Advances in numerical modeling and new observations are providing valuable information about turbulence near thunderstorms and are paving the way for the development of new turbulence avoidance and forecasting strategies for the aviation industry.
On 20 July 2010, a Boeing 777 en route from Washington, D.C. to Los Angeles encountered severe turbulence over Missouri; the aircraft was diverted to Denver to treat 17 passengers and 4 crew members who suffered injuries (National Transportation Safety Board 2010). This event provides an example of the possible consequences of unexpected turbulence encounters in the vicinity of convection, which can catch aircraft flight crews and passengers unprepared and increase the likelihood of turbulence-related injuries during flight. Although turbulence encounters with tens of injuries are uncommon, occurring on average only a few times per year, they underline the significant hazard that turbulence poses to the aviation industry. In addition to the hundreds of worldwide annual injuries, turbulence indirectly increases air travel costs because it is responsible for tens of millions of dollars in annual costs to airlines (Kauffmann and Sousa-Poza 2001; Eichenbaum 2003; Sharman et al. 2006). For these reasons, turbulence is avoided when possible using a combination of forecasts and en route tactical avoidance procedures. However, these methods are outdated and enhancements to them are stalled, partly because scientists do not fully understand the turbulence generation mechanisms. Nevertheless, significant progress is being made as a result of new turbulence observations and databases, and enhancements in numerical modeling capabilities. This article outlines this recent progress, with particular focus on turbulence near thunderstorms and provides insights for better turbulence avoidance strategies.
The specific purpose of this article is to 1) describe some of the recent findings about the dynamics underlying the generation of turbulence by thunderstorms, and identify the outstanding problems; 2) highlight the inadequacies in current methods for avoidance of thunderstorm-generated turbulence and motivate the development of new turbulence avoidance guidelines; and 3) demonstrate the capabilities of state-of-the-art forecast models that could be utilized for explicit turbulence predictions.
Of the many sources of turbulence that affect aviation (e.g., wind shear, jet streams, fronts, mountain waves, etc.), deep convective clouds are one of the most important. For example, Kaplan et al. (2005) examined 44 cases of severe turbulence and found that 86% of these were within 100 km of active deep convection. Wolff and Sharman (2008) have also illustrated the frequent occurrence of turbulence reports over regions known for convective activity (Fig. 1; see also “observing aircraft turbulence near clouds” section). The relative importance of turbulence near thunderstorms is, of course, affected by geography and the predominant meteorological conditions. For example, unlike for the United States, Kim and Chun (2011) found that over Korea only 11% of turbulence events, of moderate or greater (MOG) severity, were attributable to convection.
Convectively induced turbulence (CIT) is prevalent within convective clouds; convective updrafts, downdrafts, specific cloud features like mammatus (e.g., Schultz et al. 2006), and anvil regions are all known to be hazardous. For these turbulence processes, avoidance of cloudy air through visual identification or remote sensing with radar and satellite imagery is usually effective. Indeed, as illustrated by Fig. 2, preliminary analysis of the 20 July 2010 Missouri case identified that the turbulence encounter actually occurred directly above or just within a rapidly growing convective cell that was penetrating the anvil region of a large mesoscale convective system (MCS) from below. In this respect, the Missouri event is similar to some other events above convection described later.
Turbulence outside of cloud but ultimately caused by the cloud is also an important hazard that has been appreciated for some time (e.g., Burns et al. 1966; Prophet 1970; Keller et al. 1983; Pantley and Lester 1990). Although this near-cloud turbulence (NCT) is usually weaker than turbulence within convective cores, it is arguably more dangerous because it is invisible and undetectable by standard onboard or ground-based radar. Theories surrounding the origins of NCT are at best incomplete, relying on empirical evidence and pilot experience to determine the most hazardous regions. Nevertheless, the U.S. Federal Aviation Administration (FAA) has developed a series of guidelines for thunderstorm flying (Federal Aviation Administration 2012); the relevant guidelines are as follows:
Guideline 5. Do avoid by at least 20 miles [laterally] any thunderstorm identified as severe or giving an intense radar echo. This is especially true under the anvil of a large cumulonimbus.
Guideline 6. Do clear the top of a known or suspected severe thunderstorm by at least 1,000-ft altitude for each 10 knots of wind speed at the cloud top. This should exceed the altitude capability of most aircraft.
Both of these guidelines imply a region around thunderstorms that is hazardous. Yet, recent research has shown that turbulence that is primarily of convective origin can occur outside of those regions defined by the FAA guidelines, suggesting that the guidelines are inadequate. The guidelines do not reflect our current understanding of relevant NCT processes and should be updated.
In addition to these empirical guidelines, there is an important role for modern numerical weather prediction to assist in turbulence avoidance. State-of-the-art convection-permitting models can provide realistic representations of convective development, and we will show that they can reproduce regions of NCT with some accuracy in deterministic simulations. However, like moist convection, turbulence has low predictability and ultimately ensemble forecasts are required.
High-resolution numerical simulations are primarily responsible for the significant recent advances in the basic understanding of NCT dynamics. Such simulations are able to resolve the underlying generation processes and (if the grid spacing is small enough) explicitly resolve the scales of motion that affect aircraft.1 Our research has shown that convectively generated gravity waves (see the sidebar below) are an important cause of NCT and may ultimately be the central process responsible for turbulence remote from convection. Gravity waves can break down into turbulence but can also generate turbulence by perturbing environments that are already close to the threshold for turbulence production. The gravity waves work in concert with the cloud's broader circulations that destabilize the surrounding air by way of upper-level convective outflows and associated enhanced wind shear.
The remainder of this article describes new observational techniques and summarizes recent progress in NCT understanding using three observed and simulated examples. These three examples are chosen specifically to encompass a range of NCT locations: turbulence hundreds of kilometers from the convective portion of an MCS, tens of kilometers from an intense convective cell, and a few kilometers above a convective cloud. Gravity waves play a role in all three cases, and each of these identifies turbulence occurring beyond the minimum separation distances defined by the FAA guidelines. These examples also illustrate the current research and operational capabilities that should facilitate refinement of the FAA guidelines and the development of improved operational systems for turbulence forecasting.
OBSERVING AIRCRAFT TURBULENCE NEAR CLOUDS.
Pilots make routine reports of turbulence intensity in terms of the turbulence categories shown in Table 1. These PIREPs have been the traditional means for recording aircraft turbulence encounters, and lengthy archives of PIREPs are now available. Wolff and Sharman (2008) developed climatologies of the spatial occurrence of reported turbulence encounters over the contiguous United States and an updated example for reports associated with deep convection is provided in Fig. 1. Here, maxima in PIREP relative frequency exist over the Florida peninsula, over eastern Texas, and along the Gulf Coast, which are all regions known for frequent convective activity.
The four main turbulence categories reported by pilots along with descriptions of the corresponding aircraft and passenger responses. Also listed are the approximate atmospheric turbulence intensities (in terms of the cube root of the eddy dissipation rate, ϵ1/3) that would induce such responses in Boeing 737 and 757 aircraft in cruise [adapted from Federal Aviation Administration (2012) and Lester (1994)].
PIREPs are useful for deriving turbulence climatologies and are essential for routine tactical avoidance, but they are not research quality observations (e.g., Schwartz 1996; Sharman et al. 2006; Wolff and Sharman 2008). Uncertainties in PIREP location (median error ~ 50 km) makes them unacceptable for NCT studies, because it is almost impossible to unambiguously determine PIREP locations relative to cloud boundaries. For events associated with injuries, the flight data recorder (FDR) is often available for analysis, which circumvents some of these problems. Fortunately, most turbulence encounters do not cause injuries and then other observations are required to identify NCT episodes. NCT encounters can be identified using new in situ turbulence measurements (Cornman et al. 1995, 2004) now available from some U.S. commercial air carriers. These automated reports have horizontal and temporal accuracy of about 10 km and 1 min, respectively, and are therefore well suited for case and statistical studies of NCT. The in situ turbulence measurement and recording system provides reports of the cube root of the eddy dissipation rate (EDR; ϵ1/3 m2/3 s−1), an aircraft-independent atmospheric turbulence metric, including both the median and peak EDR encountered over one-minute time intervals during cruise. For NCT studies, the peak EDR value is preferable because it provides a good indication of the hazard to the aircraft and is better distributed over the turbulence severity reporting bins. At the time of writing this article, the EDR system is installed on all Delta Air Lines B737 and United Airlines B757 aircraft; future expansion to other carriers and aircraft types is anticipated [see also the discussion in Politovich et al. (2011) about its application to terrain-induced turbulence].
To gain a better understanding of the frequency of occurrence of NCT relative to radar-derived cloud boundaries, 7 million peak EDR reports are compared to cloud locations derived from Next Generation Weather Radar (NEXRAD) observations. These data confirm that turbulence is prevalent near convection. For example, 50% of all MOG (peak EDR > 0.3 m2/3 s−1) intensities occur within 100 km of convection, even though only 10% of all of the in situ EDR reports occur within this distance. Figure 3 depicts the distribution of MOG turbulence measurements as a function of horizontal and vertical distance to convection, defined as regions with vertically integrated liquid above 3.5 kg m−2 and echo tops above 4.6 km. The turbulence “relative risk” is defined by dividing the frequency of MOG turbulence within each of these distance bins by its overall frequency (from all sources) in the dataset (0.03%). The risk of turbulence encounters increases as the aircraft nears a thunderstorm (laterally), and the risk of MOG turbulence is almost twice the background value as far as 70 km (38 nmi) from a storm. Figure 3 also depicts the distribution of MOG turbulence encounters as a function of the aircraft's distance above the NEXRAD echo top in convective regions. Although the risk of turbulence decreases with distance above cloud, the relative risk of MOG turbulence is still 10 times the background value 3.6 km (~12,000 ft) above echo tops.
Of course, this analysis of the in situ aircraft data probably includes some turbulence encounters that are not caused by convection and events that are generated by a combination of sources. For example, it is known that long-lived organized convective systems (viz. MCSs) often occur near upper-tropospheric jet streams. Thunderstorms in mountainous regions could also generate turbulence via a combination of processes. Nonetheless, the in situ data (and the specific cases discussed later) undoubtedly highlight the hazard posed by thunderstorms.
TURBULENCE ADJACENT TO CLOUDS.
Combining automated EDR-based turbulence measurements with radar and satellite imagery has confirmed that a high frequency of NCT encounters are related to active thunderstorm regions. Two examples are presented in Fig. 4. On 17 June 2005, long-lasting and widespread turbulence occurred along the outer cirrus anvil of a MCS over the southern Great Plains of the United States (Fig. 4a), which was several hundred kilometers north of the MCS thunderstorm region (Trier and Sharman 2009). The 5 August 2005 case over the midwestern United States (Fig. 4b) consisted of a relatively isolated region of intense thunderstorms with moderate to severe turbulence 10–20 km southeast of the cloud shield. Numerical modeling has demonstrated that the two cases were likely caused by different NCT generation processes: namely, unstable upper-level storm outflows and ducted gravity waves.
17 June 2005: Unstable upper-level thunderstorm outflow.
The mesoscale environment of the 17 June 2005 turbulence encounter was investigated by Trier and Sharman (2009) using a convection-permitting Weather Research and Forecasting (WRF) model (Skamarock and Klemp 2008) simulation with a horizontal grid spacing of 3 km. The horizontal resolution of this simulation is greater than for most current operational models, but aircraft-scale turbulence is still not properly resolved and turbulence kinetic energy (TKE) is parameterized based on resolved-scale vertical shear and buoyancy productions (Janjić 1990, 1994). Simulated TKE in this case is most widespread within the MCS outflow several hundred kilometers north of the heavy rainfall region (Fig. 5), consistent with where the turbulence was recorded (Fig. 4a).
Regional gradient Richardson number Ri (see the sidebar “Gradient Richardson number”) variations across the MCS upper-level outflow occur in both the simulation and Rapid Update Cycle (RUC) model (Benjamin et al. 2004) analyses. Mesoscale zones of Ri ≤ 1 supportive of turbulence and parameterized TKE generation are prevalent on the north side of the MCS but not on its south side. These differences are related to the asymmetries in the upper-level wind shear, which itself is related to how the MCS-induced outflow superposes on the midlatitude background (westerly) flow (e.g., Fritsch and Maddox 1981).
GRADIENT RICHARDSON NUMBER
The gradient Richardson number is defined as Ri = N2/S2, where N is the Brunt–Väisälä frequency and S = |∂V/∂z| is the vertical shear magnitude. In cases where the air is saturated, N is replaced by Nm, which is the moist Brunt–Väisälä frequency (Durran and Klemp 1982). Turbulence can occur through dynamical instabilities when Ri falls below a critical value (~1 in three-dimensional flows). In a common type of dynamical instability known as Kelvin–Helmholtz instability, strong vertical shear drives the instability while the vertical stratification (N2 > 0) opposes the instability. In other situations, turbulence can arise from convective (N2 < 0) or mixed dynamic–convective instabilities.
Close inspection of the TKE north of the MCS convection (Fig. 5) reveals that the simulated turbulence, though widespread, is associated with distinct mesoscale events (see online supplemental animation of Fig. 5 at http://dx.doi.org/10.1175/BAMS-D-11-00062.2). Figure 6 depicts the onset of the later event along a vertical section oriented along the line (SW–NE) in Fig. 5. The parameterized TKE (Fig. 6) is shallow (1–2 km deep) and is primarily associated with the reduction in Ri associated with strong vertical shear and small moist static stability (implied by ∂θe/∂z → 0, where θe is the equivalent potential temperature)2 within the MCS upper-level outflow.
The reduction in stability is associated with the localized upward displacement of the θe isopleths (Fig. 6), which lags pulsations in the strength of the upstream MCS convection (Trier and Sharman 2009). Although the details of this response are not well understood, it may be associated with thermally (e.g., Pandya and Durran 1996) or mechanically (Fovell et al. 1992) forced mesoscale gravity waves. Meanwhile, the upwardly displaced θe isopleths, whose axis of maximum displacement is denoted by the thick dashed line in Fig. 6, are further steepened by differential advection on their downshear (northeast) side (Trier and Sharman 2009). That is, the sheared winds advect the θe perturbations to the northeast faster at higher altitudes (~13 km) than at lower altitudes (11–12 km). These processes within the storm outflow result in both the reduced stability and TKE generation near the outer edge of the anvil.
17 June 2005: Turbulent cirrus bands.
Further inspection of the satellite imagery for the 17 June 2005 event (Fig. 4a; 0732 UTC) reveals cloud bands extending radially outward from the MCS near its northern anvil edge. These radial bands become even more prevalent at later times (Fig. 7a; 0945 UTC) and coincide with the observations of turbulence. Turbulence is common in the vicinity of such bands associated with MCS anvils (Lenz et al. 2009) and tropical cyclone outflows and near atmospheric jet streams (Knox et al. 2010). These cloud structures are sometimes referred to as transverse bands, because they are often (though not always) oriented approximately perpendicular to jet stream winds. As discussed by Knox et al., the mechanisms responsible for these bands have remained elusive for some time.
Additional simulations of the 17 June event at even smaller horizontal grid spacing (600 m) resolve these bands (Fig. 7b; for full simulation details, see Trier et al. 2010). The simulated bands are located close to counterparts in the satellite observations (Fig. 7a) and to locations of in situ reports of turbulence (Fig. 7b). These radial bands of cold simulated infrared brightness temperature (Fig. 7b) develop in moist neutral or unstable conditions and are aligned approximately along the vertical shear vector through the depth of the anvil. Trier et al. (2010) showed that the bands are associated with shallow convection within the outer anvil and share organizational similarities with horizontal convective roll (HCR) circulations in the atmospheric boundary layer (e.g., LeMone 1973; Weckwerth et al. 1997) that often produce cumulus cloud streets.
The area of the moist static instability and the prominence of the radial cloud bands are increased by cloud radiative interactions (Trier et al. 2010). However, the regionalization of the banding appears primarily governed by the mesoscale thermodynamic destabilization mechanism discussed earlier, which manifests as the broad region of parameterized TKE within the anvil outflow in the coarser simulation.
Satellite observations show that these radial cloud bands are often aligned approximately perpendicular to high-frequency gravity waves emanating from upstream deep convection (Lenz et al. 2009). Trier et al. (2010) found a similar spatial relationship between these features in the 17 June observations (their Fig. 7) and simulations of this case (their Fig. 6). They further noted that strong vertical shear and enhanced static stability in the gravity wave region of their simulations is overlaid by nearly neutral conditions in the anvil, which supports trapping and horizontal propagation of the waves, and suggested that vertical displacements associated with these gravity waves might help excite the shallow radial convective bands in the anvil above.
The parameterized TKE in the high-resolution simulation (Fig. 7b) is much weaker and less widespread than in the coarser simulation (Fig. 5b), which is consistent with the vertical mixing in the higher-resolution simulation being largely controlled by the shallow convection resolved on the model grid. Though fine horizontal resolution is clearly required to resolve the turbulence-producing bands, the MCS-induced flows that influence the TKE parameterizations can be well simulated in coarser convection-permitting models. The new generation of operational regional models have horizontal grid spacings that are similar to the coarser-resolution simulation presented for this case (e.g., Fig. 5), suggesting they are capable of identifying portions of the MCS anvil outflow that are susceptible to turbulence through upper-level HCRs or other sources.
In this case, the timing and position of the simulated MCS showed very good agreement with the observations, which facilitated good agreement between the location of the simulated and observed turbulence. However, it is well known that forecasting the timing and organizational structure of deep convection is challenging because of the limits of predictability. Turbulence also suffers from low predictability and its prediction in this context relies inherently on the skill at convective scales. Thus, although convection-permitting models are sufficient to represent the physics of the important processes, ensemble predictions are inevitably required to capture the range of possible model outcomes.
5 August 2005: Ducted gravity waves.
At approximately 0240 UTC 5 August 2005, two commercial airliners encountered severe turbulence at cruising altitudes over northwest Indiana after having flown over or around a rapidly developing storm. According to contemporaneous satellite imagery, the planes were roughly 20 km away from any cloud having appreciable optical depth (Fig. 4b). Convection-permitting simulations of this case with the WRF model were performed by Fovell et al. (2007) and reveal that the storms provoke transient turbulence in localized areas tens of kilometers beyond the visible cloud: that is, at the margins of the horizontal separation defined by the FAA guidelines (number 5). The generation of this NCT is in part related to horizontally propagating gravity waves, which perturb an environment that is already marginally susceptible to turbulence production (i.e., small background Ri).
These simulations use 4-km horizontal grid spacing, have 100 vertical levels, and are initialized with the 0000 UTC 5 August RUC analysis.3 A (moist) “control” simulation is contrasted with a “dry” counterpart that does not permit water phase changes. The control simulation develops a relatively isolated storm at approximately the right time and location, whereas the dry simulation shows how the environment might evolve without convective influences.
During the encounters, the aircraft were flying at about 11.3–11.9 km above mean sea level (MSL), where in the control simulation significant gravity waves are emanating from the storm and propagating southeastward at ~30 m s−1 (Fig. 8). By 0230 UTC (Fig. 8a), a region of enhanced turbulence likelihood (Ri < 0.5) appears in the clear air just beyond the detectable anvil4 at approximately the same location relative to the storm as in the in situ turbulence measurements (Fig. 4b). The concurrent model vertical cross section (Fig. 8d), oriented parallel to the wave propagation vector shows that regions of locally low Ri occur within a shallow layer extending to the southeast. This layer of reduced Ri was also present in the dry simulation, although its Ri values never fell below one anywhere (Fig. 8g).
The control simulation's low Ri zone moves southeastward away from the storm, apparently phase locked with the propagating gravity waves (Figs. 8b,c; see also online supplemental animations of Fig. 8), and slowly erodes with time. The perturbations in velocity and stability associated with the waves have clearly modulated and reduced Ri (Figs. 8d–f), increasing the likelihood of turbulence in an already marginally susceptible environment (cf. Figs. 8g–i). This demonstrates that horizontally propagating gravity waves provide one mechanism by which a localized patch of turbulence can develop at a relatively large distance from an established cloud.
The control simulation's gravity wave activity adjacent to the storm arises from an excitation mechanism (unsteady convection) and a ducting layer that acts to retain wave energy in the upper troposphere. A vertical profile of the Scorer parameter (see “Gravity waves” sidebar) is calculated at 0230 UTC (Fig. 9a) near the anvil edge. The negative Scorer parameter values near 11- and 13.5-km altitudes provide conditions conducive to wave ducting in between. This duct is created mainly by the environment's jet-like wind profile (not shown). The wave duct is also present in the dry run, but no wave activity occurred because of the lack of an excitation mechanism. The ducting mechanism can provide a directional bias to the wave propagation and, as appears to be the case here, may cause certain regions around the storm to be more prone to turbulence.
GRAVITY WAVES
Gravity waves are oscillations in velocity, temperature, pressure (and other scalars) that originate from vertical displacements of stable air. Buoyancy is the restoring force of the oscillation. The best-known cause of gravity waves is airflow over mountains (i.e., mountain waves), but jet streams, fronts, and convective processes are important sources as well. Gravity waves can propagate large distances from their source and are known to play an important role in many atmospheric processes.
Altitudes or layers with negative values of l2 prohibit vertical propagation of gravity waves of all horizontal wavelengths. Such layers provide one mechanism for wave reflection, facilitating the formation of trapped or ducted waves that extend horizontally instead of vertically.
At altitudes where U approaches c due to wind shear, the Scorer parameter tends toward infinity, equivalent to the wave's vertical wavelength approaching zero. This altitude is called a critical level. Wave amplification and breaking can occur below a critical level (for upward propagating waves) or via other nonlinear effects associated with wave amplification. Wave breaking can initiate a cascade of energy to smaller scales, generating turbulence.
The net influence of the convection and gravity waves on Ri at this location is demonstrated in Figs. 9b–d. Three layers of relatively low Ri can be seen in the nonconvecting environment (i.e., the dry simulation), centered at 10, 11.25, and 13 km MSL. Convection in the moist simulation occupies the 10-km level at this location, but the other two altitudes remain in clear air. The reduced flight-level Ri is related to both enhanced vertical shear and a reduction in static stability (Figs. 9c,d), caused by a combination of the storm outflow and wave perturbations. Note the lowest Ri values at this time are located immediately above a shallower cloud, which may have played an additional role in creating the particularly low Ri values (e.g., see “enhanced shears and wake effects” subsection). Yet, the localized patch of increased turbulence likelihood subsequently moved away from its point of origin, and in general agreement with the observations establishes a hazard remote from the convection. Thus, like the 17 June 2005 case, the turbulence is ultimately caused by the deep convection yet occurs a substantial distance from the convectively active region.
TURBULENCE ABOVE CLOUDS.
On 3 August 2009, a Boeing 767 experienced severe turbulence near the Dominican Republic, en route from Rio de Janeiro to Houston. Twenty-eight passengers and five crew suffered minor injuries. Like many of these cases of turbulence near rapidly growing clouds, it is extremely difficult to determine exactly where the turbulence occurred relative to the cloud edge and uncertainties remain. Yet, a National Transportation Safety Board (NTSB) investigation states that “satellite weather imagery at the time of the incident indicates the presence of isolated, rapidly developing cumulus congestus to cumulonimbus clouds under the airplane's flight path” (National Transportation Safety Board 2011). As reflected by the FAA guidelines (guideline 6), the air above developing and mature convective clouds is known to be potentially hazardous to aircraft.
The 3 August 2009 event is reminiscent of an event that occurred over Dickinson, North Dakota, on 10 July 1997. The latter case caused 22 injuries from severe turbulence directly above a developing thunderstorm (National Transportation Safety Board 1998) and provided the motivation for a number of simulation studies (Lane et al. 2003; Lane and Sharman 2006, 2008). Although there were insufficient data to determine the exact cause of this event, detailed examination of this case using idealized modeling highlighted, among other things, the role of gravity wave breaking above convection as an important turbulence source.
Gravity wave breaking.
Numerous observational and modeling studies have documented the occurrence and characteristics of high-frequency gravity waves above convective clouds (e.g., Fovell et al. 1992; Pfister et al. 1993; Piani et al. 2000). Although such waves usually have horizontal wavelengths that are too long (>~5 km) to be felt as turbulence by aircraft, wave breaking can initiate turbulence at the subkilometer scales that do affect aircraft.
Wave breaking above convection is illustrated using an idealized three-dimensional (3D) simulation that partly resolves aircraft-scale turbulence (Fig. 10). This simulation has 150-m grid spacing in all directions and uses a thermodynamic and (unidirectional) wind environment defined using the closest sounding to the Dickinson turbulence encounter (0000 UTC 11 July 1997, Bismarck, North Dakota); full details can be found in Lane and Sharman (2006). The above-cloud perturbations in potential temperature show the gravity waves to have horizontal wavelengths of approximately 5 km and vertical displacement amplitudes of up to 500 m. Regions of wave breaking are highlighted by two thick isentropes (372 and 384 K). Specifically, the uppermost thick contour in Fig. 10a shows steepening at (x, z) ≈ (57, 14.5) km and the lowermost thick contour in Fig. 10b shows smaller-scale steepening and overturning at (x, z) ≈ (60, 13.5) km.
The wave steepening and breaking is accompanied by an enhancement in velocity perturbations at the scales of motion that affect commercial aircraft. To quantify these scales, a resolved turbulence kinetic energy per unit mass, TKE = [(u')2+(v')2+(w')2]/2, is calculated (as opposed to the parameterized TKE discussed in the “turbulence adjacent to clouds” section)5 and values outside of cloud are shown in Fig. 10. At both times, the turbulence is large in the regions of gravity wave steepening and overturning. For example, in Fig. 10b the largest TKE outside the cloud is associated with the wave overturning at (x, z) ≈ (60, 13.5) km, and a contiguous region of turbulence extends ~2 km above the uppermost cloud top. Small-scale above-cloud velocity perturbations persist throughout the cloud's evolution (e.g., online supplemental animation of Fig. 10) and are illustrated by a horizontal cross section above the cloud (Fig. 11a).
There is an extensive literature explaining gravity wave instabilities and of particular relevance here are instabilities associated with critical levels (see, e.g., Nappo 2002; “Gravity waves” sidebar). Above convection the high-frequency waves propagate at ±5–10 m s−1 relative to the cloud top wind speed and therefore a change in wind speed above the convection of only 5–10 m s−1 can incite a critical level (Lane et al. 2003). The critical level induces breaking of those waves propagating in the same direction as the above-cloud shear vector, which is consistent with the steepening and breaking waves in Fig. 10; the above-cloud shear vector points toward the left in this simulation and is only about 5 m s−1 km−1 in magnitude. Thus, even in moderate wind shear, gravity wave breaking can be an important source of NCT.
To explore the relationships between above cloud turbulence extent and background conditions, Lane and Sharman (2008) used numerous 2D and 3D simulations to examine the effects of above-cloud vertical wind shear and static stability on NCT. As expected, smaller static stabilities produce more extensive regions of turbulence. The response to changes in wind shear is more complicated, with maximum volumes of above-cloud turbulence occurring at intermediate shears. Stronger shears lead to more intense turbulence, but over smaller volumes that are confined closer to the cloud top. This confinement is related to the smaller distance between the critical level and the cloud top in conditions with stronger shear. On the other hand, weaker shears do not induce a background critical level and wave breaking is a less important generation mechanism.
Based on these results, Lane and Sharman (2008) suggest that avoidance guidelines for vertical separation should incorporate vertical shear and stability at a minimum. Yet, the FAA guideline (guideline 6) for vertical separation is based entirely on cloud-top wind speed, which is inconsistent with the underlying dynamics outlined here. Moreover, the storm shown in Fig. 10 has a cloud-top wind speed of approximately 13 m s−1 (25 kt; Lane and Sharman 2006), and the FAA guidelines would suggest a vertical separation of only 760 m (2,500 ft) above cloud top, much smaller than the vertical extent of wave breaking and turbulence in this case. Despite the simulated turbulence extending well above the cruising altitude of most commercial aircraft, the hazard is still relevant for turbulence above thunderstorms in the winter or higher latitude, when the thunderstorm tops are lower (e.g., Trier et al. 2012).
Enhanced shears and wake effects.
The generation of turbulence above convection is, of course, not limited to gravity wave effects, and the process of convective updraft growth and collapse can play an important role in turbulence generation near the cloud edge. Although much of this turbulence probably falls within the minimum vertical separation defined by the FAA guidelines, cloud-edge effects may have caused some of the recent cases that occurred close to cloud top.
For example, Grabowski and Clark (1991) describe a cloud-interfacial instability, where enhanced shear and flow deformation reduce Ri and support Kelvin-Helmholtz billows along the edge of the cloud. Lane et al.'s (2003) 2D and 3D simulations of the Dickinson case produced this shear enhancement and turbulence generation along the uppermost cloud boundaries during updraft overshoot events. Those simulations demonstrated that the enhanced shear layer was only a few hundred meters thick, a depth resolvable in Lane et al.'s 3D simulations with 16-m vertical grid spacing but unresolved in the simulation shown in Figs. 10 and 11. Other numerical simulations (Lane and Sharman 2008) showed that turbulence near the cloud edge is most prevalent in the early stages of a thunderstorm's lifetime. It follows that this mechanism is probably most relevant for incidents that occur directly above rapidly growing convective updrafts.
Other cloud-top processes may also be an important source of NCT. For example, Wang et al. (2010) presented simulations of coherent wake-like features that extend downwind of overshooting convective cores, a phenomenon sometimes observed in satellite imagery. Wang et al. liken these features to the well-known Kelvin ship wave pattern (Sharman and Wurtele 1983). Additional analysis of the idealized cloud model simulation shows that wake-like features indeed follow the collapse of the overshooting turret. Specifically, Fig. 11b shows that many coherent features are exposed by the square of the vorticity. A broad arc of enhanced vorticity (denoted “A”) occurs upstream of the underlying cloud (cf. Fig. 11a) and downstream of this arc numerous band-like structures (denoted “B”) extend laterally from the above-cloud turbulent patch. In addition to these lateral bands, the horizontal divergence (Fig. 11c) identifies flow-aligned structures that extend downwind (x ≈ 80 km) forming a turbulent wake.
The mechanisms underlying these thunderstorm wakes are not entirely reconciled. Wang et al. suggest that the cloud behaves as an obstacle that blocks the oncoming flow; indeed, the diffluent horizontal wind vectors on the upstream side of the cloud and the turbulent wake downstream seem to support that explanation (see also Fujita and Grandoso 1968; Lemon 1976). However, as described by Rotunno and Klemp (1982), the dynamics are not entirely consistent with the obstacle analogy; the cloud is porous with air flowing through the cloud edge, and the pressure gradient is opposed to the cross-updraft shear vector (which is not necessarily aligned with the storm-relative wind). Another hypothesis is that these simulated bands in Fig. 11c are an early manifestation of the radial bands described in the “17 June 2005: Turbulent cirrus bands” subsection. Regardless of the source dynamics, the spatial scale of the lateral bands and downstream structures are all approaching those that strongly influence commercial aircraft, and they may pose an additional hazard adjacent to the storm and be responsible for the known hazard downwind.
SUMMARY AND FUTURE OUTLOOK.
Fundamental understanding of near-cloud turbulence.
We have summarized recent progress made in understanding the NCT aviation hazard. These fundamental advances in our basic understanding were enabled by high-resolution numerical simulations of observed events, complemented by improved data on turbulence encounters. These studies have shown that NCT is a complex phenomenon that crucially depends on cloud characteristics, the structure of the near-cloud environment, and perturbations to that environment by cloud circulations and gravity waves. Yet, we are acutely aware that we are only beginning to scratch the surface and a variety of basic problems are still to be solved regarding the dynamics underlying NCT. Outstanding questions include the following:
What are the characteristics of NCT and how does NCT vary spatially and temporally?
How is NCT related to the mode of convective organization and its intensity?
What is the relative importance of gravity wave breaking and Kelvin–Helmholtz instability to the turbulence hazard?
What are the processes leading to the enhanced hazard near thunderstorm anvils?
What is the structure and mechanism of turbulence in thunderstorm wakes?
How common is the hazard posed by turbulence associated with ducted gravity waves?
What is the relationship between observable cloud features, the mesoscale environment, and NCT that may be useful for pilots and aviation forecasters?
What is the climatology of NCT?
These questions need to be answered, and further research on the other processes detailed in this article is necessary to advance the fundamental understanding of NCT. This progress should also provide the framework to develop new approaches for turbulence avoidance. With recent improvements in high-resolution modeling capabilities available to researchers, we believe that such much-needed advancement is achievable. Unfortunately, despite the importance of this problem and the opportunities for progress, there is relatively limited activity in this area with only a few groups around the world actively engaged in NCT research. We hope that this article has spurred additional interest in this topic and we encourage others to study this challenging problem.
The role of modern numerical weather prediction.
State-of-the art operational or real-time systems [e.g., the High-Resolution Rapid Refresh model (HRRR); Smith et al. 2008] have achieved convection-permitting model resolutions, allowing for much more realistic representations of the governing convective processes. The examples presented herein demonstrate the utility of such convection-permitting models to fully reproduce regions of turbulence near thunderstorms, even though they are unable to resolve aircraft-scale turbulence and the simulated turbulence remains parameterized. The hazard posed by rapidly growing thunderstorms underlines the need for skillful predictions and diagnoses of convective initiation and vertical development; these processes benefit from convection-permitting resolutions as well. Of course, convective initiation, vertical development, and turbulence are all processes that suffer from low predictability. Simulated turbulence is also highly sensitive to model configurations and the choice of physical parameterizations. Therefore, ensemble convection-permitting forecasts likely provide greater promise for significant advances in future turbulence prediction than do solely deterministic approaches.
Existing models like HRRR already contain a wealth of information regarding the turbulence hazard. This information is simply not being fully utilized for turbulence forecasting. Efforts focused on the development of diagnostic products tailored to convection-permitting models could provide invaluable guidance and actually take advantage of the recent advances in operational modeling capabilities. Time-lagged ensembles of these models could also be used to complement deterministic predictions and facilitate development of ensemble approaches for turbulence forecasting.
Toward improved turbulence avoidance guidelines and integrated systems.
The myriad of processes that lead to turbulence near deep convection will continue to make turbulence avoidance challenging. Simple rules, like the FAA guidelines, are a necessary first step toward rapidly assessing hazards from the cockpit. However, the simplicity of the current guidelines limits their effectiveness because they do not capture all of the important processes. We believe that more sophisticated guidelines are required that encompass ongoing improvements in our understanding of NCT. The studies presented here suggest that above-cloud wind shear and stability, mesoscale regions of low Richardson number, gravity wave ducting, and the direction of the upper-level storm outflow should all be considered as part of improved guidelines. Other properties like thunderstorm intensity and the mode of organization should be investigated as well as the changes in turbulence generation mechanisms during different stages of the cloud lifecycle (e.g., Kim and Chun 2012). Developing such guidelines should be a priority, and enhanced activity in this area would provide the critical mass of evidence to support those changes.
It is possible, however, that new turbulence guidelines that incorporate processes like wind shear, stability, and other more complicated mechanisms like wave ducting would be a challenge for pilots, air-traffic control, and aviation forecasters to use. Yet, an integrated theoretical and empirical approach that makes use of operational analyses, convection-permitting forecasts, empirical nowcasting methods and avoidance guidelines, and satellite imagery could be successful. This information could be used to construct temporally varying hazard maps for incorporation into future automatic traffic routing procedures [e.g., the Next Generation Air Transportation System (NextGen)]. If these approaches were coupled with probabilistic hazard information derived from ensemble forecasts, the possibilities for advances in turbulence avoidance and improved aviation safety would be substantial.
Todd Lane is supported by an Australian Research Council Future Fellowship (FT0990892). This work was also supported in part from NASA CAN, ASAP and ROSES grants. We thank the editor (Tom Fahey), Rebecca Morss, Rachel Badlan, Muhammad Hassim, Dragana Zovko Rajak, and three anonymous reviewers for their comments on an earlier version of the manuscript. The National Center for Atmospheric Research is sponsored by the National Science Foundation.
REFERENCES
Benjamin, S. G., and Coauthors, 2004: An hourly assimilation–forecast cycle: The RUC. Mon. Wea. Rev., 132, 495–518.
Burns, A., T. W. Harrold, J. Burnham, and C. S. Spavins, 1966: Turbulence in clear air near thunderstorms. National Severe Storms Laboratory Tech. Memo. 30, 20 pp.
Cornman, L. B., C. S. Morse, and G. Cunning, 1995: Real-time estimation of atmospheric turbulence severity from in-situ aircraft measurements. J. Aircraft, 32, 171–177.
Cornman, L. B., G. Meymaris, and M. Limber, 2004: An update on the FAA Aviation Weather Research Program's in situ turbulence measurement and report system. Preprints, 11th Conf. on Aviation, Range, and Aerospace Meteorology, Hyannis, MA, Amer. Meteor. Soc., 4.3. [Available online at http://ams.confex.com/ams/pdfpapers/81622.pdf.]
Durran, D. R., and J. B. Klemp, 1982: On the effects of moisture on the Brunt-Väisälä frequency. J. Atmos. Sci., 39, 2152–2158.
Eichenbaum, H., 2003: Historical overview of turbulence accidents and case study analysis. MCR Federal Inc. Rep. Br-M021/080–1, 82 pp. [Available from MCR Federal Inc., 175 Middlesex Turnpike, Bedford, MA 01730.]
Federal Aviation Administration, 2012: Chapter 7. FAA Aeronautical Information Manual. [Available online at www.faa.gov/air_traffic/publications/atpubs/aim.]
Fovell, R. G., D. R. Durran, and J. R. Holton, 1992: Numerical simulations of convectively generated stratospheric gravity waves. J. Atmos. Sci., 49, 1427–1442.
Fovell, R. G., R. D. Sharman, and S. B. Trier, 2007: A case study of convectively-induced clear-air turbulence. Preprints, 12th Conf. on Mesoscale Processes, Waterville Valley, NH, Amer. Meteor. Soc., 13.4. [Available online at http://ams.confex.com/ams/pdfpapers/126190.pdf.]
Fritsch, J. M., and R. A. Maddox, 1981: Convectively driven mesoscale weather systems aloft. Part I: Observations. J. Appl. Meteor., 20, 9–19.
Fujita T., and H. Grandoso, 1968: Split of a thunderstorm and cyclonic storms and their motion as determined from numerical model experiments. J. Atmos. Sci., 25, 416–439.
Grabowski, W. W., and T. L. Clark, 1991: Cloud–environment interface instability: Rising thermal calculations in two spatial dimensions. J. Atmos. Sci., 48, 527–546.
Janjić, Z. I., 1990: The step-mountain coordinate: Physical package. Mon. Wea. Rev., 118, 1429–1443.
Janjić, Z. I., 1994: The step-mountain Eta coordinate model: Further developments of the convection, viscous sublayer, and turbulent closure schemes. Mon. Wea. Rev., 122, 927–945.
Kaplan, M. L., A. W. Huffman, K. M. Lux, J. J. Charney, A. J. Riordan, and Y.-L. Lin, 2005: Characterizing the severe turbulence environments associated with commercial aviation accidents. Part 1: A 44-case study synoptic observational analyses. Meteor. Atmos. Phys., 88, 129–152.
Kauffmann, P., and A. Sousa-Poza, 2001: Market assessment of forward-looking turbulence sensing systems. NASA Rep. CR-2001-210905, 78 pp. [Available online at http://gltrs.grc.nasa.gov/reports/2001/CR-2001-210905.pdf.]
Keller, T. L., L. J. Ehernberger, and M. G. Wurtele, 1983: Numerical simulation of the atmosphere during a CAT encounter. Proc. Ninth Conf. on Aerospace and Aeronautical Meteorology, Omaha, NE, Amer. Meteor. Soc., 316–319.
Kim, J.-H., and H.-Y. Chun, 2011: Statistics and possible sources of aviation turbulence over South Korea. J. Appl. Meteor. Climatol., 50, 311–324.
Kim, J.-H., and H.-Y. Chun, 2012: A numerical simulation of convectively induced turbulence (CIT) above deep convection. J. Appl. Meteor. Climatol., 51, in press.
Knox, J. A., A. S. Bachmeier, W. M. Carter, J. E. Tarantino, L. C. Paulik, E. N. Wilson, G. S. Bechdol, and M. J. Mays, 2010: Transverse cirrus bands in weather systems: A grand tour of an enduring enigma. Weather, 65, 35–41.
Lane, T. P., and R. D. Sharman, 2006: Gravity wave breaking, secondary wave generation, and mixing above deep convection in a three-dimensional cloud model. Geophys. Res. Lett., 33, L23813, doi:10.1029/2006GL027988.
Lane, T. P., and R. D. Sharman, 2008: Some influences of background flow conditions on the generation of turbulence due to gravity wave breaking above deep convection. J. Appl. Meteor. Climatol., 47, 2777–2796.
Lane, T. P., R. D. Sharman, T. L. Clark, and H.-M. Hsu, 2003: An investigation of turbulence generation mechanisms above deep convection. J. Atmos. Sci., 60, 1297–1321.
Lemon, L., 1976: Wake vortex structure and aerodynamic origin in severe thunderstorms. J. Atmos. Sci., 33, 678–685.
LeMone, M. A., 1973: The structure and dynamics of horizontal roll vortices in the planetary boundary layer. J. Atmos. Sci., 30, 1077–1091.
Lenz, A., K. M. Bedka, W. F. Feltz, and S. A. Ackerman, 2009: Convectively induced transverse band signatures in satellite imagery. Wea. Forecasting, 24, 1362–1373.
Lester, P. F., 1994: Turbulence: A New Perspective for Pilots. Jeppesen Sanderson, 280 pp.
Lin, Y.-L., R. D. Farley, and H. D. Orville, 1983: Bulk parameterization of a snow field in a cloud model. J. Climate Appl. Meteor., 22, 1065–1092.
Nappo, C. J., 2002: An Introduction to Atmospheric Gravity Waves. Academic Press, 276 pp.
National Transportation Safety Board, 1998: Brief of incident, Ftw97Ia261. [Available at 490 L'Enfant Plaza, SW, Washington, DC 20594.]
National Transportation Safety Board, 2010: Preliminary report. DCA10IA084. [Available at 490 L'Enfant Plaza, SW, Washington, DC 20594.]
National Transportation Safety Board, 2011: Brief of incident. DCA09IA071. [Available at 490 L'Enfant Plaza, SW, Washington, DC 20594.]
Noh, Y., W. G. Cheon, S.-Y. Hong, and S. Raasch, 2003: Improvement of the k-profile model for the planetary boundary layer based on large eddy simulation data. Bound.-Layer Meteor., 107, 401–427.
Pandya, R. E., and D. R. Durran, 1996: The influence of convectively generated thermal forcing on the mesoscale circulations around squall lines. J. Atmos. Sci., 53, 2924–2951.
Pantley, K. C., and P. F. Lester, 1990: Observations of severe turbulence near thunderstorm tops. J. Appl. Meteor., 29, 1171–1179.
Pfister, L., S. Scott, M. Loewenstein, S. Bowen, and M. Legg, 1993: Mesoscale disturbances in the tropical stratosphere excited by convection: Observations and effects on the stratospheric momentum budget. J. Atmos. Sci., 50, 1058–1075.
Piani, C., D. R. Durran, M. J. Alexander, and J. R. Holton, 2000: A numerical study of three-dimensional gravity waves triggered by deep tropical convection and their role in the dynamics of the QBO. J. Atmos. Sci., 57, 3689–3702.
Politovich, M. K., R. K. Goodrich, C. S. Morse, A. Yates, R. Barron, and S. A. Cohn, 2011: The Juneau terrain-induced turbulence alert system. Bull. Amer. Meteor. Soc., 92, 299–313.
Prophet, D. T., 1970: Vertical extent of turbulence in clear air above the tops of thunderstorms. J. Appl. Meteor., 9, 320–321.
Rotunno, R., and J. B. Klemp, 1982: The influence of the shear-induced pressure gradient on thunderstorm motion. Mon. Wea. Rev., 110, 136–151.
Schultz, D. M., and Coauthors, 2006: The mysteries of mammatus clouds: observations and formation mechanisms. J. Atmos. Sci., 63, 2409–2435.
Schwartz, B., 1996: The quantitative use of PIREPs in developing aviation weather guidance products. Wea. Forecasting, 11, 372–384.
Sharman, R. D., and M. G. Wurtele, 1983: Ship waves and lee waves. J. Atmos. Sci., 40, 396–427.
Sharman, R. D., C. Tebaldi, G. Wiener, and J. Wolff, 2006: An integrated approach to mid- and upper-level turbulence forecasting. Wea. Forecasting, 21, 268–287.
Skamarock, W. C., and J. B. Klemp, 2008: A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comp. Phys., 227, 3465–3485.
Smith, T. L., S. G. Benjamin, J. M. Brown, S. S. Weygandt, T. Smirnova, and B. E. Schwartz, 2008: Convection forecasts from the hourly updated, 3-km High Resolution Rapid Refresh model. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 11.1. [Available online at http://ams.confex.com/ams/pdfpapers/142055.pdf.]
Trier, S. B., and R. D. Sharman, 2009: Convection-permitting simulations of the environment supporting widespread turbulence within the upper-level outflow of a mesoscale convective system. Mon. Wea. Rev., 137, 1972–1990.
Trier, S. B., R. D. Sharman, R. G. Fovell, and R. G. Frehlich, 2010: Numerical simulation of radial cloud bands within the upper-level outflow of an observed mesoscale convective system. J. Atmos. Sci., 67, 2990–2999.
Trier, S. B., R. D. Sharman, and T. P. Lane, 2012: Influences of moist convection on a cold-season outbreak of clear-air turbulence (CAT). Mon. Wea. Rev., 140, in press.
Wang, P. K., S.-H. Su, M. Setvak, H. Lin, and R. M. Rabin, 2010: Ship wave signature at the cloud top of deep convective storms. Atmos. Res., 97, 294–302.
Weckwerth, T. M., J. W. Wilson, R. M. Wakimoto, and N. A. Crook, 1997: Horizontal convective rolls: Determining the environmental conditions supporting their existence and characteristics. Mon. Wea. Rev., 125, 505–526.
Wolff, J. K., and R. D. Sharman, 2008: Climatology of upper-level turbulence over the contiguous United States. J. Appl. Meteor. Climatol., 47, 2198–2214.
Horizontal scales of motion between ~100 m and ~2 km are those that induce the strongest turbulent response from large commercial aircraft. For the purposes of this discussion, these scales are referred to as aircraft-scale turbulence.
Although through much of the troposphere g/θe(∂θe/∂z) is a poor approximation to the moist static stability (Durran and Klemp 1982), it is a reasonable approximation at typical commercial aircraft cruising altitudes of 11–12 km MSL, where the slopes of the moist and dry adiabats are similar.
The model domain is 800 km2, and model parameterizations include the Yonsei University (YSU) boundary layer (Noh et al. 2003), Lin et al. (1983) microphysics, and RUC land surface schemes; no cumulus parameterization is used.
Here, the cloud edge is defined in two ways: using condensation and optical depth (which is a vertical integral from above). Both of these measures identify that the region of enhanced turbulence is outside of the simulated cloud. Moreover, if radar reflectivity were used to detect the cloud boundary (as might be done by a pilot en route), the detectable cloud volume would likely be smaller and the turbulence would appear to be farther from the cloud edge.
The background velocity used to determine the perturbations is obtained by horizontally smoothing each of the velocity components (over 13 × 13 grid points using a moving average). This procedure retains horizontal scales of motion less than ~2 km in the perturbation fields: that is, aircraft-scale turbulence.