Forecasting Severe Convective Storms
Summary and Keywords
Forecasting severe convective weather remains one of the most challenging tasks facing operational meteorology today, especially in the mid-latitudes, where severe convective storms occur most frequently and with the greatest impact. The forecast difficulties reflect, in part, the many different atmospheric processes of which severe thunderstorms are a by-product. These processes occur over a wide range of spatial and temporal scales, some of which are poorly understood and/or are inadequately sampled by observational networks. Therefore, anticipating the development and evolution of severe thunderstorms will likely remain an integral part of national and local forecasting efforts well into the future.
Modern severe weather forecasting began in the 1940s, primarily employing the pattern recognition approach throughout the 1950s and 1960s. Substantial changes in forecast approaches did not come until much later, however, beginning in the 1980s. By the start of the new millennium, significant advances in the understanding of the physical mechanisms responsible for severe weather enabled forecasts of greater spatial and temporal detail. At the same time, technological advances made available model thermodynamic and wind profiles that supported probabilistic forecasts of severe weather threats.
This article provides an updated overview of operational severe local storm forecasting, with emphasis on present-day understanding of the mesoscale processes responsible for severe convective storms, and the application of recent technological developments that have revolutionized some aspects of severe weather forecasting. The presentation, nevertheless, notes that increased understanding and enhanced computer sophistication are not a substitute for careful diagnosis of the current meteorological environment and an ingredients-based approach to anticipating changes in that environment; these techniques remain foundational to successful forecasts of tornadoes, large hail, damaging wind, and flash flooding.
Forecasting severe convective storms remains one of the most challenging tasks facing operational meteorology today. This is especially true in the mid-latitudes, where severe thunderstorms occur with greatest frequency and impact. The forecast difficulties reflect, in part, the many different atmospheric processes of which severe thunderstorms are a by-product. These processes, some of which are poorly understood and/or are inadequately sampled by observational networks, occur over a wide range of spatial and temporal scales. Anticipating the development and evolution of severe thunderstorms, therefore, likely will remain central to meteorological research and forecast efforts well into the future.
This article reviews present-day approaches to the forecasting of severe local storms, updating earlier summaries by House (1963), Johns and Doswell (1992), Doswell et al. (1993), and Moller (2001). Given the excellent background material provided in the above works, only limited space is devoted to basic forecast techniques; instead, emphasis is placed on changes have occurred in the forecast process over the last two decades, spurred by supporting research and technological developments.
In agreement with the criteria presently used by the U.S. National Weather Service (NWS), a severe thunderstorm is defined as a weather event that produces one or more of the following:
2. wind gusts over 26 ms−1 (58 mph)
3. surface hail over 2.54 cm (1.00 inch) diameter
Because most flash flood events are produced by thunderstorms (Davis, 2001) and because, on average, flash flooding causes more fatalities than any other by-product of convective storms (Doswell et al., 1996), severe weather also includes excessive convective rainfall. At the same time, although lightning is a potentially costly and lethal by-product of convective storms, at present no widely accepted techniques exist for assessing the likelihood for electrically active storms. Thus, lightning production and frequency are not addressed here.
Following a brief historical overview of severe weather forecast techniques in section 2, subsequent sections present, respectively, current practices and difficulties associated with the forecasting of tornadoes, damaging wind, hail, and excessive rain. The presentation concludes with a short discussion and suggestions for future improvement.
Although attempts to forecast severe convective storms likely were made on an individual basis as far back as antiquity, the first organized efforts in the United States to anticipate and better understand severe thunderstorms were made in the late 19th century by John P. Finley and Gustavus D. Hinrichs. Finley’s and Hinrichs’s studies (Finley, 1884, 1888; Hinrichs, 1888) focused on tornadoes and widespread convective wind storms (derechos), respectively. Though not widely appreciated at the time (e.g., Bradford, 1999), their efforts provided a foundation for the more sophisticated forecast studies that followed during the last half of the twentieth century. The first part of the 20th century, in contrast, yielded only occasional work on severe convection in the United States and Europe (see, e.g., Ludlam, 1963, pp. 4–5). These sporadic efforts provided rudimentary insight into the mechanics of thunderstorms, but at best they only indirectly benefited forecasting.
The “modern era” of severe weather forecasting did not begin until the 1940s, largely in response to organized efforts to assist the military in dealing with the hazards posed by severe convective storms to aviation in World War II. At that time, and at the bequest of the War Advisory Council on Meteorology, U.S. Weather Bureau meteorologists Albert K. Showalter and Joseph R. Fulks investigated the surface and upper-air thermodynamic environment of tornadoes in the United States (Showalter & Fulks, 1943). This study, together with another by Joseph R. Lloyd (1942), served as the primary basis for the world’s first official—and successful—tornado forecast, which was made by U.S. Air Force officers Ernest J. Fawbush and Robert C. Miller for Tinker Air Force Base, Oklahoma, on March 25, 1948 (Maddox & Crisp, 1999).
In response to both the somewhat fortuitous success of Fawbush and Miller’s initial tornado forecast and similar demonstrations of forecast ability in subsequent years, the Air Force established a Severe Weather Warning Center at Tinker in 1951. The Weather Bureau followed with the creation of its Severe Weather Unit (predecessor of today’s Storm Prediction Center or SPC) in Washington, D.C., the following year (Corfidi, 1999). At the time these forecast operations were established, little was known about the physical relationship between severe weather-producing thunderstorms and their environment. Although numerous statistical, climatological, and observational (case) studies appeared during the remainder of the 1950s and in the 1960s (e.g., see references in Schaefer, 1986), forecasting during this period remained largely empirical, based on correlations and inferences between observed large-scale pressure, wind, moisture, and frontal patterns and the occurrence of tornadoes, wind, or hail (e.g., Doswell et al., 1993). This pattern recognition approach involved the preparation of “composite charts” that simultaneously displayed various meteorological parameters at different altitudes to obtain a three-dimensional view of the atmosphere, as described in Miller’s Air Force Technical Report 200 (1972). Although pattern recognition has limitations as a forecast tool relative to the methods more generally used today, pattern recognition was used with considerable success for many years and remains an important component of severe weather forecasting, especially at extended time ranges.
Focus on mesometeorology1 increased in the 1970s, with the science benefiting from numerous observational (mainly radar) and numerical studies that provided enhanced understanding of the smaller-scale processes responsible for severe local storms. Although these efforts fostered a modest shift toward more sophisticated forecast techniques, another decade would pass—a period that witnessed a burgeoning of applied and theoretical research on severe local storms and their environments—before more substantial changes in forecast approach occurred. By 1990, knowledge of the physical processes responsible for severe storm development and maintenance, and the role played by the interaction of storms with their environment, had increased to the point that forecasting achieved a more firmly scientific foundation. Corresponding improvements also occurred in the resolution, physics, and skill of the numerical model guidance that forms the primary basis of all but the shortest-term forecasts, and in the ability to display model output via increasingly sophisticated workstations. As a result, by the start of the new millennium, severe weather forecasts had begun to exhibit greater spatial and temporal detail; in particular, the availability of model thermodynamic and wind profiles for the first time allowed probabilistic forecasts of the predominant storm type (e.g., isolated supercells vs. quasi-linear convection) and hazard (e.g., tornadoes vs. high wind). Further improvements in forecast accuracy and skill continued into the 2000s, largely in response to increased understanding of the roles of convective mode and storm interactions in determining severe weather intensity, longevity, and type (e.g., Smith et al., 2012). A more detailed retrospective on severe weather forecasting practices through the mid-1980s is provided by Schaefer (1986).
General Forecast Philosophy and Methodology
Weather forecasting on any scale consists of two primary steps: diagnosis of the present atmospheric environment and prognosis of its future state. Accurate diagnosis of the existing environment is essential in establishing a good foundation onto which expected changes—the prognosis—may be applied (e.g., Uccellini et al., 1992, Moller, 2001).
Severe convective weather is a product of thunderstorms2; therefore, the diagnostic part of severe weather forecasting requires careful analysis to assess the strength and distribution of the “ingredients” (Doswell et al., 1996) required for deep, moist convection. Because the features responsible for the development and propagation of thunderstorms occur on the mesoscale3 and because mesoscale processes, in turn, are governed by the large- or “synoptic”-scale setup, analysis necessarily begins with an overview of the synoptic-scale environment before focus turns to the mesoscale. To the three ingredients necessary for storms—moisture, instability, and lift (McNulty, 1978)—a fourth ingredient, vertical wind shear,4 is required to promote the sustained, organized activity responsible for most significant severe weather. Such convection often occurs in the form of supercells–thunderstorms characterized by a single, quasi-steady rotating updraft (known as a mesocyclone) that persists for a period of time much longer than it takes individual air parcels to move through them.5 As supercells produce a disproportionate share of the most intense tornado, damaging wind, and large hail events, much of the severe weather forecast process involves identification of those areas favorable for supercell development and sustenance.
Basic meteorological fields to be examined in the analysis stage of forecasting include conventional surface and upper-air (rawinsonde)6 data. Surface observations are particularly important because they provide the highest observation density in both space and time. Some observations, such as those provided by “mesonets” (National Academy of Sciences, 2009) and moving vehicles (e.g., passenger cars and trucking fleets; see Mahoney & O’Sullivan, 2013), are now available on a nearly continuous basis, adding to their utility. Surface pressure observations are especially important as they provide an integrated assessment of mass changes occurring through the entire atmospheric column. Careful examination of regional surface data, in conjunction with animated radar and satellite imagery, is required to create a reliable surface analysis (Figure 1). Such analyses, in turn, are the primary means by which low-level mesoscale sources of lift such as fronts, convective outflow boundaries, and dry lines may be identified. These features typically are necessary to raise parcels to their levels of free convection (LFCs)7 in an organized, sustained manner, fostering longer-lived storms. Whereas surface data are routinely analyzed by automated routines for data assimilation in numerical forecast models, forecast accuracy and precision are best served by manual analysis of surface data (Uccellini et al., 1992). Especially when combined with radar and satellite data (Figure 1)—and when conducted on a regular (e.g., hourly) basis, manual mesoanalysis—the analysis of meteorological fields at the mesoscale—fosters four-dimensional visualization of how the severe weather forcing mechanisms identified are likely to evolve over space and time. A complete discussion on the application of meteorological analysis to forecasting is beyond the scope of this work, but the role of skillful thermodynamic and kinematic analysis in the forecast process cannot be overstated (e.g., see Crisp, 1979; Doswell, 1986; McGinley, 1986; Uccellini et al., 1992; Sanders & Doswell, 1995).
Rawinsonde observations also are very important, despite the fact that such data generally are available only twice daily and only at widely spaced sites (approximate average spacing of 400 km in the continental United States). Along with surface analyses, vertical profiles of thermodynamic variables and winds provided by rawinsondes and, to a limited extent, those provided by the vertically pointing radars at weather surveillance radar sites, are very helpful in discerning the predominant storm type (i.e., multicell or supercell) and mode (e.g., individual cells or linear) likely to be encountered in a severe weather event. Wind observations at a given location typically are displayed in the form of a hodograph, a polar coordinate diagram showing the vertical distribution of the horizontal wind. A hodograph and the corresponding wind profile made from a rawinsonde ascent near an ongoing tornado outbreak are shown in Figure 3. In recent years, wind and thermodynamic profiles provided by commercial aircraft during ascents and descents (Petersen, 2016) have augmented the near-term data available to severe storm forecasters. Such data have proved useful in assessing inversion (or “cap”) strength in regions characterized by elevated layers of warm air that can inhibit the development of deep convection and in providing updates between conventional rawinsonde ascents.
Although remotely sensed data from radar, satellite, and lightning-observing systems are used primarily as nowcasting8 tools, these data also supply important analysis information, especially in sparsely populated areas. When regionally mosaicked and concatenated into long animations, radar data are particularly useful in anticipating future storm coverage and organization. Animated satellite imagery provides valuable corroborating information for the placement of surface boundaries and, through interpretation of cloud structure and form, also provides subjective information regarding low-level buoyancy and inhibition. Derived fields from radar (e.g., dual-polarization fields; Kumjian, 2013) and satellite (e.g., satellite-derived precipitable water; Schmit et al., 2005) have become more important in recent years in fine-tuning analyses; such data likely will gain further value in the future as technological advances in both radar and satellite continue.
Once a thorough analysis of the current atmospheric environment has been completed, the forecast task then shifts to assessing spatial and temporal trends in the various ingredients associated with sustained, organized thunderstorms, so that the likelihood for their severe weather by-products—tornadoes, high wind, hail, and flash flooding—may be determined. Of the necessary ingredients for storm initiation: moisture, instability, and lift, lift by far is the most difficult of the three to quantify for analysis and forecasting. Numerous numerical and observational studies over the years have yielded a relatively good understanding of the background synoptic and mesoscale thermodynamic environments conducive to the formation of intense convective storms (see reviews by Doswell & Bosart, 2001; Johnson & Mapes, 2001). In particular, field studies and numerical work over the central United States have shed considerable light on the sensitivity of storm development to low-level moisture availability and to the presence of elevated mixed layers extending downstream from regions of elevated terrain. Application of this work and the ability to examine thermodynamic sounding data on interactive workstations have contributed to a significant increase in the ability to forecast storm initiation and intensity over the past two decades; as a result, completely unexpected severe weather outbreaks are now rare.
On the other hand, it remains comparatively difficult to operationally assess the strength and three-dimensional nature of the many mesoscale forcing mechanisms collectively known as “lift.” For example, while the progress of a cold front responsible for storm development might be reasonably well tracked in mesonet data, sampling of the associated frontal circulation features above the immediate surface remains inadequate to make more than qualitative statements regarding the favorability (or lack thereof) of the feature for providing the lift necessary for storm sustenance. In addition, understanding of many meso- and storm-scale processes associated with convective development—and their subsequent interaction with convection and the surrounding environment once storms have formed—remains incomplete. Factors such as these appear likely to limit forecast accuracy and precision (and result in some degree of overforecasting) until the structure of the lower tropospheric environment is more substantially resolved by data platforms of the future.
Tornadoes may be classified broadly into two groups: those directly associated with the mesocyclone of a supercell thunderstorm, known as “supercell (or mesocyclonic) tornadoes,” and those that occur elsewhere, commonly referred to as “non-supercell (or nonmesocyclonic) tornadoes.” Supercells can also produce occasional nonmesocyclonic tornadoes. Mesocyclone-associated tornadoes pose the greatest threat to lives and property, as they are the strongest and longest-lived tornadoes.
Supercells, like all thunderstorms, require a minimum amount of buoyancy, moisture, and lift for development. The deep and persistent mesocyclones associated with tornadic supercells further require the presence of strong low- to mid-level9 environmental shear, that is, appreciable change in the ambient wind speed and/or direction with height. Because most strong tornadoes are spawned by supercell mesocyclones, to a first order, tornado forecasting may be reduced to identification of those areas likely to experience enhanced low- to mid-level shear in the presence of sufficient buoyancy, moisture, and lift to support deep, moist convection. Various parameters (e.g., Hart & Cohen, 2016) based on combinations of the above ingredients have been used to delineate those areas subject to supercell storms. Typically, however, the areas conditionally favorable for long-lived mesocyclones are much larger than those that actually experience them, and most supercells do not produce tornadoes. To fine-tune forecasts, close examination of the storm-scale environment and consideration of expected storm motion vectors relative to low-level thermal and wind discontinuities is required.
Vertical wind shear implies the presence of horizontal vorticity—the tendency for local rotation about a horizontal axis. This is illustrated schematically by the thin black lines and yellow arrows in Figure 2. As shown by Davies-Jones (1984) and Davies-Jones et al. (2001), it is the tilting of ambient horizontal vorticity by storm updrafts in a strongly sheared environment that ultimately can yield mid-level rotation—that is, a mid-level mesocyclone—in a storm (red and blue arrows in Figure 2). Streamwise vorticity is that part of the environmental horizontal vorticity vector that is oriented parallel to a storm’s inflow. Environments rich in streamwise vorticity are characterized by strong lower-tropospheric speed and directional shear, with corresponding hodographs that display a strongly curved, characteristic “hook” shape (e.g., Figure 3). Modeling efforts (Davies-Jones, 1984; Klemp, 1987) and operational experience have shown that such environments are favorable for strong, long-lived supercells.
Objective analyses and forecasts of streamwise vorticity have been used since the mid-1990s to delineate geographical areas with wind profiles potentially favorable for mesocyclogenesis, demonstrating one of the more notable applications of interactive workstations in the operational meteorological community. The parameter most often used to quantify streamwise vorticity and, in particular, the favorability of the low-level shear environment for tornadoes, is storm-relative helicity (SRH). SRH is the vertical integral of the product of streamwise vorticity with the storm-relative flow in the storm’s inflow layer; on a hodograph, SRH is directly proportional to the area between the environmental wind curve and the storm motion vector (Figure 3). SRH requires the selection of a storm motion vector, either observed or forecast, to assess the magnitude of storm-relative streamwise inflow. Selection of a representative storm motion vector—and the true or “effective” layer of air feeding the updraft (Thompson et al., 2007)—is critical to accurate estimation of the SRH being ingested by a storm and its propensity to produce tornadoes. While radar and improved storm-scale numerical model output have, to some extent, mitigated the problem of anticipating correct storm motion, coarse sampling of the near-storm environment (particularly as applied to determination of storm-inflow layers) and incomplete understanding of tornadogenesis necessarily place an upper bound on the skill of forecasts based on parameters like SRH.
The literature is replete with suggested SRH thresholds for forecasting various degrees of tornado likelihood and strength (e.g., Thompson et al., 2003). Although tornado strength is positively correlated with SRH, significant events occur over a wide range of SRH values, and intense tornadoes occasionally occur with modest SRH. At the same time, sustained thunderstorms sometimes move through zones of high SRH without producing a tornado. These observations reflect, in part, the complex interplay of forcing mechanisms on various spatial and temporal scales—not to mention the influence of thermodynamic variables and other factors—that govern tornado formation and sustenance. As a consequence, strict adherence to published SRH threshold values for forecasting purposes generally is avoided. Nevertheless, recent studies (e.g., Hart & Cohen, 2016) continue to support the overall forecast utility of SRH, particularly when SRH is calculated mindful of the actual inflow layer feeding the storms. In particular, the likelihood for tornadoes of any intensity in an otherwise favorable severe storm environment increases when low-level (0–1 km) SRH exceeds values of approximately 150 m2 s−2. In addition, SRH is useful in delineating corridors of enhanced low-level shear (such as those that exist near diffuse fronts, old outflow boundaries, storm anvil shadows, and other subtle surface discontinuities) that may promote tornado development within an existing nontornadic mesocyclone as it crosses the region and is subject to enhanced vertical stretching. For this reason, the spatial geometry of the SRH field often is as important as SRH magnitude in discerning tornado potential.
To this point, little has been said regarding the thermodynamic aspects of tornado forecasting. This is not meant to imply that thermodynamic factors play a secondary role to kinematic ones; obviously, mesocyclones and tornadoes owe their existence to the parent thunderstorms and strong updrafts that spawn them. Updrafts, in turn, are a manifestation of the release of convective instability, with updraft strength augmented by the latent heat of condensation and by the presence of vertical perturbation pressure gradients that arise from an updraft’s interaction with the ambient vertical shear (see Figure 2 and Rotunno & Klemp, 1982). In areas where deep convection has been inhibited by the presence of warm layers aloft that have “capped” storm development, substantial amounts of convective instability may amass over time. Given a sufficiently focused and persistent source of lift, rapid and intense storm development may occur if the cap is breached. As the application of parcel theory, convective instability, and convective inhibition (“CIN”) to severe weather forecasting are well-documented elsewhere (e.g., Moller, 2001) and since their application has changed very little in the last quarter century, these topics are not discussed further.
Certain thermodynamic considerations that in recent years have attained notable forecast utility are worthy of mention. First, there is the vital role of elevated mixed layers (EMLs)—plumes of deeply mixed boundary layer air that originate over arid, elevated regions and subsequently move downstream with the prevailing low- to mid-tropospheric flow to areas of lower terrain. By maximizing updraft strength and vertical stretching of streamwise vorticity in a strongly sheared environment, the steep lapse rates of EMLs promote mesocyclone intensity and longevity. In addition, inhibition or “capping” associated with warm air at the base of EMLs typically is responsible for the build-up of potential instability characteristic of many tornado outbreaks; surface-based convective available potential energy (CAPE) in such situations may exceed 4000 J/kg (Figure 4). While EMLs often play a prominent role in tornado outbreaks over the central United States (given that region’s location immediately downstream from the EML source region of the western United States), EMLs also have been shown to promote significant tornado and severe weather events elsewhere in the world (e.g., Carlson & Ludlam, 1968; Banacos & Ekster, 2010). In the past, the presence of a “dry punch aloft” often appeared in pattern recognition checklists (e.g., Miller, 1972) as a factor favoring tornado development, presumably because the feature was accompanied by a swath of minimal cloud cover and enhanced surface heating. In fact, it now appears that the “dry punch” served as a proxy for the detection of unrecognized EMLs (and their role in promoting storm intensity) before the significance of EMLs was fully appreciated.
Another thermodynamic variable with demonstrated forecast skill over the last two decades is the lifted condensation level (LCL). The lifting condensation level is the altitude at which a parcel of air lifted dry-adiabatically becomes saturated; it marks the base of cumuliform clouds in a well-mixed environment. Operational experience has confirmed the findings of Rasmussen and Blanchard (1998) that tornado incidence largely is inversely related to LCL; in particular, the risk for tornadoes diminishes markedly when LCLs are above 1200 m. Although questions remain regarding the exact mechanisms involved, lower boundary layer relative humidities associated with high LCLs enhance evaporative cooling in the sub-cloud layer. This typically has a deleterious impact on tornadogenesis, as mesocyclones are more likely to be undercut by storm outflow (Markowski et al., 2002). An alternative, equally plausible hypothesis is that low LCLs imply the presence of greater near-surface buoyancy (due to the presence of increased moisture). This provides greater potential for low-level stretching of streamwise vorticity (via the development of nonhydrostatic pressure perturbations) and subsequent strengthening of mesocyclonic circulations near the surface. Operational experience supports the latter hypothesis, as environments with both low LCLs and steep low-level lapse rates are known to be to be particularly productive for tornadoes when other factors are otherwise favorable.
It has long been known that tornado incidence is correlated positively with buoyancy; in fact, intense tornadoes occasionally occur in environments that otherwise are only marginally favorable for severe weather when thermodynamic instability is extreme (e.g., Jarrell, Texas, on May 27, 1997). But a growing list of significant cool-season tornado events (most recently, southwest Kansas and the eastern Texas Panhandle on November 16, 2015) serves as a reminder that only comparatively weak vertically integrated buoyancy—perhaps as little as 500–1000 J kg−1 CAPE—need be present to support strong tornadoes when the environmental shear is strong, and vertical stretching is concentrated in the most unstable part of the buoyant layer. Narrow zones of marginal instability often are not adequately resolved in objective analyses and forecasts, and the layer(s) through which maximum stretching is likely to occur can only roughly be estimated. In addition, the degree of lift necessary to realize convective initiation in strongly sheared, weakly buoyant environments also remains difficult to assess. For these reasons, forecasts of cool season tornado events tend to be somewhat less skillful than those made during the climatologically favored spring months and likely will remain so for some time to come.
Ingredients-based forecasting based on environmental sounding data provides undeniably useful information regarding tornado potential, especially supercell tornadoes. Nevertheless, the advent of regional radar animations and numerical simulations in the 1990s highlighted a significant limitation of the “moisture-instability-lift-shear” approach, namely, that storm mode and storm interactions also play important roles in determining the intensity, longevity, and scope of a severe weather event. “Mode” here is taken to be the predominant organizational structure assumed by the convection responsible for an event. The distinction typically made is that between discrete storms and quasi-linear (squall line) convection, but other considerations include isolated versus clustered storms, supercells versus multicells, and surface-based versus elevated convection.10 “Storm interaction” refers to the degree and character of interstorm behavior—for example, how two existing storms with differing motions might change after having merged with one another.
For various reasons, tornado development is favored with discrete storms, particularly isolated, discrete supercells. This largely reflects the fact that “destructive interference” between neighboring storms is minimized with widely separated convection, and that microphysically based, storm downdraft processes that foster tornadogenesis operate most effectively when storms remain independent rather than growing upscale into clusters or lines. An example of two diagnostically similar environments that posed strikingly different severe weather threats is shown in Figure 5. The first event was associated with significant tornadoes, including an F4 that left 7 dead and 125 injured in Tuscaloosa, Alabama; the other was associated with locally damaging wind. Although space does not permit comparison of the corresponding spatial analyses, arguably the most notable difference between the two events was that in the second case the initial convection quickly assumed a linear mode, while the storms remained primarily discrete supercells in the first.
The degree to which storm mode will govern tornado development (or other severe hazard potential) in an impending event is difficult to discern, given that the relevant factors involved are neither well sampled nor well understood. Likewise, the impact of an impending storm merger is very difficult to assess in real time. In addition, the processes governing dominant convective mode are numerous and are themselves a function of the manner by which existing convection has altered the environment. In recent years, numerical simulations have provided increased understanding of how storm mode affects tornadogenesis, and convection-allowing models and ensembles have seen increased operational use in anticipating storm mode and the likely effects of storm interactions (e.g., see Kain et al., 2010). Coupled with an ingredients-based approach to forecasting that is grounded in thorough analysis of the pre-storm environment, objective tools such as these should support future increases in forecast skill. But additional research and data collection platforms of greater resolution are necessary before significant advances in the reliability of forecasts for supercell tornado likelihood, intensity, and timing may be achieved.
Non-supercell tornadoes occur in a wide variety of synoptic- and mesoscale regimes, often with minimal vertical shear. Many waterspouts and most so-called landspouts11 are of this variety, forming in response to localized stretching of weak, existing vertical vorticity along boundaries (Figure 6). Operational experience suggests that the incidence of non-supercell tornadoes is correlated with the low-level lapse rate and an absence of lower-tropospheric temperature inversions. This most likely reflects the tendency for increased vertical stretching with increasing buoyancy. Slowly moving or stationary boundaries that separate light, antiparallel flow also appear to favor development, most likely because fast-moving boundaries or boundaries with directly opposed winds tend to undercut or otherwise disrupt any nascent vertical circulations along them. Given the absence of observations of the scale necessary to resolve the small precursor circulations responsible for waterspouts and landspouts, most such events go un-forecast. A greater degree of forecast skill, however, has been achieved for landspout tornadoes that form along terrain-induced boundaries. Such boundaries appear on a regular basis in various parts of the world (e.g., the “Denver convergence-vorticity zone”; see Szoke et al., 2006) whenever a supportive large-scale pattern is present. The boundaries can be the focus for deep, diurnally driven convection and sometimes tornadoes when favorable thermodynamic conditions are present along them.
Non-supercell tornadoes, however, also occur in strongly sheared environments along squall lines and other quasi-linear convective systems (QLCS); collectively these are known as QLCS tornadoes. Although most are comparatively weak and short-lived, QLCS tornadoes occasionally attain moderate (EF-2 or 3) intensity. The many mechanisms responsible for their development—most of which involve some degree of tilting and stretching of baroclinically generated horizontal vorticity along storm outflow boundaries (e.g., see Trapp & Weisman, 2003; Atkins & St. Laurent, 2009)—occur on small spatial (~ 1–10 km) and temporal (tens of minutes) scales. The parent squall lines often are associated with strong synoptic-scale forcing for ascent and weak buoyancy; in some cases, positive CAPE appears to be nonexistent in objective analyses. These factors make QLCS tornadoes not only difficult to forecast but also problematic for real-time identification by radar. A recently introduced forecast tool (Schaumann & Przybylinski, 2012) based on the strength and orientation of the 0–3 km bulk shear vector has shown some skill in identifying those parts of a squall line most likely to spawn tornadoes and other small-scale vortices in the near term (0–20 minutes). Forecasting the potential for such events at longer time ranges, however, must await the arrival of substantially improved convection-allowing models.
Bow Echoes and Derechos
Although isolated and short-lived thunderstorms often produce damaging surface winds, the most significant and widespread damaging wind events are produced by organized mesoscale convective systems (MCSs) that occur in environments of moderate to strong vertical shear and largely unidirectional flow. Such MCSs may assume a variety of forms during their life cycle, but all contain some degree of forward-propagation, wherein there exists rapid and persistent downshear (forward) development of new storm cells along the convective system’s elongating cold pool (Figure 7). This propagation often is manifested in the form of a curving, arc-shaped radar reflectivity pattern known as a bow echo (Fujita, 1978), with the most damaging surface winds occurring in concentrated pockets of storm downdrafts known as downbursts. New cell development in many cases is fostered by the presence of rear-inflow jets (Smull & Houze, 1987) oriented perpendicular to the direction of bow echo motion. Particularly long-lived and well-organized forward-propagating convective systems are known as derechos (Johns & Hirt, 1987; Corfidi et al., 2016).
Derechos account for a significant percentage of the casualties and damage associated with convectively induced, nontornadic winds, especially over the central and eastern United States (Ashley & Mote, 2005). A derecho consists of groups or “families” of downbursts produced by a long-lived, traveling MCS. Some derecho-producing convective systems contain multiple bow echoes of various scales; in turn, bow echoes themselves often contain smaller vortices that locally enhance the destructive winds (e.g., Trapp & Weisman, 2003; Atkins & St. Laurent, 2009). Operational experience suggests that derecho-producing convective systems are not nearly as common as individual bow-echo events.
Like mesocyclonic tornadoes, severe wind-producing convective systems and derechos in particular most often occur in regions of strong thermodynamic instability, especially in areas beneath persistent EMLs where substantial potential instability has amassed. However, when considering the likelihood for damaging surface winds, it is necessary to distinguish between instability associated with updraft (storm) development from that which promotes downdrafts.
Updraft instability is associated with positive CAPE; buoyancy arises from a positive density differential between an air parcel and its environment.12 In contrast, downdrafts primarily reflect the presence of negative buoyancy originating from the evaporative cooling of precipitation particles falling through unsaturated air. Further, downdrafts are augmented by precipitation drag, but precipitation loading weakens updrafts. Because the conditions for updraft development are not the same as those for downdrafts, the environment must be carefully examined for the presence of factors that could yield a conditional risk for strong downdrafts if storms do form.
The adjective “conditional” is significant, indicating that any convective instability present may or may not be released, depending on the strength, character, and duration of ascent available to bring parcels to their levels of free convection. As previously noted, of the essential ingredients for severe thunderstorms, “lift” is the most difficult to quantify. The lift involved in derecho development can be complex, dependent not only on storm-scale interactions, but also upon microphysical factors attendant to storm precipitation cascades and their associated downdrafts and gust fronts—factors that govern the location, frequency, and intensity of new storm development. The very small spatial and temporal scales of these processes and phenomena are well below those of the operational observing systems employed today, and knowledge of how they impact downdraft development and convective initiation remains incomplete. For these reasons, derecho prediction arguably is one of the most formidable challenges facing severe weather forecasters today.
Forecasts of organized severe wind-producing convective systems have nevertheless improved over the past two decades, bolstered by increased knowledge of the requisite background thermodynamic and kinematic environments (e.g., Johns & Hirt, 1987; Wakimoto, 2001; Coniglio et al., 2004), by documentation of events throughout the world (e.g., Takemi, 1999; Gatzen, 2004; Gatzen, 2013; Gospodinov et al., 2014) and by increasingly successful convection-allowing model efforts (e.g., Weisman et al., 2013). For example, a combination of moderate to seasonably strong and largely unidirectional flow (500 hPa wind speeds at or above 25 m s−1) with abundant moisture (mixing ratios ~ 15 g kg−1), steep low- to mid-level lapse rates, and a quasi-stationary source of low-level uplift (e.g., stationary front or outflow boundary) has been found to be conducive for “progressive”-type derechos—those that are largely internally driven and favor the summer months. In contrast, lesser buoyancy but faster flow in the presence of a strong, mobile zone of ascent (often in the form of a front associated with a mid-latitude shortwave trough) are more characteristic of the serial-type events that prevail during the cool season.
Although the maximum surface winds in a derecho rarely exceed the strength of a moderately intense tornado (i.e., ~ 55 m s−1), the area affected by a typical derecho far exceeds that of the widest, longest-track tornado. Cities and suburban areas are particularly vulnerable to derecho damage because of their dense concentration of overhead electric and communication circuits that may take weeks to rebuild after a storm. For such reasons, the collective economic impact of a derecho can rival or even exceed that of a tornado outbreak or hurricane, providing incentive for further study to improve forecasts.
Severe Winds in Low-Shear Environments
The greatest number of severe and damaging wind gusts that occur each year, especially in those parts of the world not subject to frequent EMLs and derechos, are produced by local “pulse”-type severe thunderstorms. Because they occur in weakly sheared environments, such storms do not achieve a significant degree of organization or longevity. Pulse-type storms are, however, the dominant, mid-latitude severe storm mode once the westerlies have retreated poleward in summer.
Storms that produce brief or localized severe gusts (referred to as microbursts) occur in both relatively dry and moist environments. Dry microburst storms are most common over arid or semiarid regions, where the combination of strong surface heating and limited moisture promotes the development of deep, well-mixed layers and high-based convection. Downdrafts arise from the evaporation and sublimation of precipitation falling from the clouds, especially when subcloud lapse rates approach dry adiabatic (9.8 °C km−1). Deep mixed layers encourage downward acceleration of air parcels as long as sufficient precipitation is available to maintain saturated descent. The characteristic “inverted V” thermodynamic-sounding profiles of such environments have proved useful in anticipating dry microburst events in recent decades (Wakimoto, 2001, pp. 288–289, provides a good review). Occasionally, dry microburst thermodynamic environments appear in the presence of moderate unidirectional wind fields. If a moving source of low-level ascent (e.g., a cold front or elongating thunderstorm cold pool) oriented at a considerable angle to the flow also is present, microburst-producing clouds occasionally organize into short lines or arcs, extending the damaging wind risk in space and time (e.g., Corfidi et al., 2016).
Microburst-producing pulse storms in moisture-rich environments have proven more difficult to forecast. Such storms occur on a nearly daily basis during summer over mid-latitude continental regions on the western and poleward flanks of subtropical high-pressure systems. They appear to occur most often when particularly warm, moisture-rich boundary layer air is overlain by seasonably dry air in mid-troposphere. However, other factors are also likely involved, as present forecast techniques based on the decrease in equivalent potential temperature (Θe) with height result in a considerable degree of overforecasting of these events. Precipitation loading, cloud microphysics, and small-scale variables governed by vertical shear (e.g., updraft tilt and organization) also may contribute.
Hail formation requires the presence of strong updrafts to support the weight of growing hailstones. For this reason, early techniques for forecasting hail size and occurrence were based on parcel theory estimates of updraft strength from rawinsonde data. These efforts achieved only modest success, as sounding data often were not representative of the immediate storm environment, and parcel theory does not account for the effects of water loading and entrainment. More significantly, the schemes also failed to account for the perturbation pressure gradients that arise in rotating updrafts; in some supercells, rotation is estimated to boost updraft strength by 50%. Therefore, until recently, hail forecasting has retained a more climatological and pattern-recognition flavor compared with forecasts for tornadoes and damaging wind.
The many variables involved in the hail formation and hail size account, in part, for the difficulties that have been encountered in developing more advanced forecast techniques. Besides updraft strength, hail size and occurrence also are functions of the in-cloud paths and transit times of hail embryos (e.g., Knight & Knight, 2001). These factors, in turn, are dependent on updraft breadth, slope, and persistence—aspects that, at best, can be only roughly estimated in real time. Size also is influenced by the degree of melting encountered by hailstones en route to the surface. Traditionally, the wet-bulb zero level (the altitude at which the wet-bulb potential temperature is 0 °C) has been used to estimate the degree of melting. But melting rate also involves hailstone size, the freezing level, and the number density of stones.
Large hail, occasionally exceeding severe limits (≥ 2.54 cm diameter), may be produced by multicell thunderstorms, especially those that form in environments with deep EMLs. Supercells, however, exclusively are responsible for the largest hail and the longest-lived hailstorms. Nevertheless, not all supercells produce damaging hail. For example, supercells associated with the mid-latitude remnants of tropical cyclones rarely produce hail, where buoyancy through the depth of the cloud is weak (“tall, skinny CAPE”). Very large hail (larger than 5.0 cm) is uncommon with supercells that occur in modest low- to mid-level lapse rates (e.g., ~7.5 °C km−1), and hail size typically decreases as storms of any type merge into lines or clusters.
Observational and modeling studies indicate that the complex balance of in-cloud conditions that yield the largest hailstones and most prolific hailstorms exist most reliably in long-lived, isolated supercells that occur in environments of very steep low- to mid-level lapse rates. For these reasons, hailstorm forecasting focuses on discerning those situations most likely to support sustained, discrete supercells in the presence of strong instability, and forecasts for very large hail closely parallel those for strong tornadoes. One environmental characteristic that has been used operationally to identify a heightened potential for large hail-producing storms is the presence of deeply unidirectional mid- and upper-level flow. Recent modeling work by Dennis and Kumjian (2017) has, in part, confirmed this observation, and also provided supporting evidence for the observation that the risk for very large hail somewhat diminishes as wind profiles exhibit an increasing directional component in the 0–3 km layer. Additional work, however, remains to be done to better understand these relationships and to quantify them for significantly improved forecasts.
Until recently, forecasts of hail size have been largely climatological, based on reported hail occurrences with similar past meteorological conditions. One new approach to forecasting hail size that recently has achieved some degree of success is the “Hailcast” model of Jewell and Brimelow (2009). “Hailcast” employs a one-dimensional coupled cloud and hail model to predict maximum hail size, given a representative observed sounding for input. The model’s approach is more physically realistic than that of traditional techniques as hailstones are allowed to grow upon successive transits through the cloud, rather than having size determined solely by environmental thermodynamic conditions. The model produces an ensemble of 25 members and accounts for storm mode using the observed bulk shear and instability profile. Bias-corrected ensemble mean output from Hailcast has been shown to provide fairly reliable forecasts over the continental United States. The model is designed for interactive use to maximize its utility. For example, forecaster modification of input soundings ensures that the most likely storm mode and ensemble members are used to generate maximum hail size; improper modification or incorrect storm modes could yield erroneous results. In this sense, Hailcast provides a welcome alternative to traditional ensemble approaches in which the forecaster passively accepts “black box” ensemble output.
Whereas increased understanding of supercell dynamics has improved forecasts of large hail, hail size alone is not a perfect metric of potential damage or human impact. For example, significant crop losses around the world are suffered each year from extensive falls of sub-severe (i.e., less than 2.54 cm diameter) hail. Damage from hail is a function of hailstone number density and the duration of hail fall, as well as the size of the largest hailstone. To obtain better estimates of hailstorm impacts governed by storm motion, various techniques have been devised to forecast supercell movement based on observed and model-generated wind profiles. One scheme that has come into widespread use since the early 2000s is that of Bunkers et al. (2000). The Bunkers technique breaks storm motion into two components: (1) advection of the updraft by a representative mean wind, and (2) propagation away from that direction (toward either the left or right) owing to dynamically induced pressure perturbations arising from the storm’s rotation. The Bunkers technique, for example, correctly forecasts the very slow motion of supercells that sometimes appear over the United States central High Plains when hodographs are strongly curved (reflecting substantial directional change), but the mean flow is comparatively weak. Because of their slow movement and longevity, such storms occasionally produce copious amounts of modest-sized but very damaging hail.
Flash flooding remains the leading cause of convective weather-related fatalities, with death rates that annually exceed those from tornadoes and lightning. Most flash floods are caused by excessive rainfall, but they can occur in other situations, such as dam breaks, tsunamis, and volcanic lahars. Such rains most often are produced by convective storms that persist over confined areas for extended periods of time. Heavy rainfall is strongly associated with convection because stratiform uplift generally cannot process water vapor rapidly enough to support the high rainfall rates characteristic of flash flood events. The amount of runoff from convective rainfall and the potential for flash flooding are, however, also strongly influenced by nonmeteorological factors such as terrain gradients, soil permeability, vegetative cover, and land use. These hydrologic considerations, though significant, are beyond the scope of this presentation; Davis (2001) provides a valuable review.
Slowly moving or quasi-stationary supercells such as those responsible for prolific falls of hail sometimes also produce flash flooding. Notable recent examples in the United States that simultaneously produced tornadoes, large hail, and flash flooding struck southern Nebraska on June 23, 2003 (see Figure 1), and the Oklahoma City area on May 31, 2013. Although life threatening and locally very destructive, such storms fortunately are rare and affect comparatively small areas. Most excessive rain episodes, in fact, are not direct products of supercells but rather are associated with groups of multicell storms that become organized into persistent lines and clusters in the presence of rich moisture inflow.
Persistent quasi-linear convective systems can arise from many different synoptic- and mesoscale meteorological setups (e.g., Maddox et al., 1979). All of these, however, share the common attribute that individual storm motions in the system are offset by the development of new cells, yielding minimal or zero net movement of the overall convective system. Systems that exhibit such behavior are said to be back-building. Back-building is shown schematically in Figure 8, where net MCS motion is seen as the sum of two components: (1) an advective one representing the movement of individual storms by the mean cloud-layer wind, and (2) a propagational component that accounts for the development of new cells relative to existing activity (Chappell, 1986; Corfidi, 2003).
Given the presence of a moisture-rich environment with sufficient buoyancy and large-scale forcing for ascent to support deep convection, many factors govern the rate and location of new cell development relative to existing storms. Of prime importance are the distributions of buoyancy, inhibition, and moisture on the flanks of the convective system. Other factors include the location, breadth, and strength of the low-level jet, the location and strength of system-generated cold pools and gust fronts, the presence of existing meso- and synoptic-scale boundaries, and orographic influences. While convection-allowing models provide clues regarding the preferred direction, rate, and intensity of storm propagation, careful analysis of the actual mesoscale environment in the vicinity of an MCS is necessary to pinpoint the smaller-scale corridors most likely to experience persistent back-building and, potentially, excessive rainfall.
Persistent back-building convection requires that the initiating mechanism—most often an outflow boundary or, in some cases, a terrain-induced feature—remain stationary or nearly so for an extended period. Boundaries are most likely to remain quasi-stationary when the thermal contrast across them is minimal. For this reason, deeply moist thermodynamic profiles are favorable for flash-flood producing MCSs not only because they provide a ready supply of moisture and maximize precipitation efficiency (by minimizing evaporation), but also because convective downdrafts are more likely to be comparatively feeble, minimizing gust front motion. Maintenance of quasi-stationary behavior also is promoted by the presence of deeply unidirectional, weakly sheared flow oriented parallel to the initiating mechanism. Such wind profiles enhance the potential for back-building and “echo-training” (movement of successive individual storm cells along the same path) because low-level inflow and new storm development are then directed anti-parallel to system motion (e.g., as seen in Figure 8).
Occasionally, on the rear flanks of convective systems responsible for damaging winds, a nearly stationary band of thunderstorms may form. If such a band persists for an extended period of time, and/or if the line forms over a region of impervious or steeply sloped terrain, flash flooding may follow. Systems of this nature are sometimes referred to as concurrent derecho-flash flood events; the disastrous flash flood of July 20, 1977, in Johnstown, Pennsylvania, was an event of this type. Paradoxically, the unidirectional wind profiles that are conducive to downshear (or forward) propagation in convective systems—that is, the repetitive, downshear development of new thunderstorms characteristic of derechos—also are favorable for repetitive storm development in the upshear direction. As mentioned in “Bow Echoes and Derechos,” when the direction of the environmental flow around a convective system varies little with height, the system’s cold pool over time necessarily elongates downstream in the direction of the mean flow. The gust front along the downshear side of the cold pool remains progressive, while that part of the gust front aligned parallel to the flow necessarily becomes quasi-stationary. This is shown schematically in Figure 9 for a case of unidirectional westerly winds. Given favorable thermodynamic conditions along the gust front, both the progressive and stationary parts of the boundary can become the seat of repetitive storm development, sometimes with devastating results.
Advances in mesoscale convection-allowing models undoubtedly will provide improved forecasts of flash floods in the future, as will field campaigns that address both the meteorological and hydrological aspects of flash flooding (e.g., HYMEX-SOPI; see Ducrocq et al., 2014). But as with tornadoes, damaging wind, and hail, limitations of data assimilation and initialization—and inadequate sampling—suggest that progress will be slow. Flash floods associated with terrain-related updraft forcing (e.g., the Big Thompson, Colorado event of July 31, 1976) have and likely will continue to pose particular forecast difficulty because of the many complex and small-scale processes that must adequately be resolved to properly assess their roles in fostering regenerative storm development.
This article has presented an overview of current practice in severe local storm forecasting. Indirectly, information also has been presented about present-day understanding of the mesoscale processes responsible for severe convective storms. Emphasis has been placed on diagnostic analysis of the meteorological environment and on an ingredients-based approach to the forecast process, rather than on the increasingly diverse array of convection-allowing numerical model output now available to the severe storm community.
Forecasting is an exercise of the human intellect. Without regular practice, skill diminishes. Slavish use of ensemble-based numerical guidance—a trend that has been increasing in some forecast centers—diminishes the important human role in the forecast process. Review of several notable forecast failures in the recent past reveals that overreliance on model output at the expense of careful interrogation of observational data arguably contributed to subpar performance. Although diagnostic analysis is important to forecasting on all time and space scales, detailed meteorological analysis is particularly crucial to skillful forecasts of the smaller-scale weather processes responsible for severe convective weather. Understanding of many of these processes is far from complete. Research ultimately will provide greater insight, as will the improved observational platforms of the future. Maximum public benefit, however, will be realized only from forecasts that remain grounded in careful analysis—and in an ingredients-based approach—because both readily lend themselves to incorporation of new concepts and understanding.
Ashley, W. S., & Mote, T. L. (2005). Derecho hazards in the United States. Bulletin of the American Meteorological Society, 86, 1577–1592.Find this resource:
Atkins, N. T., & St. Laurent, M. (2009). Bow echo mesovortices. Part II: Their genesis. Monthly Weather Review, 137, 1514–1532.Find this resource:
Banacos, P. C., & Ekster, M. L. (2010). The association of the elevated mixed layer with significant severe weather events in the northeastern United States. Weather and Forecasting, 25, 1082–1102.Find this resource:
Bradford, M. (1999). Historical roots of modern tornado forecasts and warnings. Weather and Forecasting, 14, 484–491.Find this resource:
Bunkers, M. J., Klimowski, B. A., Zeitler, J. W., Thompson, R. L., & Weisman, M. L. (2000). Predicting supercell motion using a new hodograph technique. Weather and Forecasting, 15, 61–79.Find this resource:
Carlson, T. N., & Ludlam, F. H. (1968). Conditions for the occurrence of severe local storms. Tellus, 20, 203–226.Find this resource:
Chappell, C. F. (1986). Quasi-stationary convective events. Mesoscale meteorology and forecasting. In P. Ray (Ed.), Mesoscale meteorology and forecasting (pp. 289–310). Boston: American Meteorological Society.Find this resource:
Coniglio, M. C., Stensrud, D. J., & Richman, M. D. (2004). An observational study of derecho-producing convective systems. Weather and Forecasting, 19, 320–337.Find this resource:
Corfidi, S. F. (1999). The birth and early years of the Storm Prediction Center. Weather and Forecasting, 14, 507–525.Find this resource:
Corfidi, S. F. (2003). Cold pools and MCS propagation: Forecasting the motion of downwind-developing MCSs. Weather and Forecasting, 18, 997–1017.Find this resource:
Corfidi, S. F., Corfidi, S. J., & Schultz, D. M. (2008). Elevated convection and castellanus: Ambiguities, significance, and questions. Weather and Forecasting, 23, 1280–1303.Find this resource:
Corfidi, S. F., Coniglio, M. C., Cohen, A. E., & Mead, C. M. (2016). A proposed revision to the definition of “derecho.” Bulletin of the American Meteorological Society, 97, 935–949.Find this resource:
Corfidi, S. F., Johns, R. H., & Darrow, M. A. (2016). The Great Basin Derecho of 31 May 1994. Weather and Forecasting, 31, 917–935.Find this resource:
Crisp, C. A. (1979). Training guide for severe weather forecasters. Technical Report 79–002, Air Weather Service, 37p.Find this resource:
Davies-Jones, S. P. (1984). Streamwise vorticity: The origin of updraft rotation in supercell storms. Journal of the Atmospheric Sciences, 41, 2991–3006.Find this resource:
Davies-Jones, S. P., Trapp, R. J., & Bluestein, H. B. (2001). Tornadoes and tornadic storms. Severe Convective Storm: Meteorological Monographs, No. 50, 157–221. Boston: American Meteorological Society.Find this resource:
Davis, R. S. (2001). Flash flood forecast and detection methods. Severe Convective Storms: Meteorological Monographs, No. 50 (pp. 481–525). Boston: American Meteorological Society.Find this resource:
Dennis, E. J., & Kumjian, M. R. (2017). The impact of vertical wind shear on hail growth in simulated supercells. Journal of the Atmospheric Sciences, 74, 641–663.Find this resource:
Doswell, C. A., III. (1986). Short-range forecasting. Mesoscale meteorology and forecasting (pp. 689–719). Boston: American Meteorological Society.Find this resource:
Doswell, C. A., III, & Bosart, L. F. (2001). Extratropical synoptic-scale processes and severe convection. Severe Convective Storms: Meteorological Monographs, No. 50 (pp. 27–69). Boston: American Meteorological Society.Find this resource:
Doswell, C. A., III, Brooks, H. E., & Maddox, R. A. (1996). Flash flood forecasting: An ingredients-based methodology. Weather and Forecasting, 11, 560–581.Find this resource:
Doswell, C. A., III, Weiss, S. J., & Johns, R. H. (1993). Tornado forecasting: A review. The tornado: Its structure, dynamics, prediction, and hazards. Geophysical Monographs, No. 79 (pp. 557–571). Washington, DC: American Geophysical Union.Find this resource:
Ducrocq, V., & co-authors. (2014). HYMEX-SOPI: The field campaign dedicated to heavy precipitation and flash flooding in the northwest Mediterranean. Bulletin of the American Meteorological Society, 95, 1083–1100.Find this resource:
Finley, J. P. (1884). Tornado predictions. American Meteorological Journal, 1, 85–88Find this resource:
Finley, J. P. (1888). Tornadoes. First-prize essay, American Meteorological Journal, 5, 165–179.Find this resource:
Fujita, T. T. (1978). Manual of downburst identification for project NIMROD. Satellite and Mesometeorology Research Paper 156, Department of Geophysical Sciences, University of Chicago, 104p [NTIS PB-2860481].Find this resource:
Gatzen, C. G. (2004). A derecho in Europe: Berlin, 10 July 2002. Weather and Forecasting, 19, 639–645.Find this resource:
Gatzen, C. G. (2013). Warm-season severe wind events in Germany. Atmospheric Research, 123, 197–205.Find this resource:
Gospodinov, I., Dimitrova, T., Bocheva, L., Simeonov, P., & Dimitrov, R. (2014). Derecho-like event in Bulgaria on 20 July 2011. Atmospheric Research, 158–159, 254–273.Find this resource:
Hart, J. A., & Cohen, A. E. (2016). The statistical severe convective risk assessment model. Weather and Forecasting, 31, 1697–1714.Find this resource:
Hinrichs, G. D. (1888). Tornadoes and derechos. American Meteorological Journal, 5, 306–317, 341–349. Available at http://www.spc.noaa.gov/misc/AbtDerechos/hinrichs/amj_hinrichs.pdf.Find this resource:
House, D. C. (1963). Forecasting tornadoes and severe thunderstorms. Severe Local Storms: Meteorological Monographs, No. 27 (pp. 141–155). Boston: American Meteorological Society.Find this resource:
Jewell, R., & Brimelow, J. (2009). Evaluation of Alberta hail growth model using severe hail proximity soundings from the United States. Weather and Forecasting, 24, 1592–1609.Find this resource:
Johns, R. H., & Doswell, C. A., III. (1992). Severe local storms forecasting. Weather and Forecasting, 7, 588–612.Find this resource:
Johns, R. H., & Hirt, W. D. (1987). Derechos: Widespread, convectively induced windstorms. Weather and Forecasting, 2, 32–49.Find this resource:
Johnson, R. H., & Mapes, B. E. (2001). Mesoscale processes and severe convective weather. Severe Convective Storm: Meteorological Monographs, No. 50 (pp. 71–122). Boston: American Meteorological Society.Find this resource:
Kain, J.S., Dembek, S. R., Weiss, S. J., Case, J. L., Levit, J. J., & Sobash, R. A. (2010). Extracting unique information from high-resolution forecast models: Monitoring selected fields and phenomena every time step. Weather and Forecasting, 125, 1536–1542.Find this resource:
Klemp, J. B. (1987). Dynamics of tornadic thunderstorms. Annual Review of Fluid Mechanics, 19, 369–402.Find this resource:
Knight, C. A., & Knight, N. C. (2001). Hailstorms. Severe Convective Storms: Meteorological Monographs, No. 50, 223–254. Boston: American Meteorological Society.Find this resource:
Kumjian, M. R. (2013). Principles and applications of dual-polarization weather radar. Part I: Description of the polarimetric radar variables. Journal of Operational Meteorology, 1(19), 226–242.Find this resource:
Kumjian, M. R. (2013). Principles and applications of dual-polarization weather radar. Part II: Warm-and cold-season applications. Journal of Operational Meteorology, 1(20), 243–264.Find this resource:
Lloyd, J. R. (1942). The development and trajectories of tornadoes. Monthly Weather Review, 70, 65–75.Find this resource:
Ludlam, F. H. (1963). Severe local storms: A review. Severe Local Storms: Meteorological Monographs, No. 27, 1–30. Boston: American Meteorological Society.Find this resource:
Maddox, R. A., Chappell, C. F., & Hoxit, L. R. (1979). Synoptic and meso-α aspects of flash flood events. Bulletin of the American Meteorological Society, 60, 115–123.Find this resource:
Maddox, R. A., & Crisp, C. A. (1999). The Tinker AFB tornadoes of March 1948. Weather and Forecasting, 14, 492–499.Find this resource:
Mahoney, W. P., III, & O’Sullivan, J. M. (2013). Realizing the potential of vehicle-based observations. Bulletin of the American Meteorological Society, 94, 1007–1018.Find this resource:
Markowski, P. M., Straka, J. M., & Rasmussen, E. N. (2002). Direct surface thermodynamic observations within the rear-flank downdrafts of nontornadic and tornadic supercells. Monthly Weather Review, 130, 1692–1721.Find this resource:
McGinley, J. (1986). Nowcasting mesoscale phenomena. Mesoscale Meteorology and Forecasting (pp. 647–688). Boston: American Meteorological Society.Find this resource:
McNulty, R. P. (1978). On upper tropospheric kinematics and severe weather occurrence. Monthly Weather Review, 106, 662–672.Find this resource:
Miller, R. C. (1972). Notes on analysis and severe storms forecasting procedures of the Air Force Global Weather Central. Technical Report 200 (Rev.), Air Weather Service, 181p.Find this resource:
Moller, A. R. (2001). Severe local storms forecasting. Severe Convective Storms, Meteorological Monographs, No. 50 (pp. 433–480). Boston: American Meteorological Society.Find this resource:
National Academy of Sciences. (2009). Observing weather and climate from the ground up: A nationwide network of networks. Washington, DC: National Academies Press.Find this resource:
Petersen, R. A. (2016). On the impacts and benefits of AMDAR observations in operational forecasting. Part 1: A Review of the impact of automated aircraft wind and temperature reports. Bulletin of the American Meteorological Society, 97, 585–602.Find this resource:
Rasmussen, E. N., & Blanchard, D. O. (1998). A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 1148–1164.Find this resource:
Rotunno, R., & Klemp, J. B. (1982). The influence of the shear-induced pressure gradient on thunderstorm motion. Monthly Weather Review, 110, 136–151.Find this resource:
Sanders, F., & Doswell, C. A., III. (1995). A case for detailed surface analysis. Bulletin of the American Meteorological Society, 76, 505–521.Find this resource:
Schaefer, J. S. (1986). Severe thunderstorm forecasting: A historical perspective. Weather Forecasting, 1, 164–189.Find this resource:
Schaumann, J. S., & Przybylinski, R. W. (2012). Operational application of 0–3 km bulk shear vectors in assessing QLCS Mesovortex and Tornado Potential. Preprints, 26th Conference on Severe Local Storms, Nashville, TN: American Meteorological Society, 142p.Find this resource:
Schmit, T. J., Gunshor, M. M., Menzel, W. P., Gurka, J. J., Li, J., & Bachmeier, A. S. (2005). Introducing the next-generation advanced baseline imager on GOES-R. Bulletin of the American Meteorological Society, 86, 1079–1096.Find this resource:
Showalter, A. K., & Fulks, J. R. (1943). Preliminary report on tornadoes. U.S. Weather Bureau, Washington, DC.Find this resource:
Smith, B. T., Thompson, R. L., Grams, J. S., & Broyles, C. (2012). Convective modes for significant severe thunderstorms in the contiguous United States. Part I: Storm classification and climatology. Weather and Forecasting, 27, 1114–1135.Find this resource:
Smull, B. F., & Houze, R. A. (1987). Rear inflow in squall lines with trailing stratiform precipitation. Monthly Weather Review, 115, 2869–2889.Find this resource:
Szoke, E. J., Barjenbruch, D. R., Glancy, R., & Kleyla, R. (2006). The Denver Cyclone and tornadoes 25 years later: The continued challenge of predicting non-supercell tornadoes. Preprints, 23rd Conference. on Severe Local Storms. St. Louis, MO: American Meteorological Society, p. 87.Find this resource:
Takemi, T. (1999). Structure and evolution of a severe squall line over the arid region in northwest China. Monthly Weather Review, 127, 1301–1309.Find this resource:
Thompson, R. L., Edwards, R., Hart, J. A., Elmore, K. L., & Markowski, P. (2003). Close proximity soundings within supercell environments obtained from the Rapid Update Cycle. Weather and Forecasting, 18, 1243–1261.Find this resource:
Thompson, R. L., Mead, C. M., & Edwards, R. (2007). Effective storm-relative helicity and bulk shear in supercell thunderstorm environments. Weather and Forecasting, 22, 102–115.Find this resource:
Trapp, R. J., & Weisman, M. L. (2003). Low-level vortices within squall lines and bow echoes. Part II: Their genesis and implications. Monthly Weather Review, 131, 2804–2823.Find this resource:
Uccellini, L. W., Corfidi, S. F., Junker, N. W., Kocin, P. J., & Olson, D. A. (1992). Report on the Surface Analysis Workshop held at the National Meteorological Center, March 25–28, 1991. Bulletin of the American Meteorological Society, 73, 459–472.Find this resource:
Wakimoto, R. M. (2001). Convectively driven high-wind events. Severe Convective Storms: Meteorological Monographs, No. 50 (pp. 255–298). Boston: American Meteorological Society.Find this resource:
Wakimoto, R. M., & Wilson, J. W. (1989). Non-supercell tornadoes. Monthly Weather Review, 117, 113–1140.Find this resource:
Weisman, M. L., Evans, C. E., & Bosart, L. (2013). The 8 May 2009 superderecho: Analysis of a real-time explicit convective forecast. Weather and Forecasting, 28, 863–892.Find this resource:
(1.) “Mesometeorology” is the branch of atmospheric science that addresses phenomena having horizontal scales commonly associated with fronts, thunderstorms, and tropical storms, in contrast to “synoptic-scale” meteorology that focuses on systems of greater length and width.
(2.) Strictly speaking, lightning and thunder are not required for the production of severe weather by convective storms. However, for simplicity, “thunderstorm” will be used interchangeably with the more precise phrase, “deep moist convection.”
(3.) “Mesoscale” here refers to meteorological features on the order of 2–1000 km in length or width; “synoptic scale,” those of 1000–5000 km in size.
(4.) Vertical wind shear is a measure of the change in wind direction and/or speed with height.
(5.) An alternative definition of a supercell is a long-lived, rotating storm with a high degree of spatial correlation between the storm’s updraft and mesocyclone.
(6.) A rawinsonde is an expendable instrument package, typically balloon-borne, that transmits vertical profiles of atmospheric variables back to the surface via a radio receiving system.
(7.) The level of free convection is the altitude at which a parcel of air lifted dry-adiabatically until saturation, and moist-adiabatically thereafter, becomes warmer than its environment in a conditionally unstable atmosphere.
(8.) A nowcast is a short-term weather forecast, generally for the next few hours.
(9.) Operationally, “low-level” generally refers to the lowest 1 km AGL, while “mid-level” is taken to be approximately 3–10 km AGL.
(10.) “Elevated” generally is taken to mean convection fed by air parcels located above the boundary layer, although in the case of supercells, this distinction is not clear-cut; see Corfidi et al. (2008) for further discussion.
(11.) “Landspout” is a colloquial term for a tornado whose vorticity arises within the boundary layer and whose parent cloud does not contain a mesocyclone.
(12.) More precisely, buoyancy reflects the static part of the unbalanced, vertical pressure gradient force on a parcel due to the presence of a density difference between the parcel and its environment.