1. Introduction and motivation
Recent work performing verification of convection-allowing modeling systems (i.e., models using ~4-km grid spacing or less with explicit depiction of convection) over large sets of cases has focused on the ability of these models to forecast convective storms and rainfall systems (e.g., Kain et al. 2008, 2010a,b; Lean et al. 2008; Roberts and Lean 2008; Weisman et al. 2008; Clark et al. 2009, 2011, 2012; Schwartz et al. 2009, 2010; Johnson et al. 2011a, 2011b; Sobash et al. 2011; Marsh et al. 2012; Sun et al. 2012 and many others). Exceptions include Coniglio et al. (2010), who examined environmental fields like 2-m and 850-hPa temperature and dewpoint, and Clark et al. (2010), who examined spread growth for similar environmental fields. It is intuitive to emphasize verification of convective storms and rainfall for convection-allowing models because explicitly depicting convection should make the most difference in realism relative to coarser convection-parameterizing models for these phenomena. However, the location, timing, and evolution of simulated storms are still connected to large-scale environmental parameters and weather systems. Thus, verification of large-scale fields is also critical to a robust evaluation of convection-allowing models, especially since an increase in resolution can sometimes adversely impact model performance given that not all model components work well for all scales of motion (e.g., Wolff et al. 2012).
With these considerations in mind, the present study evaluates forecasts of a common feature of the southern high plains during spring that is tied to large-scale environmental parameters and weather systems: the dryline. Specifically, 24-h dryline forecasts from a 4-km grid-spacing configuration of the Weather Research and Forecasting Model (WRF; Skamarock et al. 2008) run in real time at the National Severe Storms Laboratory (NSSL-WRF, hereafter) are compared to corresponding dryline forecasts from the 12-km grid-spacing North American Mesoscale Model (NAM; Rogers et al. 2009)—the operational mesoscale model run by the National Centers for Environmental Prediction (NCEP). Daily 0000 UTC initializations of the NSSL-WRF and NAM are compared over the 5-yr period of 2007–11 for the months April–June. These months were chosen because they cover the approximate period with the highest frequency of dewpoint temperature gradients and drylines over the southern U.S. high plains (e.g., Dodd 1965; Hoch and Markowski 2005).
The main goals of this study are to document and compare average errors in forecast dryline position from the NSSL-WRF and NAM forecast systems. Because of differences in model configurations, definitive conclusions regarding the impact of any single aspect of model configuration, for example, boundary layer parameterization or grid resolution, cannot be reached. However, because both modeling systems share the same initial conditions—0000 UTC NAM analyses—we can at least conclude that forecast differences are not related to the initial conditions. The remainder of the study is organized as follows. Section 2 presents relevant background information on drylines and their role in convection initiation, section 3 includes descriptions of the data sources and dryline identification methods, section 4 presents results, and section 5 provides a summary and conclusions.
2. Background
The dryline is the intersection where relatively warm, moist air originating from the Gulf of Mexico meets a relatively hot, dry air mass that originates over the elevated terrain of the southwestern United States and northern Mexico (e.g., Fujita 1958; Beebe 1958; Schaefer 1986). This intersection is generally oriented north–south (orthogonal to the elevation gradient). Rising terrain from the eastern Great Plains to the foothills of the Rocky Mountains impedes the easterly component of the generally southerly flow, causing the moist air to have a greater longitudinal than zonal component (Schaefer 1974a). Afternoon dewpoint gradients of 10 K (100 km)−1 are commonly found normal to the dryline. Although early studies suggested that surface virtual temperature was practically constant across the boundary (e.g., McGuire 1962), a body of recent work including cloud-scale modeling and more detailed observations demonstrates conclusively that virtual temperature frequently experiences pronounced gradients across the dryline. The intensity of the dryline has been found to be highly correlated with the confluence along the dryline, which itself is a function of surface lee cyclogenesis (Schultz et al. 2007). Frontogenetical dryline dynamics and horizontal water vapor accumulation can help lead to sharp dewpoint gradients by solenoidally forcing vertical circulation at the dryline (Hoskins and Bretherton 1972; Weiss and Bluestein 2002; Markowski and Richardson 2010). Additionally, because westerly winds are generally present above the low-level moist layer, a capping inversion forms, preventing the low-level moisture from vertically mixing (Schaefer 1974b). These factors create sharp surface dewpoint gradients where the top of the subtropical boundary layer intersects the rising terrain (Bluestein 1993). West of particularly strong drylines, dewpoints have been known to fall below zero degrees Fahrenheit, with values of relative humidity in single digits. Pietrycha and Rasmussen (2001) documented an extreme moisture gradient associated with a dryline of 10 K (1 km)−1.
Daytime movement of the dryline is dictated both by cross-dryline variations in vertical mixing (Markowski and Richardson 2010) and by cloud-scale, dry-convective boundary layer roll circulations to which the dryline tends to locally attach (Ziegler et al. 1997; Ziegler and Rasmussen 1998). In the absence of strong synoptic-scale forcing, the dryline will move eastward during the day and retreat to the west at night. Because the depth of the subtropical boundary layer typically decreases toward the west, the amount of surface sensible heating needed to destabilize the stable layer decreases toward the west as well (Markowski and Richardson 2010). Throughout the day, vertical mixing increases proportionally to the incoming solar radiation and the associated surface heat fluxes. Static stability in the moist sector is reduced by increasing solar radiation until the lapse rate becomes dry adiabatic just east of the dryline. If the moist sector east of the dryline is warmed and destabilized enough to break the capping inversion, the moist layer will then vertically mix with the dry elevated mixed layer aloft, the surface dewpoint will drop, westerly momentum will be mixed to the ground, and the dryline will “propagate” eastward (Bluestein 1993; Markowski and Richardson 2010). The more vertical mixing present, the faster the dryline can propagate (Schaefer 1974b). Based on their main method of propagation, drylines can be classified into two groups: quiescent and active. Quiescent dryline movement is described above; however, in active cases, strong cyclogenesis will advect the dryline eastward with the surface cyclone more like a classical front (Bluestein 1993).
The dryline is an important airmass boundary for convection initiation over the southern high plains (Fujita 1958; Rhea 1966; Schaefer 1986). Strong localized convergence and abundant moisture often make the dryline a preferred location for initiation; however, forecasting convection initiation is limited by the lack of boundary layer and lower-tropospheric observations on the sub-1-km scales (cloud scales) that “force” convection initiation (Ziegler et al. 1997; Ziegler and Rasmussen 1998; Richter and Bosart 2002). Because of the presence of a strong capping inversion above the moist sector, the dryline is often a strong inhibitor of convection initiation (Schaefer 1974b). However, dryline-induced vertical circulation patterns can effectively erode these capping inversions allowing convection initiation to occur. Several mechanisms for convection initiation along the dryline have been observed, including localized forcing from circulations associated with horizontal convective rolls, as well as gravity waves propagating eastward on top of the subtropical boundary layer (Ziegler et al. 1997; Ziegler and Rasmussen 1998; Wakimoto et al. 2006; Xue and Martin 2006).
Given the importance of drylines to convection initiation, the forecast of the dryline location is a critical component to a convection forecast. Despite this importance, a review of the literature does not find any systematic evaluations of forecast dryline position from operational models. Past research has found that sensible weather parameters directly linked to boundary layer mixing processes are often associated with large forecast errors (e.g., Coniglio et al. 2013 and references therein), but these studies have not explicitly examined forecast drylines. Hane et al. (2001) evaluated dryline motion from a limited-area mesoscale model for one case. Ziegler et al. (1997) anticipated that mesoscale model output could assist forecasters in preparing refined forecasts of the east–west location of drylines and the potential for dryline convection; however, the study mainly emphasized simulated convection initiation processes associated with drylines. The lack of systematic forecast dryline evaluations may be because a lack of finescale observations makes identifying the precise location of the dryline difficult. Also, only within the last decade or so have operational mesoscale models had adequate grid spacing to fully resolve the sharp moisture gradients associated with drylines. Other difficulties associated with locating and tracking drylines occur because drylines can “jump” (Hane et al. 2001), sometimes more than one moisture gradient is present (Hane et al. 1993; Crawford and Bluestein 1997), and other types of airmass boundaries (e.g., outflow boundaries and cold fronts) have some characteristics similar to those of drylines, making objective dryline identification more difficult. Finally, manually identifying drylines is quite labor intensive.
3. Data and methodology
Dryline forecast datasets include the 4-km grid-spacing NSSL-WRF, which is initialized daily at 0000 UTC and integrated over 36 h (e.g., Kain et al. 2010b). Before 9 June 2009, WRF version 2.2 was used for the NSSL-WRF with a domain encompassing most of the United States except for portions of the west. After 9 June 2009, the domain was expanded to encompass the entire CONUS and the model version was updated to version 3.1.1. Other model specifications during the 5-yr analysis period were unchanged. Physics parameterizations include the WRF single-moment six-class (WSM-6; Hong and Lim 2006) microphysics, Mellor–Yamada–Janjić (MYJ) boundary layer scheme (Mellor and Yamada 1982; Janjić 2002), the Noah land surface model (Chen and Dudhia 2001), and the Dudhia (1989) and Rapid Radiative Transfer Model (RRTM; Mlawer et al. 1997) for shortwave and longwave radiation, respectively. Initial and lateral boundary conditions (3-h updates) are from 0000 UTC initializations of the NAM model and interpolated onto a 40-km grid.
The other forecast dryline dataset is from the NAM. During the April–June 2007–11 period, the NAM used the Nonhydrostatic Mesoscale Model (WRF-NMM; Janjić 2003) dynamics core with initial conditions derived from the Gridpoint Statistical Interpolation (GSI; Wu et al. 2002). NAM physics packages included the MYJ boundary layer parameterization, the Noah land surface model, Ferrier et al.'s (2002) microphysics parameterization, the Betts–Miller–Janjić (BMJ; Betts 1986; Betts and Miller 1986; Janjić 1994) cumulus parameterization, and Geophysical Fluid Dynamics Laboratory (GFDL) shortwave (Lacis and Hansen 1974) and longwave (Fels and Schwarzkopf 1975; Schwarzkopf and Fels 1991) radiation parameterizations. Throughout the 5-yr analysis period, numerous minor bug fixes and other changes intended to improve the model physics and data assimilation packages were implemented.1
The dataset used for identifying the approximate observed dryline locations was from 20-km grid-spacing Rapid Update Cycle (RUC) model analyses provided by NCEP (Benjamin et al. 2004a,b). These analyses are generated using hourly intermittent three-dimensional variational data assimilation (3DVAR) cycles in which recent observations from various sources [e.g., wind profiler, radar, aircraft, surface aviation routine weather report (METAR), satellite, etc.] are assimilated using the previous 1-h RUC model forecasts as the background field.
In the forecast and analysis datasets, dryline positions were determined using a manual identification procedure. The main criterion for dryline classification was the existence of an unambiguous boundary between relatively moist and dry air with along-boundary length scales of O(100) km. The specific humidity field was used to identify these boundaries because, as noted by Hoch and Markowski (2005), specific humidity is not sensitive to elevation differences, unlike dewpoint temperature. At some point along the unambiguous moisture boundary, a specific humidity gradient magnitude of at least 3 g kg−1 (100 km)−1 was required for classification as a dryline (Hoch and Markowski 2005). To avoid misclassifying cold fronts as drylines, temperature fields were examined, and any segment of the moisture boundary associated with a noticeable temperature drop was eliminated from consideration as a dryline. Because of the change in moisture, there is often a noticeable temperature rise after dryline passage, and because these temperature rises are consistent with drylines, they helped confirm whether a boundary in question should be considered a dryline. Additionally, moisture gradients that were clearly the result of convective outflow were not considered. These moisture gradients were most often evident in the NSSL-WRF as approximately circular regions of lower specific humidity emanating from convection. In a few cases for the NSSL-WRF, simulated storms ahead of the dryline generated outflow that impinged upon the dryline causing an apparent westward “bulge.” Similar behavior has been documented in observed drylines interacting with westward-propagating cold pools using West Texas Mesonet data (Geerts 2008). If a relatively small portion of the dryline was affected by convective outflow, unaffected portions of the dryline were simply connected during manual identification. If relatively large portions of the dryline were affected by convective outflow, the case was not considered, which was a subjective decision made on a case-by-case basis. Finally, a wind shift from a dry source region to a moist source region was required for dryline classification, which was determined through examination of 10-m wind vectors in the forecasts and analyses.
For manual dryline identification, the Grid Analysis and Display System (GrADS; http://www.iges.org/grads/) was used to display relevant fields for a domain centered over the southern high plains. Manual dryline identification was restricted to the area shown in Fig. 1, which illustrates an example of fields examined for RUC analyses valid 14 April 2011. For each case in which a dryline was deemed present, a GrADS script allowed the user to manually draw a series of points outlining the dryline, and the coordinates of these points were output to a file for subsequent analyses. Straight-line segments connecting each of these points composed the dryline. The points were drawn along the axis of the maximum specific humidity gradient magnitude. An example of a manually drawn dryline with the corresponding specific humidity gradient is shown in Fig. 1c; note that the northward extension of the manually drawn dryline stops in central Oklahoma because, north of this point, the specific humidity gradient also coincides with a temperature drop consistent with a cold front (Fig. 1b).
The 0000 UTC 15 Apr 2011 RUC analysis fields: (a) 2-m dewpoint (shaded, °F) and 2-m specific humidity gradient magnitude greater than 4 g kg−1 (100 km)−1 (hatched); (b) mean sea level pressure (contours, hPa), 2-m temperature (shaded, °F), and 10-m wind barbs; (c) 2-m specific humidity gradient magnitude [shaded; g kg−1(100 km)−1] with the manually defined dryline denoted by the black line; and (d) 2-m specific humidity.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
The 0000 UTC 15 Apr 2011 RUC analysis fields: (a) 2-m dewpoint (shaded, °F) and 2-m specific humidity gradient magnitude greater than 4 g kg−1 (100 km)−1 (hatched); (b) mean sea level pressure (contours, hPa), 2-m temperature (shaded, °F), and 10-m wind barbs; (c) 2-m specific humidity gradient magnitude [shaded; g kg−1(100 km)−1] with the manually defined dryline denoted by the black line; and (d) 2-m specific humidity.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
The 0000 UTC 15 Apr 2011 RUC analysis fields: (a) 2-m dewpoint (shaded, °F) and 2-m specific humidity gradient magnitude greater than 4 g kg−1 (100 km)−1 (hatched); (b) mean sea level pressure (contours, hPa), 2-m temperature (shaded, °F), and 10-m wind barbs; (c) 2-m specific humidity gradient magnitude [shaded; g kg−1(100 km)−1] with the manually defined dryline denoted by the black line; and (d) 2-m specific humidity.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
To compute average dryline positions for evaluation of mean dryline position errors, an average longitude was computed for each dryline. This average was obtained by finding the longitude at the midpoint of each individual line segment that composed the dryline. Weights were then assigned to the midpoint longitude of each segment based on the ratio of the segment length to that of the entire dryline. Finally, the average dryline longitude was simply calculated by taking the weighted average of the midpoint longitudes.
To account for the differences in the resolved scales among the three datasets, a General Meteorological Package (GEMPAK) function, GWFS, which applies a Gaussian weighted filter, was used to dampen (by at least 67%) wavelengths below 120 km in the specific humidity field before gradient magnitudes were computed. Applying this filter was a very important step, especially in the NSSL-WRF. Not only did applying the filter allow an equitable comparison of NSSL-WRF to the other datasets, it also helped smooth out finescale structures in the gradient field while retaining and oftentimes emphasizing the dryline location, making it easier to identify. Examples of the specific humidity gradient magnitude before and after application of the filter to the NSSL-WRF are shown in Figs. 2a and 2b, respectively. Smaller-scale features, some of which are associated with convective outflow, are much less apparent after the filter is applied.
NSSL-WRF 24-h forecast valid 25 May 2011 of specific humidity gradient magnitude [g kg−1 (100 km)−1, grayscale shading] and composite reflectivity (dBZ, colored shading) (a) without and (b) with the filter applied to the specific humidity field prior to calculation of the gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
NSSL-WRF 24-h forecast valid 25 May 2011 of specific humidity gradient magnitude [g kg−1 (100 km)−1, grayscale shading] and composite reflectivity (dBZ, colored shading) (a) without and (b) with the filter applied to the specific humidity field prior to calculation of the gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
NSSL-WRF 24-h forecast valid 25 May 2011 of specific humidity gradient magnitude [g kg−1 (100 km)−1, grayscale shading] and composite reflectivity (dBZ, colored shading) (a) without and (b) with the filter applied to the specific humidity field prior to calculation of the gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
As noted by a reviewer, it is possible that the filtering applied to the NSSL-WRF could artificially shift the dryline location as resolved at the 4-km grid scale if the modeled moisture profile across the dryline boundary is not approximately a step function. For example, if moisture is approximately constant behind the dryline, but moisture gradually increases in front of the dryline, the filter could induce an eastward bias. To test whether the filter applied to the NSSL-WRF data induced any systematic biases, a simple test was performed. First, a point was defined every 0.25° in latitude along each manually defined dryline from the NSSL-WRF forecasts. Then, at each point, the filtered and unfiltered specific humidity gradients were extracted along the manually defined dryline and 0.75° in longitude to the west and east of the dryline at 0.05°-longitude increments (15 points to the west and 15 points to the east of each dryline point). Examples of this procedure for the dryline manually defined in the NSSL-WRF at 0000 UTC 23 May 2007 are shown in Figs. 3a and 3b. Finally, averages were computed for each of these 31 points over each set of points along each manually defined dryline. For both the filtered and unfiltered NSSL-WRF datasets (Figs. 3c and 3d, respectively), the maximum of the averages occurred at the point along the manually defined dryline indicating that the filter did not induce a bias outside the range ±0.05° longitude.
(a) Unfiltered and (b) filtered specific humidity gradient magnitude [g kg−1 (100 km)−1, shading] from the 24-h NSSL-WRF forecast valid 0000 UTC 23 May 2007. A thin black line marks the location of the manually defined dryline, and the black horizontal lines (black circles mark the midpoint) denote the axes along which specific humidity gradient magnitude values are extracted at 0.05° longitude increments. (c) Average along each horizontal axis from each dryline of unfiltered specific humidity gradient magnitude. The vertical black line marks the midpoint, which corresponds to the point along the manually defined dryline. (d) As in (c), but for the filtered specific humidity gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
(a) Unfiltered and (b) filtered specific humidity gradient magnitude [g kg−1 (100 km)−1, shading] from the 24-h NSSL-WRF forecast valid 0000 UTC 23 May 2007. A thin black line marks the location of the manually defined dryline, and the black horizontal lines (black circles mark the midpoint) denote the axes along which specific humidity gradient magnitude values are extracted at 0.05° longitude increments. (c) Average along each horizontal axis from each dryline of unfiltered specific humidity gradient magnitude. The vertical black line marks the midpoint, which corresponds to the point along the manually defined dryline. (d) As in (c), but for the filtered specific humidity gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
(a) Unfiltered and (b) filtered specific humidity gradient magnitude [g kg−1 (100 km)−1, shading] from the 24-h NSSL-WRF forecast valid 0000 UTC 23 May 2007. A thin black line marks the location of the manually defined dryline, and the black horizontal lines (black circles mark the midpoint) denote the axes along which specific humidity gradient magnitude values are extracted at 0.05° longitude increments. (c) Average along each horizontal axis from each dryline of unfiltered specific humidity gradient magnitude. The vertical black line marks the midpoint, which corresponds to the point along the manually defined dryline. (d) As in (c), but for the filtered specific humidity gradient magnitude.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Drylines were first identified in RUC analyses. Then, for each day a dryline was present in the RUC analyses, the NSSL-WRF was examined for drylines. Finally, for the days in which a dryline was present in both RUC analyses and NSSL-WRF forecasts, NAM forecasts were examined. For comparisons among the three datasets, statistics were only computed for days in which drylines were identified in all three datasets. To obtain an observed dryline climatology, separate analyses were conducted using every dryline case identified in the RUC analyses, regardless of whether a dryline was defined in the NSSL-WRF and/or NAM datasets.
Out of 455 days within the analysis period, drylines were identified in the RUC analyses on 134 days (Table 1). This fraction of dryline days (29%) is very near what previous studies have found (e.g., Rhea 1966; Schaefer 1974a; Peterson 1983). For example, Hoch and Markowski (2005) found that 32% of days during a 30-yr period for the April–June period had drylines. Out of the 134 days in which drylines were identified in RUC analyses, drylines were identified on 116 days in both the NSSL-WRF and NAM datasets. For the 18 days in which drylines were only manually defined in the RUC analyses, the reasons for not defining drylines in the NSSL-WRF and NAM are listed in Table 2. For 11 of these days, the NSSL-WRF forecasts contained too much convective outflow interacting with the dryline to allow for meaningful manual identification; for 1 day, NAM and NSSL-WRF data were missing; for 4 days, either the NSSL-WRF or NAM did not contain a dryline; and for 2 days, the RUC analyses contained too much convective outflow to allow for meaningful comparisons to NSSL-WRF and NAM forecasts (for these two cases, RUC analysis dryline positions were estimated as best as possible and were included in subsequent analyses that do not involve comparisons among the three datasets).
Dates on which drylines were manually identified in at least one of the datasets (134 total dates). For the 116 dates indicated by lightface, drylines were manually identified in each of the NSSL-WRF, NAM, and RUC analysis datasets. For the 18 dates in boldface, drylines were manually identified in the RUC analyses, but not in the 24-h NSSL-WRF and/or NAM forecasts. The drylines were present at 0000 UTC for each day indicated.
A list of each date on which a dryline was defined in the RUC analysis, but not defined in the NSSL-WRF and NAM datasets, and the reason that each of these dates was not considered.
4. Results
a. Example cases
Figure 4 illustrates three representative cases in which drylines were identified. The 24 May 2011 case (Figs. 4a–d) could be considered an “active” dryline case, since it was associated with a high-amplitude short-wave trough that had moved rapidly eastward into the southern high plains by the afternoon of 24 May. An extremely unstable air mass ahead of the dryline, as well as strong upper-level flow that veered with height, created a favorable environment for strong/violent long-track tornadoes. Accordingly, the Storm Prediction Center (SPC) expected tornadic storms to be initiated along the dryline and issued a “high risk” warning for severe weather. Despite the high degree of certainty that a significant severe weather outbreak would take place, at the time the first “day 1” severe weather outlook was issued (0600 UTC), there was still significant uncertainty regarding the precise placement of the dryline and thus where convection initiation was likely (S. Goss, SPC, 2012, personal communication). Wording in the SPC outlook2 issued at 0600 UTC suggested that by late afternoon the dryline would extend from south-central Kansas through central Oklahoma and into north-central Texas [the NSSL-WRF forecasts would not have been available for the 0600 UTC day 1 outlook (J. Grams, SPC, 2012, personal communication)]. However, to reflect the uncertainty in the forecast dryline position, a relatively loose gradient from the high to general thunderstorm risk areas was depicted from south-central Kansas through west-central Oklahoma and into north Texas (Fig. 5a). It was not until the 1630 UTC outlook was issued that it had become clear from more recently updated model guidance and surface observations that the dryline would be located farther west than previously thought. The tightening of the gradient from high to general thunderstorm risk areas and general westward shift in the slight to high risk areas contained in the 1630 UTC issued severe weather outlook (Fig. 5b) reflected the increased degree of certainty in dryline location and likely convection initiation (J. Hart, SPC, 2012, personal communication). These types of adjustments from earlier to later outlooks are fairly typical because SPC forecasters are aware that boundary layer mixing and resulting dryline positioning within models is often not well depicted (S. Goss, SPC, 2012, personal communication). Indeed, the average longitude of the 24-h forecast dryline position from the NSSL-WRF was about 1.36° farther east than the RUC analysis. While the 24-h NAM forecast was better than the NSSL-WRF, it still contained a noticeable eastward displacement relative to the observed dryline of about 0.25°. For reference, at the northern-most, middle, and southern-most latitudes of the analysis domain, 1° longitude corresponds to 81.3, 89.4, and 96.3 km, respectively.
Specific humidity (shaded, g kg−1) valid 0000 UTC 25 May 2011 with manually defined dryline locations denoted by black lines for (a) RUC analysis, (b) 24-h NSSL-WRF forecast, and (c) 24-h NAM forecast. (d) Dryline locations for (a)–(c) are shown together. The rest of the panels show conditions similar to those in (a)–(d), but for (e)–(h) 0000 UTC 7 Jun 2007 and (i)–(l) 0000 UTC 7 Apr 2010.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Specific humidity (shaded, g kg−1) valid 0000 UTC 25 May 2011 with manually defined dryline locations denoted by black lines for (a) RUC analysis, (b) 24-h NSSL-WRF forecast, and (c) 24-h NAM forecast. (d) Dryline locations for (a)–(c) are shown together. The rest of the panels show conditions similar to those in (a)–(d), but for (e)–(h) 0000 UTC 7 Jun 2007 and (i)–(l) 0000 UTC 7 Apr 2010.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Specific humidity (shaded, g kg−1) valid 0000 UTC 25 May 2011 with manually defined dryline locations denoted by black lines for (a) RUC analysis, (b) 24-h NSSL-WRF forecast, and (c) 24-h NAM forecast. (d) Dryline locations for (a)–(c) are shown together. The rest of the panels show conditions similar to those in (a)–(d), but for (e)–(h) 0000 UTC 7 Jun 2007 and (i)–(l) 0000 UTC 7 Apr 2010.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Day 1 severe weather outlooks (valid until 1200 UTC the following day) for 24 May 2011 produced by SPC at (a) 0600 and (b) 1630 UTC.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Day 1 severe weather outlooks (valid until 1200 UTC the following day) for 24 May 2011 produced by SPC at (a) 0600 and (b) 1630 UTC.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Day 1 severe weather outlooks (valid until 1200 UTC the following day) for 24 May 2011 produced by SPC at (a) 0600 and (b) 1630 UTC.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
During late afternoon on 24 May, numerous tornadoes occurred with supercells that developed along and just east of the dryline in Kansas, Oklahoma, and Texas (Fig. 5). These tornadoes included two that were rated EF4 on the enhanced Fujita scale and one rated EF5 in central Oklahoma. Considering the high population density along the I-35 corridor including the Oklahoma City, Oklahoma, metropolitan area, accurately forecasting the dryline's position was crucial in determining how many people were at risk during this severe weather event.
The second example dryline case, 6 June 2007 (Figs. 4e–h), was not associated with nearly as extreme environmental parameters for severe weather as was 24 May 2011. However, strong midlevel flow that veered with height and moderate instability led the SPC to issue a “moderate risk” warning for much of the northern plains. Convection initiation was forecast to occur along a dryline extending northward through the Texas and Oklahoma Panhandle and into western Kansas and Nebraska. Similar to 24 May 2011, 24-h dryline forecasts from the NAM had relatively small average longitudinal errors (0.07° westward error), and the NSSL-WRF 24-h dryline forecast was east of the NAM. However, for 6 June, NSSL-WRF only had an average eastward displacement of 0.17°. Despite the small average longitudinal errors in both models for 6 June, at different points along the dryline, there were much larger errors. For example, in western Kansas, the NSSL-WRF had a noticeable eastward error, while in the Texas Panhandle the NAM had a relatively large westward error (Fig. 4h).
In the Texas Panhandle, there was a noticeable eastward “bulge” in the analyzed dryline (Fig. 4e). The NSSL-WRF hints at this bulge with relatively lower values of specific humidity in the Texas Panhandle (Fig. 4f), but because the specific humidity gradient magnitude was stronger farther west of this relative minimum (not shown), the manually defined dryline did not include it. This case illustrates that using the average dryline longitude, while helpful for diagnosing mean east–west errors, may “mask” some of the errors along more localized segments of drylines.
The last example dryline case, 6 April 2010 (Figs. 4i–l), was associated with a lower severe weather risk than the first two cases. For this case, the archived SPC severe weather outlooks indicated that a narrow corridor of moderate instability was expected along a stationary frontal boundary extending from Illinois into southern Iowa–northern Missouri to northeast Kansas and ahead of a dryline expected to extend from Kansas through central Oklahoma and Texas. Because the expected orientation of the vertical shear vectors favored upscale growth of convection, rather than long-lived supercells, a “slight risk” was issued by SPC for much of the region ahead of the dryline and along the stationary boundary to the north. Similar to the 24 May 2011 case, both the NAM and NSSL-WRF had eastward dryline displacement errors relative to the RUC analysis (0.80° and 1.13° eastward errors in the NAM and NSSL-WRF, respectively).
b. Model and analysis dryline climatology
To illustrate the spatial distribution of drylines, the analysis domain was divided into 0.5° × 0.5° grid boxes, and the frequency that a dryline intersected each grid box at 0000 UTC over the 116 cases with manually defined drylines in the RUC analyses and NSSL-WRF and NAM forecasts was computed (Figs. 6a–d). These frequencies were also computed for the cases that occurred during each month (Figs. 6e–p). Additionally, the domain was divided into 0.5° latitudinal bands, and a composite dryline position was computed based on the average dryline longitude for drylines intersecting each latitudinal band. In Fig. 6, these composite dryline positions are indicated by black lines within the latitudinal bands intersected by at least 10 drylines (for all months; Figs. 6a–d) or 4 drylines (for individual months; Figs. 6e–p).
Dryline frequency (shaded) with average dryline location indicated by black line for all months in (a) RUC analyses, (b) NSSL-WRF, and (c) NAM. (d) Average dryline location for all months in RUC, NSSL-WRF, and NAM shown together. The rest of the panels show conditions similar to those in (a)–(d), but for cases in (e)–(h) April, (i)–(l) May, and (m)–(p) June. The number of cases for each month is shown in parentheses after the month label in each row.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Dryline frequency (shaded) with average dryline location indicated by black line for all months in (a) RUC analyses, (b) NSSL-WRF, and (c) NAM. (d) Average dryline location for all months in RUC, NSSL-WRF, and NAM shown together. The rest of the panels show conditions similar to those in (a)–(d), but for cases in (e)–(h) April, (i)–(l) May, and (m)–(p) June. The number of cases for each month is shown in parentheses after the month label in each row.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Dryline frequency (shaded) with average dryline location indicated by black line for all months in (a) RUC analyses, (b) NSSL-WRF, and (c) NAM. (d) Average dryline location for all months in RUC, NSSL-WRF, and NAM shown together. The rest of the panels show conditions similar to those in (a)–(d), but for cases in (e)–(h) April, (i)–(l) May, and (m)–(p) June. The number of cases for each month is shown in parentheses after the month label in each row.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Examination of the RUC analysis dryline frequencies over these 116 cases (Fig. 6a) shows that drylines occurred most frequently over the western halves of Texas and Oklahoma as well as southwest Kansas. The RUC analysis composite dryline (black line in Fig. 6d) stretches from southwest Texas through western Oklahoma and into southwest Kansas, extending farther east with increasing latitude within the corridor from −100° to −102° longitude. The longitude of the composite analysis dryline is consistent with the preferred dryline longitude of −101° found by Hoch and Markowski (2005), who examined drylines occurring over a 30-yr period. In general, the spatial distribution of dryline frequencies over all cases for the NSSL-WRF (Fig. 6b) and NAM (Fig. 6c) is quite similar to that of the RUC analysis. However, from the overlay of composite drylines (Fig. 6d), there are clearly some differences in average dryline placement. On average, the composite dryline in the 24-h NSSL-WRF forecasts is farthest east, especially over Texas. For the 24-h NAM forecasts, the composite dryline over Texas lines up quite well with the RUC analysis, but farther north, the composite NAM dryline has a slight westward bias.
For the monthly spatial distributions of dryline frequencies (Figs. 6e–p), the RUC analyses for April (Fig. 6e) and May (Fig. 6i) are quite similar. For both of these months, the RUC composite dryline appears to fall between one relative maximum in dryline frequency over west Texas and another over central Texas. This bimodal dryline position frequency in April and May appears to be better discriminated in the NSSL-WRF than in the NAM. It is quite possible that these two maxima separate quiescent drylines that tend not to move very far to the east from active drylines that do tend to move east. For both April and May, the composite dryline location in the NSSL-WRF has a clear eastward bias. For the NAM during April and May, the composite dryline lines up quite well with the RUC analysis (Figs. 6h and 6l).
During June (Fig. 6m), the spatial distribution of dryline frequencies in the RUC analyses is noticeably different from that of April and May. Instead of two maxima, there is only one maximum in west Texas, and the corridor over which drylines occur has a clear southwest-to-northeast orientation, which is also reflected in the composite dryline. This orientation may be related to the general northward shift during June in the upper-tropospheric jet stream that would result in only the northern portions of drylines being active (i.e., advected east). The single June mode in west Texas has a very similar location and orientation to the western mode of the bimodal April and May distribution. The spatial distributions in the NSSL-WRF and NAM simulations replicate this southwest-to-northeast orientation fairly well, but extend higher dryline frequencies too far west into the Texas Panhandle, which results in the northern section of the NSSL-WRF and NAM composite drylines being farther west than the RUC analysis (Figs. 6m–p). However, in general, June seems to be the month in which the NAM and NSSL-WRF composite drylines are most closely clustered around the composite dryline in the RUC analysis.
To obtain an observed dryline climatology based on the RUC analyses, the RUC analysis dryline frequencies are recomputed with the additional 18 cases that did not have corresponding NSSL-WRF and NAM drylines (Fig. 7). Not surprisingly, the spatial dryline distributions are similar to those using the 116 cases, and the average dryline positions between the two sets of cases (denoted by the black and gray lines in Fig. 7 do not meaningfully differ.
(a)–(d) As in Figs. 5a, 5e, 5i, and 5m, respectively, except for all cases in which drylines were manually defined in the RUC analyses regardless of whether drylines were identifiable in the NAM or NSSL-WRF datasets. The number of cases for each month is shown in parentheses after the month label along the y axis. The black line indicates the average dryline position from all RUC cases and the gray line indicates the average dryline position from only the dates in which drylines were identified in all three datasets (i.e., same as the black lines in Figs. 5a, 5e, 5i, and 5m).
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
(a)–(d) As in Figs. 5a, 5e, 5i, and 5m, respectively, except for all cases in which drylines were manually defined in the RUC analyses regardless of whether drylines were identifiable in the NAM or NSSL-WRF datasets. The number of cases for each month is shown in parentheses after the month label along the y axis. The black line indicates the average dryline position from all RUC cases and the gray line indicates the average dryline position from only the dates in which drylines were identified in all three datasets (i.e., same as the black lines in Figs. 5a, 5e, 5i, and 5m).
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
(a)–(d) As in Figs. 5a, 5e, 5i, and 5m, respectively, except for all cases in which drylines were manually defined in the RUC analyses regardless of whether drylines were identifiable in the NAM or NSSL-WRF datasets. The number of cases for each month is shown in parentheses after the month label along the y axis. The black line indicates the average dryline position from all RUC cases and the gray line indicates the average dryline position from only the dates in which drylines were identified in all three datasets (i.e., same as the black lines in Figs. 5a, 5e, 5i, and 5m).
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
c. Statistical dryline comparisons
For a more robust statistical analysis of average dryline position errors, Fig. 8 shows box plots constructed using the distributions of average longitudinal dryline errors in the NSSL-WRF and NAM simulations for various subsets of the data. Longitudinal errors were computed simply by taking the difference between the average dryline longitude in the 24-h NAM or NSSL-WRF forecast and the RUC analysis for each case. The computation of the average dryline longitudes is described in section 3. Positive (negative) errors indicate eastward (westward) dryline bias. For each distribution depicted by a box plot, a one-sample Student's t test was applied using the R statistical software package (R Development Core Team 2012) to determine whether the mean differed significantly (at level α = 0.05) from zero (i.e., whether there were significant east–west biases). In other words, the null hypothesis is that the mean errors are equal to zero. Rejection of the null hypothesis (i.e., significance) indicates that the difference in the mean errors from zero was likely not due to random chance. In addition, for each pair of NSSL-WRF and NAM distributions depicted by box plots in Fig. 8, a Student's t test was applied to determine whether the difference in the means between NSSL-WRF and NAM differed significantly from zero (i.e., whether the mean errors between NSSL-WRF and NAM were significantly different). The P values from these significance tests are shown at the bottom of Fig. 8.
Box plots for the distributions of average longitudinal dryline errors in the NAM (red) and NSSL-WRF (blue) datasets over all cases, and different months, synoptic patterns, and years. The number of cases in each category is indicated by the value in parentheses below each category label. The median is indicated by the straight black line through each box, the box encompasses the IQR, outliers defined by values outside of 1.5 × IQR are marked by crosses, and horizontal lines (whiskers) denote the smallest and largest values that are not outliers. The P values displayed at the bottom are from a Student's t test of whether the means from each distribution were different from zero (second from bottom row), and whether the difference in means from each pair of NSSL-WRF and NAM distributions was different from zero (bottom row). The P values indicating significance (α = 0.05) are indicated by black text.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Box plots for the distributions of average longitudinal dryline errors in the NAM (red) and NSSL-WRF (blue) datasets over all cases, and different months, synoptic patterns, and years. The number of cases in each category is indicated by the value in parentheses below each category label. The median is indicated by the straight black line through each box, the box encompasses the IQR, outliers defined by values outside of 1.5 × IQR are marked by crosses, and horizontal lines (whiskers) denote the smallest and largest values that are not outliers. The P values displayed at the bottom are from a Student's t test of whether the means from each distribution were different from zero (second from bottom row), and whether the difference in means from each pair of NSSL-WRF and NAM distributions was different from zero (bottom row). The P values indicating significance (α = 0.05) are indicated by black text.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Box plots for the distributions of average longitudinal dryline errors in the NAM (red) and NSSL-WRF (blue) datasets over all cases, and different months, synoptic patterns, and years. The number of cases in each category is indicated by the value in parentheses below each category label. The median is indicated by the straight black line through each box, the box encompasses the IQR, outliers defined by values outside of 1.5 × IQR are marked by crosses, and horizontal lines (whiskers) denote the smallest and largest values that are not outliers. The P values displayed at the bottom are from a Student's t test of whether the means from each distribution were different from zero (second from bottom row), and whether the difference in means from each pair of NSSL-WRF and NAM distributions was different from zero (bottom row). The P values indicating significance (α = 0.05) are indicated by black text.
Citation: Weather and Forecasting 28, 3; 10.1175/WAF-D-12-00092.1
Over all of the 116 cases, the median dryline error in the NAM was very close to zero and the Student's t test confirmed that the mean dryline error in the NAM was not significantly different than zero. Additionally, the interquartile range (IQR) of mean dryline error fell well within ±0.5°. In contrast, over all of the 116 cases for the NSSL-WRF, the median dryline error was near 0.5° and the entire IQR was above 0.0. The difference from zero in the NSSL-WRF mean dryline errors was highly significant, indicating a systematic eastward bias, which was not surprising given the comparative locations of the composite drylines in Fig. 6.
Comparing dryline errors by month, the lack of a significant east–west bias in 24-h forecast NAM drylines is consistent across all months. For the NSSL-WRF, all months have the eastward bias, but the mean errors are smaller during June. For all months, the difference in mean dryline errors between NSSL-WRF and NAM was found to be statistically significant. April and May had the most dryline cases (43 and 47, respectively), while June had the fewest (26).
The cases were also categorized according to synoptic pattern with the idea that mean errors may be a function of the type of regime present (middle section of Fig. 8). The synoptic patterns were determined by matching each dryline case to a basic upper-level pattern (trough, ridge, or zonal) using archived 500-hPa geopotential heights from the SPC's hourly mesoscale analysis (Bothwell et al. 2002). In some cases, the 500-hPa geopotential height tendency for the 24-h period immediately preceding the dryline analysis was calculated using the North American Regional Reanalysis (NARR) dataset (Mesinger et al. 2006) in order to distinguish between an approaching trough and a building ridge. For these cases, if the majority of the RUC analysis dryline coincided with height falls (rises) greater than 10 geopotential meters (gpm), the patterns were classified as troughs (ridges). The majority of cases were associated with a trough (55), followed by ridge (39) and zonal (22) cases. Similar to the breakdown of cases by month, the mean dryline errors in the NAM for each pattern were fairly consistent, with none of the patterns having mean dryline errors significantly different than zero. For the NSSL-WRF, each pattern was associated with a significant eastward bias, and the mean dryline errors for trough cases were slightly greater than ridge and zonal cases.
Finally, the cases were divided according to year (right section of Fig. 8). There was a notable difference in the number of cases identified during the different years, with 2011 having the most dryline cases (37), followed by 2008 (28), 2010 (19), 2007 (17), and 2009 (15). The year with the most drylines, 2011, was the year with the most severe weather in the April–June period, while the year with the fewest drylines, 2009, had the least amount of severe weather based on storm reports in the National Climatic Data Center (NCDC) publication Storm Data, which are available from the SPC web site (http://www.spc.noaa.gov/climo/historical.html). For the NAM mean dryline errors, the year 2008 had the largest westward mean errors and 2009 had the largest eastward errors, but none of the years had mean errors that were significantly different than zero. For the NSSL-WRF mean dryline errors, 2010 and 2011 had the largest eastward biases, although when 2010 and 2011 are compared to each of the other years, only the differences with 2009 are significant (not shown).
5. Summary and discussion
This study evaluated 24-h forecasts of drylines from an experimental 4-km grid-spacing version of the WRF initialized daily at 0000 UTC at NSSL (NSSL-WRF), as well as the 12-km grid-spacing NAM run operationally at NCEP/Environmental Modeling Center (EMC). For verification, drylines as depicted by 0000 UTC analyses of the Rapid Update Cycle (RUC) model were used. Using a manual identification procedure that considered the specific humidity gradient magnitude, temperature, and wind fields, 116 cases containing drylines in all three datasets (NSSL-WRF, NAM, and RUC) were identified for the period 1 April–30 June 2007–11.
For the NAM over all 116 cases, the mean dryline errors were very close to zero, indicating no systematic east–west biases in dryline placement, and most of the errors (~75%) fell within ±0.5° longitude. The lack of a systematic bias was generally evident across all subgroups of dryline cases.
In contrast, a significant eastward bias was found in the 24-h forecast mean dryline position of the NSSL-WRF. This eastward bias was consistent across all months, synoptic regimes, and years. However, the eastward bias was larger in April and May than in June, it was larger for trough cases than for ridge and zonal cases, and it was larger during 2010 and 2011 than in the other years. One thing the subgroups with larger eastward biases have in common is that they are all likely to be associated with a larger portion of “active” drylines, which tend to move much farther east than “quiescent” drylines (Hane 2004). Active drylines are typically linked with an eastward-moving synoptic-scale weather system that helps advect the dryline eastward in addition to vertical mixing processes, while quiescent dryline movement is dominated by processes associated with vertical mixing. Therefore, most trough cases in this study likely include drylines influenced by synoptic-scale weather systems. Both 2010 and 2011 were years with frequent passages of short-wave troughs across the central U.S. high plains. Climatologically, during June, the upper-tropospheric jet stream begins to move northward from the central to the northern U.S. high plains. Thus, mid- to upper-level short-wave troughs and the associated enhanced flow tend to influence drylines in the southern U.S. high plains less frequently during June than in April and May. So, it seems very plausible that the NSSL-WRF exaggerates the eastward movement of active drylines.
To diagnose the physical mechanism causing this eastward bias would require much more in-depth analyses of individual simulations, as well as high spatial and temporal resolution observational data within the boundary layer. However, because dryline movement occurs from boundary layer vertical mixing processes, in subsequent work it would be intuitive to examine aspects of the boundary layer parameterization schemes and how they influence dryline movement. Both the NAM and NSSL-WRF use the MYJ boundary layer parameterization, but given the different dynamics cores and grid spacings used for the NAM and NSSL-WRF, one would not necessarily expect the MYJ scheme to behave similarly in each of the models (e.g., Wyngaard 2004). For example, because the vertical mixing ahead of the dryline is directly related to dry-convective boundary layer circulations, the dryline in the higher-resolution NSSL-WRF will uniquely react to these circulation patterns that are not resolved in the NAM. This mesoscale turbulence in the NSSL-WRF should make a large contribution to the total vertical transport of heat and moisture in the boundary layer relative to parameterized subgrid-scale vertical fluxes (Pielke et al. 1991). Thus, it is possible that the MYJ, combined with resolved mesoscale fluxes, may be producing too much vertical mixing in the convection-allowing NSSL-WRF.
Clearly, the systematic eastward bias in 24-h dryline forecasts in the NSSL-WRF will impact the model-forecasted location of storms that initiate from dryline-related processes. Thus, an awareness of this bias should be valuable information for severe weather forecasters, and should be of use to model developers—especially, researchers working on boundary layer parameterization. Another important impact of this study to operational forecasters is the influence dryline position has on fire weather parameters. Given the abrupt moisture gradient along active drylines, any position errors would be highly relevant to fire weather forecasting because dry conditions and high winds, which can create a dangerous fire weather environment, are consequences of dryline passages. Finally, the methodology developed to identify drylines worked quite well, but was very labor intensive. Future work is recommended to develop automated methods for identifying, tracking, and visualizing drylines in high-resolution model data. With the increasing use of convection-allowing ensemble systems, automated approaches for identifying phenomena like drylines will become increasingly valuable.
Acknowledgments
Much of this research was completed by the first three authors as part of an undergraduate Capstone course within the University of Oklahoma's School of Meteorology. AJC was supported by the NOAA/Office of Oceanic and Atmospheric Research under NOAA–University of Oklahoma Cooperative Agreement NA17RJ1227, U.S. Department of Commerce. This paper benefitted from discussions with Conrad Ziegler, Mike Coniglio, Stephen Goss, Jeremy Grams, John Hart, and Matt Parker (as well as the rest of the Convective Storms Group at NCSU). We also thank the two anonymous reviewers for a thorough review of this paper, including many helpful and insightful comments.
REFERENCES
Beebe, R. G., 1958: An instability line development as observed by the tornado research airplane. J. Meteor., 15, 278–282.
Benjamin, S. G., and Coauthors, 2004a: An hourly assimilation–forecast cycle: The RUC. Mon. Wea. Rev., 132, 495–518.
Benjamin, S. G., Grell G. A. , Brown J. M. , Smirnova T. G. , and Bleck R. , 2004b: Mesoscale weather prediction with the RUC hybrid isentropic–terrain-following coordinate model. Mon. Wea. Rev., 132, 473–494.
Betts, A. K., 1986: A new convective adjustment scheme. Part I: Observational and theoretical basis. Quart. J. Roy. Meteor. Soc., 112, 677–691.
Betts, A. K., and Miller M. J. , 1986: A new convective adjustment scheme. Part II: Single column tests using GATE wave, BOMEX, ATEX, and Arctic air-mass data sets. Quart. J. Roy. Meteor. Soc., 112, 693–709.
Bluestein, H., 1993: Observations and Theory of Weather Systems. Vol. 2, Synoptic–Dynamic Meteorology in Midlatitudes, Oxford University Press, 594 pp.
Bothwell, P. D., Hart J. A. , and Thompson R. L. , 2002: An integrated three-dimensional objective analysis scheme in use at the Storm Prediction Center. Preprints, 21st Conf. on Severe Local Storms/19th Conf. on Weather Analysis and Forecasting/15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., JP3.1. [Available online at https://ams.confex.com/ams/pdfpapers/47482.pdf.]
Chen, F., and Dudhia J. , 2001: Coupling an advanced land-surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model description and implementation. Mon. Wea. Rev., 129, 569–585.
Clark, A. J., Gallus W. A. Jr., Xue M. , and Kong F. , 2009: A comparison of precipitation forecast skill between small convection-allowing and large convection-parameterizing ensembles. Wea. Forecasting, 24, 1121–1140.
Clark, A. J., Gallus W. A. Jr., Xue M. , and Kong F. , 2010: Growth of spread in convection-allowing and convection-parameterizing ensembles. Wea. Forecasting, 25, 594–612.
Clark, A. J., and Coauthors, 2011: Probabilistic precipitation forecast skill as a function of ensemble size and spatial scale in a convection-allowing ensemble. Mon. Wea. Rev., 139, 1410–1418.
Clark, A. J., and Coauthors, 2012: An overview of the 2010 Hazardous Weather Testbed Experimental Forecast Program Spring Experiment. Bull. Amer. Meteor. Soc., 93, 55–74.
Coniglio, M. C., Elmore K. L. , Kain J. S. , Weiss S. J. , Xue M. , and Weisman M. L. , 2010: Evaluation of WRF model output for severe-weather forecasting from the 2008 NOAA Hazardous Weather Testbed Spring Experiment. Wea. Forecasting, 25, 408–427.
Coniglio, M. C., Correia J. , Marsh P. T. , and Kong F. , 2013: Verification of convection-allowing WRF model forecasts of the planetary boundary layer using sounding observations. Wea. Forecasting, 28, 842–862.
Crawford, T. M., and Bluestein H. B. , 1997: Characteristics of dryline passage. Mon. Wea. Rev., 125, 463–477.
Dodd, A. V., 1965: Dewpoint distribution in the contiguous United States. Mon. Wea. Rev., 93, 113–122.
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.
Fels, S. B., and Schwarzkopf M. D. , 1975: The simplified exchange approximation: A new method for radiative transfer calculations. J. Atmos. Sci., 32, 1475–1488.
Ferrier, B. S., Jin Y. , Lin Y. , Black T. , Rogers E. , and DiMego G. , 2002: Implementation of a new grid-scale cloud and rainfall scheme in the NCEP Eta Model. Preprints, 19th Conf. on Weather Analysis and Forecasting/15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., 280–283.
Fujita, T. T., 1958: Structure and movement of a dry front. Bull. Amer. Meteor. Soc., 39, 574–582.
Geerts, B., 2008: Dryline characteristics near Lubbock, Texas, based on radar and West Texas Mesonet data for May 2005 and May 2006. Wea. Forecasting, 23, 392–406.
Hane, C. E., 2004: Quiescent and synoptically-active drylines: A comparison based upon case studies. Meteor. Atmos. Phys., 86, 195–211.
Hane, C. E., Ziegler C. L. , and Bluestein H. B. , 1993: Investigation of the dryline and convective storms initiated along the dryline: Field experiments during COPS-91. Bull. Amer. Meteor. Soc., 74, 2133–2145.
Hane, C. E., Baldwin M. E. , Bluestein H. B. , Crawford T. M. , and Rabin R. M. , 2001: A case study of severe storm development along a dryline within a synoptically active environment. Part I: Dryline motion and an Eta Model forecast. Mon. Wea. Rev., 129, 2183–2204.
Hoch, J., and Markowski P. , 2005: A climatology of springtime dryline position in the U.S. Great Plains region. J. Climate, 18, 2132–2137.
Hong, S.-Y., and Lim J.-O. J. , 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129–151.
Hoskins, B. J., and Bretherton F. P. , 1972: Atmospheric frontogenesis models: Mathematical formulation and solution. J. Atmos. Sci., 29, 11–37.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945.
Janjić, Z. I., 2002: Nonsingular implementation of the Mellor–Yamada level 2.5 scheme in the NCEP Meso Model. NCEP Office Note 437, 61 pp.
Janjić, Z. I., 2003: A nonhydrostatic model based on a new approach. Meteor. Atmos. Phys., 82, 271–285.
Johnson, A., Wang X. , Kong F. , and Xue M. , 2011a: Hierarchical cluster analysis of a convection-allowing ensemble during the Hazardous Weather Testbed 2009 Spring Experiment. Part I: Development of the object-oriented cluster analysis method for precipitation fields. Mon. Wea. Rev., 139, 3673–3693.
Johnson, A., Wang X. , Xue M. , and Kong F. , 2011b: Hierarchical cluster analysis of a convection-allowing ensemble during the Hazardous Weather Testbed 2009 Spring Experiment. Part II: Ensemble clustering over the whole experiment period. Mon. Wea. Rev., 139, 3694–3710.
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.
Kain, J. S., and Coauthors, 2010a: Assessing advances in the assimilation of radar data and other mesoscale observations within a collaborative forecasting–research environment. Wea. Forecasting, 25, 1510–1521.
Kain, J. S., Dembek S. R. , Weiss S. J. , Case J. L. , Levit J. J. , and Sobash R. A. , 2010b: Extracting unique information from high-resolution forecast models: Monitoring selected fields and phenomena every time step. Wea. Forecasting, 25, 1536–1542.
Lacis, A. A., and Hansen J. E. , 1974: A parameterization for the absorption of solar radiation in the earth's atmosphere. J. Atmos. Sci., 31, 118–133.
Lean, H. W., Clark P. A. , Dixon M. , Roberts N. M. , Fitch A. , Forbes R. , and Halliwell C. , 2008: Characteristics of high-resolution versions of the Met Office Unified Model for forecasting convection over the United Kingdom. Mon. Wea. Rev., 136, 3408–3424.
Markowski, P. M., and Richardson Y. P. , 2010: Mesoscale Meteorology in Midlatitudes. Wiley-Blackwell, 407 pp.
Marsh, P. T., Kain J. S. , Lakshmanan V. , Clark A. J. , Hitchens N. M. , and Hardy J. , 2012: A method for calibrating deterministic forecasts of rare events. Wea. Forecasting, 27, 531–538.
McGuire, E. L., 1962: The vertical structure of three drylines as revealed by aircraft traverses. National Severe Storms Laboratory Rep. 7, 10 pp.
Mellor, G. L., and Yamada T. , 1982: Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys., 20, 851–875.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360.
Mlawer, E. J., Taubman S. J. , Brown P. D. , Iacono M. J. , and Clough S. A. , 1997: Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the long-wave. J. Geophys. Res., 102 (D14), 16 663–16 682.
Peterson, R. E., 1983: The west Texas dryline: Occurrence and behavior. Preprints, 13th Conf. on Severe Local Storms, Tulsa, OK, Amer. Meteor. Soc., J9–J11.
Pielke, R. A., Dalu G. A. , Snook J. S. , Lee T. J. , and Kittel T. G. F. , 1991: Nonlinear influence of mesoscale land use on weather and climate. J. Climate, 4, 1053–1069.
Pietrycha, A. E., and Rasmussen E. N. , 2001: Observations of the Great Plains dryline utilizing mobile mesonet data. Preprints, Ninth Conf. on Mesoscale Processes, Fort Lauderdale, FL, Amer. Meteor. Soc., P5.23. [Available online at https://ams.confex.com/ams/pdfpapers/22296.pdf.]
R Development Core Team, cited 2012: R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. [Available online at http://www.R-project.org.]
Rhea, J. O., 1966: A study of thunderstorm formation along dry lines. J. Appl. Meteor., 5, 59–63.
Richter, H., and Bosart L. F. , 2002: The suppression of deep moist convection near the southern Great Plains dryline. Mon. Wea. Rev., 130, 1665–1691.
Roberts, N. M., and Lean H. W. , 2008: Scale-selective verification of rainfall accumulations from high-resolution forecasts of convective events. Mon. Wea. Rev., 136, 78–97.
Rogers, E., and Coauthors, 2009: The NCEP North American Mesoscale modeling system: Recent changes and future plans. Preprints, 23rd Conf. on Weather Analysis and Forecasting/19th Conf. on Numerical Weather Prediction, Omaha, NE, Amer. Meteor. Soc., 2A.4. [Available online at https://ams.confex.com/ams/pdfpapers/154114.pdf.]
Schaefer, J. T., 1974a: The life cycle of the dryline. J. Appl. Meteor., 13, 444–449.
Schaefer, J. T., 1974b: A simulative model of dryline motion. J. Atmos. Sci., 31, 956–964.
Schaefer, J. T., 1986: The dryline. Mesoscale Meteorology and Forecasting, P. S. Ray, Ed., Amer. Meteor. Soc., 549–572.
Schultz, D. M., Weiss C. C. , and Hoffman P. M. , 2007: The synoptic regulation of dryline intensity. Mon. Wea. Rev., 135, 1699–1709.
Schwartz, C. S., and Coauthors, 2009: Next-day convection-allowing WRF model guidance: A second look at 2-km versus 4-km grid spacing. Mon. Wea. Rev., 137, 3351–3372.
Schwartz, C. S., and Coauthors, 2010: Toward improved convection-allowing ensembles: Model physics sensitivities and optimizing probabilistic guidance with small ensemble membership. Wea. Forecasting, 25, 263–280.
Schwarzkopf, M. D., and Fels S. B. , 1991: The simplified exchange method revisited—An accurate, rapid method for computation of infrared cooling rates and fluxes. J. Geophys. Res., 96 (D5), 9075–9096.
Skamarock, W. C., and Coauthors, 2008. A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3.pdf.]
Sobash, R. A., Kain J. S. , Bright D. R. , Dean A. R. , Coniglio M. C. , and Weiss S. J. , 2011: Probabilistic forecast guidance for severe thunderstorms based on the identification of extreme phenomena in convection-allowing model forecasts. Wea. Forecasting, 26, 714–728.
Sun, J., Trier S. B. , Xiao Q. , Weisman M. L. , Wang H. , Ying Z. , Xu M. , and Zhang Y. , 2012: Sensitivity of 0–12-h warm-season precipitation forecasts over the central United States to model initialization. Wea. Forecasting, 27, 832–855.
Wakimoto, R. M., Murphey H. V. , Browell E. V. , and Ismail S. , 2006: The “triple point” on 24 May 2002 during IHOP. Part I: Airborne Doppler and LASE analyses of the frontal boundaries and convection initiation. Mon. Wea. Rev., 134, 231–249.
Weisman, M. L., Davis C. , Wang W. , Manning K. W. , and Klemp J. B. , 2008: Experiences with 0–36-h explicit convective forecasts with the WRF-ARW model. Wea. Forecasting, 23, 407–437.
Weiss, C. C., and Bluestein H. B. , 2002: Airborne pseudo–dual Doppler analysis of a dryline–outflow boundary intersection. Mon. Wea. Rev., 130, 1207–1226.
Wolff, J. K., Ferrier B. S. , and Mass C. F. , 2012: Establishing closer collaboration to improve model physics for short-range forecasts. Bull. Amer. Meteor. Soc., 93, ES51–ES53.
Wu, W. S., Purser R. J. , and Parrish D. F. , 2002: Three-dimensional variational analysis with spatially inhomogeneous covariances. Mon. Wea. Rev., 130, 2905–2916.
Wyngaard, J. C., 2004: Toward numerical modeling in the “terra incognita.” J. Atmos. Sci., 61, 1816–1826.
Xue, M., and Martin W. J. , 2006: A high-resolution modeling study of the 24 May 2002 dryline case during IHOP. Part II: Horizontal convective rolls and convective initiation. Mon. Wea. Rev., 134, 172–191.
Ziegler, C. L., and Rasmussen E. N. , 1998: The initiation of moist convection at the dryline: Forecasting issues from a case study perspective. Wea. Forecasting, 13, 1106–1131.
Ziegler, C. L., Lee T. J. , and Pielke R. A. , 1997: Convective initiation at the dryline: A modeling study. Mon. Wea. Rev., 125, 1001–1026.
Online documentation of changes to operational models run by NCEP/Environmental Modeling Center can be found online (http://www.emc.ncep.noaa.gov/mmb/mmbpll/eric.html). In October 2011, after the period of study, NCEP implemented a major change to the operational NAM, switching the dynamics core from the Nonhydrostatic Mesoscale Model (NMM) to a new version now known as the Nonhydrostatic Multiscale Model on B-grid (NMM-B). Additional documentation can be found online as well (http://www.nws.noaa.gov/os/notification/tin11-16nam_changes_aad.htm).
Archived SPC severe weather outlooks can be retrieved online (http://www.spc.noaa.gov/products/outlook/).
