results, both approaches generally indicate similar areas of increased frequency during the same time of the year. Indeed, seasonal spatial patterns between NDSEV and synthetic HCW correlate well (see Table 4), a good indication that the large-scale environment strongly modulates the timing and location of ensuing synthetic HCW events. Moreover, the association between monthly mean NDSEV and monthly mean synthetic HCW days reveals that NDSEV explains over 80% of the variance of synthetic HCW, with coefficient of determination R2 values of 0.82 and 0.87 for the historical and future periods, respectively (Fig. 10). It is readily apparent, however, that the slope of each line differs, with a 30% decline in the future period, suggestive that the future response of synthetic HCW is weaker relative to the increase in NDSEV.
Pattern correlation between seasonal mean NDSEV from GFDL CM3 and WRF synthetic HCW for both the historical (1971–2000) and future (RCP8.5; 2071–2100) periods.
Linear association between regional monthly mean (CONUS east of 105°W; land points only) NDSEV from GFDL CM3 and WRF synthetic HCW days. Monthly mean values were computed for each of the 30 yr of historical (1971–2000; blue) and future (RCP8.5, 2071–2100; red), and bootstrapped 95% confidence intervals are shaded. The least squares regression equation (and 95% confidence intervals for the slope) and coefficients of determination are displayed in the bottom-right corner.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
To further pursue this finding, we present analyses of the environmental bias and conditional probability. Here, environmental bias refers to the ratio between NDSEV and synthetic HCW days, whereas conditional probability describes the probability of a synthetic HCW day given the occurrence of NDSEV. Monthly mean environmental bias and conditional probability, averaged over the eastern two-thirds of the CONUS domain, are shown in Figs. 11a and 11b. The environmental bias increases throughout the entire annual cycle, though most significantly over the months from May to September, a result that points to an increase in the rate at which NDSEV overestimates synthetic HCW. Spatially, the mean changes in environmental bias increase for most locations during all four seasons, but regional variability is noted (Fig. 12a). As for the conditional probability, the occurrence of NDSEV in the future period is more likely to produce synthetic HCW in DJF, MAM, and SON, though less robustly during the autumn months (Fig. 12b). The opposite is true during JJA, where the mean conditional probability tends to decrease significantly; this is true for a large corridor of the central United States (Fig. 12b), though the opposite is true for areas in the lee of the Rocky Mountains and in the U.S. Northeast.
Mean monthly (a) environmental bias and (b) conditional probability of synthetic HCW given NDSEV. Values represent the mean areal average over the CONUS domain (east of 105°W) for both historical (1971–2000; blue) and future (RCP8.5, 2071–2100; red) periods. Bootstrapped 95% confidence intervals are shaded.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
Mean change in seasonal (a) environmental bias and (b) conditional probability of synthetic HCW given NDSEV in the future (2071–2100) relative to the historical (1971–2000) period. Stippling indicates where the distribution of seasonal means between the two periods are statistically significant from one another at the 95% confidence level using the Mann–Whitney U test.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
A question that could be raised regarding these inconsistencies between the future response from the GCM and WRF is whether the differences are a result of a discrepancy in the representation of the larger-scale environment by the two models or if changes in the processes that enable the environments to be realized play a larger role. For the former, mean changes in NDSEV from WRF agree favorably with that from GFDL CM3 (see Fig. 13 and Table 5), though a distinction is noted during JJA along the Gulf Coast from Texas to Mississippi where WRF NDSEV far exceeds that from GFDL CM3. This appears to result from an underestimation of mean CIN in WRF future simulations in the south-central United States, thereby increasing the frequency of NDSEV. Other parameters (CAPE, S06, and surface specific humidity) show relatively good agreement in terms of spatial pattern and magnitude of change (Table 5), and the overall large-scale environment represented in WRF remains largely consistent with that of the parent GCM. In line with previous studies (e.g., Trapp et al. 2007a, 2009; Diffenbaugh et al. 2013), the increase in NDSEV is primarily driven by increases in CAPE, in large part attributed to robust increases in near-surface specific humidity. Both CAPE and specific humidity are shown to undergo significant increases in all seasons by the late twenty-first century within the GFDL CM3 projections, but most significantly during JJA, with seasonal mean CAPE increasing from 500 to over 1000 J kg−1 for some locations in the central United States (Fig. 14). While mean S06 does tend to decrease overall, it varies regionally and by season, with the most significant weakening during JJA and SON. Despite this, the subdaily combination of CAPE and S06 still support an increased frequency of NDSEV, largely driven by the robust increase in CAPE. Parcel theory relates the theoretical maximum updraft speed wmax to CAPE as
Comparison of the mean seasonal response in NDSEV from (left) GFDL CM3 and (right) WRF in the future (2071–2100) relative to the historical (1971–2000) period. NDSEV from WRF is estimated on the native 4-km grid using the same 6-hourly time intervals as from GFDL CM3 and then averaged within the 1° grid boxes.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
RMSE and PC between WRF and GCM seasonal mean NDSEV (days), 2-m q (g kg−1), CAPE (J kg−1), CIN (J kg−1), and S06 (m s−1) for the historical (1971–2000) and future (RCP8.5; 2071–2100) periods.
Seasonal mean differences 2-m q (g kg−1), CAPE (J kg−1), CIN (J kg−1), and S06 (m s−1) in the future (2071–2100) relative to the historical baseline mean (1971–2000) from GFDL CM3. Stippling indicates where the distribution of seasonal means between the two periods are statistically significant from one another at the 95% confidence level using the Mann–Whitney U test.
Citation: Journal of Climate 30, 24; 10.1175/JCLI-D-16-0885.1
While the presence of favorable large-scale conditions for severe thunderstorms does not guarantee the occurrence of HCW, the change in the incidence of synthetic HCW relative to NDSEV frequency in the future is compelling. The changing response between the two periods suggests that processes that promote (or inhibit) initiation of convection, whether on the synoptic and/or mesoscale, may be modulated in a future climate. The role of increased CIN in suppressing convection may influence synthetic HCW occurrence; analysis from GFDL CM3 indicates that mean CIN magnitude increases across all seasons, but most strongly during JJA, upward of 40–80 J kg−1 in the central United States, and in the same areas that exhibited little change or slight decreases in synthetic HCW (Fig. 14). Indeed, Trapp and Hoogewind (2016) found that convective storms failed to initiate in numerical modeling experiments under pseudoglobally warmed conditions, largely due to both increased CIN and reduced parcel lifting. Modifying the CIN criteria in the formulation of NDSEV from −100 to −50 J kg−1 does little to change the overall results; this suggests that CIN alone is unlikely the primary contributor to changes in storm initiation. Rather, a reduction in parcel lifting may play a larger role. One such possibility involves the incidence of extratropical cyclones. Ideally, cyclone-tracking procedures could be implemented to directly address this problem, though this is currently beyond the scope of this study. Instead, we point to the analysis of Chang (2013), who revealed that the frequency of extratropical cyclones over the CONUS within the GFDL CM3 simulations demonstrated statistically significant decreases for MAM, JJA, and SON over the period 2081–2100 (relative to 1980–99), with DJF remaining relatively unchanged. The largest reductions are shown to occur during JJA (−24.5%). While this requires further study, the reduction in extratropical cyclone frequency appears plausible as a contributor to the reduced conditional probability during JJA.
5. Conclusions
The traditional approach for assessing the potential impact of climate change on HCW has relied upon analysis of projected changes in frequency of favorable environmental conditions from global climate model projections. This approach, however, fails to account for whether or not storms will be realized within such favorable environments. In this work, dynamical downscaling at convection-allowing scales allows the high-resolution model to intrinsically simulate the initiation, evolution, and intensity of convective weather systems within a driving global climate model. Results from this high-resolution dynamical downscaling approach show HCW increases in both frequency and intensity by the end of the twenty-first century. Both the environmental results from GFDL CM3 and dynamically downscaled synthetic HCW from WRF largely agree on the seasonal timing and locations of enhanced frequency. Likewise, each approach suggests that the severe weather season may lengthen, perhaps extended by more than one month. However, the high-resolution downscaling provides evidence that the proportion of favorable environments resulting in HCW may find significant changes in the future. For example, results show that the conditional probability of HCW declines during summer for much of the central United States, which may be attributed to both an increase in the magnitude of convective inhibition and decreased forcing for ascent. Such an outcome supports the motivation for continued use of dynamical downscaling to overcome the limitations of GCM-based environmental analysis.
By and large, this study adds to the growing body of evidence to support the conclusion that unabated warming of the climate system may lead to a greater frequency of HCW in the future. While not the first to use WRF at convective-permitting resolution for the purpose of investigating the HCW–climate change connection, this work is novel in that the regional climate simulations encompass the entire annual cycle and simulations have been extended beyond that of previous studies (to 30 yr). Admittedly, the trajectory of change in environments favorable to HCW varies between GCMs, especially during JJA (Diffenbaugh et al. 2013; Seeley and Romps 2015); therefore, an ensemble-based approach is desired to determine the robustness of the climate change signal. Because of the immense computational expense of such an undertaking, this approach is currently too costly. Thus, the work presented herein represents only one possible solution of the mean spatiotemporal changes of HCW. Despite this fact, some insight is gained as to how daily HCW frequency may be modified given potential changes in large-scale environmental conditions in an anthropogenically altered climate.
Not addressed within this study is the subdaily frequency and variability of simulated HCW within the regional climate model simulations, which will be the topic of forthcoming work. Additional analyses of variables that may be closely tied to individual severe hazards (i.e., tornadoes, strong winds, and large hail) will also be pursued.
Acknowledgments
The authors thank the Rosen Center for Advanced Computing (RCAC) at Purdue University for providing computational resources and technical support and also the Purdue Climate Change Research Center for providing student financial support for K.A.H. Additionally, the authors thank the three anonymous reviewers for providing valuable comments, which greatly improved the manuscript.
REFERENCES
Allen, J. T., D. J. Karoly, and K. J. Walsh, 2014a: Future Australian severe thunderstorm environments. Part I: A novel evaluation and climatology of convective parameters from two climate models for the late twentieth century. J. Climate, 27, 3827–3847, https://doi.org/10.1175/JCLI-D-13-00425.1.
Allen, J. T., D. J. Karoly, and K. J. Walsh, 2014b: Future Australian severe thunderstorm environments. Part II: The influence of a strongly warming climate on convective environments. J. Climate, 27, 3848–3868, https://doi.org/10.1175/JCLI-D-13-00426.1.
Ashley, W. S., S. Strader, T. Rosencrants, and A. J. Krmenec, 2014: Spatiotemporal changes in tornado hazard exposure: The case of the expanding bull’s-eye effect in Chicago, Illinois. Wea. Climate Soc., 6, 175–193, https://doi.org/10.1175/WCAS-D-13-00047.1.
Brooks, H. E., 2013: Severe thunderstorms and climate change. Atmos. Res., 123, 129–138, https://doi.org/10.1016/j.atmosres.2012.04.002.
Brooks, H. E., C. A. Doswell III, and M. P. Kay, 2003: Climatological estimates of local daily tornado probability for the United States. Wea. Forecasting, 18, 626–640, https://doi.org/10.1175/1520-0434(2003)018<0626:CEOLDT>2.0.CO;2.
Carley, J. R., B. R. J. Schwedler, M. E. Baldwin, R. J. Trapp, J. Kwiatkowski, J. Logsdon, and S. J. Weiss, 2011: A proposed model-based methodology for feature-specific prediction for high-impact weather. Wea. Forecasting, 26, 243–249, https://doi.org/10.1175/WAF-D-10-05008.1.
Chang, E. K. M., 2013: CMIP5 projection of significant reduction in extratropical cyclone activity over North America. J. Climate, 26, 9903–9922, https://doi.org/10.1175/JCLI-D-13-00209.1.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129, 569–585, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Craven, J. P., and H. E. Brooks, 2004: Baseline climatology of sounding derived parameters associated with deep moist convection. Natl. Wea. Dig., 28, 13–24.
Davies, H. C., and R. E. Turner, 1977: Updating prediction models by dynamical relaxation: An examination of the technique. Quart. J. Roy. Meteor. Soc., 103, 225–245, https://doi.org/10.1002/qj.49710343602.
Del Genio, A. D., M.-S. Yao, and J. Jonas, 2007: Will moist convection be stronger in a warmer climate? Geophys. Res. Lett., 34, L16703, https://doi.org/10.1029/2007GL030525.
Diffenbaugh, N. S., M. Scherer, and R. J. Trapp, 2013: Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proc. Natl. Acad. Sci. USA, 110, 16 361–16 366, https://doi.org/10.1073/pnas.1307758110.
Donner, L. J., and Coauthors, 2011: The dynamical core, physical parameterizations, and basic simulation characteristics of the atmospheric component AM3 of the GFDL global coupled model CM3. J. Climate, 24, 3484–3519, https://doi.org/10.1175/2011JCLI3955.1.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107, https://doi.org/10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Gensini, V. A., and W. S. Ashley, 2011: Climatology of potentially severe convective environments from the North American regional reanalysis. Electron. J. Severe Storms Meteor., 6 (8), http://www.ejssm.org/ojs/index.php/ejssm/article/view/85/68.
Gensini, V. A., and T. L. Mote, 2014: Estimations of hazardous convective weather in the United States using dynamical downscaling. J. Climate, 27, 6581–6589, https://doi.org/10.1175/JCLI-D-13-00777.1.
Gensini, V. A., and T. L. Mote, 2015: Downscaled estimates of late 21st century severe weather from CCSM3. Climatic Change, 129, 307–321, https://doi.org/10.1007/s10584-014-1320-z.
Gensini, V. A., C. Ramseyer, and T. L. Mote, 2014: Future convective environments using NARCCAP. Int. J. Climatol., 34, 1699–1705, https://doi.org/10.1002/joc.3769.
Giorgi, F., and L. O. Mearns, 1999: Introduction to special section: Regional climate modeling revisited. J. Geophys. Res., 104, 6335–6352, https://doi.org/10.1029/98JD02072.
Gula, J., and W. R. Peltier, 2012: Dynamical downscaling over the Great Lakes basin of North America using the WRF regional climate model: The impact of the Great Lakes system on regional greenhouse warming. J. Climate, 25, 7723–7742, https://doi.org/10.1175/JCLI-D-11-00388.1.
Hall, S. G., and W. S. Ashley, 2008: Effects of urban sprawl on the vulnerability to a significant tornado impact in northeastern Illinois. Nat. Hazards Rev., 9, 209–219, https://doi.org/10.1061/(ASCE)1527-6988(2008)9:4(209).
Harding, K. J., P. K. Snyder, and S. Liess, 2013: Use of dynamical downscaling to improve the simulation of central U.S. warm season precipitation in CMIP5 models. J. Geophys. Res. Atmos., 118, 12 522–12 536, https://doi.org/10.1002/2013JD019994.
Hong, S.-Y., and M. Kanamitsu, 2014: Dynamical downscaling: Fundamental issues from an NWP point of view and recommendations. Asia-Pac. J. Atmos. Sci., 50, 83–104, https://doi.org/10.1007/s13143-014-0029-2.
Kain, J. S., and Coauthors, 2008: Some practical considerations regarding horizontal resolution in the first generation of operational convection-allowing NWP. Wea. Forecasting, 23, 931–952, https://doi.org/10.1175/WAF2007106.1.
Kain, J. S., S. R. Dembek, S. J. Weiss, J. L. Case, J. J. Levit, and R. A. Sobash, 2010: Extracting unique information from high-resolution forecast models: Monitoring selected fields and phenomena every time step. Wea. Forecasting, 25, 1536–1542, https://doi.org/10.1175/2010WAF2222430.1.
Kendon, E. J., N. M. Roberts, C. A. Senior, and M. J. Roberts, 2012: Realism of rainfall in a very high-resolution regional climate model. J. Climate, 25, 5791–5806, https://doi.org/10.1175/JCLI-D-11-00562.1.
Kotlarski, S., S. Hagemann, P. Krahe, R. Podzun, and D. Jacob, 2012: The Elbe River flooding 2002 as seen by an extended regional climate model. J. Hydrol., 472–473, 169–183, https://doi.org/10.1016/j.jhydrol.2012.09.020.
Lauer, A., C. Zhang, O. Elison-Timm, Y. Wang, and K. Hamilton, 2013: Downscaling of climate change in the Hawaii region using CMIP5 results: On the choice of the forcing fields. J. Climate, 26, 10 006–10 030, https://doi.org/10.1175/JCLI-D-13-00126.1.
Lee, C. C., 2012: Utilizing synoptic climatological methods to assess the impacts of climate change on future tornado-favorable environments. Nat. Hazards, 62, 325–343, https://doi.org/10.1007/s11069-011-9998-y.
Lo, J. C.-F., Z.-L. Yang, and R. A. Pielke Sr., 2008: Assessment of three dynamical climate downscaling methods using the Weather Research and Forecasting (WRF) model. J. Geophys. Res., 113, D09112, https://doi.org/10.1029/2007JD009216.
Lucas-Picher, P., F. Boberg, J. H. Christensen, and P. Berg, 2013: Dynamical downscaling with reinitializations: A method to generate finescale climate datasets suitable for impact studies. J. Hydrometeor., 14, 1159–1174, https://doi.org/10.1175/JHM-D-12-063.1.
Mahoney, K., M. A. Alexander, G. Thompson, J. J. Barsugli, and J. D. Scott, 2012: Changes in hail and flood risk in high-resolution simulations over Colorado’s mountains. Nat. Climate Change, 2, 125–131, https://doi.org/10.1038/nclimate1344.
Mahoney, K., M. A. Alexander, J. D. Scott, and J. J. Barsugli, 2013: High-resolution downscaled simulations of warm-season extreme precipitation events in the Colorado Front Range under past and future climates. J. Climate, 26, 8671–8689, https://doi.org/10.1175/JCLI-D-12-00744.1.
Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2007: Assessment of the severe weather environment in North America simulated by a global climate model. Atmos. Sci. Lett., 8, 100–106, https://doi.org/10.1002/asl.159.
Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2009: Preliminary investigation into the severe thunderstorm environment of Europe simulated by the Community Climate System Model 3. Atmos. Res., 93, 607–618, https://doi.org/10.1016/j.atmosres.2008.09.014.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, https://doi.org/10.1029/97JD00237.
Moss, R. H., and Coauthors, 2010: The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756, https://doi.org/10.1038/nature08823.
Nakanishi, M., and H. Niino, 2004: An improved Mellor–Yamada level-3 model with condensation physics: Its design and verification. Bound.-Layer Meteor., 112, 1–31, https://doi.org/10.1023/B:BOUN.0000020164.04146.98.
Nakanishi, M., and H. Niino, 2006: An improved Mellor–Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397–407, https://doi.org/10.1007/s10546-005-9030-8.
NOAA, 2014: Billion-dollar weather and climate disasters: Overview. NCEI, https://www.ncdc.noaa.gov/billions.
Pan, Z., E. Takle, W. Gutowski, and R. Turner, 1999: Long simulation of regional climate as a sequence of short segments. Mon. Wea. Rev., 127, 308–321, https://doi.org/10.1175/1520-0493(1999)127<0308:LSORCA>2.0.CO;2.
Paquin, D., R. de Elía, and A. Frigon, 2014: Change in North American atmospheric conditions associated with deep convection and severe weather using CRCM4 climate projections. Atmos.–Ocean, 52, 175–190, https://doi.org/10.1080/07055900.2013.877868.
Prein, A. F., G. J. Holland, R. M. Rasmussen, J. Done, K. Ikeda, M. P. Clark, and C. H. Liu, 2013: Importance of regional climate model grid spacing for the simulation of heavy precipitation in the Colorado headwaters. J. Climate, 26, 4848–4857, https://doi.org/10.1175/JCLI-D-12-00727.1.
Prein, A. F., and Coauthors, 2015: A review on regional convection-permitting climate modeling: Demonstrations, prospects, and challenges. Rev. Geophys., 53, 323–361, https://doi.org/10.1002/2014RG000475.
Qian, J.-H., A. Seth, and S. Zebiak, 2003: Reinitialized versus continuous simulations for regional climate downscaling. Mon. Wea. Rev., 131, 2857–2874, https://doi.org/10.1175/1520-0493(2003)131<2857:RVCSFR>2.0.CO;2.
Rasmussen, R., and Coauthors, 2011: High-resolution coupled climate runoff simulations of seasonal snowfall over Colorado: A process study of current and warmer climate. J. Climate, 24, 3015–3048, https://doi.org/10.1175/2010JCLI3985.1.
Rasmussen, R., and Coauthors, 2014: Climate change impacts on the water balance of the Colorado headwaters: High-resolution regional climate model simulations. J. Hydrometeor., 15, 1091–1116, https://doi.org/10.1175/JHM-D-13-0118.1.
Robinson, E. D., R. J. Trapp, and M. E. Baldwin, 2013: The geospatial and temporal distributions of severe thunderstorms from high-resolution dynamical downscaling. J. Appl. Meteor. Climatol., 52, 2147–2161, https://doi.org/10.1175/JAMC-D-12-0131.1.
Seeley, J. T., and D. M. Romps, 2015: The effect of global warming on severe thunderstorms in the United States. J. Climate, 28, 2443–2458, https://doi.org/10.1175/JCLI-D-14-00382.1.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132, 3019–3032, https://doi.org/10.1175/MWR2830.1.
Sobash, R., J. S. Kain, D. R. Bright, A. R. Dean, M. C. Coniglio, S. J. Weiss, and J. J. Levit, 2009: Forecast guidance for severe thunderstorms based on identification of extreme phenomena in convection-allowing model forecasts. 23rd Conf. on Weather Analysis and Forecasting, Omaha, NE, Amer. Meteor. Soc., 4B.6, https://ams.confex.com/ams/23WAF19NWP/techprogram/paper_154328.htm.
Sobash, R., J. S. Kain, D. R. Bright, A. R. Dean, M. C. Coniglio, and S. J. Weiss, 2011: Probabilistic forecast guidance for severe thunderstorms based on the identification of extreme phenomena in convection-allowing model forecasts. Wea. Forecasting, 26, 714–728, https://doi.org/10.1175/WAF-D-10-05046.1.
Strader, S., and W. Ashley, 2015: The expanding bull’s-eye effect. Weatherwise, 68, 23–29, https://doi.org/10.1080/00431672.2015.1067108.
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, https://doi.org/10.1175/BAMS-D-11-00094.1.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 5095–5115, https://doi.org/10.1175/2008MWR2387.1.
Trapp, R. J., and K. A. Hoogewind, 2016: The realization of extreme tornadic storm events under future anthropogenic climate change. J. Climate, 29, 5251–5265, https://doi.org/10.1175/JCLI-D-15-0623.1.
Trapp, R. J., N. S. Diffenbaugh, H. E. Brooks, M. E. Baldwin, E. D. Robinson, and J. S. Pal, 2007a: Changes in severe thunderstorm environment frequency during the 21st century caused by anthropogenically enhanced global radiative forcing. Proc. Natl. Acad. Sci. USA, 104, 19 719–19 723, https://doi.org/10.1073/pnas.0705494104.
Trapp, R. J., B. A. Halvorson, and N. S. Diffenbaugh, 2007b: Telescoping, multimodel approaches to evaluate extreme convective weather under future climates. J. Geophys. Res., 112, D20109, https://doi.org/10.1029/2006JD008345.
Trapp, R. J., N. S. Diffenbaugh, and A. Gluhovsky, 2009: Transient response of severe thunderstorm forcing to elevated greenhouse gas concentrations. Geophys. Res. Lett., 36, L01703, https://doi.org/10.1029/2008GL036203.
Trapp, R. J., E. D. Robinson, M. E. Baldwin, N. S. Diffenbaugh, and B. R. J. Schwedler, 2011: Regional climate of hazardous convective weather through high-resolution dynamical downscaling. Climate Dyn., 37, 677–688, https://doi.org/10.1007/s00382-010-0826-y.
Van Klooster, S. L., and P. J. Roebber, 2009: Surface-based convective potential in the contiguous United States in a business-as-usual future climate. J. Climate, 22, 3317–3330, https://doi.org/10.1175/2009JCLI2697.1.
Van Vuuren, D. P., and Coauthors, 2011: The representative concentration pathways: An overview. Climatic Change, 109, 5–31, https://doi.org/10.1007/s10584-011-0148-z.
Weisman, M. L., W. C. Skamarock, and J. B. Klemp, 1997: The resolution dependence of explicitly modeled convective systems. Mon. Wea. Rev., 125, 527–548, https://doi.org/10.1175/1520-0493(1997)125<0527:TRDOEM>2.0.CO;2.