SEARCH

SEARCH BY CITATION

Keywords:

  • tornado;
  • severe thunderstorm;
  • climate change

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[1] Understanding of the possible response of severe convective precipitating storms to elevated greenhouse gas concentrations remains elusive. To address this problem, telescoping, multimodel approaches are proposed, which allow representation of a broad range of processes that could regulate convective storm behavior. In the global-cloud approach (G-C), the NCEP-NCAR Reanalysis Project (NNRP) global data set provides initial and boundary conditions for short-term integrations of a mesoscale model and nested convective-cloud-permitting domain. In the global-regional-cloud approach (G-R-C), the NNRP data set provides initial and boundary conditions for long-term integrations of a regional climate model, which in turn forces short-term integrations of a mesoscale model and nested convective-cloud-permitting domain. Upon applying these approaches to historical extreme convective storm events, it was found that the global-scale data could be dynamically downscaled to produce realistic convective-scale solutions. In particular, tornado proxies computed from the model-simulated winds were shown to compare well in relative numbers to those of tornado observations on many of the days considered. This supports the telescoping modeling concept as a viable means to address effects of elevated greenhouse gas concentrations on convective-scale phenomena. In an evaluation of the two approaches, it was also found that simulations of the historical events by the G-C were superior to those by the G-R-C. Sensitivity of the convective-scale processes to details in the downscaled synoptic-scale flow, and to the placement of the mesoscale model domain within the regional climate model, reduced the effectiveness of the G-R-C.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[2] Climate models consistently project increases in the frequency of annual, seasonal, and daily precipitation extremes, particularly over North America, in response to anthropogenic increases in greenhouse gas concentrations [e.g., Diffenbaugh et al., 2005; Meehl et al., 2005; Tebaldi et al., 2006; see also Intergovernmental Panel on Climate Change, 2001]. However, conclusions on whether, and/or how, the projected precipitation will manifest as changes in locally intense, convective precipitating storms (CPSs) with several-hour timescales remain elusive. In addition to torrential rainfall, CPSs are capable of producing destructive surface winds, hail, lightning, and tornadoes. The societal and economic impacts of these local-scale hazardous weather phenomena can be significant [e.g., NOAA, 2005], and hence increases in their occurrence could have substantial implications.

[3] The potential response of convective storms and associated phenomena to increases in greenhouse gas concentrations remains unconstrained, largely because of an incomplete understanding of how global-scale changes in radiative forcing are realized at local scales (and in turn how the local-scale changes feed back to the global scale). Any viable means to address this problem must account for the fact that the organization of cumulus clouds into intense storms is governed in part by larger- (synoptic) scale or ambient distributions of temperature, moisture, and winds.

[4] Consider an adaptation of the approach taken by weather forecasters, who assess the likelihood of intense convective storms on a given day by examining parameters such as the convective available potential energy (CAPE), and the vertical change or shear of the ambient horizontal wind vector (equation image / ∂z). For example, it is generally accepted that when the ambient CAPE and the ambient vertical wind shear over a surface to ∼4–7 km layer jointly exceed approximate threshold values, severe, rotating cumulonimbi can develop [Weisman and Klemp, 1982; Johns et al., 1993]. These so-called supercell thunderstorms [Browning, 1964] can have durations of several hours, and are almost always associated with some form of hazardous weather, including tornadoes and large hail. When ambient vertical wind shear is large over a shallower layer (surface to ∼2–3 km), the predominant convective organization is often linear-shaped thunderstorm systems, or squall lines, whose main threat is damaging winds and heavy rainfall, although hail and tornadoes are possible [e.g., Trapp et al., 2005a].

[5] The forecasting procedure can be extended to climate applications by similarly examining these parameters of CAPE, shear, etc., as derived from (global and regional) climate model output. Brooks [2006] has begun such an effort for the modern climate, using the National Center for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) Reanalysis Project (NNRP) global data set [Kalnay et al., 1996]. He has found, for example, that the NNRP data reveal a global maximum in tornadic storm environment conditions over the central United States that corresponds very well with the observed maximum in tornado reports. We are, in a separate study, also applying this procedure, using the Diffenbaugh et al. [2005] simulations of current and future United States regional climate (R. Trapp, et al., Changes in severe thunderstorm frequency during the 21st century due to anthropogenically enhanced global radiative forcing, submitted to Proceedings of the National Academy of Sciences of the United States of America, 2007).

[6] This approach of using environmental parameters as proxies for convective storms and associated phenomena is not without its limitations, however. Indeed, one must make the tenuous assumption that storms have or will initiate in these environments. In an attempt to avoid this limitation, we pursue herein an alternative yet complementary approach in which convective storms are treated explicitly. By design, we require that the processes that could regulate convective storm behavior in response to changes in global radiative forcing be captured over a broad range of spatial and temporal scales. Accordingly, we propose a telescoping approach that culminates in 3D, convection-permitting model simulations (Figure 1).

image

Figure 1. Schematic of a telescoping, multimodel approach using a global climate model, a regional climate model, and a cloud-resolving or convection-permitting model.

Download figure to PowerPoint

[7] Given a number of possible ways to implement this approach, the objective of the current study is to develop a telescoping model strategy that is accurate in terms of the antecedent conditions on the synoptic and mesoscale, the organizational mode of the CPSs, and the representation of severe weather associated with these storms. Two specific strategies, the global-cloud (G-C) and the global-regional-cloud (G-R-C), are described and evaluated on the basis of simulations of historical convective storm events. The latter of the two is motivated by the fact that regional climate model applications have been shown to improve the simulation of regional-scale extreme event frequencies over that of the driving GCM [Bell et al., 2004].

2. Telescoping Modeling

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

2.1. Components

[8] The basic components of the telescoping modeling technique are a global climate model or global data set, a regional climate model, and a convection-permitting model.

2.1.1. Global Data

[9] Global data from the NNRP currently serve as the primary source of initialization and boundary information for the limited area models in both modeling approaches. NNRP data are available from 1948 to present on a 2.5 × 2.5 degree latitude/longitude grid at 6-hour intervals.

2.1.2. Regional Climate Model

[10] For the G-R-C we utilize the Abdus Salam Institute for Theoretical Physics Regional Climate Model, version 3 (RegCM3) [Pal et al., 2007]. RegCM3 is a hydrostatic, compressible, finite difference model that uses vertical sigma coordinates [e.g., Giorgi et al., 1993; Giorgi and Mearns, 1999; Pal et al., 2000]. RegCM3 is based on earlier versions of RegCM developed at NCAR from the NCAR-Pennsylvania State University mesoscale model, version 4.

[11] For the current study, RegCM3 is applied on a domain centered at 37.581°N, 95°W (Figure 2). The RegCM3 grid and parameterizations follow Pal et al. [2000]. The horizontal grid consists of 128 (80) points in longitude (latitude), with a 55.6-km grid point spacing; the vertical grid has 18 sigma levels. In this configuration, RegCM3 is able to capture the seasonal and spatial structure of temperature and precipitation in the continental United States [Diffenbaugh et al., 2006]. Given the NNRP data as initial and boundary conditions, the model is continuously integrated in time for the full model year prior to the year of each event, specifically, 1 January 1973 to 5 April 1974, and 1 January 2000 to 10 May 2001. The longer-term integrations of the regional climate model allow the modeled land surface processes to be well equilibrated with the overlying atmosphere, during the time periods of interest [e.g., Giorgi and Mearns, 1999]. Greenhouse gas concentrations are taken from Schlesinger and Malyshev [2001].

image

Figure 2. Computational domain for RegCM3.

Download figure to PowerPoint

2.1.3. Convection-Permitting Model

[12] The advanced research version of the Weather Research and Forecasting (WRF) model Version 2.1.2, a fully compressible, nonhydrostatic model [Michalakes et al., 2001; Skamarock et al., 2005], is used here as a mesoscale model and (two-way nested) convection-permitting model. Table 1 provides specific details of our application of the WRF model in the G-C and G-R-C configurations. Parameterizations of physical processes follow those used in experimental, high-resolution, WRF-model predictions of convective storms in the United States [e.g., Weisman et al., 2004; Kain et al., 2006]. With “cold-start” initial conditions as well as boundary conditions supplied by a regional forecast model (the operational Eta model), these 24 to 36 hour time integrations were performed over computational domains that spanned more than two thirds of the continental United States. These initial and boundary conditions are presumed to sufficiently resolve the mechanisms, fronts and other heterogeneities in the synoptic-scale and mesoscale fields, that help initiate and force CPS events. In turn, the WRF grids with horizontal grid points spaced ≤ 4 km are presumed to sufficiently resolve the evolution and basic structure of CPSs, and hence obviate the need for convective parameterization [Weisman et al., 1997]. Evaluated subjectively in terms of CPS initiation, evolution, and morphology or organizational mode, the predictions have compared well to observations on many (though certainly not all) days [Kain et al., 2006], and are considered to have some skill.

Table 1. WRF Model Physics Parameterization Schemesa
ParameterizationScheme
MicrophysicsLFO
Radiation (SW/LW)Dudhia/RRTM
Land surface modelNoah
Planetary boundary layerYSU
Convective clouds and precipitationKain-Fritsch

2.2. Two Approaches

[13] In all experiments, the G-C utilizes the NNRP global data set as initial and boundary conditions for the WRF model; the NNRP soil moisture and soil temperature are also used to initialize the WRF land surface model (LSM). The WRF model is then time-integrated over 30 hours; for the multiday event considered, the WRF is reinitialized each day and then time-integrated over 30 hours. The G-C employs three nested domains (hereinafter, d01, d02, d03) with 27 km, 9 km, and 3 km horizontal grid point spacing, respectively (Figure 3); the stretched vertical grid has 31 levels. Hence the WRF is used as a traditional mesoscale model that is then nested down to a convection-permitting domain (d03). Neither this domain nor d02 make use of a cumulus parameterization scheme; on the d01 domain, however, the Kain-Fritsch cumulus parameterization scheme [Kain, 2004] is implemented. Two-way interaction between d03 and d02 is afforded, as it is between d02 and d01. The domains are judiciously placed according to the location of each event (Figure 3).

image

Figure 3. Computational domains for the WRF model used in the (a) 3–4 April 1974 simulations and (b) 2–8 May 2001 simulations. Domain 1 (d01) has a horizontal grid point spacing of 27 km, domain 2 (d02) has a spacing of 9 km, and domain 3 (d03) has a spacing of 3 km.

Download figure to PowerPoint

[14] The G-R-C uses the NNRP global data set to drive the RegCM3 model. The RegCM3 is run continuously over the domain shown in Figure 2, for the time periods described above. The RegCM3 output is then used as initial and boundary conditions for the WRF model, which is again time-integrated for 30 hours. (For consistency with the G-C, the NNRP soil moisture and soil temperature are used to initialize the WRF LSM. Initial experiments show little sensitivity of the G-R-C solutions presented herein to the LSM initialization data, however.) The three nested WRF domains are the same in the G-R-C as in the G-C, and two-way interaction between d03 (d02) and d02 (d01) is again afforded. So, in the G-R-C, the regional climate model drives the mesoscale model (d01), which is then nested down to a convection-permitting domain (d03).

[15] Although investigations of the response to 21st century climate forcing will require atmosphere-ocean general circulation model-simulated fields as part of both configurations, here we investigate the modeling approach performance given “perfect boundary conditions.” Use of the global reanalysis data set in the both configurations allows us to provide a most favorable constraint on their performance.

[16] It is important to mention here that the interaction between the RegCM3 and WRF is one-way, as obviously are the interactions between the NNRP and RegCM3, and the NNRP and WRF. This lack of two-way coupling is acknowledged as a limitation. Efforts such as those at NCAR to develop a two-way nested regional climate model using the WRF and the Community Climate System Model (CCSM) will eventually remove this limitation; see: http://www.mmm.ucar.edu/modeling/nrcm/index.php.

3. G-C and G-R-C Testing

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

3.1. Case Studies

[17] Two case studies of extreme CPS events are used to test the telescoping modeling approaches. The first is the Tornado Super Outbreak of 3–4 April 1974, considered to be one of the most devastating tornado outbreaks of the 20th century [Hoxit and Chappell, 1975; Brooks and Doswell, 2001]. The second case consists of the sequence of severe convective storms observed throughout the southern Great Plains during the period 2–8 May 2001. Well within the climatologically favored time for tornadoes and severe storms in Oklahoma and Texas [e.g., Brooks et al., 2003], this case is chosen to represent a “typical” occurrence of extreme CPSs.

3.2. Evaluation

[18] In addition to a qualitative assessment of the model solutions, we seek an objective or quantitative means to compare model output to observations, and hence to evaluate the modeling approaches. Besides rain gauge and weather radar data, the only relevant severe convective storm observations are eyewitness reports of tornadoes, severe winds, and hail, which incidentally form the current basis for the climate statistics of these phenomena. Here, tornado reports are chosen, since these reports are, in many ways, more reliable than those of hail and wind [e.g., Trapp et al., 2006]. Tornado-scale motions are not explicitly simulated by our modeling methodology, however, and so a model-based proxy of a tornado is required.

[19] Consider the use of a horizontal spatial correlation between the vertical components of velocity and vorticity. This quantifies the propensity of a convective storm to be tornadic when it possesses a deep, rotating updraft or mesocyclone [e.g., Weisman and Klemp, 1984; Davies-Jones, 1984; Lilly, 1986; Droegemeier et al., 1993]. A form of this correlation (see Attachment H, written by L. Wicker, J. Kain, S. Weiss, and D. Bright, in the 2005 SPC/NSSL Spring Program Overview and Operations Plan document by Weiss et al. [2005]) was used in 2005 during the NOAA Storm Prediction Center (SPC)/National Severe Storms Laboratory (NSSL) Spring Program [e.g., Kain et al., 2006] to locate possible supercells in high-resolution experimental forecasts. We have adapted it for our study, defining a parameter S at each grid point within domain d03 as

  • equation image

where w is vertical velocity, ζ is vertical vorticity, and the overbar indicates a vertical average, over each grid column, from the lowest model level to approximately 5 km. To eliminate insignificant and/or spurious values, the w and ζ are required to exceed 5 m s−1 and 0.001 s−1 thresholds, respectively, at each grid level involved in the calculation.

[20] On the basis of previous research and recent applications, we assume here that S > 0.6 indicates a rotating convective storm and potential tornado. Specifically, each maxima (with S > 0.6) in the 2D field of S is used as a proxy for a tornado-producing CPS (e.g., see Figure 4), with the understanding that not all supercells or mesocyclonic storms spawn tornadoes [e.g., Trapp et al., 2005b]. (Trapp et al. [2005b] considered the percentage of tornadic mesocyclones as a function of mesocyclone “base.” When the height of this base was at or below 1000 m, ∼40% of the mesocyclones considered were associated with a tornado.) Counts of maxima are determined over 1.5-hour intervals, and then summed to arrive at a total number of assumed independent “tornado” detections over a 12-hour period, within domain d03. (Our experimentation shows that evaluation of S maxima over 1.5-hour intervals should reveal independent supercells and thus tornado detections. This is consistent with the study by Burgess et al. [1982] who determined the core evolution of a mesocyclone to be 60 to 90 min.) The tornado detections are then compared to the number of tornado reports over the same period and geographical domain.

image

Figure 4. Telescoping model simulation of the 3–4 April 1974 event. (a–d) Radar reflectivity (at 0.5-km altitude; dBZ) and S from the G-C at 2100 UTC, 3 April 1974, and 0000 UTC, 4 April 1974; (e–g) radar reflectivity and S from the G-R-C at 2100 UTC, 3 April 1974, and 0000 UTC, 4 April 1974. All fields are shown on domain d03.

Download figure to PowerPoint

4. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[21] The 30-hour WRF simulations in both approaches and for both cases commence at 0000 UTC, which is roughly 12–15 hours prior to convection initiation within d03. This initial ∼12 hours of simulation time is considered primarily to be the interval over which the model generates or “spins-up” the mesoscale portion of the atmospheric kinetic energy spectrum. Our discussion therefore focuses on the remaining time in the integration, specifically on the evolution of the salient convective storm-scale features within d03. To be clear, we do not expect that specific storms of a given case will be simulated. Hence, in the assessments, errors in time and (especially) space are acceptable as long as the simulations generate storms that are comparable in number, intensity, and morphology.

4.1. G-C: 3–4 April 1974

[22] Weakly precipitating storms are found in domain d03 of the G-C solution by 1500 UTC on 3 April 1974 (hereinafter, the day and time are expressed following the convention 3/1500). As suggested in the radar summary charts presented by Hoxit and Chappell [1975] (not shown), this early evolution in the G-C solution lags that of the observations by several hours. However, intense CPSs with deep, rotating updrafts (and S > 0.6) are located in Indiana and Ohio by 3/2100 (Figures 4a and 4b), consistent with the tornado observations on this day (Figure 5). Such CPSs are most numerous at ∼4/0000 (Figures 4c and 4d), but continue to be identified in eastern Illinois, Indiana, Ohio, and Kentucky through 4/0600.

image

Figure 5. Tracks of all tornadoes that occurred during the period 1200 UTC, 3 April 1974, to 1200 UTC, 4 April 1974.

Download figure to PowerPoint

[23] As summarized in Table 2, 64 tornado proxies or detections are determined from the G-C simulation over the period 3/1800 to 4/0600, and within domain d03. This number of detections compares favorably with the actual number of tornado reports (76), over the same period and geographical domain. Additional G-C simulations demonstrate the robustness of this solution.

Table 2. Quantitative Summary for G-C Simulations of the 3–4 April 1974 Casea
Simulation NameTornado Detections
  • a

    The actual number of tornado reports over the period 3/1800 to 4/0600 and within the geographical domain described by d03 is 76.

G-C64
Sensitivity: spin up test84
Sensitivity: resolution ratio test66
Sensitivity: larger domain test72
Sensitivity: no nested domains64

[24] In the spirit of the regional climate modeling “Big-Brother” experiments of Denis et al. [2002], we examined solution sensitivity to (1) model spin-up time, by initializing the WRF model at 2/1200 instead of at 3/0000; (2) domain size and placement, by extending domain d01 to include the entire contiguous United States; and (3) resolution “jump,” by including an additional WRF domain with 81-km grid spacing, so that the ratio of the NNRP grid spacing to that of the outermost WRF domain is ∼2.5 instead of ∼7. Qualitatively, these model solutions have differences in the details of individual storm characteristics and evolution (although not in morphology). Quantitatively, however, they still result in comparatively high numbers of tornado detections (Table 2), which is most relevant to an eventual goal of using the modeling methodology to generate climate statistics of tornadoes.

[25] One final sensitivity to be considered regards the use of nested computational domains. The basic G-C configuration allows convective precipitation that is generated through a parameterization scheme on the coarse domain (d01) to move into the fine domain (d03), wherein the convection is explicitly represented. Although such domain nesting is still widely practiced (owing primarily to computational resource limitations), questions remain about the impacts of resolution-dependent model adjustments across nested domain boundaries, especially those involving convective precipitation [see, e.g., Warner and Hsu, 2000]. To address this, we reconfigured the G-C such that the WRF model has only a single domain (Figure 6a). This domain has 4-km horizontal grid point spacing, and is considered convection permitting.

image

Figure 6. As in Figure 2 except for the G-C sensitivity experiment with the single WRF domain and 4-km grid point spacing. Radar reflectivity at 0000 UTC, 4 April 1974, over the (a) complete and (b) magnified section of domain and (c) S over the magnified section.

Download figure to PowerPoint

[26] The high-resolution G-C configuration produces storms that are comparable in number, intensity, and morphology to the G-C control run (e.g., Figures 6b, 6c, 4c, and 4d). In fact, an identical number of tornado detections (64) is determined within the same geographical domain as d03, over the period 3/1800 to 4/0600 (Table 2). We do find that the temporal evolution of the CPSs was delayed by a few hours, compared to that in the G-C control run. The CPSs also tended to be geographically located farther the north. These sensitivities are considered to be acceptable, recalling the objective of our study. We can conclude that although it would be preferable to utilize a modeling approach that culminates in a single, large, convection-permitting domain as in Figure 6a, the use of nested domains in this study appears to be satisfactory. A challenge of such a nested approach in investigating the CPS response to elevated greenhouse forcing will be identifying the locations and time periods for the convection-permitting model simulations, as these will obviously not be identifiable from the observational record.

[27] The results so far demonstrate that the relatively coarse global data (2.5 × 2.5 degree latitude/longitude grid) of the NNRP contain sufficient information such that when dynamically downscaled, the synoptic and mesoscale processes that help initiate and force extreme CPSs are represented. We now examine whether inclusion of processes attendant with a well-equilibrated land surface–atmosphere, a potential benefit offered by long-term integrations of the regional climate model, can improve upon this convective-scale solution.

4.2. G-R-C: 3–4 April 1974

[28] In stark contrast to the G-C results, only 9 tornado detections are determined from the G-R-C simulation (Table 3), owing simply to the relative lack of simulated CPSs (Figure 4). Indeed, rotating convective storms were never as numerous and widespread in the G-R-C as in the G-C (and in the observations).

Table 3. As in Table 2 Except for the G-R-C
Simulation NameTornado Detections
G-R-C9
Sensitivity: 55 km RCM without d0130
Sensitivity: larger domain d0139

[29] Inspection of the domain d01 fields suggests that the prominent synoptic-scale features (500 hPa trough, surface cyclone, etc.) of this event appear to be comparably simulated by the G-R-C and G-C. On the other hand, the structure of the low-level thermodynamic fields is not. In particular, the low-level air advected into Kentucky-Indiana-Ohio (d03) during the hours centered about 3/1800 is relatively cool and dry in the G-R-C solution. As a consequence, the CAPE in this region is quite low (less than ∼100 J kg−1 when computed using the “most unstable” parcel) and unsupportive of widespread CPSs. An explanation for this reduced CAPE can be gained from a diagnosis of the RegCM3 output used as initial and boundary conditions, which reveals that the convective parameterization scheme in RegCM3 had activated in the vicinity of the southern boundary of the WRF domain (Figure 7). In other words, air that had been cooled and stabilized was supplied to the WRF through its southern boundary condition. Eventually, the parameterized convection dissipated, but the heat and moisture fluxes over the time remaining in the event were sufficient only to support modest numbers of storms in d03.

image

Figure 7. Convective precipitation (mm day−1) and horizontal wind vectors at the lowest model (sigma) level, from the RegCM3 simulation, at 1800 UTC, 3 April 1974. Approximate location of the WRF domain d01 is outlined in gray.

Download figure to PowerPoint

[30] The implication of this G-R-C case study is that simulation of convective storm events using the G-R-C is sensitive to the chance placement of the WRF d01 within the RegCM3 domain. To test this, experiments were conducted (1) with a much larger d01 that extended well into the Gulf of Mexico and (2) with the exclusion of the original d01, and hence, with only two nested domains (d02 and d03). Table 3 confirms the considerable sensitivity of CPS simulation to the WRF domain placement, with both experiments resulting in an increase in the number of tornado detections by a factor of 3 to 4.

[31] Interestingly, the numbers of detections (and hence CPSs) from these experiments are still well below those from the G-C. One suggestion is that, for this event, the RegCM3 downscaled the NNRP data in some way that was relatively less conducive to a storm outbreak. For more insight, consider both the RegCM3 and NNRP data interpolated to a grid equivalent to d01. At 3/1800, the NNRP data resolve a midlatitude cyclone that is centered in northeast Missouri/southwest Iowa at 850 hPa (Figure 8a). The cyclone center is displaced slightly northward in the RegCM3 data. More significantly, the associated south-southwesterly winds are stronger in the NNRP data throughout the Gulf Coast states northward to Indiana (Figure 8b). These differences in the wind and height fields help explain the comparatively drier boundary layer and lower CAPE in the RegCM3 data (Figures 8c and 8d) over this broad region, and therefore suggest a dependence of the details of the convective scale solution on the details of the synoptic-scale forcing imposed as boundary conditions.

image

Figure 8. Meteorological conditions at 1800 UTC, 3 April 1974, based on the NNRP and RegCM3 data interpolated to the WRF domain d01. (a and b) Geopotential heights (interval of 30 m) and wind vectors at 850 hPa, from the NNRP and RegCM3, respectively. For reference, the magnitude of a vector that exactly reaches the tail of the next adjacent vector is 10 m s−1. The shaded contours in Figure 8b are negative values of the RegCM3–NNRP wind magnitude difference (interval of 2.5 m s−1) at 850 hPa. (c and d) RegCM3 – NNRP water vapor mixing ratio difference (interval of 1 g kg−1) at 850 hPa and RegCM3–NNRP CAPE difference (interval of 250 J kg−1). The CAPE is calculated using the “most unstable” parcel. Shaded contours indicate negative differences.

Download figure to PowerPoint

[32] Additionally, we note that the differences shown in Figure 8 developed while the RegCM3 is integrated a full model year prior to the year of the event: A short-term simulation (not shown) of this event by RegCM3 without such prior integration results in a solution that is negligibly different than that of the NNRP. This suggests that the longer-term integration of RegCM3 degraded the simulation of atmospheric features critical for the WRF simulation of CPSs. This long-term degradation has implications about the value of a regional climate model in this modeling application, as discussed further in section 5.

4.3. G-C: 2–8 May 2001

[33] During this convectively active week, tornadoes were reported within the geographical area defined by d03 on 4, 5, and 6 May 2001. Each of these three days is now considered. Recall that daily 30-hour simulations are generated using the WRF model in both approaches. The daily initialization time is 0000 UTC.

[34] On the afternoon and evening of 4 May, an extensive squall line was observed in Oklahoma through Texas (Figure 9), with numerous incidents of severe weather and tornadoes reported in association with this line. The G-C solution 5/0000 consists of a line of cells extending from southwest Oklahoma to the southern Texas Panhandle (Figure 9). Although this is not exactly the same organization displayed in the radar composite, 6 tornado detections were found in d03 over the 12-hour period 4/1800 to 5/0600, as compared to 11 tornado reports (Table 4).

image

Figure 9. (a) G-C simulation at 0000 UTC, 5 May 2001, depicted through radar reflectivity on domain d02. (b) Corresponding observed radar reflectivity, at 0001 UTC.

Download figure to PowerPoint

Table 4. Quantitative Summary for Daily G-C Simulations Over the Period 2–8 May 2001a
Simulation NameTornado DetectionsActual Tornado Reports
  • a

    Note from the text that the G-R-C failed generate any tornado detections during this period.

2 May 200100
3 May 200150
4 May 2001611
5 May 2001443
6 May 20013311
7 May 2001460
8 May 200100

[35] On 5 May, isolated supercells as well as lines of convective storms were observed during the afternoon and evening (Figure 10). The G-C was successful in simulating intense storms in Texas and Oklahoma on this day (Figure 10), but produced too many with rotating updrafts in d03, as evidenced by the 44 tornado detections and the 3 tornado reports in d03 (Table 4). The G-C similarly overproduced tornado detections on 6 May and 7 May, even though the simulated storm evolution was generally consistent with that observed. The overdetections reveal a shortcoming in our quantitative evaluation procedure on these days: We found many instances of single convective storms possessing multiple mesocyclones, albeit weak and relatively small ones. Since these can be physically realistic, and are represented in the model output as maxima in S, we included them in our counts. The minor rotating cores are much less apt to spawn tornadoes however, and at best are nominally resolved. Hence, in future applications of our evaluation procedure, the calculation of S should include spatial filtering and otherwise be improved to isolate the most physically significant rotating updrafts.

image

Figure 10. As in Figure 9 except at 0400 UTC, 6 May 2001 (simulated), and 0402 UTC (observed).

Download figure to PowerPoint

4.4. G-R-C: 2–8 May 2001

[36] As with the 3–4 April 1974 case, the G-R-C solutions for this case are dramatically different than the G-C solutions. Notably, the G-R-C failed on all days during the period to generate intense, rotating storms, and therefore tornado proxies. We again turn to the RegCM3 and NNRP data interpolated to a common grid for an explanation.

[37] A significant synoptic-scale feature present in the observations (and in the NNRP data) near the beginning of the period was a midtropospheric “cutoff” low, centered in northeastern Arizona and southeastern Utah on 4/0000 (Figure 11a). Downscaling of NNRP data by the RegCM3 resulted in a progressive, small-amplitude trough rather than a cutoff low (Figure 11b). The simple consequence of this difference (and those reflected in the lower-tropospheric fields) was a large-scale environment that lacked sufficient CAPE and shear, and hence could not support intense CPSs throughout the southern Great Plains. As with the April 1974 event, a short-term simulation (not shown) of this event by RegCM3 without the 1-year prior integration results in a synoptic-scale flow that is negligibly different from that present the NNRP. This again suggests that the errors in the RegCM3 synoptic circulation result not from the ability of RegCM3 to downscale the NNRP data, but rather from intrinsic features that develop over the longer-term RegCM3 integration.

image

Figure 11. The 500 hPa geopotential heights and wind vectors, at 0000 UTC, 4 May 2001, from (a) NNRP and (b) RegCM3. Contour interval is 60 m. The data are interpolated to the WRF domain d01.

Download figure to PowerPoint

5. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[38] This study is motivated by the need to investigate possible changes in the frequency, intensity, and geographical distribution of severe convective precipitating storms and associated hazards under enhanced greenhouse forcing. Our objective here was to test methods for such investigations. An experimental design requirement was that the processes that could regulate CPS behavior in response to changes in global radiative forcing be captured over a broad range of spatial and temporal scales. Critically, the method must ultimately account for the fact that investigations of future climate changes will lack observational and reanalysis data for initial and boundary conditions. Accordingly, telescoping modeling approaches were proposed and evaluated.

[39] In global-cloud modeling approach (G-C), the NNRP global data set provided initial and boundary conditions for short-term integrations of the WRF model and nested convection-permitting domain. In the global-regional-cloud modeling approach (G-R-C), the NNRP global data set provided initial and boundary conditions for long-term integrations of a regional climate model (RegCM3), which in turn forced short-term integrations of the WRF model and nested convection-permitting domain. The G-C and G-R-C were quantitatively and qualitatively evaluated on the basis of their respective simulations of historical extreme CPS events.

[40] The first conclusion reached from the current study is that the relatively coarse global data for the historical events could be dynamically downscaled to produce realistic convective-scale solutions. In particular, tornado proxies computed from the model-simulated winds were shown to quantitatively compare well with tornado observations on many of the days considered. Most importantly, many more proxies were found in the simulation of a significant tornado outbreak than in a more typical tornado event. This motivates us to continue to pursue the telescoping modeling concept as a viable means to address effects of elevated greenhouse gas concentrations on extreme convective storms.

[41] The second conclusion regards the model solution accuracy as a function of the downscaling approach. Regional climate model applications have been shown to improve the simulation of regional-scale extreme event frequencies over that of the driving GCM [Bell et al., 2004]. Our hypothesis was that the RegCM3 would add value to the G-R-C methodology through its ability to simulate processes sensitive to a well-equilibrated land surface–atmosphere, such as mesoscale circulations driven by horizontal gradients in soil moisture/temperature [e.g., Weaver and Avissar, 2001]. However, on the basis of subjective and objective evaluations, the solutions from the G-R-C were inferior to those from the G-C. The poor performance of the G-R-C was due, in one case, to the large sensitivity of the placement of the WRF domain within the RegCM3. Domain reconfigurations led to better, albeit still relatively less accurate solutions. Ultimately, the simulation of the convective scale was limited in both test cases by synoptic-scale inaccuracies in the initial and boundary conditions provided by the RegCM3. These resulted not from the ability of RegCM3 to downscale the NNRP data, but rather from intrinsic features that developed over the long-term RegCM3 integrations. Hence, for these specific CPS events, this disadvantage of the long-term integrations overwhelmed any advantages.

[42] To be fair, accurate synoptic-scale representations, on specific days, is beyond the expected capability of any regional climate model that has been continuously integrated for more than a year. Hence, now that we have demonstrated the modeling concept using the historical events, we will next subject the modeling approaches to the test of generating daily convective-scale solutions over multiple-week periods, over multiple years, rather than for specific historical events. This will allow us to begin to evaluate the climate statistics of CPSs and associated phenomena, which is the ultimate application of this research.

[43] In addition to performing this multiple realization test, we have a number of model issues that need to be addressed. These include, but are not limited to (1) the extraction of compatible land-surface fields from GCM and/or RCM output that can be used to initialize the LSM in the WRF model; (2) the use of model ensembles to help determine solution robustness; (3) the acceptability of use of multiple nests in the WRF model for the current application, or if a single, large domain at 3–4 km grid point spacing is necessary; and (4) the reliability and sensitivity of the tornado proxy technique, and the development of analogous proxies for other hazardous phenomena.

[44] Although we have assessed the ability of these two telescoping modeling approaches to simulate historical events, it is necessary to consider that in order to robustly investigate CPS behavior in response to elevated greenhouse gas forcing, any successful approach must ultimately capture critical processes independent of observed or reanalysis data. Thus far, the performance of the G-C indicates that a modeling system consisting of a GCM driving a convection-permitting model could be applied for studying CPS dynamics under elevated greenhouse gas forcing. However, in order for this configuration to be viable, the global model must accurately capture both the large-scale features and the initial surface conditions represented in the reanalysis data set. Further, fine-scale climate processes not resolved by the global model must not force CPS behavior. Evaluation of these requirements will require further testing of both configurations, with an atmosphere-ocean general circulation model as the primary driver.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information

[45] We have benefited from conversations with J. Pal, F. Giorgi, W. Gutowski, and A. Seth and appreciate the constructive comments made by the three anonymous reviewers. This work stems from the Climate and Extreme Weather initiative within the Department of Earth and Atmospheric Sciences at Purdue University and was supported in part by NSF ATM-0541491. This is PCCRC paper 0705.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
  • Bell, J. A., L. C. Sloan, and M. A. Snyder (2004), Regional changes in extreme climatic events: A future climate scenario, J. Clim., 17, 8187.
  • Brooks, H. E. (2006), A global view of severe thunderstorms: Estimating the current distribution and possible future changes, Preprints, AMS Severe Local Storms Special Symposium, no. J4.2, Am. Meteorol. Soc., Atlanta, Ga., 30 Jan. – 2 Feb. (Available at http://www.nssl.noaa.gov/users/brooks/public_html/papers/AMS2K6.pdf).
  • Brooks, H. E., and C. A. Doswell III (2001), Normalized damage from major tornadoes in the United States: 1890–1999, Weather Forecasting, 16, 168176.
  • Brooks, H. E., C. A. Doswell III, and M. P. Kay (2003), Climatological estimates of local daily tornado probability for the United States, Weather Forecasting, 18, 626640.
  • Browning, K. A. (1964), Airflow and precipitation trajectories within severe local storms which travel to the right of the winds, J. Atmos. Sci., 21, 634639.
  • Burgess, D. W., V. T. Wood, and R. A. Brown (1982), Mesocyclone evolution statistics, Preprints, in 12th Conference on Severe Local Storms, Am. Meteorol. Soc., pp. 422424, San Antonio, Tx.
  • Chen, F., and J. Dudhia (2001), Coupling an advanced land-surface/hydrology model with the Penn State/NCAR MM5 modeling system. Part I: Model description and implementation, Mon. Weather Rev., 129, 569585.
  • Davies-Jones, R. P. (1984), Streamwise vorticity: The origin of updraft rotation in supercell storms, J. Atmos. Sci., 41, 29913006.
  • Denis, B., R. Laprise, D. Caya, and J. Côté (2002), Downscaling ability of one-way nested regional climate models: The Big-Brother Experiment, Clim. Dyn., 18, 627646.
  • Diffenbaugh, N. S., J. S. Pal, R. J. Trapp, and F. Giorgi (2005), Fine-scale processes regulate the response of extreme events to global climate change, Proc. Natl. Acad. Sci. U. S. A., 102, 15,77415,778.
  • Diffenbaugh, N. S., M. Ashfaq, B. Shuman, and J. W. Williams (2006), Summer aridity in the United States: Response to mid-Holocene changes in insolation and sea surface temperature, Geophys. Res. Lett., 33, L22712, doi:10.1029/2006GL028012.
  • Droegemeier, K. K., S. M. Lazarus, and R. P. Davies-Jones (1993), The influence of helicity on numerically simulated convective storms, Mon. Weather Rev., 121, 20052029.
  • Dudhia, J. (1989), Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model, J. Atmos. Sci., 46, 30773107.
  • Giorgi, F., and L. O. Mearns (1999), Introduction to special section: Regional climate modeling revisited, J. Geophy. Res., 104(D6), 63356352.
  • Giorgi, F., G. Bates, and S. Nieman (1993), The multi-year surface climatology of a regional atmospheric model over the western United States, J. Clim., 6, 7595.
  • Hoxit, L. R., and C. F. Chappell (1975), Tornado outbreak of April 3–4, 1974: Synoptic analysis, NOAA Tech. Rep. ERL-338-APCL-37, Natl. Oceanic. and Atmos. Admin., Boulder, Colo. (Available at http://www.ncdc.noaa.gov/oa/climate/extremes/1999/april/TornOut.pdf).
  • Iacono, M. J., E. J. Mlawer, S. A. Clough, and J.-J. Morcrette (2000), Impact of an improved longwave radiation model, RRTM, on the energy budget and thermodynamic properties of the NCAR Community Climate Model, CCM3, J. Geophys. Res., 105, 14,87314,890.
  • Intergovernmental Panel on Climate Change (2001), Climate Change 2001: The Scientific Basis: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, edited by J. T. Houghton et al., 881 pp., Cambridge Univ. Press, New York.
  • Johns, R. H., J. M. Davies, and P. W. Leftwich (1993), Some wind and instability parameters associated with strong and violent tornadoes: 2. Variations in the combinations of wind and instability parameters, in The Tornado: Its Structure, Dynamics, Prediction and Hazards, Geophys. Monogr., vol.79., edited by C. Church et al., pp. 583590, AGU, Washington, D. C.
  • Kain, J. S. (2004), The Kain–Fritsch convective parameterization: An update, J. Appl. Meteorol., 43, 170181.
  • Kain, J. S., S. J. Weiss, J. J. Levit, M. E. Baldwin, and D. R. Bright (2006), Examination of convection-allowing configurations of the WRF model for the prediction of severe convective weather: The SPC/NSSL spring program 2004, Weather Forecasting, 21, 167181.
  • Kalnay, E. (1996), The NCEP/NCAR 40-Year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437471.
  • Lilly, D. K. (1986), The structure, energetics and propagation of rotating convective storms. Part I: Energy exchange with the mean flow, J. Atmos. Sci., 43, 113124.
  • Lin, Y.-L., R. D. Farley, and H. D. Orville (1983), Bulk parameterization of the snow field in a cloud model, J. Clim. Appl. Meteorol., 22, 10651092.
  • Meehl, G. A., J. M. Arblaster, and C. Tebaldi (2005), Understanding future patterns of increased precipitation intensity in climate model simulations, Geophys. Res. Lett., 32, L18719, doi:10.1029/2005GL023680.
  • Michalakes, J., S. Chen, J. Dudhia, L. Hart, J. Klemp, J. Middlecoff, and W. Skamarock (2001), Development of a next generation regional weather research and forecast model, in Developments in Teracomputing: Proceedings of the Ninth ECMWF Workshop on the Use of High Performance Computing in Meteorology, edited by W. Zwieflhofer, and N. Kreitz, pp. 269276, World Sci. Hackensack, N. J.
  • 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, 16,66316,682.
  • NOAA (2005), Natural hazard statistics (Cited 2005), Off. Clim., Water, Weather Svc., Silver Spring, Md. (Available at http://www.nws.noaa.gov/om/hazstats.shtml).
  • Noh, Y., W. G. Cheon, S.-Y. Hong, and S. Raasch (2003), Improvement of the K-profile model for the planetary boundary layer based on large eddy simulation data, Boundary Layer Meteorol., 107, 401427.
  • Pal, J., E. Small, and E. Eltahir (2000), Simulation of regional-scale water and energy budgets: Representation of subgrid cloud and precipitation processes within RegCM, J. Geophy. Res., 105(D24), 29,57929,594.
  • Pal, J. S., et al. (2007), Regional climate modeling for the developing world: The ICTP RegCM3 and RegCNET, Bull. Am. Meteorol. Soc., 88, 13951409.
  • Schlesinger, M. E., and S. Malyshev (2001), Changes in near-surface temperature and sea-level for the Post-SRES CO2-stabilization scenarios, Integrated Assess., 2, 95199.
  • Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers (2005), A description of the Advanced Research WRF Version 2, NCAR Tech. Note, NCAR/TN-468+STR, 88 pp., Natl. Cent. for Atmos. Res., Boulder, Colo.
  • Tebaldi, C., K. Hayhoe, J. M. Arblaster, and G. A. Meehl (2006), Going to the extremes: An intercomparison of model-simulated historical and future changes in extreme events, Clim. Change, 79, 185211, doi:10.1007/s10584-006-9051-4.
  • Trapp, R. J., S. A. Tessendorf, E. S. Godfrey, and H. E. Brooks (2005a), Tornadoes from squall lines and bow echoes. Part I: Climatological distribution, Weather Forecasting, 20, 2334.
  • Trapp, R. J., G. J. Stumpf, and K. L. Manross (2005b), A reassessment of the percentage of tornadic mesoscyclones, Weather Forecasting, 20, 680687.
  • Trapp, R. J., D. M. Wheatley, N. T. Atkins, R. W. Przybylinski, and R. Wolf (2006), Buyer beware: Some words of caution on the use of severe wind reports in post-event assessment and research, Weather Forecasting, 21, 408415.
  • Warner, T. T., and H.-M. Hsu (2000), Nested-model simulation of moist convection: The impact of coarse-grid parameterized convection on fine-grid resolved convection, Mon. Weather Rev., 128, 22112231.
  • Weaver, C. P., and R. Avissar (2001), Atmospheric disturbances caused by human modification of the landscape, Bull. Am. Meteorol. Soc., 82, 269281.
  • Weisman, M. L., and J. B. Klemp (1982), The dependence of numerically simulated convective storms on vertical wind shear and buoyancy, Mon. Weather Rev., 110, 504520.
  • Weisman, M. L., and J. B. Klemp (1984), The structure and classification of numerically simulated convective storms in directionally varying wind shears, Mon. Weather Rev., 112, 24792498.
  • Weisman, M. L., W. C. Skamarock, and J. B. Klemp (1997), The resolution dependence of explicitly modeled convective systems, Mon. Weather Rev., 125, 527548.
  • Weisman, M. L., C. Davis, J. Done, W. Wang, and J. Bresch (2004), Real-time explicit convective forecasts using the WRF model during the BAMEX field program, Preprints, 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, no. 12.4, Amer. Meteor. Soc., Seattle, Wash. (Available at http://ams.confex.com/ams/84Annual/techprogram/paper_72169.htm).
  • Weiss, S., J. Kain, J. Levit, M. Baldwin, D. Bright, G. Carbin, and J. Hart (2005), SPC/NSSL Spring Program 2005: Program overview and operations plan, NOAA, Silver Spring, Md. (Available at http://www.spc.noaa.gov/exper/Spring_2005/2005_ops_plan.pdf).

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Telescoping Modeling
  5. 3. G-C and G-R-C Testing
  6. 4. Results
  7. 5. Discussion and Conclusions
  8. Acknowledgments
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jgrd13744-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
jgrd13744-sup-0002-t02.txtplain text document0KTab-delimited Table 2.
jgrd13744-sup-0003-t03.txtplain text document0KTab-delimited Table 3.
jgrd13744-sup-0004-t04.txtplain text document0KTab-delimited Table 4.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.