Clim Dyn

DOI 10.1007/s00382-013-1713-0

Recent intense hurricane response to global climate change

Greg Holland • Cindy L. Bruyère

Received: 5 September 2012 / Accepted: 21 February 2013

 The Author(s) 2013. This article is published with open access at Springerlink.com

Abstract An Anthropogenic Climate Change Index 1 Introduction

(ACCI) is developed and used to investigate the potential
global warming contribution to current tropical cyclone Recent community consensus (Knutson et al. 2010; Inter-
activity. The ACCI is defined as the difference between the governmental Panel on Climate Change—IPCC 2012) has
means of ensembles of climate simulations with and concluded that it is likely that the frequency of intense
without anthropogenic gases and aerosols. This index hurricanes will increase with future anthropogenic climate
indicates that the bulk of the current anthropogenic change. The increases are substantial, approaching a dou-
warming has occurred in the past four decades, which bling in frequency of Cat 4 and 5 hurricanes [based on the
enables improved confidence in assessing hurricane chan- Saffir–Simpson classification, Simpson and Rielh (1981)]
ges as it removes many of the data issues from previous for each C in global warming (Bender et al. 2010; IPCC
eras. We find no anthropogenic signal in annual global 2012; Done et al. 2012). IPCC (2007) also concluded that
tropical cyclone or hurricane frequencies. But a strong the current ‘warming of the climate system is unequivocal’.
signal is found in proportions of both weaker and stronger If we accept these two statements, then it logically follows
hurricanes: the proportion of Category 4 and 5 hurricanes that there should already be an observable increase in
has increased at a rate of *25–30 % per C of global intense hurricanes. Yet IPCC (2012) concluded that ‘There
warming after accounting for analysis and observing sys- is low confidence in any observed long-term (i.e., 40 years
tem changes. This has been balanced by a similar decrease or more) increases in tropical cyclone activity’, based lar-
in Category 1 and 2 hurricane proportions, leading to gely on potential errors in the observed data. Here we
development of a distinctly bimodal intensity distribution, investigate this apparent anomaly and find that there has
with the secondary maximum at Category 4 hurricanes. been an increase in the proportion of intense hurricanes
This global signal is reproduced in all ocean basins. The relative to all hurricanes, and that is strongly related to an
observed increase in Category 4–5 hurricanes may not Anthropogenic Climate Change Index (ACCI). The index
continue at the same rate with future global warming. The is developed in Sect. 2; Sect. 3 examines the recent hur-
analysis suggests that following an initial climate increase ricane signal, and discusses some of the consequences; and
in intense hurricane proportions a saturation level will be our conclusions are in Sect. 4.
reached beyond which any further global warming will
have little effect.
2 An index of anthropogenic climate change
Keywords Hurricane
Climate change
Global warming
It is often assumed that anthropogenic climate change has
occurred as a quasi-linear warming trend from around the
turn of the twentieth century. This causes considerable
G. Holland (&)
C. L. Bruyère
problems with assessing the impact of climate change on
National Center for Atmospheric Research,
NESL/NCAR, P.O. Box 3000, Boulder, CO 80307, USA severe weather events such as hurricanes (e.g. Knutson
e-mail: gholland@ucar.edu et al. 2010). However, comparison of global climate model

123
 G. Holland, C. L. Bruyère

simulations with and without anthropogenic forcing this paper are robust to the details of the derivation of the
(Fig. 1) indicates that the net anthropogenic climate forc- ACCI, for example by replacing the anthropogenic simu-
ing—including cooling as well as warming agents—was lations with the observed temperatures. Some earlier
indistinguishable from natural variability until the mid- warming is seen in the more recent CCSM4 simulation
twentieth century (Meehl et al. 2004, 2007, 2012; IPCC (Community Climate System Model, Figs. 1b, 2a), which is
2007). The ensembles of model simulations with and considered to be due to the lack of an indirect effect for
without anthropogenic emissions closely track each other sulfates in the model (Meehl et al. 2012). We also examined
and the observed global surface temperature until around defining ACCI from ensemble information, including tak-
1970. After then, the anthropogenic simulations continue to ing the ensemble extremes and other combinations. Our
track the observed global warming, but those with only conclusions are robust to all of these approaches.
natural forcing indicate a cooling trend back to nineteenth The ACCI data in Fig. 2a indicate that consideration of
century temperature levels. This change has been attributed the impact of anthropogenic climate change should be
to reductions in emissions of sulfate aerosols, which limited to the last four decades. This is quite different to a
counteracted the warming component due to greenhouse number of other studies that have assumed an essentially
gases before 1960 (Meehl et al. 2004, 2012; Mann and linear warming over the past 100 years or so [see, e.g.
Emanuel 2006; Thompson et al. 2010). Knutson et al. (2010) for hurricane implications]. If
We therefore define an ACCI by the difference between greenhouse gases alone are considered then an extended
the model simulations with and without anthropogenic period of nearly linear warming is justified, but anthropo-
gases. When applied to the ensemble means from Fig. 1a, genic impacts on the climate also arise from a range of
the ACCI is essentially zero until around 1960 (Fig. 2a), other contributors that have counteracted the pure green-
after which an exponential change to a quasi-linear increase house gas warming until recently [see Meehl et al. (2012)
of around 0.35 C per decade commenced. The results in for a detailed discussion].

Fig. 1 Ensemble simulations of annual-mean global surface temper-

ature with (red) and without (blue) anthropogenic gas forcing, Fig. 2 a ACCI calculated from the differences between the ensemble
together with the observed global surface temperatures (black); thin annual means in Fig. 1 with the solid line being a cubic fit to the
lines indicate ensemble extremes: a Based on CMIP3 (adapted from annual values; b Relationship between the ACCI and annual global
Fig. 2d of Meehl et al. 2004); b based on CCSM4 (from Fig. 2 of tropical SST anomalies (±30 Lat) together with linear trends, both of
Meehl et al. 2012) the indicated variances have p \ 0.01

123
Recent intense hurricane

For the global hurricane analysis we choose to focus on 2010) but there remain issues with the approaches used to
the slightly shorter period from 1975 to 2010, as this is a convert from the satellite pattern code to a surface wind
reasonably homogeneous period of global satellite data; speed.
going back to 1960 does not affect our conclusions. CMIP5 These issues are particularly evident in the western
(Coupled Model Intercomparison Project 5) data were not North Pacific, where there are four different archival cen-
available at the analysis time, so to enable use of the full ters: the Joint Typhoon Warning Center (JTWC), the Hong
tropical cyclone record, we use the CMIP3 ACCI up to Kong Observatory (HKO), the Shanghai Typhoon Institute
2000 and extend this to 2010 using the annual changes in (STI) and the WMO Regional Specialized Meteorological
CCSM4. A test using linear extrapolation of the CMIP3 Center, Tokyo (RSMC). Maximum winds for each of these
data to 2010 produced essentially the same results. archives are based on: 1-min mean at JTWC, 10-min mean
Figure 2b compares the ACCI with the observed annual at HKO and RSMC, and 2-min mean at STI. Furthermore,
tropical (30 north and south) sea surface temperature (SST) aircraft reconnaissance provided a substantial basis for
anomalies relative to the mean for 1975–2010 from the intensity estimates until 1988 after which the Dvorak
merged Hadley Centre and NOAA’s optimum interpolation technique became the major approach. Varying approaches
(OI) SST data set (Hurrell et al. 2008). They track each other to converting the Dvorak analysis to intensity has resulted
closely with the ACCI explaining [50 % of the observed in a divergence of intensity trends amongst the centers,
tropical SST variance (p \ 0.01). Note that the slope in with the JTWC data containing an increasing intensity
Fig. 2b is \1, since the ACCI values have an additional trend and the HKO and RSMC a decreasing one (Wu et al.
component arising from the net cooling in the natural forcing 2006; Song et al. 2010).
simulations (Fig. 1) compared to the observations. Wu and Zhao (2012) examined these discrepancies
Focusing on the relationship with ACCI from 1975 using the Emanuel (2008) statistical-dynamical downscal-
removes much of the hurricane analysis difficulty that has ing technique. They found an increasing intensity trend that
been experienced with issues such as changing observing also is supported by observed relationships with SST and
systems and analysis approaches prior to 1975 (e.g. vertical shear, and is in broad agreement with the observed
Landsea et al. 2006; Knutson et al. 2010; Holland and increasing tendency for equatorial developments and
Webster 2007). The approach of using a global index also longer time spent in favorable conditions. This lends sup-
is consistent with the findings of Grinsted et al. (2012), port to the JTWC archive, though the Wu and Zhao trends
who showed that North Atlantic landfalling hurricane were less than those for JTWC.
variations are more closely related to global mean SST than Based on this analysis, we adopt the following proce-
they are to local SST or climate indices. We shall show that dure for assessing hurricane trends:
using a shorter period of record is not a concern as the
• Except for the western North Pacific, we use the
signal is large and significant.
IBTrACS data without change. IBTrACS has been
converted to a uniform 10-min mean wind definition
3 Global tropical cyclone changes (Knapp et al. 2010). For the western North Pacific, we
use the JTWC data set without correction aside from
3.1 Data conversion to 10-min mean.
• To help minimize uncertainties, including the intensity
The tropical cyclone analysis is based on the global definition variations, we bin all hurricanes into the
IBTrACS data set (Knapp et al. 2010). A homogeneous Saffir–Simpson category ranges (Simpson and Rielh
satellite reanalysis by Kossin et al. (2007) found close 1981). This is for consistency with other studies. The
agreement of the archived intensities in the North Atlantic Saffir–Simpson ranges are not constant, varying
and western North Pacific, but there were inconsistencies slightly from one category to another, but a check
for more intense storms in the other regions. We utilize the using fixed-width bins found no notable differences.
recent update to this analysis (Kossin 2012) as an inde- Even though the Cat 4–5 interval has no explicit upper
pendent check on the IBTrACS results. bound, it is effectively bounded by available energy as
Archived intensities are somewhat of a hodge-podge of defined by the potential intensity (Emanuel 1987) and is
different approaches and definitions. Even at the most basic approximately the same width as Cat 1–2.
level, the definition of wind speed varies considerably • All variance numbers use the 5-years smoothed annual
around the globe, with 1, 10, 3 and even 2-min averaging time series to remove ENSO type variability. These
periods in use (Harper et al. 2008). Some homogenization include contributions from both trend and multi-year
is provided by the universal use of the Dvorak satellite variance. The p values are calculated from the raw
interpretation technique (Velden et al. 2006; Knaff et al. annual data to ensure no serial correlation.

123
 G. Holland, C. L. Bruyère

We also conduct several additional data quality checks:

We recalculate our main conclusions from the IBTrACS
data using:
• The recently upgraded satellite data set and auto-
mated analysis from Kossin (2012), and
• A land-proximity dataset derived from the global
IBTrACS data. These are interpolated to add nine
points between each 6-h position in the archive;
land proximity is then defined by whenever the
tropical cyclone moves to within 0.5 lat/long of
land, with the land mask having 1 min resolution
(*2 km);
• An independent global landfall data set used in a
recent study (Weinkle et al. 2012).
We further compare the observed changes to recent
climate simulations with the dynamical downscaling
results of Bender et al. (2010), Done et al. (2012), and
Wu and Zhao (2012).

3.2 Tropical cyclone frequency

The global annual frequency of all tropical cyclones

Fig. 3 Anthropogenic influence on: a annual frequency of global
(Fig. 3a) has remained essentially constant with ACCI,
tropical cyclones and hurricanes; b hurricane proportions in each of
aside from a general variability with amplitude around the Saffir–Simpson hurricane categories
10 % from the mean for the whole period. This remarkable
degree of global consistency has been noted in several
previous studies [see Frank and Young (2007) for a sum- decrease of Cat 1 hurricanes for increasing ACCI; Cat 2
mary]. Note that the lack of trend does not carry over to hurricanes have little relationship for lower ACCI values,
individual ocean basins, some of which have experienced but decrease towards higher ACCI; little real change is
trends since 1975 with the standouts being a substantial observed in Cat 3 aside from an initial slight increase and
decrease in the western North Pacific and an increase in the then similar decline; and Cat 4 and 5 hurricanes both
North Atlantic (Webster et al. 2005). We imply no causes increase markedly with ACCI. Based on these results we
for these in this study; they could be due to either or both of further bin the hurricane proportions into Cat 1–2 and Cat
natural variability and climate change. 4–5 for analysis.
A sustained upward trend is found between the global
3.3 Hurricane proportion proportion of Cat 4–5 hurricanes and ACCI (Fig. 4), bal-
anced by a similar decrease in Cat 1–2 hurricanes. The
Hurricanes as a family also have experienced no global results are independent of the choice of models to calculate
trend with anthropogenic warming (Fig. 3a), but have a the ACCI as can be seen by comparing Fig. 4a and b. In
marked low-frequency variation with a maximum around both cases the ACCI explains 80–85 % of the variance in
the middle of the analysis period. By contrast, previous the smoothed annual hurricane proportions with p \ 0.01
studies have reported a marked upward trend in intense (using unsmoothed data). This finding is consistent with the
hurricanes (Webster et al. 2005; Emanuel 2005, 2007; SST-related increases in Cat 4–5 and decreases in Cat 1–2
Elsner et al. 2008), one that is closely related to increasing found by Kishtawal et al. (2012), the relationship of intense
SST (Hoyos et al. 2010). This trend in intense hurricanes is hurricanes with SST found by Hoyos et al. (2010), and the
the focus of the remainder of our analysis. Atlantic landfall hurricane changes noted by Grinsted et al.
We bin all hurricanes into the five Saffir–Simpson cat- (2012).
egories and take annual proportions of each relative to the The global relationship also is consistently reproduced
total number of hurricanes. These are smoothed with a in each ocean basin, although the eastern North Pacific
5-year running mean to remove short-term variability. The and North Atlantic increases have p [ 0.05 (Fig. 5). Note
resulting relationship with ACCI (Fig. 3b) has a sustained that the global ACCI has been used for these regional

123
Recent intense hurricane

Fig. 4 Relationship of anthropogenic change defined from: a CMIP3

and, b CCSM4 with annual proportions of Cat 1–2 and Cat 4–5
hurricanes. Note the different scales for the ACCI. A 5-year running
mean smoother has been used (indicated by the SM in the legend), Fig. 5 Relationship between global ACCI and annual Cat 4–5
thin solid (dashed) lines indicate linear (quadratic) trends, and all hurricane proportions by ocean basin. A 5-year running mean
variances have p \ 0.01 (using unsmoothed data) smoothing has been used. The variances in a have p \ 0.01 and the
North Indian trend has p \ 0.05 (using unsmoothed data)
comparisons to ensure comparison with the global climate
change without contamination from regional variations.
This relationship between ACCI and regional Cat 4–5 whole but the slope is smaller. Since 1995 there has been a
proportions is consistent with the SST-related findings for marked increase in the annual frequency of North Atlantic
the globe by Hoyos et al. (2010), for the North Pacific by tropical cyclones, including intense hurricanes and these
Emanuel (2005), and for the North Atlantic by Elsner are closely aligned with increasing North Atlantic SST [see
(2006) and Grinsted et al. (2012). It is notable that the Bruyère et al. (2012) for a detailed discussion], the global
regional Cat 4–5 relationship is independent of the trend balance (Fig. 3a) has been maintained by a corresponding
in all tropical cyclones; for example both the western decrease for annual tropical cyclones in the western North
North Pacific and North Atlantic have increasing trends in Pacific. This leads to the intriguing possibility that for
Cat 4–5 proportions even though the total number of individual regions the degree of proportional change is
hurricanes has gone down in the Pacific and up in the inversely related to concomitant change in the total num-
Atlantic. bers. Further investigation of this potential relationship is
By using proportions, we are thus assessing overall ongoing.
changes in the probability distribution function (PDF)
without contamination by changes in total numbers that can 3.4 Discussion
easily mask the true intense hurricane signal. This explains
the apparent difference with the findings of a negative Cat The strong and consistent relationship between increasing
4–5 trend in the western North Pacific and some other proportions of intense hurricanes and the ACCI is striking
ocean basins by Klotzbach (2006). and points to a substantial observed anthropogenic increase
The North Atlantic is an interesting situation. The pro- in intense hurricanes. But could there be an alternative
portions have changed consistently with the globe as a explanation?

123
 G. Holland, C. L. Bruyère

3.4.1 Analysis issues hurricanes for both current and future climate. They found
a significant increase at a factor of 2–7 in ‘Katrina-like’
One possible trend error arises from the universal use of the hurricanes for each 1 C increase in global SST associated
Dvorak technique. There may have been a consistent with global warming.
change in analysis practice and/or satellite data that led to Hurricanes in proximity to land have better observa-
increasing analysis of intense systems over the time period tions than those well out to sea and provide another check
(Knaff et al. 2010). We check this potential impact by to the quality of the overall results. The global landfall
using the objective satellite-based reanalysis undertaken by results (as defined in Sect. 3.1) contain the same overall
Kossin et al. (2007). As shown in Fig. 6 and Table 1, the trends (Fig. 7) as were found for the entire data set in
homogenized data of Kossin (2012) clearly contain a Fig. 4a. Cat 4–5 proportions are lower than those for the
similar signal, though at a reduced trend compared to global hurricane data, an expected result since high
IBTrACS. The weaker trend may be attributed to two intensities are only maintained for a relatively short period
factors: the automated Dvorak analysis used by Kossin may of the overall cyclone lifetime and could easily be missed
underestimate the most intense storms, or there may be an by the similarly short period spent within landfall range.
artificial analysis trend in the archived data. The slope of the Cat 4–5 trend for this land-proximity
On a regional basis, Wu and Zhao (2012) applied the subset is *27 % per C, essentially the same as the
Emanuel (2008) dynamical downscaling approach to *29 % per C from the Kossin (2012) homogenized
reanalysis data for the western North Pacific and found a satellite data in Fig. 6a.
significant increase in the frequency of Cat 4–5 hurricanes General agreement with Fig. 7 is found for landfall in
(Table 1) that is similar to that in Fig. 4b after taking each ocean basin except the North Indian Ocean and the
proportions, but at around half the rate. Grinsted et al. western North Pacific. The landfall rarity of intense North
(2012) used a homogeneous index of storm surge on the US Indian Ocean hurricanes means the data are too noisy for
coast to examine potential changes in landfalling analysis. For the western North Pacific the trend is very
weak and slightly negative, which we attribute to a general
eastward and equatorward movement of the main genesis
location in recent decades (Wu and Zhao 2012). This
promotes the development of a higher proportion of intense
hurricanes, but also means that fewer of them will make
landfall.
Weinkle et al. (2012) examined the global number of
hurricanes that actually make landfall in each of the Saffir–
Simpson categories. The proportion of Cat 4–5 at landfall
to all landfall hurricanes in their data set has increased with
ACCI at a rate of *21 % per C (p \ 0.01) since 1975.
This study leads us to the following observational con-
clusions. Over the past few decades there has been an
observed *40 % per C trend relative to the ACCI in the
proportion of IBTrACS Cat 4–5 hurricanes (p \ 0.01) and
a compensating decrease in Cat 1–2 proportions. Com-
parison with the Kossin (2012) homogenized satellite data
and with global sets of tropical cyclones in proximity to
land indicates that changing analysis and observing prac-
tices may have contributed 10–15 % of the IBTrACS trend.
This leaves a real trend in the proportion of Cat 4–5 hur-
ricanes of 25–30 % per C that is directly attributed to
anthropogenic warming as defined by the ACCI. This is in
agreement with that derived from other studies (Table 1).
If annual mean tropical SST is used instead of the ACCI,
the trend in proportion of Cat 4–5 hurricanes after adjusting
Fig. 6 Relationship between: a anthropogenic change and annual Cat for analysis errors is *40 % per C SST warming (see
4–5 hurricane proportions from Kossin (2012) reanalysis; b compar-
Fig. 2b and associated discussion). These represent an
ison of IBTrACS (abscissa) and Kossin Cat 4–5 proportions. A 5-year
running mean smoothing has been used and all variances have increase of more than double in the proportion of intense
p \ 0.01 (using unsmoothed data) hurricanes, which is in agreement with, but less than, the

123
Recent intense hurricane

Table 1 Comparison of the observed changes in the proportion of Cat 4–5 hurricanes per C of global warming compared with those derived
from the indicated sources
Method When Source Increase in proportion per C global warming (%)

Global observed Past IBTrACS *40

Global observed Past Kossin (2012) *29
Global observed, in proximity to land Past IBTrACS *27
Global observed landfall Past Weinkle et al. (2012) *21
Western North Pacific dynamical downscaling Past Wu and Zhao (2012) *25 (43)
North Atlantic dynamical downscaling Future Bender et al. (2010) *11 (19)
North Atlantic statistical downscaling Future Done et al. (2012) *15 (19)
All global changes have p \ 0.01 and for the lower three rows parentheses indicate observed regional changes using IBTrACS

Done et al. (2012) applied Weibull and Generalized

Pareto distributions to assess the expected changes in
future North Atlantic hurricane extremes based on changes
to the mean and variance of the truncated distribution
predicted by the NCAR Nested Regional Climate Model
(Done et al. 2012). Both approaches showed that relatively
modest increases in the mean (3 %) and standard deviation
of (7 %) produced an increase in the proportion of Cat 4–5
hurricanes of *15 % per C of global warming.

3.4.3 Impact of internal variability

Fig. 7 Relationship of anthropogenic change to annual proportions of There is a possibility that some of the observed trend arises
Cat 1–2 and Cat 4–5 hurricanes in proximity to land. A 5-year running
mean smoothing has been used, thin solid lines indicate linear trends, from internal tropical cyclone variability that has aliased
and all variances have p \ 0.01 (using unsmoothed data) into the 35-year period used here. However, we consider
that a substantial contamination from internal variability is
2–7 times increase found in Katrina-like storms by Grin- highly unlikely. By using global mean temperatures in the
sted et al. (2012) for the North Atlantic. development of the ACCI, we have explicitly excluded all
but external forcing factors from this aspect of the analysis.
3.4.2 Comparison with model simulations There could be an influence of 11-year sun cycles or
impulsive events such as volcanoes, but we suggest that the
The observed trend in Cat 4–5 hurricanes is consistent with period chosen is too long for these to have a cumulative
independent modeling studies. Oouchi et al. (2006) found a effect on the trend.
general increase in the most intense hurricanes and a For global hurricane proportions, we are aware of no
decrease in weaker ones for a time-slice simulation forced association with internal variability such as the Pacific Dec-
by current and future surface temperatures and greenhouse adal Oscillation or similar. Figure 4 contains potentially
gases; similar findings were made by Bengtsson et al. interesting short-period variability but as shown by the qua-
(2007). Bender et al. (2010) examined the potential climate dratic fit to the Cat 4–5 proportions the trend appears to be very
changes in North Atlantic hurricanes using two versions of close to linear as would be expected from a response to the
the GFDL hurricane model as a downscaling tool applied global warming trend. By using global data we also have
to tropical cyclones generated in the GFDL climate model. removed any potential issues from recent discussion over
The technique for the climate simulation was to use current whether relative or in situ SSTs are the driving influence in
climate boundary conditions (horizontal and surface) plus a specific ocean basins (Vecchi and Soden 2007). The consis-
constant change in thermodynamic parameters (Knutson tent regional Cat 4–5 relationship with global ACCI argues for
et al. 2007). Bender et al. (2010) found that for each C of the regional changes observed here being unaffected by rel-
global warming the proportion of Cat 4–5 hurricanes ative processes and thus due largely to global changes.
increased by *11 % and the proportion of Cat 1–2 hur- Further evidence for the lack of a contribution from
ricanes decreased by *7 %. internal variability is provided by the modeling results for

123
 G. Holland, C. L. Bruyère

Table 2 Comparison North Atlantic changes for Cat 4–5 hurricanes,

showing that the differing study results arise from the use of different
base periods, and that a final, saturation proportion may be reached at
around 30–35 % of Cat 4–5 hurricanes
Source Base Base Increase in Final
period proportion proportion per proportion
(%) C (%) (%)

IBTrACS 1973–1977 12 19 31
Bender 1980–2007 22 11 33
et al.
(2010)
Done 1980–1994 16 15 31
et al.
(2012) Fig. 8 Evolving proportions of global tropical cyclone proportions
(relative to all hurricanes) binned into tropical storms (indicated by 0)
and the five Saffir–Simpson categories over the indicted decades
the North Atlantic in Table 2. Because of the experimental
configuration for these modeling studies, the simulated 1998). This leads to a suggested causal chain: global
changes contained no contribution from internal variability. warming provides conditions generally more conducive to
intensification through longer periods spent in favorable
3.4.4 Comparison with extreme value analysis conditions, with more rapid intensification rates (in
agreement with Kishtawal et al. 2012); the increased
The generally agreed expected increase in overall tropical intensification rate enables some weaker hurricanes to
cyclone intensity with global warming is around 5 % per reach higher intensities; the constraint to the maximum
C, or an increase of around 2–3 m s-1 (Emanuel 1987; intensity cap then results in a developing bimodal distri-
Henderson-Sellers et al. 1998; Knutson et al. 2010). It is bution with a secondary peak the Cat 4 level.
often assumed that this means negligible change in hur-
ricane extremes, but this can easily be shown to be 3.4.5 Where is the limit?
incorrect by fitting an extreme distribution to the tropical
cyclone PDF and examining the changes in extremes One curious feature of the observed changes has been their
resulting from small bulk changes. For example, by fitting quasi-linear nature (e.g. Fig. 4). Global Cat 4–5 hurricanes
a Weibull distribution to the current tropical cyclone PDF have increased in proportion from *20 % (a third of the
[see Done et al. (2012) for discussion on the method] it proportion of Cat 1–2) in the late 1970’s to be equal to Cat
can readily be shown that a change of 2–3 m s-1 in both 1–2 at *40 %. Obviously this linear increase cannot go on
the mean and standard deviation—similar to the resolution indefinitely as the proportion of Cat 4–5 hurricanes cannot
of the IBTrACS—produces a *20 % increase in the exceed one.
proportion (equivalent to doubling) of Cat 4–5 hurricanes. Table 2 compares the different values and base periods
Such a sensitivity of the extremes to relatively small for the North Atlantic observations with the downscaling
changes in the mean and standard deviation is well known by Bender et al. (2010) and Done et al. (2012). No
and has been emphasized by the IPCC (2012, their adjustment downward was made to the observed Atlantic
Fig. SPM3). trend due to analysis issues since North Atlantic intensity
However, there seems to be much more happening than archives are reasonably accurate over the period (Kossin
the straightforward ‘wagging of the tail’ implied by et al. 2007). When the base period is taken into account, the
extreme value analysis. Rather, the characteristics of the differing increases in column four converge to a remark-
global intensity distribution have changed consistently over ably consistent net increase in Cat 4–5 proportion of
the past 30 years (Fig. 8); an initially purely exponential *30–35 %. This suggests that the North Atlantic may
tail has developed bimodal characteristics with a secondary saturate at 30–35 % proportion of Cat 4–5 hurricanes after
peak at Cat 4 growing consistently from 1965 to the which no further increases with global warming will occur.
present. If true, a further maximum increase in North Atlantic Cat
We hypothesize that this bimodal development arises 4–5 proportion of *10 % from the current level of *25 %
from the bounded nature of the hurricane intensity distri- (Fig. 5b) may be expected.
bution. Since the maximum wind intensity is only expected For the globe the saturation limit may be around
to increase by around 5 % or \5 m s-1 per C of global 40–50 % proportion of Cat 4–5 hurricanes. Figure 9a
warming (e.g. Emanuel 1987; Henderson-Sellers et al. shows the variation with annual mean tropical SST of the

123
Recent intense hurricane

Pacific). This and the ultimate level of potential saturation

under continued global warming is subject of a separate
study. Questions that arise include: what, if anything, is the
relationship with the development of the bimodal intensity
distribution (Fig. 8); why is there a variation in the
apparent saturation level across different basins; and is
there some external climatic factor that determines both
this saturation and the remarkably constant annual fre-
quency of tropical cyclones across the globe?

4 Conclusions

An ACCI is developed as the difference between the global

surface temperatures from ensemble means of model sim-
ulations with and without anthropogenic gases included.
From this perspective the global warming signal appeared
around 1960 and has increased to a current level of
*0.8 C [as has previously been implied by Meehl et al.
(2004, 2007, 2012) and IPCC (2007)].
We find an observed change in the proportion of global
Cat 4–5 hurricanes (relative to all hurricanes) at a rate of
*40 % increase in proportion per C increase in ACCI
(Figs. 4, 5, 6; Table 1) using the IBTrACS global data set.
This global trend is consistent with the observed changes in
each tropical cyclone basin (Fig. 5) and the trends are
significant at p \ 0.01 for the globe and p \ 0.05 for the
majority of the cyclone basins. By comparison, a homog-
Fig. 9 Relationship of annual Cat 4–5 proportions to annual varia- enized satellite data set and automated Dvorak analysis,
tions in tropical SST anomalies relative to the 1975–2010 mean: together with two landfall hurricane data sets indicate
a global proportions, b western North Pacific proportions. The data trends of 20–30 % and these observed changes are con-
are binned into 0.1 C intervals from which average, maximum, and sistent with a number of independent modeling studies
minimum proportions are derived
(Table 1). We conclude that since 1975 there has been a
substantial and observable regional and global increase in
mean, maximum and minimum proportions of Cat 4–5. the proportion of Cat 4–5 hurricanes of 25–30 % per C of
The data have been extended back to 1950 and binned into anthropogenic global warming.
0.1 C intervals. Note that the tropical SST values contain The increasing proportion of intense hurricanes has been
components of both natural and anthropogenic variability. accompanied by a similar decrease in weaker hurricanes
SST was used because the ACCI is derived from long-term and the development of a distinctly bimodal distribution in
model simulations that cannot reproduce the observed the proportions of hurricanes in each Saffir–Simpson cat-
interannual variations. The maxima in Cat 4–5 proportions egory (Fig. 8). We suggest that this arises from the capped
climbs steeply with increasing SST, then appears to reach a nature of tropical cyclones to a maximum value defined by
saturation level of 40–50 % while both the means and the the potential intensity, which increases only slightly with
minima continue to increase. A similar rapid increase to global warming.
saturation is apparent for the western North Pacific in An important finding is that the proportion of intense
Fig. 9b, except that here the saturation level is higher at hurricanes appears to initially increase in response to
60–65 %. The other ocean basins are too noisy for explicit warming oceans, but then approach a saturation level after
analysis, but they appear to fall between the North Atlantic which no further increases occur. There is tentative evi-
and western North Pacific. dence that the saturation level will differ across the tropical
These observations suggest that there may be an areal cyclone basins and that the global proportion of Cat 4–5
component to the saturation level; small basins (e.g. North hurricanes may already be near it’s saturation level of
Atlantic) saturate at a lower proportions than large basins *40–50 %. This has considerable societal implications
with their associated larger warm pools (e.g. western North that are being examined in a companion study.

123
 G. Holland, C. L. Bruyère

Acknowledgments The National Center for Atmospheric Research Holland GJ, Webster PJ (2007) Heightened tropical cyclone activity
is sponsored by the National Science Foundation (NSF) and support in the North Atlantic: natural variability or climate trend? Phil
for this work was also provided by the Willis Research Network, the Trans R Soc A 365:2695–2716
Research Program to Secure Energy for America and NSF EASM Hoyos CD, Agudelo PA, Webster PJ, Curry JA (2010) Deconvolution
grant S1048841. We thank James Done, Asuka Suzuki-Parker and of the factors contributing to the increase in global hurricane
Tom Galarneau for helpful discussions, Ming Ge for extracting the intensity. Science 312:94–97
land proximity data, Roger Pielke Jr for providing the Weinkle et al. Hurrell JW, Hack JJ, Shea D, Caron JM, Rosinski J (2008) A new sea
(2012) landfall data, Jerry Meehl and Julie Arblaster for providing the surface temperature and sea ice boundary dataset for the
climate simulation data from Meehl et al. (2004, 2007, 2012), and Jim community atmosphere model. J Clim 21:5145–5153
Kossin and Chris Velden for providing the homogenized satellite IPCC (2007) Fourth Assessment Report, www.ipcc.ch/ipccreports/
data. ar4-wg1.htm
IPCC (2012) Managing the risks of extreme events and disasters to
Open Access This article is distributed under the terms of the advance climate change adaptation. http://ipcc-wg2.gov/SREX/,
Creative Commons Attribution License which permits any use, dis- p 582
tribution, and reproduction in any medium, provided the original Kishtawal CM, Jaiswal N, Singh R, Niyogi D (2012) Tropical cyclone
author(s) and the source are credited. intensification trends during satellite era (1986–2010). Geophys
Res Lett 39:L10810. doi:10.1029/2012GL051700
Klotzbach PJ (2006) Trends in global tropical cyclone activity over
the past twenty years (1986–2005). Geophys Res Lett 33:
L10805. doi:10.1029/2006GL025881
References Knaff JA, Brown DP, Courtney J, Gallina GM, Beven JL II (2010) An
evaluation of Dvorak technique-based tropical cyclone intensity
Bender MA, Knutson TR, Tuleya RE, Sirutis JJ, Vecchi GA, Garner estimates. Weather Forecast 25:1362–1379. doi:10.1175/2010
ST, Held IM (2010) Modeled impact of anthropogenic warming WAF2222375.1
on the frequency of intense Atlantic hurricanes. Science Knapp KR, Kruk MC, Levinson DH, Diamond HJ, Neumann CJ
327:454–458 (2010) The international best track archive for climate steward-
Bengtsson L, Hodges KI, Esch M, Keenlyside N, Kornblueh L, Luo ship (IBTrACS). Bull Am Meteor Soc 91:363–376. www.ncdc.
J-J, Yamagata T (2007) How may tropical cyclones change in a noaa.gov/oa/ibtracs/
warmer climate? Tellus 59A:539–561 Knutson TR, Sirutis JJ, Garner ST, Held IM, Tuleya RE (2007)
Bruyère CL, Holland GJ, Towler E (2012) Investigating the use of a Simulation of the recent multidecadal increase of Atlantic
genesis potential index for tropical cyclones in the North hurricane activity using an 18-km-grid regional model. Bull Am
Atlantic Basin. J Clim 25:8611–8626 Meteorol Soc 88:1549–1565
Done JM, Holland GJ, Bruyère CL, Leung LR, Suzuki-Parker A Knutson TR, McBride JL, Chan J, Emanuel K, Holland GJ, Landsea
(2012) Modeling high-impact weather and climate: lessons CW, Held CI, Kossin JP, Srivastava AK, Sugi M (2010) Tropical
from a tropical cyclone perspective. NCAR/TN-490 ? STR, cyclones and climate change. Nat Geosci 3:157–163
p 28. [Available online at http://nldr.library.ucar.edu/repository/ Kossin JP (2012) Personal communication
collections/TECH-NOTE-000-000-000-854] Kossin JP, Knapp KR, Vimont DJ, Murnane RJ, Harper BA (2007) A
Elsner JB (2006) Evidence in support of the climate change-Atlantic globally consistent reanalysis of hurricane variability and trends.
hurricane hypothesis. Geophys Res Lett 33:L16705. doi: Geophys Res Lett 34:L04815. doi:10.1029/2006GL028836
10.1029/2006GL026869 Landsea CW, Harper BA, Hoarau K, Knaff JA (2006) Can we detect
Elsner JB, Kossin JP, Jagger YH (2008) The increasing intensity of trends in extreme tropical cyclones? Science 313:452–454
the strongest tropical cyclones. Nature 455:92–95 Mann M, Emanuel K (2006) Atlantic hurricane trends linked to
Emanuel KA (1987) The dependence of hurricane intensity on climate change. EOS 87:233–241
climate. Nature 326:483–485. doi:10.1038/326483a0 Meehl GA, Washington WM, Ammann CM, Arblaster JM, Wigley
Emanuel KA (2005) Increasing destructiveness of tropical cyclones TML, Tebaldi C (2004) Combinations of natural and anthropo-
over the past 30 years. Nature 436:686–688 genic forcings in twentieth-century climate. J Clim 17:3721–
Emanuel KA (2007) Environmental factors affecting tropical cyclone 3727
power dissipation. J Clim 20:5497–5509 Meehl GA, Arblaster JM, Tebaldi C (2007) Contributions of natural
Emanuel KA (2008) The Hurricane—climate connection. Bull Am and anthropogenic forcing to changes in temperature extremes
Meteor Soc 89:ES10–ES20 over the United States. Geophys Res Lett 34. doi:10.1029/
Frank WM, Young GS (2007) The interannual variability of tropical 2007GL030948
cyclones. Mon Weather Rev 135:3587–3598. doi:10.1175/MWR Meehl GA, Washington WM, Arblaster JM, Hu A, Teng H, Tebaldi
3435.1 C, Sanderson B, Lamarque JF, Conley A, Strand WG, White JB
Grinsted A, Moore JC, Jevreva S (2012) Homogeneous record of III (2012) Climate system response to external forcings and
Atlantic hurricane surge threat since 1923. Proc Nat Acad Sci. climate change projections in CCSM4. J Clim 25:3661–3683.
Published online before print October 15, 2012. doi:10.1073/ doi:10.1175/JCLI-D-11-00240.1
pnas.1209542109 Oouchi K, Yosimura J, Mizuta R, Kusonoki S, Noda A (2006)
Harper BA, Kepert JD, Ginger JD (2008) Guidelines for converting Tropical cyclone climatology in a global-warming climate as
between various wind averaging periods in tropical cyclone simulated in a 20 km mesh global atmospheric model: frequency
conditions. World Meteorological Organization, TD 1555, p 51. and wind intensity analyses. J Meteorol Soc Jpn 84:259–276.
World Meteorological Organization, Geneva doi:10.2151/jmsj.84.259
Henderson-Sellers A, Zhang H, Berz G, Emanuel K, Gray W, Simpson RH, Rielh H (1981) The hurricane and its impact. Louisiana
Landsea C, Holland G, Lighthill J, Shieh S-L, Webster P, State University Press, Baton Rouge, p 398
McGuffie K (1998) Tropical cyclones and global climate Song J-J, Wang Y, Wu L (2010) Trend discrepancies among three
change: a post-IPCC assessment. Bull Am Meteor Soc 79: best track data sets of western North Pacific tropical cyclones.
19–38 J Geophys Res 115:D12128. doi:10.1029/2009JD013058

123
Recent intense hurricane

Thompson DWJ, Wallace JM, Kennedy JJ, Jones PD (2010) An Webster PJ, Holland GJ, Curry JA, Chang H-R (2005) Changes in
abrupt drop in Northern Hemisphere sea surface temperature tropical cyclone number, duration and intensity in a warming
around 1970. Nature 467:444–447. doi:10.1038/nature09394 environment. Science 309:1844–1846
Vecchi GA, Soden BJ (2007) Effect of remote sea surface temper- Weinkle J, Maue R, Pielke R Jr (2012) Historical global tropical
ature change on tropical cyclone potential intensity. Nature cyclone landfalls. J Clim. doi:10.1175/JCLI-D-11-00719.1 (early
450:1066–1070 release)
Velden C, Harper B, Wells F, Beven JL II, Zehr R, Olander T, Wu L, Zhao H (2012) Dynamically derived tropical cyclone intensity
Mayfield M, Guard C, Lander M, Edson R, Avila L, Burton A, changes over the Western North Pacific. J Clim 25:89–98. doi:
Turk M, Kikuchi A, Christian A, Caroff P, McCrone P (2006) 10.1175/2011JCLI4139.1
The Dvorak tropical cyclone intensity estimation technique: a Wu M-C, Yeung K-H, Chang W-L (2006) Trends in western North
satellite-based method that has endured for over 30 years. Bull Pacific tropical cyclone intensity. EOS Trans Am Geophys
Am Meteor Soc 87:1195–1210 Union 87:537–538

123