ADVANCES IN ATMOSPHERIC SCIENCES, VOL.

34, MAY 2017, 576–586

• Original Paper •

Increased Light, Moderate, and Severe Clear-Air Turbulence in

Response to Climate Change
Paul D. WILLIAMS∗

Department of Meteorology, University of Reading, Reading RG6 6BB, United Kingdom

(Received 31 October 2016; revised 20 February 2017; accepted 28 February 2017)

ABSTRACT
Anthropogenic climate change is expected to strengthen the vertical wind shears at aircraft cruising altitudes within
the atmospheric jet streams. Such a strengthening would increase the prevalence of the shear instabilities that generate
clear-air turbulence. Climate modelling studies have indicated that the amount of moderate-or-greater clear-air turbulence
on transatlantic flight routes in winter will increase significantly in future as the climate changes. However, the individual
responses of light, moderate, and severe clear-air turbulence have not previously been studied, despite their importance
for aircraft operations. Here, we use climate model simulations to analyse the transatlantic wintertime clear-air turbulence
response to climate change in five aviation-relevant turbulence strength categories. We find that the probability distributions
for an ensemble of 21 clear-air turbulence diagnostics generally gain probability in their right-hand tails when the atmospheric
carbon dioxide concentration is doubled. By converting the diagnostics into eddy dissipation rates, we find that the ensemble-
average airspace volume containing light clear-air turbulence increases by 59% (with an intra-ensemble range of 43%–68%),
light-to-moderate by 75% (39%–96%), moderate by 94% (37%–118%), moderate-to-severe by 127% (30%–170%), and
severe by 149% (36%–188%). These results suggest that the prevalence of transatlantic wintertime clear-air turbulence will
increase significantly in all aviation-relevant strength categories as the climate changes.

Key words: turbulence, climate change, aviation, jet stream

Citation: Williams, P. D., 2017: Increased light, moderate, and severe clear-air turbulence in response to climate change.
Adv. Atmos. Sci., 34(5), 576–586, doi: 10.1007/s00376-017-6268-2.

1. Introduction increase the probability of flight disruptions and delays. At

cruising altitudes, shifting wind patterns may modify optimal
The climate is changing—not just where we live at
flight routes and affect travel times (Karnauskas et al., 2015;
ground level, but also where we fly at 30 000–40 000 feet.
Irvine et al., 2016; Williams, 2016), and stronger jet-stream
Climate change may have important consequences for avi-
wind shears may increase clear-air turbulence (Williams and
ation, because the meteorological characteristics of the at-
Joshi, 2013, 2016).
mosphere influence airport operations, flight routes, journey
In the field of climate science, the midlatitude jet streams
times, and the safety and comfort of passengers and crew.
in both the northern and southern hemispheres are expected to
The contribution of aviation to climate change has long been
strengthen at aircraft cruising altitudes as the climate changes
recognised (Lee et al., 2009). In contrast, the impacts of cli-
(Lorenz and DeWeaver, 2007; Solomon and Polvani, 2016).
mate change on aviation have only recently begun to emerge,
The strengthening of these thermal winds occurs because,
as discussed by Puempel and Williams (2016) in a recent arti-
despite polar amplification of warming in the lower tropo-
cle for the International Civil Aviation Organization (ICAO).
sphere, increased carbon dioxide (CO2 ) is enhancing the
In terms of airport operations, rising sea levels and storm
column-averaged pole-to-equator temperature gradient in the
surges threaten coastal airports, many of which are located
midlatitudes, through the combined effect of tropospheric
at altitudes of only a few metres above the mean sea level
warming and stratospheric cooling (Shine et al., 2003; Del-
(Burbidge, 2016). Warmer air on runways is imposing in-
cambre et al., 2013; Goessling and Bathiany, 2016). The in-
creasingly frequent take-off weight restrictions (Coffel and
crease in the magnitude of the meridional temperature gra-
Horton, 2015). More extreme weather, such as an increase in
dient at aircraft cruising altitudes stems from robust thermo-
the frequency of lightning strikes (Romps et al., 2014), may
dynamic effects and is unlikely to be significantly abated by
feedbacks from the dynamics (Vallis et al., 2015). In the
∗ Corresponding author: Paul D. WILLIAMS lower troposphere, the result of the temperature changes is
Email: [email protected] weaker zonal winds, which may cause more extreme weather

© The Author [2017]. This article is published with open access at link.springer.com.
MAY 2017 WILLIAMS 577

(Francis and Vavrus, 2012). However, in the upper tropo- and one of its four engines (Clark et al., 2000).
sphere and lower stratosphere, the result is stronger zonal The economic costs of turbulence arise from injuries
winds (Delcambre et al., 2013; Simpson, 2016). to passengers and crew, damage to airframes and cabins,
In the field of aviation turbulence, observational stud- flight delays, inspections, repairs, and post-accident inves-
ies generally support the hypothesis that clear-air turbulence tigations. Estimates of the total economic cost vary from
is generated by the breaking of unstable Kelvin–Helmholtz around US$100 million annually (Riddaway, 1998) to nearly
waves in highly sheared regions of the atmosphere (Endlich, US$200 million annually for United States carriers alone
1964; Atlas et al., 1970; Watkins and Browning, 1973). This (Williams, 2014). The operational challenges associated with
hypothesis concurs with evidence that the jet stream contains any future increase in turbulence are compounded by the
about three times more clear-air turbulence than the rest of the projected future growth of the aviation sector. Historically,
atmosphere (Reiter, 1963, Section 9.2). The hypothesis also global air traffic (measured in passenger–kilometers) has ex-
concurs with evidence that eastbound transatlantic flights, be- perienced an average long-term growth rate of 5% per year,
ing closer to the strong tailwinds in the core of the jet stream, which corresponds to a doubling period of about 14 years.
encounter more clear-air turbulence than westbound transat- According to Boeing’s market outlook, this trend is expected
lantic flights (Kim et al., 2016). Gravity waves, including to continue for at least the next 20 years (Warner, 2013).
those generated as the atmosphere adjusts to a loss of balance Even present-day encounters with vertical and horizontal
(Williams et al., 2003, 2005, 2008), may play a role by initi- high-altitude wind shear may infringe current aircraft certi-
ating the Kelvin–Helmholtz instability in an otherwise stable fication envelopes, according to accident and incident inves-
part of the atmosphere (Knox et al., 2008; McCann et al., tigations (ICAO, 2015). New evidence indicates that only
2012). 14% of aircraft encounters with turbulence occur in the vicin-
It follows from the above discussion that anthropogenic ity of convection (Meneguz et al., 2016). If true, this evi-
climate change, by strengthening the vertical wind shears dence suggests that the proportion of encounters that are at-
at aircraft cruising altitudes within the jet streams, may be tributable to clear-air turbulence and mountain waves is 86%,
increasing the frequency and intensity of clear-air turbu- which is much higher than previously thought. Pilots are
lence. As reviewed by Williams and Joshi (2016), evidence known to be more likely to attempt evasive manoeuvres as
is emerging of upward trends in recent turbulence statistics the likelihood of encountering moderate-to-severe turbulence
(FAA, 2006; Jaeger and Sprenger, 2007; Wolff and Sharman, increases (Krozel et al., 2011). Therefore, in a more turbu-
2008; Kim and Chun, 2011), although the interpretation of lent atmosphere, flight paths could become more convoluted,
these trends requires caution in some cases. Climate change lengthening journey times and increasing fuel consumption
may continue to increase the prevalence of clear-air turbu- and emissions. Turbulence causes wear-and-tear to airframes,
lence in the coming decades. Climate modelling studies have which could require more time out of service while undergo-
indicated that the volume of airspace containing moderate- ing routine inspections and maintenance. Finally, air-to-air
or-greater clear-air turbulence on transatlantic flight routes in refuelling has been proposed as a method to reduce fuel con-
winter will increase by 40%–170%, relative to pre-industrial sumption and emissions, but it requires mid-air contact be-
times, when the CO2 is doubled (Williams and Joshi, 2013). tween the cruiser and feeder planes, and turbulence poses a
An intensification of clear-air turbulence could have im- hazard to such operations (Lee, 2013).
portant consequences for aviation. Turbulence is hazardous In their study on the future response of clear-air turbu-
to aircraft and is the underlying cause of many people’s fear lence to climate change, Williams and Joshi (2013) focused
of air travel (Sharman et al., 2012). The median length of a on turbulence in the moderate-or-greater strength category.
patch of turbulence is currently about 60 km, which equates This approach has two limitations. First, light and light-to-
to about five minutes of flying time, and the median thick- moderate turbulence are excluded from consideration. Al-
ness is currently about 1 km, which limits the efficacy of though turbulence in these categories is not generally a safety
altitude changes as an evasive manoeuvre (Sharman et al., risk to passengers or crew, it does nevertheless distress ner-
2014). Conservative estimates indicate that there are 790 tur- vous passengers and cause mild wear-and-tear to airframes.
bulence encounters annually for scheduled United States car- The second limitation of focussing on moderate-or-greater
riers, resulting in 687 minor injuries to flight attendants, 38 turbulence is that moderate-to-severe and severe turbulence
serious injuries to flight attendants, 120 minor injuries to pas- are included together in the same category as moderate tur-
sengers, and 17 serious injuries to passengers (Kauffmann, bulence. Unlike turbulence in the weaker categories, severe
2002). However, the actual injury rates are probably much turbulence does pose a safety risk to passengers and crew, be-
higher because of under-reporting, with other estimates indi- cause it causes aircraft to execute random motions with ver-
cating that there are over 63 000 encounters with moderate- tical accelerations that exceed the gravitational acceleration
or-greater turbulence and 5000 encounters with severe-or- (Lane et al., 2012). Therefore, severe turbulence arguably
greater turbulence annually (Sharman et al., 2006). In ad- warrants separate consideration.
dition to injuring passengers and crew, turbulence can also For the above reasons, the present paper extends the
cause structural damage to aircraft. For example, a plane fly- analysis of Williams and Joshi (2013), by analysing the
ing over Colorado on 9 December 1992 encountered extreme transatlantic wintertime clear-air turbulence response to cli-
clear-air turbulence, which tore off about 6 m of its left wing mate change in five individual aviation-relevant turbulence
578 INCREASED TURBULENCE IN RESPONSE TO CLIMATE CHANGE VOLUME 34

strength categories. The methodology and results are pre- spheric CO2 concentration is doubled from its pre-industrial
sented in Section 2. The paper concludes with a summary value of 280 ppm to 560 ppm, probability generally shifts
and discussion in Section 3. to the right in the histograms of Fig. 1, causing the diagnos-
tics to gain probability in their right-hand tails. These shifts
imply a reduction in the prevalence of smooth air and an in-
2. Methodology and results crease in the prevalence of turbulent air. Exceptionally, the
To extend the analysis of Williams and Joshi (2013), for Brown index and the magnitude of the potential vorticity gain
consistency we use the same geographic area (50◦ –75◦ N, probability in their left-hand tails as well as their right-hand
10◦ –60◦ W) and altitude (200 hPa). The geographic area is tails. Aviation-affecting turbulence lies far into the right-hand
chosen because it lies within the North Atlantic flight corri- tails, which are the parts of the probability distributions that
dor, which is the busiest oceanic airspace in the world. It con- we wish to analyse in detail.
tains the majority of transatlantic traffic, as indicated by grid- To proceed, it is advantageous to convert the 21 clear-
ded global inventories of fuel burn and emissions obtained air turbulence diagnostics into eddy dissipation rates. The
from the Federal Aviation Administration’s Aviation Envi- eddy dissipation rate is a natural measure for quanti-
ronment Design Tool (Kim et al., 2005; Malwitz et al., 2005; fying the strength of turbulence. For a given aircraft
Wilkerson et al., 2010; Wilcox et al., 2012). The altitude is type, aircraft weight, airspeed, and altitude, the root-
chosen because it is within the range of permitted flight levels mean-square vertical acceleration of the aircraft in tur-
for the North Atlantic flight corridor (Irvine et al., 2013). Ac- bulence is proportional to the cube-root of the eddy
cording to a special report by the Intergovernmental Panel on dissipation rate (MacCready, 1964). For a large com-
Climate Change, the cruising altitudes of civil aircraft are not mercial aircraft, cube-rooted eddy dissipation rates of
expected to increase significantly over the next few decades, 0.0–0.1 m2/3 s−1 generate null turbulence, 0.1–0.2 m2/3 s−1
because of physical limitations and costs (Penner et al., 1999, generate light turbulence, 0.2–0.3 m2/3 s−1 generate light-
Section 7.2.2). to-moderate turbulence, 0.3–0.4 m2/3 s−1 generate moder-
For consistency, we also use the same season (winter), ate turbulence, 0.4–0.5 m2/3 s−1 generate moderate-to-severe
climate model (GFDL-CM2.1), anthropogenic forcing sim- turbulence, 0.5–0.6 m2/3 s−1 generate severe turbulence,
ulations (pre-industrial and doubled-CO2 ), and 21-member 0.6–0.7 m2/3 s−1 generate severe-to-extreme turbulence, and
ensemble of clear-air turbulence diagnostics as Williams and values greater than 0.7 m2/3 s−1 generate extreme turbulence
Joshi (2013). Winter is chosen because it is the season in (Lane et al., 2012; Williams, 2014; Sharman et al., 2014).
which the prevalence of clear-air turbulence peaks in the According to atmospheric measurements and models, the
North Atlantic sector (Jaeger and Sprenger, 2007). The cli- eddy dissipation rate and its cube root are log-normally dis-
mate model is chosen because the simulated upper-level tributed (Frehlich and Sharman, 2004). The log-normal dis-
winds in the northern extra-tropics agree well with reanal- tribution has two degrees of freedom, because it is completely
ysis data, and because the spatial pattern of clear-air turbu- specified by the mean and standard deviation of the underly-
lence over the North Atlantic diagnosed from reanalysis data ing normal distribution. Any two independent constraints on
is successfully captured by the model (Williams and Joshi, the cumulative distribution function of the log-normal distri-
2013). The numerical resolution of the atmosphere is 2.5◦ bution will uniquely define the values of these two parame-
in longitude, 2.0◦ in latitude, and 50 hPa in pressure alti- ters. Here, we require that the probability of light-or-greater
tude around the 200 hPa level. The doubled-CO2 simulation turbulence (with a cube-rooted eddy dissipation rate greater
is chosen because CO2 is projected to reach twice its pre- than 0.1 m2/3 s−1 ) is 3.0% (Watkins and Browning, 1973;
industrial concentration by the middle of this century, accord- Kim and Chun, 2016) and that the probability of moderate-or-
ing to moderate scenarios for future emissions (Meehl et al., greater turbulence (with a cube-rooted eddy dissipation rate
2007). Weighted linear combinations of the clear-air turbu- greater than 0.3 m2/3 s−1 ) is 0.4% (Sharman et al., 2006; Kim
lence diagnostics calculated from numerical weather predic- and Chun, 2016). This definition of moderate-or-greater tur-
tion models have been found to have significant skill when bulence is slightly different from the one used by Williams
verified against pilot reports (PIREPs), and these combina- and Joshi (2013). These two constraints on the cumulative
tions are currently being used for operational turbulence pre- distribution function allow us to infer that the natural log-
dictions (Sharman et al., 2006). arithm of the cube-rooted eddy dissipation rate (divided by
Probability distributions for the ensemble of 21 clear-air 1 m2/3 s−1 to produce a dimensionless numerical value) is
turbulence diagnostics are shown as histograms in Fig. 1. The normally distributed with a mean and standard deviation of
diagnostics are calculated from daily mean temperature and −5.00 and 1.43, respectively. These values agree with the re-
wind fields over 20 winters in each simulation. The diagnos- sults of other studies (e.g., Sharman et al., 2014; Sharman and
tics are described in detail by Sharman et al. (2006). They Pearson, 2017) to within a factor of about two, with any dis-
include the Colson and Panofsky (1965) index, the Brown crepancies presumably attributable to the different empirical
(1973) index, and the Ellrod and Knapp (1992) indices. For input data used to constrain the estimates.
each diagnostic, greater values are associated with stronger The above log-normal distribution for the cube-rooted
turbulence, where greater here means closer to plus infinity eddy dissipation rate yields percentile ranges for each tur-
rather than simply greater in magnitude. When the atmo- bulence strength category. These percentile ranges and their
MAY 2017 WILLIAMS 579
8 8 8
PRE−INDUSTRIAL
Probability (%)

Probability (%)

Probability (%)
6 DOUBLED CO2 6 6

4 4 4

2 2 2

0 0 0
−300 −200 −100 0 0 2 4 6 −45 −40 −35 −30 −25
Negative Richardson number Magnitude of vertical shear of horizontal wind (10−3 s−1) Colson−Panofsky index (103 kt2)

20 10 15
Probability (%)

Probability (%)

Probability (%)
15
10
10 5
5
5

0 0 0
0 100 200 300 40 60 80 100 120 0 200 400 600
Frontogenesis function (10−9 m2 s−3 K−2) Brown index (10−6 s−1) Brown energy dissipation rate (10−6 J kg−1 s−1)

10 10 8
Probability (%)

Probability (%)

Probability (%)
6

5 5 4

0 0 0
0 50 100 150 0 50 100 150 0 20 40 60
Variant 1 of Ellrod’s turbulence index (10−9 s−2) Variant 2 of Ellrod’s turbulence index (10−9 s−2) Flow deformation (10−6 s−1)

15 30 15
Probability (%)

Probability (%)

Probability (%)
10 20 10

5 10 5

0 0 0
2 4 6 8 10 0 0.2 0.4 0.6 0.8 1 0 5 10 15 20
−9 −2 −6 −1
Magnitude of potential vorticity (PVU) Relative vorticity squared (10 s ) Magnitude of horizontal temperature gradient (10 K m )

10 15 15
Probability (%)

Probability (%)

Probability (%)

10 10
5
5 5

0 0 0
0 20 40 60 0 1 2 3 4 5 0 0.5 1 1.5
Wind speed (m s−1) Wind speed × directional shear (10−3 rad s−1) Flow deformation × wind speed (10−3 m s−2)

20 20 20
Probability (%)

Probability (%)

Probability (%)

15 15 15

10 10 10

5 5 5

0 0 0
0 10 20 30 40 0 200 400 600 800 1000 0 5 10 15 20
−9 −1 −1
Flow deformation × vertical temperature gradient (10 K m s ) Magnitude of residual of NBE (10−12 s−2) −6 −1
Magnitude of horizontal divergence (10 s )

30 20 15
Probability (%)

Probability (%)

Probability (%)

15
20 10
10
10 5
5

0 0 0
0 10 20 30 40 50 0 0.5 1 1.5 0 0.5 1 1.5
−18 −3
Version 1 of North Carolina State University index (10 s ) Negative absolute vorticity advection (10−9 s−2) Magnitude of relative vorticity advection (10−9 s−2)

Fig. 1. Probability distributions for 21 clear-air turbulence diagnostics within the North Atlantic flight corridor at 200 hPa in
winter. The distributions are calculated from a pre-industrial control simulation and a doubled-CO2 simulation, as described in
the text. The histograms show the probability (%) for turbulence to occur within each finite-width bin. NBE is the nonlinear
balance equation. In the units, kt is knots and 1 PVU is 10−6 m2 s−1 K kg−1 .

corresponding probabilities are listed in Table 1. Severe-to- light-or-greater turbulence is 3.0% and the probability of
extreme and extreme turbulence are neglected because of moderate-or-greater turbulence is 0.4%. The probabilities
their rarity. Note that, by construction, the probability of for each turbulence strength category agree reasonably well
580 INCREASED TURBULENCE IN RESPONSE TO CLIMATE CHANGE VOLUME 34

with the relative frequencies at which the categories appear centile between the (100 × 0.5/N)th and [100 × (N − 0.5)/N]th
in automated in-flight measurements (Williams, 2014) and percentiles.
in PIREPs in the United States (Schwartz, 1996) and South In our case, there are 22 × 14 = 308 model gridpoints
Korea (Kim and Chun, 2011). Exact quantitative agreement in the geographic area being studied, and there are 20 ×
cannot be expected, because of inconsistent PIREP reporting (31 + 31 + 28) = 1800 days in 20 winters (from 1 Decem-
practices and because automated measurements and PIREPs ber to 28 February inclusive). Therefore, we have a total of
contain a substantial avoidance bias, which is caused by pi- N = 308 × 1800 = 554 400 data points in each of the prob-
lots attempting to evade the strongest turbulence (Sharman et ability distributions shown in Fig. 1. It follows that each
al., 2014). of the pre-industrial probability distributions contains ap-
We may now apply the percentile ranges listed in Table 1 proximately 12 000 light turbulence data points, 2800 light-
to the pre-industrial probability distributions shown in Fig. 1. to-moderate data points, 1100 moderate data points, 550
By doing so, we may infer the onset threshold for each moderate-to-severe data points, and 550 severe data points.
strength category and each turbulence diagnostic. For N data Therefore, each turbulence strength category is well sam-
points, we compute the value of a given percentile using the pled. The onset thresholds calculated in the above manner
method of linear interpolation between closest ranks, as fol- are listed in Table 2. Note that the thresholds are dependent
lows. First, the N data points are sorted into ascending or- on the grid resolution of the atmospheric model. Therefore,
der and taken to be the values of the (100 × 0.5/N)th, (100 × the values listed in Table 2 may differ from those computed
1.5/N)th, . . ., [100 × (N − 0.5)/N]th percentiles. Then, linear in other studies (e.g., Sharman et al., 2006). For example,
interpolation is used to compute the value of any given per- the Colson–Panofsky thresholds in Table 2 are negative, be-

Table 1. The defining characteristics of six turbulence strength categories for a large commercial aircraft. EDR is the eddy dissipation
rate and g is the acceleration due to gravity. The vertical acceleration assumes proportionality to EDR1/3 , subject to the onset of severe
turbulence occurring at 1.0g. The percentile ranges and probabilities are calculated using an assumed log-normal probability distribution,
as described in the text.

Turbulence strength category Null Light Light-to-moderate Moderate Moderate-to-severe Severe

EDR1/3 range (m2/3 s−1 ) 0–0.1 0.1–0.2 0.2–0.3 0.3–0.4 0.4–0.5 >0.5
Vertical acceleration range (g) 0–0.2 0.2–0.4 0.4–0.6 0.6–0.8 0.8–1.0 >1.0
Percentile range (%) 0–97.0 97.0–99.1 99.1–99.6 99.6–99.8 99.8–99.9 99.9–100
Probability (%) 97.0 2.1 0.5 0.2 0.1 0.1

Table 2. Onset thresholds for each turbulence strength category and each clear-air turbulence diagnostic. The thresholds are calculated by
applying the percentile ranges listed in Table 1 to the pre-industrial probability distributions shown in Fig. 1, as described in the text. The
thresholds are for turbulence diagnosed from the GFDL-CM2.1 climate model and apply to large commercial aircraft. In the units column,
kt is knots and 1 PVU is 10−6 m2 s−1 K kg−1 .

Light-to- Moderate-to-
Diagnostic Units Light moderate Moderate severe Severe
Negative Richardson number — −15.4 −9.8 −7.9 −6.7 −5.9
Magnitude of vertical shear of horizontal wind 10−3 s−1 5.3 6.6 7.4 7.9 8.4
Colson–Panofsky index 103 kt2 −29.3 −27.0 −25.2 −23.7 −22.2
Frontogenesis function 10−9 m2 s−3 K−2 770 1280 1660 1980 2340
Brown index 10−6 s−1 99 106 110 113 118
Brown energy dissipation rate 10 J kg−1 s−1
−6 870 1370 1730 2030 2330
Variant 1 of Ellrod’s turbulence index 10−9 s−2 195 292 360 419 472
Variant 2 of Ellrod’s turbulence index 10−9 s−2 184 282 356 419 477
Flow deformation 10−6 s−1 50.9 60.9 66.9 71.8 76.3
Magnitude of potential vorticity PVU 8.33 8.73 8.98 9.19 9.41
Relative vorticity squared 10−9 s−2 2.46 3.74 4.70 5.50 6.24
Magnitude of horizontal temperature gradient 10−6 K m−1 14.7 17.6 19.4 20.8 22.0
Wind speed m s−1 40.9 48.4 52.4 55.3 58.5
Wind speed × directional shear 10 rad s−1
−3 3.21 3.94 4.39 4.72 5.08
Flow deformation × wind speed 10−3 m s−2 1.65 2.29 2.76 3.17 3.54
Flow deformation × vertical temperature gradient 10−9 K m−1 s−1 53 84 106 127 151
Magnitude of residual of nonlinear balance equation 10−12 s−2 1230 1840 2270 2610 2960
Magnitude of horizontal divergence 10−6 s−1 11.9 15.7 18.2 20.4 22.5
Version 1 of North Carolina State University index 10−18 s−3 1200 3600 6300 9300 13 000
Negative absolute vorticity advection 10−9 s−2 1.33 1.86 2.23 2.56 2.93
Magnitude of relative vorticity advection 10−9 s−2 1.44 1.99 2.34 2.66 3.00
MAY 2017 WILLIAMS 581

cause the Colson–Panofsky index is proportional to 1−Ri/0.5, examination of the right-hand tails of the probability distri-
and the Richardson number (Ri) is rarely less than 0.5 in the butions shown in Fig. 1. Close-ups of the right-hand tails are
GFDL-CM2.1 model. shown in Fig. 2. As anticipated, the probability distributions
The onset thresholds listed in Table 2 facilitate a detailed for the ensemble of 21 clear-air turbulence diagnostics gen-
0
10 0
10
0
Probability (%)

Probability (%)
10

Probability (%)
−1
10
−1
−2 10 −1
10 10
PRE−INDUSTRIAL
−3 DOUBLED CO2
10 −2 −2
10 10
−20 −15 −10 −5 5 6 7 8 9 −30 −28 −26 −24 −22 −20
−3 −1 3 2
Negative Richardson number Magnitude of vertical shear of horizontal wind (10 s ) Colson−Panofsky index (10 kt )

0 0
10 0
10
Probability (%)

Probability (%)

Probability (%)
10

−1 −1
−1
10 10 10

−2
10 −2
−2 10
10
1000 1500 2000 2500 90 100 110 120 130 1000 1500 2000 2500
Frontogenesis function (10−9 m2 s−3 K−2) Brown index (10−6 s−1) Brown energy dissipation rate (10−6 J kg−1 s−1)

0
0 0 10
10
Probability (%)

Probability (%)

Probability (%)
10

−1
−1 −1
10 10
10

−2 −2
−2 10
10 10

200 300 400 500 200 300 400 500 50 60 70 80

−9 −2 −9 −2 −6 −1
Variant 1 of Ellrod’s turbulence index (10 s ) Variant 2 of Ellrod’s turbulence index (10 s ) Flow deformation (10 s )

0 0 0
10 10 10
Probability (%)

Probability (%)

Probability (%)

−1
10 −1
10
−1
10

−2
10 −2 −2
10 10
8 8.5 9 9.5 10 2 3 4 5 6 7 12 14 16 18 20 22 24
−9 −2 −6 −1
Magnitude of potential vorticity (PVU) Relative vorticity squared (10 s ) Magnitude of horizontal temperature gradient (10 K m )

0 0 0
10 10 10
Probability (%)

Probability (%)

Probability (%)

−1
10 −1 −1
10 10

−2
10 −2 −2
10 10
35 40 45 50 55 60 65 3 3.5 4 4.5 5 5.5 1.5 2 2.5 3 3.5 4
−1
Wind speed (m s ) Wind speed × directional shear (10−3 rad s−1) Flow deformation × wind speed (10−3 m s−2)

0
0 0 10
10 10
Probability (%)

Probability (%)

Probability (%)

−1 −1
10 −1 10
10

−2 −2
10 −2 10
10
50 100 150 1000 1500 2000 2500 3000 3500 10 15 20 25
−9 −1 −1
Flow deformation × vertical temperature gradient (10 K m s ) Magnitude of residual of NBE (10−12 s−2) Magnitude of horizontal divergence (10−6 s−1)

2
10
0 0
Probability (%)

Probability (%)

Probability (%)

1
10 10 10
0
10 −1 −1
10 10
−1
10
−2 −2 −2
10 10 10

0 5000 10000 15000 20000 1 1.5 2 2.5 3 3.5 1 1.5 2 2.5 3 3.5
−18 −3 −9 −2 −9 −2
Version 1 of North Carolina State University index (10 s ) Negative absolute vorticity advection (10 s ) Magnitude of relative vorticity advection (10 s )

Fig. 2. Same as Fig. 1, but showing enlarged views of the right-hand tails of the probability distributions. The probability axes
are logarithmic. On each turbulence diagnostic axis, the five markers indicate the onset thresholds for light, light-to-moderate,
moderate, moderate-to-severe, and severe turbulence, as listed in Table 2.
582 INCREASED TURBULENCE IN RESPONSE TO CLIMATE CHANGE VOLUME 34

erally gain probability in their right-hand tails when the CO2 than weaker turbulence, but also implying a higher degree
is doubled. The onset thresholds for each turbulence strength of uncertainty.
category are indicated by markers on the turbulence diagnos- A geographic map of the spatial distribution of the chang-
tic axes in Fig. 2. Reference to the markers indicates that the ing prevalence of light-or-greater clear-air turbulence in the
increase in probability occurs in nearly all of the diagnostics North Atlantic region is shown in Fig. 4. Similar maps
and nearly all of the strength categories. The increased prob- for moderate-or-greater and severe-or-greater clear-air turbu-
abilities represent increases in the volume of airspace con- lence are shown in Fig. 5 and Fig. 6, respectively. Within
taining turbulence. the regions delineated by the black dashed lines in these fig-
The probability distributions shown in Fig. 2 may be in- ures, the agreement between the 21 diagnostics on the sign
tegrated between each pair of adjacent markers, enabling the of the change is statistically significant at the 90% level. The
probability gains to be calculated per strength category and significance here is assessed using the binomial distribution,
per turbulence diagnostic. These gains, expressed as per- under the null hypothesis that each of the 21 diagnostics is
centage increases in the doubled-CO2 climate relative to the independent and equally likely to increase or decrease, fol-
pre-industrial climate, are shown in Fig. 3. Most of the 21 lowing Williams and Joshi (2013). Within most of the North
diagnostics show increased turbulence in all five strength Atlantic region, a statistically significant majority of the di-
categories. Specifically, all 21 diagnostics show increases agnostics agree on an increased frequency of light-or-greater
in the amount of light and light-to-moderate turbulence, turbulence. For moderate-or-greater turbulence, the area of
and at least 16 of the 21 diagnostics show increases in the significant agreement is reduced, but it still occupies most
amount of moderate, moderate-to-severe, and severe turbu- of the North Atlantic region north of 50◦ N. For severe-or-
lence. To summarise the 21 different estimates of the per- greater turbulence, the area of significant agreement is fur-
centage increase within each strength category, we calcu- ther reduced, but it still occupies most of the region between
late the median (50th percentile) and 25th–75th percentiles, Canada, Greenland, and the UK, which is a very busy part of
which respectively indicate an ensemble-average value and the North Atlantic flight corridor.
an intra-ensemble range. By these measures, the prevalence
of light turbulence increases by 59% (43%–68%), light-to-
moderate by 75% (39%–96%), moderate by 94% (37%– 3. Summary and discussion
118%), moderate-to-severe by 127% (30%–170%), and se- This paper has used climate model simulations to analyse
vere by 149% (36%–188%). The averages and ranges both the transatlantic wintertime clear-air turbulence response to
increase substantially from light to severe turbulence, sug- climate change in five aviation-relevant turbulence strength
gesting greater percentage increases in stronger turbulence categories. We have found that the probability distributions

400
Negative Richardson number
Magnitude of vertical shear of horizontal wind
Brown energy dissipation rate
Variant 2 of Ellrod’s turbulence index
350 Variant 1 of Ellrod’s turbulence index
Magnitude of horizontal temperature gradient
Relative vorticity squared
Version 1 of North Carolina State University index
Increase in amount of turbulence (%)

300 Magnitude of residual of nonlinear balance equation

Brown index
Wind speed
Flow deformation
250 Flow deformation × wind speed
Magnitude of horizontal divergence
Magnitude of potential vorticity
Frontogenesis function
200 Flow deformation × vertical temperature gradient
Magnitude of relative vorticity advection
Wind speed × directional shear
Negative absolute vorticity advection
150 Colson−Panofsky index

100

50

−50
Light Light−to−moderate Moderate Moderate−to−severe Severe

Fig. 3. Bar charts showing the percentage increase in the amount of light, light-to-moderate, moderate, moderate-to-
severe, and severe clear-air turbulence within the North Atlantic flight corridor at 200 hPa in winter. The increase
refers to the change in a doubled-CO2 simulation compared to a pre-industrial simulation. The 21 clear-air turbulence
diagnostics are ordered within each strength category according to the magnitude of the change in the light category.
MAY 2017 WILLIAMS 583

80N 21 80N 21
20 20
19 19
18 18
17 17
16 16
15 15
60N 14 60N 14
13 13
12 12
11 11
10 10
9 9
8 8
40N 7 40N 7
6 6
5 5
4 4
3 3
2 2
1 1
20N 0 20N 0
80W 60W 40W 20W 0 20E 80W 60W 40W 20W 0 20E

Fig. 4. Spatial distribution of the number of the 21 clear-air tur- Fig. 6. Same as Fig. 4, but for severe-or-greater turbulence.
bulence diagnostics to show an increase in the frequency of
light-or-greater clear-air turbulence at 200 hPa in winter. The depends on a number of other factors in addition to the
increase refers to the change in a doubled-CO2 simulation com- strength and frequency of atmospheric turbulence. One factor
pared to a pre-industrial simulation. Red shading indicates that is the skill of operational clear-air turbulence forecasts, such
most of the diagnostics show an increase, and blue shading indi- as the Graphical Turbulence Guidance system (Sharman et
cates that most show a decrease. The black dashed lines, which
al., 2006). Any future improvements in these forecasting al-
are contours at 7 and 14, delineate the regions in which at least
two-thirds of the diagnostics agree on the sign of the change.
gorithms would improve the ability of pilots to divert around
patches of clear-air turbulence instead of unexpectedly en-
countering them. Another factor is the dynamical response
80N 21 of aircraft to turbulence. Some modern aircraft are fitted with
20
19 an accelerometer in their nose cone. If the accelerometer reg-
18 isters a sudden change in altitude, which is large enough to
17
16 be indicative of turbulence, then the wing flaps are rapidly ad-
15
60N 14 justed in an attempt to damp the vertical motion and reduce
13 the acceleration. It is only in the absence of these changing
12
11 extraneous factors that the quantitative turbulence increases
10 calculated herein would be expected to translate proportion-
9
8 ately into stronger and more frequent turbulence encounters
40N 7
6 by aircraft.
5 Another potentially relevant factor is the development of
4
3 on-board technology to detect clear-air turbulence. In prin-
2
1 ciple, the increased risk of clear-air turbulence encounters in
20N 0 future could be mitigated by equipping aircraft with Light
80W 60W 40W 20W 0 20E Detection and Ranging (LIDAR) ultra-violet laser systems
(Vrancken et al., 2016). Forward-looking LIDAR could fore-
Fig. 5. Same as Fig. 4, but for moderate-or-greater turbulence. warn pilots of any invisible density perturbations indicative
of clear-air turbulence up to 10–15 km ahead, potentially with
for an ensemble of 21 clear-air turbulence diagnostics gener- enough lead time to alert passengers and crew or even to at-
ally gain probability in their right-hand tails when the atmo- tempt an evasive manoeuvre. Kauffmann (2002) has calcu-
spheric CO2 concentration is doubled. By converting the di- lated that the business case for installing LIDAR technology
agnostics into eddy dissipation rates, we have found substan- is currently negative. In other words, the economic costs of
tial increases in the airspace volume containing light, light-to- retro-fitting aircraft currently exceed the economic benefits
moderate, moderate, moderate-to-severe, and severe clear-air of avoiding clear-air turbulence. However, it is likely that the
turbulence. We conclude that the amount of transatlantic win- business case will improve in future, as the LIDAR technol-
tertime clear-air turbulence in the atmosphere will increase ogy becomes less expensive and clear-air turbulence becomes
significantly in all aviation-relevant strength categories as the more prevalent.
climate changes. In addition to the impacts on aviation, any increase in
The projected increases in the prevalence of clear-air tur- clear-air turbulence could also have important consequences
bulence do not necessarily imply more in-flight injuries or for atmospheric dynamics and thermodynamics. Clear-air
greater levels of passenger discomfort. Aircraft bumpiness turbulence mixes atmospheric constituents in the stratosphere
584 INCREASED TURBULENCE IN RESPONSE TO CLIMATE CHANGE VOLUME 34

(Lilly et al., 1974) and across the tropopause (Shapiro, 1980; REFERENCES
Traub and Lelieveld, 2003; Karpechko et al., 2007). There- Atlas, D., J. I. Metcalf, J. H. Richter, and E. E. Gossard, 1970: The
fore, any increase in clear-air turbulence would increase birth of “CAT” and microscale turbulence. J. Atmos. Sci., 27,
the troposphere–stratosphere exchange of radiatively signif- 903–913.
icant constituents such as ozone and water vapour. This Brown, R., 1973: New indices to locate clear-air turbulence. Me-
increased exchange could modify the temperature structure teor. Mag., 102, 347–361.
of the tropopause region (Forster and Shine, 2002), which Burbidge, R., 2016: Adapting European airports to a changing cli-
could have dynamical consequences throughout the tropo- mate. Transportation Research Procedia, 14, 14–23.
sphere (Maycock et al., 2013). These possible impacts of Clark, T. L., W. D. Hall, R. M. Kerr, D. Middleton, L. Radke, F.
increased clear-air turbulence on the large-scale atmospheric M. Ralph, P. J. Neiman, and D. Levinson, 2000: Origins of
circulation have not yet been explored but should be a priority aircraft-damaging clear-air turbulence during the 9 December
1992 Colorado downslope windstorm: Numerical simulations
for future research.
and comparison with observations. J. Atmos. Sci., 57, 1105–
Future work should extend our study by examining other 1131.
seasons, flight levels, and geographic regions. The sensi- Coffel, E., and R. Horton, 2015: Climate change and the impact
tivity of our results to the climate model resolution should of extreme temperatures on aviation. Weather, Climate, and
also be explored. It would be informative to compare tur- Society, 7, 94–102.
bulence diagnosed from climate models with in-flight turbu- Colson, D., and H. A. Panofsky, 1965: An index of clear air turbu-
lence measurements and also with turbulence diagnosed from lence. Quart. J. Roy. Meteor. Soc., 91, 507–513.
reanalysis data. These comparisons are likely to require in- Delcambre, S. C., D. J. Lorenz, D. J. Vimont, and J. E. Martin,
terdisciplinary collaborations between climate scientists, tur- 2013: Diagnosing northern hemisphere jet portrayal in 17
bulence scientists, and airlines. Scientific and socioeconomic CMIP3 global climate models: Twenty-first-century projec-
uncertainties should be quantified by using different climate tions. J. Climate, 26, 4930–4946.
Ellrod, G. P., and D. I. Knapp, 1992: An objective clear-air turbu-
models and greenhouse-gas scenarios, respectively. Another
lence forecasting technique: Verification and operational use.
source of uncertainty is that future lower-stratospheric tem- Wea. Forecasting, 7, 150–165.
perature changes, which influence the meridional tempera- Endlich, R. M., 1964: The mesoscale structure of some regions of
ture gradient at aircraft cruising altitudes in the midlatitudes, clear-air turbulence. J. Atmos. Sci., 3, 261–276.
depend on warming from ozone recovery in addition to cool- FAA, 2006: Preventing injuries caused by turbulence. Advisory
ing from greenhouse gases (Shine et al., 2003). Convection Circular 120-88A, Federal Aviation Administration, Wash-
may interact with clear-air turbulence (Trier et al., 2012; Trier ington, DC.
and Sharman, 2016) and the response of this interaction to Forster, P. M. de F., and K. P. Shine, 2002: Assessing the climate
climate change should be investigated. Finally, future stud- impact of trends in stratospheric water vapor. Geophys. Res.
ies should investigate the response of severe-to-extreme and Lett., 29, 1086.
extreme clear-air turbulence to climate change. Turbulence Francis, J. A., and S. J. Vavrus, 2012: Evidence linking Arctic am-
plification to extreme weather in mid-latitudes. Geophys. Res.
in these strength categories is so rare that very long climate
Lett., 39, L06801.
model simulations will be required in order to generate robust Frehlich, R., and R. Sharman, 2004: Estimates of turbulence from
statistics, presenting a technical challenge that will need to be numerical weather prediction model output with applications
overcome. to turbulence diagnosis and data assimilation. Mon. Wea. Rev.,
132, 2308–2324.
Acknowledgements. The author is financially supported Goessling, H. F., and S. Bathiany, 2016: Why CO2 cools the mid-
through a University Research Fellowship from the Royal Society dle atmosphere—A consolidating model perspective. Earth
(reference UF130571). He thanks Jenny LIN for her invitation and System Dynamics, 7, 697–715.
encouragement to write an article on aviation turbulence. The au- ICAO, 2015: International Civil Aviation Organiza-
thor acknowledges the modelling groups, the Program for Climate tion (ICAO) Second High-Level Safety Conference
Model Diagnosis and Intercomparison, and the World Climate Re- (HLSC), Montreal, Canada, 2–5 February 2015, Work-
ing Paper on Extreme Meteorological Conditions
search Programme’s Working Group on Coupled Modelling for their
(HLSC/15-WP/36). [Available online from www.icao.int/
roles in making available the climate model data. Support of this
Meetings/HLSC2015/Documents/WP/wp036 en.pdf].
dataset is provided by the Office of Science, US Department of En- Irvine, E. A., B. J. Hoskins, K. P. Shine, R. W. Lunnon, and
ergy. The constructive comments of four reviewers are gratefully C. Froemming, 2013: Characterizing North Atlantic weather
acknowledged. patterns for climate-optimal aircraft routing. Meteorological
Applications, 20, 80–93.
Open Access. This article is distributed under the terms of the Irvine, E. A., K. P. Shine, and M. A. Stringer, 2016: What are the
Creative Commons Attribution 4.0 International License (http://cre- implications of climate change for trans-Atlantic aircraft rout-
ativecommons.org/licenses/by/4.0/), which permits unrestricted use, ing and flight time? Transportation Research Part D: Trans-
distribution, and reproduction in any medium, provided you give ap- port and Environment, 47, 44–53.
propriate credit to the original author(s) and the source, provide a Jaeger, E. B., and M. Sprenger, 2007: A Northern Hemispheric cli-
matology of indices for clear air turbulence in the tropopause
link to the Creative Commons license, and indicate if changes were
region derived from ERA40 reanalysis data. J. Geophys. Res.,
made.
MAY 2017 WILLIAMS 585

112(D20), D20106. provement in clear-air turbulence forecasting based on spon-

Karnauskas, K. B., J. P. Donnelly, H. C. Barkley, and J. E. Martin, taneous imbalance theory: The ULTURB algorithm. Meteo-
2015: Coupling between air travel and climate. Nature Clim. rological Applications, 19, 71–78.
Change, 5, 1068–1073. Meehl, G. A., and Coauthors, 2007: Global climate projections.
Karpechko, A., A. Lukyanov, E. Kyrö, S. Khaikin, L. Kor- Climate Change 2007: The Physical Science Basis. Contri-
shunov, R. Kivi, and H. Vömel, 2007: The water vapour bution of Working Group I to the Fourth Assessment Report
distribution in the Arctic lowermost stratosphere during the of the Intergovernmental Panel on Climate Change, Solomon
LAUTLOS campaign and related transport processes includ- et al., Eds., Cambridge University Press, 747–846.
ing stratosphere–troposphere exchange. Atmos. Chem. Phys., Meneguz, E., H. Wells, and D. Turp, 2016: An automated sys-
7, 107–119. tem to quantify aircraft encounters with convectively induced
Kauffmann, P., 2002: The business case for turbulence sensing sys- turbulence over Europe and the Northeast Atlantic. J. Appl.
tems in the US air transport sector. Journal of Air Transport Meteor. Climatol., 55, 1077–1089.
Management, 8, 99–107. Penner, J. E., D. H. Lister, D. J. Griggs, D. J. Dokken, and M. Mc-
Kim, B., and Coauthors, 2005: System for assessing Aviation’s Farland, 1999: Intergovernmental Panel on Climate Change
Global Emissions (SAGE), Version 1.5, Technical Manual (IPCC) special report: Aviation and the global atmosphere.
FAA-EE-2005-01, Federal Aviation Administration, Wash- Cambridge University Press, 373 pp.
ington, DC. Puempel, H., and P. D. Williams, 2016: The impacts of climate
Kim, J.-H., and H.-Y. Chun, 2011: Statistics and possible sources change on aviation: Scientific challenges and adaptation path-
of aviation turbulence over South Korea. J. Appl. Meteor. Cli- ways. ICAO Environmental Report 2016: On Board A Sus-
matol., 50, 311–324. tainable Future, 205–207.
Kim, J.-H., W. N. Chan, B. Sridhar, R. D. Sharman, P. D. Williams, Reiter, E. R., 1963: Jet-stream Meteorology. University of Chicago
and M. Strahan, 2016: Impact of the North Atlantic Oscilla- Press, 515 pp.
tion on transatlantic flight routes and clear-air turbulence. J. Riddaway, R. W., 1998: Notes and news. Meteorological Applica-
Appl. Meteor. Climatol., 55, 763–771. tions, 5, 183–188.
Kim, S.-H., and H.-Y. Chun, 2016: Aviation turbulence encounters Romps, D. M., J. T. Seeley, D. Vollaro, and J. Molinari, 2014: Pro-
detected from aircraft observations: Spatiotemporal charac- jected increase in lightning strikes in the United States due to
teristics and application to Korean Aviation Turbulence Guid- global warming. Science, 346, 851–854.
ance. Meteorological Applications, 23, 594–604. Schwartz, B., 1996: The quantitative use of PIREPs in develop-
Knox, J. A., D. W. McCann, and P. D. Williams, 2008: Applica- ing aviation weather guidance products. Wea. Forecasting, 11,
tion of the Lighthill–Ford theory of spontaneous imbalance 372–384.
to clear-air turbulence forecasting. J. Atmos. Sci., 65, 3292– Shapiro, M. A., 1980: Turbulent mixing within tropopause folds
3304. as a mechanism for the exchange of chemical constituents be-
Krozel, J., V. Klimenko, and R. Sharman, 2011: Analysis of tween the stratosphere and troposphere. J. Atmos. Sci., 37,
clear-air turbulence avoidance maneuvers. Air Traffic Control 994–1004.
Quarterly, 19, 147–168. Sharman, R., C. Tebaldi, G. Wiener, and J. Wolff, 2006: An inte-
Lane, T. P., R. D. Sharman, S. B. Trier, R. G. Fovell, and J. K. grated approach to mid- and upper-level turbulence forecast-
Williams, 2012: Recent advances in the understanding of ing. Wea. Forecasting, 21, 268–287.
near-cloud turbulence. Bull. Amer. Meteor. Soc., 93, 499–515. Sharman, R. D., and J. M. Pearson, 2017: Prediction of energy dis-
Lee, D. S., D. W. Fahey, P. M. Forster, P. J. Newton, R. C. N. sipation rates for aviation turbulence. Part I: Forecasting non-
Wit, L. L. Lim, B. Owen, and R. Sausen, 2009: Aviation and convective turbulence. J. Appl. Meteor. Climatol., 56, 317–
global climate change in the 21st century. Atmos. Environ., 337.
43, 3520–3537. Sharman, R. D., S. B. Trier, T. P. Lane, and J. D. Doyle, 2012:
Lee, L., 2013: A climatological study of clear air turbulence over Sources and dynamics of turbulence in the upper troposphere
the North Atlantic. Master’s thesis, Dept. of Earth Sciences, and lower stratosphere: A review. Geophys. Res. Lett., 39,
Uppsala University. L12803.
Lilly, D. K., D. E. Waco, and S. I. Adelfang, 1974: Strato- Sharman, R. D., L. B. Cornman, G. Meymaris, J. Pearson, and
spheric mixing estimated from high-altitude turbulence mea- T. Farrar, 2014: Description and derived climatologies of au-
surements. J. Appl. Meteor., 13, 488–493. tomated in situ eddy-dissipation-rate reports of atmospheric
Lorenz, D. J., and E. T. DeWeaver, 2007: Tropopause height and turbulence. J. Appl. Meteor. Climatol., 53, 1416–1432.
zonal wind response to global warming in the IPCC scenario Shine, K. P., and Coauthors, 2003: A comparison of model-
integrations. J. Geophys. Res., 112, D10119. simulated trends in stratospheric temperatures. Quart. J. Roy.
MacCready, P. B., 1964: Standardization of gustiness values from Meteor. Soc., 129, 1565–1588.
aircraft. J. Appl. Meteor., 3, 439–449. Simpson, I. R., 2016: Climate change predicted to lengthen
Malwitz, A., and Coauthors, 2005: System for assessing Avi- transatlantic travel times. Environ. Res. Lett., 11, 031002.
ation’s Global Emissions (SAGE), Version 1.5, Validation Solomon, A., and L. M. Polvani, 2016: Highly significant re-
Assessment, Model Assumptions and Uncertainties. Tech. sponses to anthropogenic forcings of the midlatitude jet in the
Rep. FAA-AA-EE-2005-03, Federal Aviation Administra- Southern Hemisphere. J. Climate, 29, 3463–3470.
tion, Washington, DC. Traub, M., and J. Lelieveld, 2003: Cross-tropopause transport over
Maycock, A. C., M. M. Joshi, K. P. Shine, and A. A. Scaife, 2013: the eastern Mediterranean. J. Geophys. Res., 108, 4712.
The circulation response to idealized changes in stratospheric Trier, S. B., and R. D. Sharman, 2016: Mechanisms influencing
water vapor. J. Climate, 26, 545–561. cirrus banding and aviation turbulence near a convectively
McCann, D. W., J. A. Knox, and P. D. Williams, 2012: An im- enhanced upper-level jet stream. Mon. Wea. Rev., 144, 3003–
586 INCREASED TURBULENCE IN RESPONSE TO CLIMATE CHANGE VOLUME 34

3027. Williams, P. D., 2016: Transatlantic flight times and climate

Trier, S. B., R. D. Sharman, and T. P. Lane, 2012: Influences of change. Environ. Res. Lett., 11, 024008.
moist convection on a cold-season outbreak of Clear-Air Tur- Williams, P. D., and M. M. Joshi, 2013: Intensification of win-
bulence (CAT). Mon. Wea. Rev., 140, 2477–2496. ter transatlantic aviation turbulence in response to climate
Vallis, G. K., P. Zurita-Gotor, C. Cairns, and J. Kidston, 2015: Re- change. Nature Clim. Change, 3, 644–648.
sponse of the large-scale structure of the atmosphere to global Williams, P. D., and M. M. Joshi, 2016: Clear-air turbulence in a
warming. Quart. J. Roy. Meteor. Soc., 141, 1479–1501. changing climate. Aviation Turbulence: Processes, Detection,
Vrancken, P., and Coauthors, 2016: Flight tests of the DELICAT Prediction, R. Sharman and T. Lane, Eds., Springer Interna-
airborne LIDAR system for remote clear air turbulence detec- tional Publishing, 465–480.
tion. EPJ Web of Conferences, 119, 14003. Williams, P. D., P. L. Read, and T. W. N. Haine, 2003: Spontaneous
Warner, M., 2013: Boeing: Current market outlook 2013–2032. generation and impact of inertia–gravity waves in a stratified,
Boeing Commercial Airplanes, Seattle, WA. two-layer shear flow. Geophys. Res. Lett., 30, 2255.
Watkins, C. D., and K. A. Browning, 1973: The detection of clear Williams, P. D., T. W. N. Haine, and P. L. Read, 2005: On the
air turbulence by radar. Physics in Technology, 4, 28–61. generation mechanisms of short-scale unbalanced modes in
Wilcox, L. J., K. P. Shine, and B. J. Hoskins, 2012: Radiative forc- rotating two-layer flows with vertical shear. J. Fluid Mech.,
ing due to aviation water vapour emissions. Atmos. Environ., 528, 1–22.
63, 1–13. Williams, P. D., T. W. N. Haine, and P. L. Read, 2008: Inertia–
Wilkerson, J. T., M. Z. Jacobson, A. Malwitz, S. Balasubramanian, gravity waves emitted from balanced flow: Observations,
R. Wayson, G. Fleming, A. D. Naiman, and S. K. Lele, 2010: properties, and consequences. J. Atmos. Sci., 65, 3543–3556.
Analysis of emission data from global commercial aviation: Wolff, J. K., and R. D. Sharman, 2008: Climatology of upper-level
2004 and 2006. Atmos. Chem. Phys., 10, 6391–6408. turbulence over the contiguous United States. J. Appl. Meteor.
Williams, J. K., 2014: Using random forests to diagnose aviation Climatol., 47, 2198–2214.
turbulence. Machine Learning, 95, 51–70.