Clim Jcli D 15 0129.1

15 SEPTEMBER 2015 KNUTSON ET AL.
7203
Global Projections of Intense Tropical Cyclone Activity for the Late

Twenty-First Century from Dynamical Downscaling of
CMIP5/RCP4.5 Scenarios
THOMAS R. KNUTSON, JOSEPH J. SIRUTIS, AND MING ZHAO

NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey
ROBERT E. TULEYA
Center for Coastal Physical Oceanography, Old Dominion University, Norfolk, Virginia
MORRIS BENDER AND GABRIEL A. VECCHI

NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey
GABRIELE VILLARINI
IIHR-Hydroscience and Engineering, The University of Iowa, Iowa City, Iowa
DANIEL CHAVAS
Department of Civil and Environmental Engineering, Princeton University,
Princeton, New Jersey
(Manuscript received 11 February 2015, in final form 30 June 2015)
ABSTRACT
Global projections of intense tropical cyclone activity are derived from the Geophysical Fluid Dynamics
Laboratory (GFDL) High Resolution Atmospheric Model (HiRAM; 50-km grid) and the GFDL hurricane
model using a two-stage downscaling procedure. First, tropical cyclone genesis is simulated globally using
HiRAM. Each storm is then downscaled into the GFDL hurricane model, with horizontal grid spacing near
the storm of 6 km, including ocean coupling (e.g., ‘‘cold wake’’ generation). Simulations are performed using
observed sea surface temperatures (SSTs) (1980–2008) for a ‘‘control run’’ with 20 repeating seasonal cycles
and for a late-twenty-first-century projection using an altered SST seasonal cycle obtained from a phase 5 of
CMIP (CMIP5)/representative concentration pathway 4.5 (RCP4.5) multimodel ensemble. In general
agreement with most previous studies, projections with this framework indicate fewer tropical cyclones
globally in a warmer late-twenty-first-century climate, but also an increase in average cyclone intensity,
precipitation rates, and the number and occurrence days of very intense category 4 and 5 storms. While these
changes are apparent in the globally averaged tropical cyclone statistics, they are not necessarily present in
each individual basin. The interbasin variation of changes in most of the tropical cyclone metrics examined is
directly correlated to the variation in magnitude of SST increases between the basins. Finally, the framework
is shown to be capable of reproducing both the observed global distribution of outer storm size—albeit with a
slight high bias—and its interbasin variability. Projected median size is found to remain nearly constant
globally, with increases in most basins offset by decreases in the northwest Pacific.
Corresponding author address: Thomas R. Knutson, NOAA/Geophysical Fluid Dynamics Laboratory, Climate Impacts and Extremes
Group, 201 Forrestal Road, Princeton, NJ 08542.
E-mail: [email protected]
DOI: 10.1175/JCLI-D-15-0129.1
Ó 2015 American Meteorological Society

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC
7204 JOURNAL OF CLIMATE VOLUME 28
1. Introduction climate change despite the coarse-resolution limitations

1 of the CMIP3 and CMIP5 climate models.
The question of how tropical cyclone activity could
In this study, the main focus will be on using a dy-
be altered by future anthropogenic warming (Knutson
namical downscaling approach to investigate the re-
et al. 2010; Christensen et al. 2013) is complicated by a
sponse of tropical cyclones in all basins to a climate
number of factors. First, past observational data are of
change scenario obtained from a multimodel ensemble
limited use for enhancing confidence in future pro-
of CMIP5 models (RCP4.5 scenario). Our approach
jections because it remains uncertain whether past
closely follows that of Bender et al. (2010) and Knutson
changes in any tropical cyclone metric exceed the level
et al. (2013) for the Atlantic basin, but extends this ap-
expected from natural processes alone (Knutson et al.
proach to all ocean basins, with some methodological
2010). One exception to this assessment discussed by
adjustments as discussed in the model description sec-
IPCC (2013) concerns the likely role of anthropogenic
tion. Among other published studies of possible future
aerosol forcing (along with natural variability) in con-
global tropical cyclone activity, that of Murakami et al.
tributing to a temporary decrease in Atlantic hurricane
(2012b) appears exceptional in terms of the dynamical
activity during the 1970s and 1980s (e.g., Villarini and
model’s capability of simulating very intense (category 4
Vecchi 2012, 2013; Dunstone et al. 2013). A second issue
and 5) tropical cyclones. However, even in that case, the
is that uncertainties in the patterns of future sea surface
model used by Murakami et al. did not include ocean
temperature (SST) changes can lead also to uncer-
coupling, in which a storm generates a ‘‘cold wake’’ in
tainties in the associated tropical cyclone projections
the SST field, particularly when traveling over regions
(Zhao et al. 2009; Sugi et al. 2009; Knutson et al. 2013).
with a relatively shallow mixed layer. This process can
There is also an important dependence of tropical cy-
be an important negative feedback on tropical cyclone
clone intensity on the vertical structure of atmospheric
intensity (e.g., Bender and Ginis 2000; Lin et al. 2013),
temperature changes, for which at least some historical
which could possibly alter the tropical cyclone response
data/reanalyses are likely to have important shortcom-
to climate change [although see Knutson et al. (2001) for
ings (Vecchi et al. 2013). Finally, numerous studies have
an early assessment of this mechanism showing little
found that the relatively coarse grid global models, such
effect]. Therefore the present study appears to be
as those used in phase 3 of the Coupled Model Inter-
unique in comparison to previously published tropical
comparison Project (CMIP3; Meehl et al. 2007) or
cyclone global projection studies in that the models can
phase 5 (CMIP5; Taylor et al. 2012) usually do not
simulate very intense (category 4 and 5) storms in vari-
produce very realistic climatological tropical cyclone
ous basins using relatively high spatial resolution near
activity (Camargo et al. 2013) and are particularly de-
the storm (as fine as 6 km) while including ocean cou-
ficient at simulating very intense (category 4 and 5)
pling as a negative feedback on intensity.
tropical cyclones. Our previous studies of tropical cy-
As a preliminary test of the downscaling approach, we
clones for the Atlantic basin (Bender et al. 2010;
first simulate global tropical cyclone activity for the
Knutson et al. 2013) found that the simulated frequency
years 1980–2008 using observed interannually varying
response of very intense tropical cyclones to climate
SST and sea ice distributions. Each of these 3081 simu-
warming differs from that of weaker tropical cyclones:
lated tropical cyclones (compared to 2518 tropical
the intense tropical cyclones tend to increase in fre-
cyclones for the same period in observations) is then
quency whereas the weaker tropical cyclones tend to
downscaled into the Geophysical Fluid Dynamics
decrease in frequency in these simulations, using a va-
Laboratory (GFDL) hurricane model so that higher-
riety of CMIP3 and CMIP5 model scenarios as boundary
intensity storms (category 4 and 5) can be simulated and
forcings. Thus, some form of downscaling, for example
compared with observations in each basin. For observed
by regional or global model time-slice experiments
storm tracks and intensity, we use the HURDAT and
(Murakami et al. 2012b; Knutson et al. 2013), statistical
Joint Typhoon Warning Centers (JTWC) tropical cy-
downscaling (Zhao and Held 2010), or statistical/
clone datasets, except as noted in the text. We then
deterministic downscaling (Emanuel et al. 2008; Emanuel
simulate and compare present-day and late-twenty-first-
2013), has commonly been used to attempt to gain in-
century climate conditions using two sets of 20-yr runs
sight into the response of tropical cyclone activity to
based on a repeating seasonal cycle design (no inter-
annual variability) as described in section 2. Thus, in
1 summary, we perform three basic sets of experiments:
We use the term ‘‘tropical cyclone’’ in this report to refer collec-
tively to tropical storms and hurricanes (i.e., to tropical cyclones with historical (with interannually varying SSTs), present-
intensities exceeding 17.5 m s21), with the latter having intensities day control (repeating seasonal cycle), and late-twenty-
greater than 33 m s21 regardless of their basin of occurrence. first century (repeating seasonal cycle).

15 SEPTEMBER 2015 KNUTSON ET AL. 7205
TABLE 1. Summary of the 13 CMIP5 (Taylor et al. 2012) global climate models used in this study to create the multimodel anomalies in
HiRAM (using SST and sea ice concentration). (Expansions of acronyms are available online at http://www.ametsoc.org/PubsAcronymList.)
Modeling center (or group) Model name

Canadian Centre for Climate Modeling and Analysis CanESM2
Centre National de Recherches Meteorologiques/Centre Europeen de Recherche et Formation Avancees en CNRM-CM5
Calcul Scientifique
Commonwealth Scientific and Industrial Research Organization in collaboration with Queensland Climate CSIRO-Mk3.6.0
Change Centre of Excellence
Chinese Academy of Sciences, State Key Laboratory Numerical Modeling for Atmospheric Sciences and FGOALS-g2
Geophysical Fluid Dynamics
NOAA Geophysical Fluid Dynamics Laboratory GFDL-CM3
NOAA Geophysical Fluid Dynamics Laboratory GFDL-ESM2G
NOAA Geophysical Fluid Dynamics Laboratory GFDL-ESM2M
Met Office Hadley Centre HadGEM2-ES
Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environmental MIROC5
Studies, and Japan Agency for Marine-Earth Science and Technology
Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The MIROC-ESM
University of Tokyo), and National Institute for Environmental Studies
Max Planck Institute for Meteorology MPI-ESM-LR
Meteorological Research Institute MRI-CGCM3
Norwegian Climate Centre NorESM1-M
2. Model description based on the time average (1982–2005) of the HadISST

SST dataset (see Zhao et al. 2009). Thus we are not at-
We use a two-step downscaling procedure to simu-
tempting to model the effect of any climate change–
late tropical cyclones in our study. First we use a global
induced changes in interannual variability (e.g., El
atmospheric model (50-km grid spacing) forced by a
Niño–Southern Oscillation) in our cyclone simulations.
present-day or future climate SSTs, sea ice, and ap-
For the late-twenty-first-century climate change runs,
propriate greenhouse gas levels to obtain a distribution
the changes in SSTs, sea ice, and greenhouse gas con-
of tropical cyclone genesis. Each tropical cyclone gen-
centrations from the control run conditions are based
erated using this global model is then downscaled into
on the difference between 20-yr mean climatologies
the GFDL hurricane model, which is a regional model
from the multimodel mean of the CMIP5 RCP4.5 sce-
that uses multiple layers of nesting to focus resolution
nario for years 2081–2100 (i.e., late twenty-first century)
on a single tropical cyclone (grid spacing as fine as 6 km)
and years 2001–20 (i.e., present day). Note there is a
for the duration of the cyclone, beginning from the time
15-yr difference between the midpoint year of our con-
the cyclone first reaches tropical storm intensity in the
trol condition SSTs (1994) and the midpoint year of the
global model.
baseline ‘‘present-day’’ period for the RCP4.5 climate
change calculations (2010). Therefore if one assumes
a. HiRAM C180 global model
the control case to be representative of conditions for
The GFDL High Resolution Atmospheric Model 1994, rather than 2010, then the so-called late-twenty-
(HiRAM C180; 50-km grid) used to simulate the genesis first-century scenarios would correspond approximately
of tropical cyclones is described in Zhao et al. (2009). to the 20-yr period centered on 2075 (rather than 2090)
Given the global SST distribution and evolution, this as obtained from the CMIP5/RCP4.5 scenarios. For
model has a demonstrated ability to simulate the in- these climate change perturbation experiments the
terannual variability of tropical cyclone occurrence SST increase in the various tropical storm basins ranges
quite realistically, especially in the Atlantic basin, but from 1.18 to 1.78C as discussed in more detail later in
also to a reasonable degree in the northeast and north- this report.
west Pacific basins (Zhao et al. 2009). The 13 CMIP5 climate models (RCP4.5 scenario) that
To reduce the confounding influence of interannual are used in our study are listed in Table 1. Our selection
variability (‘‘noise’’) on our climate change sensitivity of these 13 models was based on the availability of sea
runs, we run control (present day) and late-twenty-first- ice concentration data at the time our multimodel en-
century downscaling experiments using 20 repeating semble was constructed for our earlier study (Knutson
identical seasonal cycles of SST and external forcings for et al. 2013), and we have continued to use the same
each climate regime, with the control seasonal cycle models in the present study.

b. GFDL hurricane model that the peak wind speed and radius of maximum wind
in GFDL hurricane model at the initial time is the same
The individual tropical cyclones detected in the as that found in HiRAM. It is possible that some in-
HiRAM C180 global model [see Zhao et al. (2009) for cipient disturbances that failed to develop in the
details of the detection scheme] are rerun at enhanced HiRAM simulation would have developed into at least
resolution, and with ocean coupling, using a version of tropical storms had they been initialized as weak dis-
the operational GFDL hurricane model. The hurricane turbances in the GFDL hurricane model. Thus we are
model consists of a triply nested moveable mesh atmo- relying on the HiRAM solution to define the number of
spheric model coupled to the Princeton Ocean Model tropical cyclone cases to downscale. For the climate
(POM; Bender et al. 2007). The 58 latitude by 58 longi- change experiments, we assumed that the ocean mixed
tude inner nest of the regional atmospheric model has, in layer depth (defined here as the maximum depth at
the version used in the present study, a horizontal grid which the ocean temperature is no more than 0.58C
spacing of about 6 km (i.e., 1/ 188). The middle nest lower than at the surface) was unchanged in the
covers an 118 3 118 region with a grid spacing of 1/ 68. warmer climate. That is, the SST change from the cli-
The high-resolution (inner and middle) nests move mate models was applied through the entire mixed
along with the storm to maintain enhanced resolution in layer. Below the mixed layer, this warming perturba-
the vicinity of the tropical cyclone. The stationary outer tion was tapered to zero, with the tapering adjusted so
domain spans 1158 in the east–west direction and 508 in that the vertical temperature gradient did not exceed
the north–south direction with a grid spacing of 1/ 28 and 0.048C m21 (following Yablonsky and Ginis 2008). A
is positioned for each tropical storm basin. The atmo- recent study Huang et al. (2015) suggests that in-
spheric model physics has been modified from that used cluding the enhanced vertical gradient of upper ocean
in Bender et al. (2010) and Knutson et al. (2013) by surface temperature from CMIP5 late-twenty-first-
implementing the physics upgrades in the GFDL oper- century simulations can reduce substantially the in-
ational hurricane model adopted in 2012 (including tensification of TCs with climate change relative to a
upgrades in cloud microphysics, cumulus parameteri- case with no change in vertical gradient. While our
zations, and boundary layer and surface physics). (See experiments include an enhanced vertical gradient
http://www.nws.noaa.gov/os/notification/tin12-18gfdl_ beneath the mixed layer, a more detailed comparison
aaa.htm for more details). The only exception is that with the findings of Huang et al. is a subject of a future
the inner-nest resolution in our experiments is set as 1/188 study. Note that the cold wakes generated by the
(;6 km) and the criterion for large-scale condensation storms are only affecting the high-resolution nested
was appropriately modified. The hurricane model ex- experiments. The storms do not feed back onto SSTs in
periments were run for up to 15 days, which allowed the HiRAM. The original climate models would have in-
entire tropical cyclone lifetime of almost all storms to be cluded some impact of simulated tropical cyclones on
simulated, including the landfalling stages. the ocean, although those storms were much less in-
The ocean coupling included in our experiments tense and distributed differently in time than those in
provides an important physical process for the simula- the dynamical downscaling framework analyzed here;
tions, allowing the tropical cyclone to influence the thus, we expect any impact of this effect on our results
ocean and generating a cold wake in the SSTs, which to be small.
can in turn affect the cyclone’s intensity. As in the op-
erational prediction version of the model, the GFDL
3. Simulated versus observed tropical cyclone
hurricane model’s atmospheric component was coupled
activity
to a one-dimensional version of the POM model for
all basins except the North Atlantic, where a three- In this section, we compare simulated and observed
dimensional version was used. The ocean model was tropical cyclone activity in terms of annual counts,
initialized by the atmospheric analysis SST at the surface tracks, intensity distribution, precipitation distribution,
and observed [U.S. Navy Generalized Digital Environ- and storm size. Tropical cyclone activity as simulated for
mental Data (GDEM)] climatology below, using an the years 1980–2008 using interannually varying ob-
assimilation methodology that prevents convective in- served SSTs is first summarized in Fig. 1. The storm
stability in any model layers (Yablonsky and Ginis tracks from the HiRAM C180 global model (Fig. 1b) are
2008). The integrations were initiated at the first di- quite realistic in their geographical distribution, al-
agnosed time of tropical storm intensity for the HiRAM though there are some deficiencies apparent in the
global model and initialized by interpolation without model simulation such as too few hurricane-strength
any synthetic vortex replacement (or bogusing), such tropical cyclones over the northeast Pacific basin, Bay of

FIG. 1. (a) Tracks of observed tropical cyclones for the years 1980–2008. (b) Simulated
tropical cyclone tracks for 1980–2008 obtained using HiRAM C180 running over observed
interannually varying SSTs (1980–2008). (c) Simulated tropical cyclone tracks obtained using
the GFDL hurricane model to resimulate tropical cyclone cases at higher resolution. These
cases in (c) used the HiRAM C180 simulation to provide initial conditions and boundary
conditions for the individual storm cases to the higher-resolution model. Storm categories on
the Saffir–Simpson scale are depicted by the track colors, varying from tropical storm (blue) to
category 5 (black; see legend). The numbers in parentheses above each panel denote the total
number of storms found.
Bengal, and the southeast Indian Ocean basins. A well- total tropical cyclones (categories 0–5, or at least
known deficiency of the HiRAM C180 model is that the tropical storm strength) than observed (106 vs 87 per
upper-limit intensity for the simulated tropical cyclones year), despite having criteria for counting as a tropical
is effectively about category 1 or 2 depending on the cyclone that are stricter for the model (i.e., 3-day du-
basin. While this has been partly addressed at least for ration at tropical storm strength required) than for
the Atlantic basin, using a statistical refinement pro- observations. There is a slight reduction in the number
cedure (Zhao and Held 2010), in the present paper we of tropical cyclones in the GFDL hurricane model
use dynamical downscaling into the GFDL hurricane downscaling simulations compared to the host HiRAM
model to simulate a more realistic intensity distribution model (104 vs 106 per year), which is mostly caused by
for the present-day climate. The map of simulated tracks tropical cyclone cases failing to develop or lasting
following the second downscaling step (Fig. 1c) shows a fewer than three days at tropical storm strength in the
much better agreement with observations for intense hurricane model downscaling runs. A few downscaling
hurricanes than the HiRAM simulation, although even cases failed because the case occurred too close to the
for the higher-resolution model a clear deficiency (un- boundary of the regional model. The hurricane model
derestimate) of the number of category 5 tropical cy- simulates more hurricanes (i.e., total hurricanes or
clones remains. categories 1–5, also called cat 1–5; 89 vs 48 per year)
Tropical cyclone annual count statistics globally and and major hurricanes (defined as categories 3–5, or cat
by basin for the model versus observations are sum- 3–5; 42 vs 23 per year) than observed, but fewer very
marized in Table 2. The HiRAM model simulates more intense hurricanes (i.e., categories 4 and 5 or cat 4–5; 13

TABLE 2. Tropical cyclone frequency statistics from downscaling experiments using observed SSTs for 1980–2006. Means are the annual
counts for various storm categories. Correlations are for the observed vs modeled interannual variations. Bold correlation entries are
significantly greater than zero at the 0.05 level (one-sided test), assuming individual years are temporally independent (i.e., exceed
a critical value of 0.32). ‘‘Cat’’ refers to the Saffir–Simpson intensity category (1–5) with ‘‘cat 0’’ signifying less than hurricane strength.
North Northeast Northwest North South Southwest

Variable Global Atlantic Pacific Pacific Indian Indian Pacific
No. tropical cyclones (cat 0–5)
Observed mean 86.7 11.9 16.5 26.6 4.8 16.8 10.1
C180 mean 105.5 13.0 18.6 35.1 6.0 18.1 14.6
Hurr. model mean 104.1 12.7 18.2 34.9 5.8 18.0 14.5
Correlation: C180 vs obs. 0.08 0.48 0.51 0.49 0.06 0.09 0.23
Correlation: Hurr. model vs obs. 0.06 0.47 0.50 0.49 0.15 0.09 0.21
No. hurricanes (cat 1–5)
Observed mean 48.0 6.6 9.1 17.0 1.4 8.7 5.2
C180 mean 34.8 4.6 3.5 16.9 1.0 4.2 4.5
Hurr. model mean 88.6 10.0 15.1 31.0 4.8 15.2 12.5
Observed mean 23.4 2.6 4.3 9.2 0.6 4.5 2.3
C180 mean 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Hurr. model mean 42.4 3.5 5.8 18.2 2.1 6.4 6.4
Observed mean 15.3 1.6 2.7 6.9 0.4 2.6 1.2
C180 mean 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Hurr. model mean 12.7 1.1 0.9 7.6 0.5 1.0 1.7
vs 15 per year) than observed. It is notable that the (cat 4–5), only the North Atlantic is significantly corre-
hurricane model has a slight low bias in cat 4–5 hurri- lated with observations. The highest annual correlation
cane frequency despite an overall high bias in the total found in our experiments was r 5 0.75 for major hurri-
number of tropical cyclones and hurricanes in the canes in the northeast Pacific basin.
simulation. [Note that for convenience, we use the term The simulated frequency distributions of tropical cy-
‘‘hurricane’’ here to describe tropical cyclones with clone intensity (Fig. 2) show reasonable agreement with
intensities greater than 33 m s21 regardless of their observations for the North Atlantic and northwest Pa-
basin of occurrence.] cific basins, except for wind speeds greater than 65 m s21.
Table 2 also contains a summary of the correlation There is some disagreement between two different
between the observed and simulated time series of an- observational datasets for the northwest Pacific basin,
nual storm counts of various classes, globally and for with the JTWC dataset having a greater occurrence of
each basin. For both HiRAM and the GFDL hurricane intense typhoons (.55 m s21) than the Japanese Me-
model, the total tropical cyclone (defined as categories teorological Agency (JMA) dataset, as has been noted
0–5, or cat 0–5) simulated numbers are significantly in previous studies (e.g., Song et al. 2010). The model’s
correlated (r . 0.31) to observed in the North Atlantic, distribution is closer to the JMA estimates, especially
northeast Pacific, and northwest Pacific basins, but this is for storms exceeding 50 m s21 maximum intensity.
not the case for the remaining three basins or the global Nonetheless, on the basis of these comparisons and
mean series. For higher-intensity classes (e.g., hurri- available information on the observed datasets, we
canes, major hurricanes, and cat 4–5 hurricanes) the suspect that our model is underestimating the occur-
GFDL hurricane model simulations are significantly rence of very intense typhoons in the northwest Pacific
correlated with observations for the same three basins basin. The modeled intensity distribution is more
for (cat 1–5) hurricanes (North Atlantic, northeast Pa- peaked than the observed in the northeast Pacific,
cific, and northwest Pacific); for major (cat 3–5) hurri- south Indian, north Indian, and southwest Pacific ba-
canes, the North Atlantic and northeast Pacific are sins, with an unrealistically peaked distribution mode
significantly correlated; and for very intense hurricanes around 50 m s21. The pronounced bimodal distribution

FIG. 2. Comparison of observed (black) and simulated (red) distributions of tropical cyclone
intensity (maximum surface winds speed in m s21) based on one value per storm at the time of
the storm’s maximum intensity (1980–2008 observations or SST conditions for the model).
Distributions are shown for various tropical cyclone basins. Simulated results were obtained
using the GFDL hurricane model for the final downscaling step. For the northwest Pacific
basin, observed distributions from two alternative data sources are included [Joint Typhoon
Warning Centers (JTWC) and Japanese Meteorological Agency (JMA)]. Distributions are
normalized to relative frequency by dividing by the number of storms in each histogram bin by
the total number of storms observed or simulated for the basins.
for observations in the north Indian basin is not re- as low as are simulated by the higher-resolution GFDL
produced by our model. hurricane model (aqua dots)—the HiRAM C180 does
Figure 3 shows the joint distribution of simulated in- not simulate maximum surface wind speeds above about
tensities as a scatterplot of central pressures and maxi- 55 m s21 in our experiments. This contrasts with the
mum wind speeds. While the HiRAM C180 model (green hurricane model, which simulates maximum surface wind
dots) simulates relatively low central pressures—at least speeds in some cases of over 70 m s21 for present-day

model overestimating rainfall—and the TRMM under-

estimating rainfall—near the storm center compared to
gauge values. The precipitation distribution for a com-
posite of tropical cyclones from HiRAM (not shown)
has a peak of about 200–270 mm day21.
Following from recent work exploring outer storm
size (Chavas and Emanuel 2010; Kim et al. 2014), we
estimate the radius of 12 m s21 surface winds (r12) from
the radial profile of the azimuthal wind at the lowest
model level for each time step. The azimuthal compo-
nent of the wind is chosen because it is the quantity
predicted by theory to decay to zero at some finite radius
(Emanuel 2004; Chavas et al. 2015); this property also
offers practical benefit when estimating the radius of
FIG. 3. Scatterplot of tropical cyclone maximum wind speeds relatively small wind speeds in the outer circulation.
(m s21) vs central minimum pressure (hPa or mb). Dots denote
For the purposes of direct comparison with observa-
single occurrences for observations (purple), GFDL hurricane
model control runs (aqua) and late-twenty-first-century projection tions, r12 is estimated following the methodology un-
runs (black), or HiRAM C180 control runs (green) and late- derlying the QuikSCAT-based radial wind profile
twenty-first-century projections (orange). Observations are shown database (QSCAT-R) detailed in Chavas and Vigh
by the purple dots and curve. The smooth curves are least squares (2014) and summarized here. Storm latitude, longitude,
quadratic best-fit lines through the data for the various cases. Note
intensity, and translation vector are taken from track
that the observations are shown for the period 2001–12, while the
‘‘Control’’ simulations shown are based on climatological SST data (model: collocated 6-hourly output; observations:
conditions for 1982–2005. HURDAT2 in the Atlantic and northeast Pacific
basins, JTWC in all other basins interpolated to each
QuikSCAT pass time). To match QuikSCAT observa-
conditions. The purple and blue solid curves in Fig. 3 tions, which are only available over water, model data
show the nonlinear (quadratic polynomial) best-fit curves over land are masked out. Next, from each available
for the observed and simulated (GFDL hurricane model) tropical cyclone fix, radial wind profiles are calculated by
present-day wind–pressure data. These curves show that first subtracting from all data points an estimate of the
the observed (global) wind–pressure relationship is well background flow vector (Lin and Chavas 2012) and then
captured in the GFDL hurricane model. decomposing the flow field into its azimuthal and radial
Figure 4 compares the modeled and observed area- components in a polar coordinate system relative to
averaged rainfall, as a function of distance to the track the storm center. All data are then averaged within an-
center, for the rainiest 10% of storms, following the nuli of width Dr 5 18.55 km (i.e., 1/ 68) for the model and
methodology of Villarini et al. (2014), which focused on Dr 5 12.5 km for observations, moving radially outwards
storms within 308 of the equator to reduce the influence from the center in increments of Dr/4. Finally, r12 is es-
of extratropical stages on the results. Under this ap- timated moving radially outwards from the radius of
proach, the 10% of storms having the highest average maximum winds.
daily rainfall accumulation within a 58 radius around the Additionally, because accurate radial profile estima-
storm center were composited for both models and tion depends on having sufficient azimuthal data cov-
observations, a procedure that tends to focus on the erage, for both observations and model simulations we
storms with relatively high potential rainfall impact. The define a data asymmetry parameter as a function of ra-
model composite is compared to observations from dius, denoted j, as the magnitude of the vector mean of
the Tropical Rainfall Measuring Mission (TRMM) all grid point distance vectors from the center. For small
Multisatellite Precipitation Analysis (TMPA; Huffman Dr/r, j 5 0 for data with perfect azimuthal symmetry
et al. 2007). The modeled distribution exceeds the TRMM (lower uncertainty) and j 5 1 in the case of a single data
estimate, with a peak value of up to 350 mm day21 or point (higher uncertainty). This parameter may be used
more at a radius of about 50–100 km from the storm; this to select subsets of r12 values with sufficient azimuthal
compares to a peak of 150 to 250 mm day21 for TRMM. coverage and thus lower uncertainty.
For North Atlantic landfalling cases, the GFDL hurri- For a consistent comparison between model and ob-
cane model rainfall has been compared with TRMM servations, we apply an identical set of filters to both
estimates and observed rain gauge values (not shown; datasets: 1) Vmax . 25 m s21 to avoid particularly weak
Tuleya et al. 2007), which also indicated the hurricane storms, 2) j(r12) # 0.5, and 3) storm center latitude

FIG. 4. Profiles of tropical cyclone precipitation rates (mm day21), averaged from the storm
center to the radius indicated on the abscissa (in degrees) based on the lifetime-average pre-
cipitation rate for the 10% rainiest tropical cyclones (see text). Results are compared for ob-
servations from TRMM satellite measurements (dashed) and the GFDL hurricane model
(solid) for downscaling experiments based initially on HiRAM C180 global atmospheric sim-
ulations obtained using observed interannually varying SSTs (1980–2008). Results are shown
for the (a) Northern and (b) Southern Hemisphere for various tropical cyclone basins (colors;
see legend).
u # 308 to remove cases undergoing extratropical tran- not sensitive to variations in these parameters. The re-
sition. Additionally, cases are removed if the storm center sulting sample sizes are N 5 1324 (observations), N 5
is over land or if r12 , 50 km; the latter is imposed because 40 282 (control), and N 5 32 998 (late twenty-first cen-
accurate decomposition of the flow vectors at small radii is tury). Finally, observational data are binned into basins
increasingly sensitive to the accuracy of the storm center defined by identical boundaries as used for the model.
position, which is subject to errors particularly in obser- Figure 5 displays the probability distributions of r12
vational analysis. The conclusions presented below are for observations and both model simulations, shown

FIG. 5. Relative frequency of tropical cyclone size, globally and for various tropical cyclone basins (AL 5 North
Atlantic; EP 5 northeast Pacific; WP 5 northwest Pacific; NI 5 north Indian; SI 5 south Indian; and SP 5 southwest
Pacific). The size metric, r12, is the radius at which the azimuthal-mean azimuthal wind speed decreases to 12 m s21.
Black curves depict observed estimates based on QuikSCAT satellite measurements; blue and red curves depict
distributions based on model simulations for control (present day; blue) or warm climate (late twenty-first century;
red) conditions. The ‘‘X’’ marks on each diagram denote median values. The numbers listed on each diagram denote
the number of cases analyzed. The top right panel shows the global and interbasin variation of median tropical
cyclone sizes for this metric. Control runs are based on climatological SSTs for 1982–2005. See text for further details.
both globally and within each basin. First and foremost, 8.9% in the northwest Pacific to 32% in the north Indian
the control (present-day) simulation performs well in Ocean. Moreover, the model also adequately captures
reproducing the observations, both in terms of distri- intrabasin variability as measured by the coefficient of
bution shape, which is approximately lognormal [similar variation (CV 5 s/m), whose magnitude of 0.53 globally
to the results of Chavas and Emanuel (2010)], and var- compares well with the observed value of 0.49, in-
iability. Indeed, the interbasin variability in median r12 is dicating significant variability in storm size within basins
surprisingly well captured in the model (Fig. 5, upper as is seen in observations.
right panel), albeit with a slightly high bias that is con- The above comparisons of simulated storm tracks,
sistent across all basins; the model median overestimates intensity distribution, rainfall, and storm size distribu-
the observed value by 9.3% globally, with a range from tions indicate that for all metrics there is room for

TABLE 3. Tropical cyclone activity (percent change) statistics from downscaling experiments for CMIP5 multimodel ensembles (future
vs present day). The future scenarios use RCP4.5 averaged conditions for late twenty-first century and are compared to the ‘‘present-day’’
simulations for 1982–2005 climatological SST conditions. ‘‘Cat’’ refers to Saffir–Simpson intensity category (1–5) with ‘‘cat 0’’ signifying
less than hurricane strength. Rain rate is the average rain rate within 100 km of the storm center, including all tropical cyclones (not just
10% rainiest). PDI is power dissipation index in units of 109 m3 s22. ‘‘Hur (wind . 65)’’ refers to hurricanes with surface wind speeds
greater than or equal to 65 m s21. ‘‘Maxwnd_tc’’ and ‘‘maxwnd_hur’’ are percent changes of mean lifetime-maximum intensities for all
tropical cyclones (wind speed . 17.5 m s21) or hurricanes (wind speed . 33 m s21). The p values, for a null hypothesis of no change from
present to future, are given in the line below each percent change entry. These use a two-sided Mann–Whitney–Wilcoxon test for all
frequency or days-of-occurrence metrics, and a one-sided test (for increase) for the intensity and rain rate metrics. Bold values indicate
significance at the p , 0.05 level. ‘‘Inf’’ refers to cases where no occurrences were simulated in the present-day run while some were
simulated in the future runs, indicating an infinite percent increase.
North Northeast Northwest North South Southwest

Variable Global Atlantic Pacific Pacific Indian Indian Pacific
No. of TC (cat 0–5) 216.4 29.4 16.3 234.5 19.5 226.1 236.6
p value ,0.01 0.39 0.02 ,0.01 0.07 ,0.01 ,0.01
No. of hur (cat 1–5) 216.6 217.5 19.3 231.6 25.6 228.4 240.6
p value ,0.01 0.16 0.01 ,0.01 0.04 ,0.01 ,0.01
No. of hur (cat 3–5) 1.8 2.7 83.7 216.9 21.7 28.3 250.6
p value 0.72 0.76 ,0.01 ,0.01 0.14 0.42 ,0.01
No. of hur (cat 4–5) 28.3 42.1 337.5 26.5 200.0 63.6 258.3
p value ,0.01 0.63 ,0.01 0.59 20.05 0.07 0.01
No. of hur (wind . 65) 59.3 125.0 Inf 16.7 Inf Inf 0.0
p value 0.01 0.36 0.63 1.00
ACE 213.1 29.7 44.2 226.9 23.2 228.8 241.7
p value 0.15 0.29 ,0.01 ,0.01 0.13 ,0.01 ,0.01
PDI 29.7 23.1 52.7 222.7 28.6 226.6 243.9
p value ,0.01 0.53 ,0.01 ,0.01 0.08 ,0.01 ,0.01
maxwnd_ts 3.6 0.4 8.2 7.4 3.4 1.8 25.6
p value ,0.01 0.41 ,0.01 ,0.01 0.07 0.20 0.98
maxwnd_hur 4.1 4.5 7.8 5.5 1.6 3.3 23.1
p value ,0.01 0.04 ,0.01 ,0.01 0.23 0.03 0.88
Cat 4–5 days 34.5 175.4 478.1 9.7 405.0 55.4 252.5
p value ,0.01 0.14 ,0.01 0.50 0.04 0.39 0.03
Rain rate_tc (cat 0–5) 14.3 17.3 17.2 20.8 10.5 8.5 21.2
p value ,0.01 ,0.01 ,0.01 ,0.01 0.01 0.03 0.66
Rain rate hur (cat 1–5) 13.4 20.5 14.4 15.5 12.8 11.1 3.5
p value ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 ,0.01 0.58
Rain rate hur (cat 3–5) 8.8 14.0 9.7 9.7 10.8 6.1 10.6
p value ,0.01 0.02 ,0.01 ,0.01 ,0.01 ,0.01 0.65
Rain rate hur (cat 4–5) 7.7 9.4 11.4 8.7 21.4 21.5 15.3
p value ,0.01 0.13 ,0.01 ,0.01 0.06 0.54 0.93
Delta SST (8C) 1.546 1.685 1.567 1.551 1.384 1.145
improvement, but the results suggest that the model (see also Fig. 1). Globally there is a reduction in tropical
simulation of present-day tropical cyclone activity is cyclone genesis (216%), consistent with previous global
adequate to justify an exploration of our downscaling warming studies using the HiRAM C180 model (e.g.,
framework’s response to climate change scenarios for Zhao et al. 2009), and with numerous pre-CMIP5 studies
the late twenty-first century, keeping in mind the de- by various groups using global atmospheric or coupled
ficiencies noted in this section. models (e.g., Knutson et al. 2010; Christensen et al.
2013). The projected reduction is generally consistent
with a more recent analysis of CMIP5 climate model
4. Late-twenty-first-century climate change projections using an empirical tropical cyclone detection
scenarios method (Tory et al. 2013), who found a 7%–28% re-
duction in tropical cyclone frequency across eight
a. Tropical cyclone genesis frequency changes
CMIP5 models that had reasonable present-day inferred
The changes in tropical cyclone genesis frequency for tropical cyclone climatology. However, our model’s
the CMIP5/RCP4.5 late-twenty-first-century scenario simulated decrease is in contrast to the 10%–40% global
are summarized for each basin and globally in Table 3 increase in late-twenty-first-century tropical cyclone

frequency projected by Emanuel (2013) using a statistical–

deterministic framework with CMIP5 RCP8.5 projec-
tions. The projections in Emanuel (2013) and Tory et al.
for the late twenty-first century used subsets of CMIP5
models that were different from those included in our
study, and both of those studies were based on a
downscaling of the stronger climate forcing (RCP8.5)
scenario.
As shown by Held and Zhao (2011), the reduction in
tropical cyclone frequency in HiRAM includes a con-
tribution from the direct effect of increasing atmo-
spheric CO2 (with SST held fixed). Their finding and
that of Zhao et al. (2013) confirm the earlier findings of
Yoshimura and Sugi (2005). Zhao et al. (2013) show that
for the seven models they examined, the CO2-induced
reduction in genesis frequency is even more robust
across models than the reduction in genesis frequency
associated with a 2-K uniform SST warming alone. The
physical mechanism producing the global reduction in
tropical cyclone frequency in the various model pro-
jections is still unclear, but possible mechanisms
include a slowing of the large-scale tropical circulation
(Sugi et al. 2002; Bengtsson et al. 2007) due to mecha-
nisms as discussed in Knutson and Manabe (1995) and
Held and Soden (2006) or increases in the saturation
deficit between the surface and middle troposphere
FIG. 6. As in Fig. 2, but comparing (a) total or (b) normalized
(Emanuel et al. 2008). A recent discussion of the fre- frequency distributions of maximum surface wind intensities
quency reduction issue is given in Sugi et al. (2012). (m s21, one value per storm) for simulations using present-day
The reduced global genesis frequency in the RCP4.5 (control) or CMIP5/RCP4.5 late-twenty-first-century (warming)
experiments is reflected in a reduction in four of the six climate conditions (see legend) for all TC basins. Simulation results
are shown for HiRAM C180 (purple, aqua) and for the GFDL
major tropical cyclone basins (North Atlantic, north-
hurricane model downscaling (red, dark blue).
west Pacific, south Indian, and southwest Pacific). An
increase is simulated for the northeast Pacific and north
Indian Ocean basins. An increase is also simulated for figure compares the globally aggregated distributions of
the South Atlantic basin (Fig. 1), although this basin tropical cyclone intensities for the present-day and
is a special case where the historical occurrence is quite warm-climate conditions, for both the HiRAM C180
rare (e.g., Catarina in 2004); moreover, the HiRAM and GFDL hurricane models. Despite the global re-
C180 seems to overestimate present-day genesis rate duction in number of tropical cyclones for the warmer
there (cf. Figs. 1a and 1b), implying that quantitative climate, the GFDL hurricane model shows an increase
projections of increases for this basin should be viewed in the number of very intense tropical cyclones, with a
with caution. ‘‘cross-over point’’ from decreased to increased fre-
quency at an intensity of around 55–60 m s21. This in-
b. Storm intensity changes
creased occurrence of higher maximum intensities is
Figure 3 shows that both HiRAM and the GFDL difficult to discern in the HiRAM C180 distribution,
hurricane model have a slightly expanded range of high even after the distributions are normalized by the total
intensities for the warmer climate conditions, consider- number of storms (Fig. 6b). On the other hand, for the
ing all basins as an aggregate global distribution. This is GFDL hurricane model, such a normalization of the
shown by the occurrence of black dots and orange dots distribution makes the increased relative occurrence at
on Fig. 3 at the upper right edges of the scatterplot re- high intensity much more apparent. The average in-
gion, extending beyond the green and aqua-colored dot tensity of all tropical cyclones (cat 0–5) in the GFDL
regions, respectively. hurricane model increases by 3.6% globally (Table 3),
The influence of climate warming on the simulated while the average intensity of all tropical cyclones (cat
intensity distributions is seen more clearly in Fig. 6. This 1–5) that exceed hurricane intensity (33 m s21) increases

FIG. 7. Tracks of simulated cat 4–5 tropical cyclones for (a) present-day or (b) late-twenty-
first-century (RCP4.5; CMIP5 multimodel ensemble) conditions. Simulated tropical cyclone
tracks were obtained using the GFDL hurricane model to resimulate (at higher resolution) the
tropical cyclone cases originally obtained from the HiRAM C180 global mode. Storm cate-
gories or intensities are shown over the lifetime of each storm, according to the Saffir–Simpson
scale. The categories are depicted by the track colors, varying from tropical storm (blue) to
category 5 (black; see legend).
by 4.1%. For HiRAM, the average intensity of all c. Storm track and occurrence changes
tropical cyclones increases by 0.4% globally, while the
average intensity of all tropical cyclones that exceed While tropical cyclone genesis rates are important,
hurricane intensity increases by 0.7%. The larger in- metrics that are even more closely tied to impacts in-
tensity increases we find for the GFDL hurricane model clude the storm tracks and maps of storm occurrence
(compared to HiRAM) are relatively more consistent rates. Additionally, a small number of very intense storms
with previous studies for the Atlantic basin (Hill and have caused a disproportionate amount of damage his-
Lackmann 2011; Knutson et al. 2013) and for the globe torically (e.g., Pielke et al. 2008). The tracks of all storms
(Oouchi et al. 2006; Murakami et al. 2012b) that used that reach at least category-4 intensity (.58 m s21) in the
relatively high-resolution modeling frameworks (2–9-km present-day and late-twenty-first-century simulations
grid spacing for the Atlantic studies; 20-km grid spacing are shown in Fig. 7. The global increase in the number of
in Oouchi et al. and Murakami et al.). these storms is 128% (244 storms in the control versus
The results in Table 3 indicate that the projected in- 313 storms in the late-twenty-first-century simulation).
tensity increases noted above for the global distributions A reduction is apparent in the southwest Pacific basin
do not occur in all basins. For tropical cyclones ex- (258%; Table 3). The basin with the largest fractional
ceeding tropical storm or hurricane intensity (greater increase is the northeast Pacific (1338%) although sub-
than 17.5 or 33 m s21, respectively), the average intensity stantial increases are also projected in the Atlantic
decreases in the southwest Pacific basin by 25.6% (142%), north Indian (1200%), and south Indian
and 23.1%, respectively. The average intensity of hur- (164%) basins. Focusing on the strongest category of
ricanes increases in the Atlantic by 14.5%, although the storm in our simulations (maximum surface wind speeds
average intensity of all tropical cyclones combined exceeding 65 m s21), Table 3 shows that the global fre-
barely changes (10.4%) in the Atlantic basin. The basin quency increases by 59%, with increases in all basins except
with the most pronounced intensity increases is the the southwest Pacific, where there is no change in projected
northeast Pacific (18% for all tropical cyclones and also frequency; however, the annual number of storms of this
for hurricanes). The northwest Pacific basin also has intensity is very limited in our simulated study (increasing
statistically significant projected maximum intensity in- from 2.7 to 4.3 yr21).
creases of 17% for all tropical cyclones and 16% for An alternative metric that incorporates the duration
hurricanes. of storms at category 4 and 5 intensity is the number of

FIG. 8. Simulated occurrence of all tropical storms (tropical cyclones with winds exceeding
17.5 m s21) for (a) present-day or (b) late-twenty-first-century (RCP4.5; CMIP5 multimodel
ensemble) conditions; unit: storms per decade. Simulated tropical cyclone tracks were obtained
using the GFDL hurricane model to resimulate (at higher resolution) the tropical cyclone cases
originally obtained from the HiRAM C180 global mode. Occurrence refers to the number of
days, over a 20-yr period, in which a storm exceeding 17.5 m s21 intensity was centered within the
108 3 108 grid region. (c) Difference in occurrence rate between late twenty-first century and
present day [(b) minus (a)]. White regions are regions where no tropical storms occurred in the
simulations [in (a) and (b)] or where the difference between the experiments is zero [in (c)].
cat 4–5 days. This metric is projected to increase by 35% most of the north Indian Ocean, northeast Pacific, North
globally (Table 3), in comparison to the 24% increase in Atlantic, and southwest Indian Ocean regions show an
number, indicating that the average time duration per increase. A small region off the coast of Brazil also
cat 4–5 storm that is spent at intensities above the shows an increase but this is due to a single intense storm
category-4 threshold is also increasing in the warm cli- in the late-twenty-first-century runs (Fig. 7). The areas
mate simulation. The sample of 244 (control) or 313 of decreased cat 4–5 occurrence (Fig. 9) include most of
(late twenty-first century) cat 4–5 storms in our simula- the southwest Pacific basin and parts of the eastern In-
tions is a large enough sample for us to investigate some dian Ocean basins. In the northwest Pacific basin, the
of the regional patterns of change in occurrence. Thus, southern section of the basin has some areas of decrease,
the regional patterns of changes in tropical cyclone oc- while the northern and eastern sections of this basin
currence and of cat 4–5 storm occurrence are shown in show increases suggestive of an expansion of the range
Figs. 8 and 9, respectively. Most of the northeast Pacific of cat 4–5 occurrence. In terms of percent changes in cat
and north Indian Ocean have increased tropical cyclone 4–5 occurrence, there are also large percent increases
occurrence in the late-twenty-first-century simulation projected for a number of basins (Table 3), especially in
(Fig. 8), while most of the remainder of the globe has a some basins where the rate of occurrence is relatively
decrease. Considering cat 4–5 storm occurrence (Fig. 9), small in the control climate. The absolute changes of cat

FIG. 9. As in Fig. 8, but for tropical cyclones of at least category-4 intensity (surface winds of at
least 59 m s21).
4–5 storm occurrence (in storms per decade; Fig. 9) are For this analysis, relative SST is defined as the local SST
more evenly distributed across the tropics. A number of change at a grid point for the appropriate season (July–
regions show projected increases in cat 4–5 occurrence November or January–May) compared to the average
but decreases in overall tropical cyclone occurrence SST change over 308N–308S for that same season. The
(Figs. 8 and 9). Among these are most of the North relation of relative SST to tropical cyclone potential
Atlantic basin and much of the northwest Pacific basin. intensity has been previously illustrated in Vecchi and
The north Indian Ocean shows a significant projected Soden (2007). The grid point by grid point correlation of
increase in hurricane-force tropical cyclone frequency hurricane occurrence changes in the hurricane model
and cat 4–5 occurrence days (Table 3). versus relative SST changes for the season is 0.63. For
Figure 10 shows the distribution of projected SST tropical cyclone occurrence (not shown) the correlation
changes used in our downscaling experiments for the is 0.45. Thus relative SST changes appear to be impor-
main tropical cyclone activity months for the Northern tant in statistically describing the spatial patterns of
and Southern Hemisphere basins, taken as the July– tropical cyclone and hurricane occurrence changes in
November and January–May seasons, respectively. The the GFDL hurricane model downscaling.
SST increases are relatively less in regions with strong Aggregate measures of tropical cyclone activity such
decreases in tropical cyclone and cat 4–5 occurrence as accumulated cyclone energy (ACE) or the power
(e.g., southwest Pacific basin) and are relatively greater dissipation index (PDI) incorporate frequency, in-
in regions with strong increases, such as the northeast tensity, and duration characteristics. These are formed
Pacific basin. This is illustrated quantitatively in the by summing the maximum surface wind intensity of a
bottom panel of Fig. 10, which contains a scatterplot of storm—raised to the second power for ACE and to
relative SST change versus percent change in hurricane the third power for PDI—at each 6-hourly period,
occurrence at each grid point in the domain 308N–308S. then summing over all storms occurring during the

FIG. 10. SST difference (CMIP5/RCP4.5 late twenty-first century minus present day) used in the tropical cyclone dynamical down-
scaling experiments for the (a) July–November or (b) January–May seasons (units: 8C). (c) Scatterplot of percent change in hurricane (cat
1–5) occurrence frequency vs relative SST difference for each 108 3 108 grid box (308N–308S) that had nonzero hurricane occurrence in
both control and warm climate simulations. Relative SST differences are from (a) for all Northern Hemisphere basin points (red) and from
(b) for all Southern Hemisphere basin points (blue) shown in (c), and are relative to the average SST change over 308N–308S.

FIG. 11. Profiles of tropical cyclone precipitation rates (mm day21), for the 10% rainiest
tropical cyclones (308N–308S) as in Fig. 4, but for the control runs (present day; blue curve) or
the warming runs (late-twenty-first-century RCP4.5 projection; red curve).
period/region of consideration. For our earlier downscaling Figure 12 shows the percent change in precipitation
studies with the GFDL hurricane model (e.g., Bender rate as a function of averaging radius (i.e., precipita-
et al. 2010; Knutson et al. 2013), we did not assess the tion averaged over a disk of the given radius) for
ACE or PDI changes for the high-resolution runs be- the various basins, considering all tropical cyclones.
cause the experiments were limited to 5 days duration. The dotted line in each panel gives an indication of the
Since we have now expanded the runs to up to 15 days in change in large-scale environmental water vapor in
length per storm we can more reasonably analyze these the basin, here approximated as 7% times the average
metrics. Both ACE and PDI decrease globally, by 213% SST change in the basin, based on the approximate rate
and 210%, respectively (Table 3). This reflects the par- of increase of saturation vapor pressure with temper-
tially offsetting influences of decreased overall storm ature at surface temperatures characteristic of the
frequency and increase in average storm intensity, with lower troposphere (e.g., Held and Soden 2006). The
the reduction in frequency apparently being the domi- enhanced percentage increase in precipitation rate
nating influence. At the basin scale, four of the six major near the storm center, which was prominently seen for
basins have a simulated decrease while two (the northeast our Atlantic tropical cyclone downscaling experiments
Pacific and north Indian basins) have a projected increase. in Knutson et al. (2013), is not seen in all of the indi-
vidual basins in this study, nor was it evident in future
d. Storm-related precipitation rate changes
projected tropical cyclone precipitation rates for
Figure 11 shows the changes in precipitation in landfalling storms over the eastern United States
simulated tropical cyclones as a function of radius, (Wright et al. 2015). The percent change in pre-
considering the 10% rainiest storms globally in each cipitation rate is typically similar to the water vapor
set of experiments. A clear increase is seen in the warm scaling (7% times the SST change). Notable exceptions
climate storms. Table 3 presents some summary sta- include the northeast Pacific and near the storm center
tistics for tropical cyclone rain rates, averaged within (within 150 km) in the northwest Pacific, where the
100 km of the storm center, considering tropical cy- precipitation rate increase exceeds the water vapor
clones of various intensity classes and including all scaling, and the southwest Pacific, where the pre-
tropical cyclones—not just the 10% rainiest storms. cipitation rate increase is systematically less than the
Globally, this metric increases by 14% for tropical water vapor scaling. These two basins are where the
cyclones of at least tropical storm intensity and 13% SST increases are largest and smallest, respectively,
for hurricanes. Comparable statistics for individual and where the intensity increase is greatest or decrease
basins are also shown in Table 3. Considering all is greatest, respectively. The results suggest that in
tropical cyclones, the precipitation rate increases in all addition to the background water vapor content, the
basins except for the southwest Pacific basin, which precipitation rate increase is enhanced by dynamical
shows a slight decrease (21.2%). Notably, this basin is (convergence) increases, particularly in regions with
also the single basin that has a decrease in average large relative SST changes, and the precipitation rate
storm intensity (25.6%) and the largest percentage can decrease in regions where the intensity change is
decreases in tropical cyclone, hurricane, and intense negative and the relative SST change is negative. The
hurricane occurrence (Table 3). relation of precipitation rate changes to relative SST

FIG. 12. Percent change (warm climate minus con-

trol) in tropical cyclone precipitation as a function of
averaging radius from the storm center, for all tropical
cyclones in each basin. The dotted line, computed as
the SST change over the basin multiplied by 7% 8C21,
approximates the increase in atmospheric water vapor
content in the basin associated with the SST warming,
assuming negligible change in relative humidity.
changes in the various basins will be examined further Welch’s two-sample t test. Although the change in me-
later in this report. dian size is negligible globally, this is not the case when
looking across basins, where a decrease is found in the
e. Storm size changes
northwest Pacific (28%), with no statistically significant
In addition to intensity and track, storm size also plays change in the north Indian basin and small increases in
an important role in modulating tropical cyclone dam- all other basins (17% to 115%); the differences in
age (Zhai and Jiang 2014). Changes in size due to the sample size allow the northwest Pacific signal to balance
warming scenario simulated here are shown in Fig. 5. those of the other basins. Notably, the two largest shifts
Globally, median storm size stays nearly constant are found in the northeast Pacific (115%) and North
(11%); the change in mean of the log-transformed data Atlantic basins (111%). The finer-grain details of
is not statistically significant at the 5% level based on a changes in size will be explored in a future study.

5. Discussion and conclusions In terms of regional distribution, the increase in pro-

jected cat 4–5 occurrence is fairly widespread in the
Using a two-step downscaling framework and the Northern Hemisphere basins (Fig. 9). Tropical cyclones in
GFDL hurricane model, we have improved the intensity general (cat 0–5), while decreasing in occurrence in the
distributions of simulated tropical cyclones compared to majority of regions as well as the global average, still show
the host HiRAM model. HiRAM C180 provides a fairly some regions of increase, particularly in the central and
realistic tropical cyclone genesis distribution as the ini- eastern North Pacific including in the vicinity of Hawaii,
tial step in the downscaling but fails to capture tropical which is qualitatively similar to climate change projection
cyclones with winds of category 4 and 5 intensity, which results reported in previous studies (Murakami et al. 2013).
we believe is important for late-twenty-first-century There is also decreased cat 4–5 occurrence projected in
climate change projection studies. The inclusion of some areas, but these are more limited regions—notably
ocean coupling in our framework provides an important the southwest Pacific and eastern Indian Ocean basins, and
additional physical process that is not included in most parts of the northwest Pacific basin. Interestingly, the re-
other climate impacts studies. Moreover, this frame- gion of the southwest Pacific near northeast Australia,
work reproduces fairly well both the observed global which is one area of projected decrease in tropical cyclone
distribution of storm intensity and outer storm size and activity, including cat 4–5 storms, is also a region where
their interbasin variability. The dynamical downscaling past data analysis indicates a downward trend in intense
framework then provides a scenario-based assessment, tropical cyclone landfalls over the past century (Callaghan
including for very intense tropical cyclones, of changes and Power 2011). While results from the RCP4.5 forcing
in tropical cyclone frequency, precipitation, size, and scenario cannot be straightforwardly compared with an
intensity characteristics in response to the large-scale historical forcing scenario, this finding may suggest the
climate change (late twenty-first century) projection as possibility of detecting a human influence on tropical cy-
simulated by a CMIP5 13-model ensemble using the clone activity (i.e., a decrease) in this region on the basis
RCP4.5 scenario. of a long (century scale) record of intense tropical cyclone
Salient features of the late-twenty-first-century pro- landfalls in northeast Australia. Projected median storm
jected changes include a substantial reduction in global size increases in most basins, but this is offset by a decrease
tropical cyclone frequency (216%), but an increase in in the northwest Pacific, resulting in a negligible increase
the frequency of the most intense storms (124% for cat globally (11%). Notable increases occur in the northeast
4–5 and 159% for tropical cyclones with maximum winds Pacific (115%) and North Atlantic basins (111%) in
exceeding 65 m s21). The tropical cyclone frequency re- these simulations.
duction is a relatively robust projection identified in The tropical cyclone precipitation results shown here,
previous studies (Knutson et al. 2010; Murakami et al. with a pronounced increase in the warmer climate, are
2012a,b; Mallard et al. 2013; Zhao et al. 2013). The consistent with earlier studies (e.g., Knutson et al. 2010,
projected increased frequency of very intense tropical 2013; Villarini et al. 2014). We do not find strong evi-
cyclones is consistent with the results from earlier down- dence to support the enhanced rate nearer the storm
scaled studies using the GFDL hurricane model that were center as found by Knutson et al. (2013) for the Atlantic
confined to only the Atlantic basin (Bender et al. 2010; basin. Additionally, our results (e.g., Table 3) suggest a
Knutson et al. 2013) and with the global model study of link between average tropical cyclone precipitation rate
Murakami et al. (2012b) .There is a larger projected change at the basin level and the simulated change in TC
fractional increase in the number of cat 4–5 days (135%) intensity. A physical mechanism suggested by the results
compared to their frequency increase. Global averaged (see also Wang et al. 2015) is that enhanced tropospheric
intensity of tropical cyclones and hurricanes increased by water vapor in the warmer climate enhances moisture
about 4%. Global aggregate activity measures (ACE convergence and thus rainfall rates, but that in a basin
and PDI) show decreases of 213% and 210%, indicating where the average intensity of tropical cyclones de-
the dominating influence of the overall tropical cyclone creases, the reduced tropical cyclone circulation in-
frequency reduction over the influence of increased tensity can offset the higher water vapor content and
average intensity. Our results are quite different from even produce a small decrease in tropical cyclone pre-
the statistical–deterministic downscaling projections of cipitation rates in that basin.
Emanuel (2013) for CMIP5 models. Emanuel’s pro- The projected tropical cyclone metric changes pre-
jections include a large increase in global PDI and trop- sented here vary by basin, with the interbasin spread being
ical cyclone frequency over the twenty-first century. explained to a large extent by variation between the re-
Projected median storm size is found to remain nearly gions in the magnitude of SST change. This is illustrated
constant globally in our projections. and summarized in Fig. 13, which contains scatterplots of

FIG. 13. Scatterplots and linear correlations (r) of the percent change in various tropical cyclone metrics (warm
climate vs control) plotted against the average change in SST (warm minus control, in 8C) for each basin. Metrics
analyzed in each panel are identified in the text above each panel. Rainfall metrics are for all tropical cyclones for
the categories stated. See main text or the caption of Table 3 for further details.
the percent change in a number of tropical cyclone metrics predictor of the interbasin variation of response in storm
by basin versus the average SST increase in the basins. frequency for various category storms, of ACE and PDI,
The size of SST increase in each basin (essentially the of maximum storm intensity, of cat 4–5 days, and of pre-
relative SST change) appears to be a reasonable statistical cipitation rates within 100 km of storm center—the latter

at least for tropical cyclones collectively and hurricanes Camargo, S., M. Ting, and Y. Kushnir, 2013: Influence of
(cat 1–5). With the exception of the case for precipitation local and remote SST on North Atlantic tropical cyclone
potential intensity. Climate Dyn., 40, 1515–1529, doi:10.1007/
rate for cat 3–5 hurricanes, which has a negligible corre-
s00382-012-1536-4.
lation, all of the other metrics have a substantial correla- Chavas, D. R., and K. A. Emanuel, 2010: A QuikSCAT climatol-
tion with the basinwide SST changes, ranging from 0.64 to ogy of tropical cyclone size. Geophys. Res. Lett., 37, L18816,
0.9. This finding highlights the importance of reliable doi:10.1029/2010GL044558.
projections of the interbasin variations or patterns of fu- ——, and J. Vigh, 2014: QSCAT-R: The QuikSCAT tropical
cyclone radial structure dataset. NCAR Tech. Note TN-
ture SST change from climate models, in addition to the
5131STR, 27 pp.
overall tropical mean magnitude of the warming, for fu- ——, N. Lin, and K. A. Emanuel, 2015: A model for the complete
ture tropical cyclone changes (see also Sugi et al. 2009). radial structure of the tropical cyclone wind field. Part I: Com-
Future research will continue to address the issue of ro- parison with observed structure. J. Atmos. Sci., doi:10.1175/
bustness of projections to the use of different downscaling JAS-D-15-0014.1, in press.
frameworks such as Emanuel (2013). Nonetheless, our Christensen, J. H., and Coauthors, 2013: Climate phenomena and
their relevance for future regional climate change. Climate
results suggest that narrowing uncertainties of the large- Change 2013: The Physical Science Basis, T. F. Stocker et al.,
scale climate change inputs to the downscaling frame- Eds., Cambridge University Press, 1217–1308.
works (e.g., SST change patterns) is an important research Dunstone, N. J., D. M. Smith, B. B. B. Booth, L. Hermanson, and
task for reducing uncertainty in future tropical cyclone R. Eade, 2013: Anthropogenic aerosol forcing on Atlantic
projections worldwide. tropical storms. Nat. Geosci., 6, 534–539, doi:10.1038/ngeo1854.
Emanuel, K. A., 2004: Tropical cyclone energetics and structure.
Atmospheric Turbulence and Mesoscale Meteorology, Cam-
Acknowledgments. We thank Tim Marchok and an
bridge University Press, 165–191.
anonymous reviewer for helpful comments on our work. ——, 2013: Downscaling CMIP5 climate models shows increased
We acknowledge PCMDI and the modeling groups tropical cyclone activity over the 21st century. Proc. Natl. Acad.
contributing to the CMIP3 and CMIP5 model archives Sci. USA, 110, 12219–12224, doi:10.1073/pnas.1301293110.
for generously making their model output available to ——, R. Sundararajan, and J. Williams, 2008: Hurricanes and
the community. This material is based in part upon work global warming—Results from downscaling IPCC AR4 sim-
ulations. Bull. Amer. Meteor. Soc., 89, 347–367, doi:10.1175/
supported by the National Science Foundation under BAMS-89-3-347.
Grant AGS-1262099 (Gabriele Villarini and Gabriel Held, I. M., and B. J. Soden, 2006: Robust responses of the hy-
A. Vecchi) and AGS-1331362 (Daniel R. Chavas). We drological cycle to global warming. J. Climate, 19, 5686–5699,
acknowledge Hyeong-Seog Kim, sponsored by the doi:10.1175/JCLI3990.1.
Willis Research Network, for early developmental work ——, and M. Zhao, 2011: The response of tropical cyclone statistics
to an increase in CO2 with fixed sea surface temperatures.
on the global simulation framework. We acknowledge
J. Climate, 24, 5353–5364, doi:10.1175/JCLI-D-11-00050.1.
the assistance of Isaac Ginis and Richard Yablonsky of Hill, K. A., and G. M. Lackmann, 2011: The impact of future climate
the University of Rhode Island on the hurricane model change on TC intensity and structure: A downscaling approach.
ocean coupling components. J. Climate, 24, 4644–4661, doi:10.1175/2011JCLI3761.1.
Huang, P., I.-I. Lin, C. Chou, and R.-H. Huang, 2015: Change in
ocean subsurface environment to suppress tropical cyclone
REFERENCES
intensification under global warming. Nat. Commun., 6, 7188,
Bender, M. A., and I. Ginis, 2000: Real-case simulations of hurricane– doi:10.1038/ncomms8188.
ocean interaction using a high-resolution coupled model: Effects Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite
on hurricane intensity. Mon. Wea. Rev., 128, 917–946, doi:10.1175/ Precipitation Analysis (TMPA): Quasi-global, multiyear,
1520-0493(2000)128,0917:RCSOHO.2.0.CO;2. combined-sensor precipitation estimates at fine scales.
——, ——, R. E. Tuleya, B. Thomas, and T. Marchok, 2007: The J. Hydrometeor., 8, 38–55, doi:10.1175/JHM560.1.
operational GFDL coupled hurricane–ocean prediction sys- IPCC, 2013: Climate Change 2013: The Physical Science Basis. T. F.
tem and a summary of its performance. Mon. Wea. Rev., 135, Stocker et al., Eds., Cambridge University Press, 1535 pp.
3965–3989, doi:10.1175/2007MWR2032.1. Kim, H.-S., G. A. Vecchi, T. R. Knutson, W. G. Anderson, T. L.
——, T. R. Knutson, R. E. Tuleya, J. J. Sirutis, G. A. Vecchi, S. T. Delworth, A. Rosati, F. Zeng, and M. Zhao, 2014: Tropical
Garner, and I. M. Held, 2010: Modeled impact of anthropo- cyclone simulation and response to CO2 doubling in the
genic warming of the frequency of intense Atlantic hurricanes. GFDL CM2.5 high-resolution coupled climate model.
Science, 327, 454–458, doi:10.1126/science.1180568. J. Climate, 27, 8034–8054, doi:10.1175/JCLI-D-13-00475.1.
Bengtsson, L., K. I. Hodges, M. Esch, N. Keenlyside, L. Kornblueh, Knutson, T. R., and S. Manabe, 1995: Time-mean response over the
J.-J. Luo, and T. Yamagata, 2007: How may tropical cyclones tropical Pacific to increased CO2 in a coupled ocean–
change in a warmer climate? Tellus, 59A, 539–561, doi:10.1111/ atmosphere model. J. Climate, 8, 2181–2199, doi:10.1175/
j.1600-0870.2007.00251.x. 1520-0442(1995)008,2181:TMROTT.2.0.CO;2.
Callaghan, J., and S. Power, 2011: Variability and decline in the ——, R. E. Tuleya, W. Shen, and I. Ginis, 2001: Impact of CO2-
number of severe tropical cyclones making land-fall over induced warming on hurricane intensities simulated in a hur-
eastern Australia since the late nineteenth century. Climate ricane model with ocean coupling. J. Climate, 14, 2458–2468,
Dyn., 37, 647–662, doi:10.1007/s00382-010-0883-2. doi:10.1175/1520-0442(2001)014,2458:IOCIWO.2.0.CO;2.

——, and Coauthors, 2010: Tropical cyclones and climate change. Tory, K. J., S. S. Chand, J. L. McBride, H. Ye, and R. A. Dare, 2013:
Nat. Geosci., 3, 157–163, doi:10.1038/ngeo779. Projected changes in late-twenty-first-century tropical cyclone
——, and Coauthors, 2013: Dynamical downscaling projections of frequency in 13 coupled climate models from phase 5 of the
twenty-first-century Atlantic hurricane activity: CMIP3 and Coupled Model Intercomparison Project. J. Climate, 26, 9946–
CMIP5 model-based scenario. J. Climate, 26, 6591–6617, 9959, doi:10.1175/JCLI-D-13-00010.1.
doi:10.1175/JCLI-D-12-00539.1. Tuleya, R. E., M. DeMaria, and R. J. Kuligowski, 2007: Evaluation
Lin, I.-I., and Coauthors, 2013: An ocean coupling potential in- of GFDL and simple statistical model rainfall forecasts for
tensity index for tropical cyclones. Geophys. Res. Lett., 40, U.S. landfalling tropical storms. Wea. Forecasting, 22, 56–70,
1878–1882, doi:10.1002/grl.50091. doi:10.1175/WAF972.1.
Lin, N., and D. R. Chavas, 2012: On hurricane parametric wind and Vecchi, G. A., and B. J. Soden, 2007: Effect of remote sea surface
applications in storm surge modeling. J. Geophys. Res., 117, temperature change on tropical cyclone potential intensity.
D09120, doi:10.1029/2011JD017126. Nature, 450, 1066–1070, doi:10.1038/nature06423.
Mallard, M. S., G. M. Lackmann, A. Aiyyer, and K. Hill, 2013: ——, S. Fueglistaler, I. M. Held, T. R. Knutson, and M. Zhao, 2013:
Atlantic hurricanes and climate change. Part I: Experimental Impacts of atmospheric temperature trends on tropical
design and isolation of thermodynamic effects. J. Climate, 26, cyclone activity. J. Climate, 26, 3877–3891, doi:10.1175/
4876–4893, doi:10.1175/JCLI-D-12-00182.1. JCLI-D-12-00503.1.
Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, Villarini, G., and G. A. Vecchi, 2012: Twenty-first-century pro-
J. F. B. Mitchell, R. J. Stouffer, and K. E. Taylor, 2007: The jections of North Atlantic tropical storms from CMIP5
WCRP CMIP3 multimodel dataset: A new era in climate models. Nat. Climate Change, 2, 604–607, doi:10.1038/
change research. Bull. Amer. Meteor. Soc., 88, 1383–1394, nclimate1530.
doi:10.1175/BAMS-88-9-1383. ——, and ——, 2013: Projected increases in North Atlantic tropical
Murakami, H., R. Mizuta, and E. Shindo, 2012a: Future changes in cyclone intensity from CMIP5 models. J. Climate, 26, 3231–
tropical cyclone activity projected by multi-physics and multi- 3240, doi:10.1175/JCLI-D-12-00441.1.
SST ensemble experiments using the 60-km-mesh MRI- ——, D. A. Lavers, E. Scoccimarro, M. Zhao, M. F. Wehner, G. A.
AGCM. Climate Dyn., 39, 2569–2584, doi:10.1007/ Vecchi, T. R. Knutson, and K. A. Reed, 2014: Sensitivity of
s00382-011-1223-x. tropical cyclone rainfall to idealized global-scale forcings.
——, and Coauthors, 2012b: Future changes in tropical cyclone J. Climate, 27, 4622–4641, doi:10.1175/JCLI-D-13-00780.1.
activity projected by the new high-resolution MRI-AGCM. Wang, C.-C., B.-X. Lin, C.-T. Chen, and S.-H. Lo, 2015: Quanti-
J. Climate, 25, 3237–3260, doi:10.1175/JCLI-D-11-00415.1. fying the effects of long-term climate change on tropical cy-
——, B. Wang, T. Li, and A. Kitoh, 2013: Projected increase in clone rainfall using a cloud-resolving model: Examples of two
tropical cyclones near Hawaii. Nat. Climate Change, 3, 749– landfall typhoons in Taiwan. J. Climate, 28, 66–85, doi:10.1175/
754, doi:10.1038/nclimate1890. JCLI-D-14-00044.1.
Oouchi, K., J. Yoshimura, H. Yoshimura, R. Mizuta, S. Kusunoki, Wright, D. B., T. R. Knutson, and J. A. Smith, 2015: Regional cli-
and A. Noda, 2006: Tropical cyclone climatology in a global- mate model projections of rainfall from U.S. landfalling tropical
warming climate as simulated in a 20 km-mesh global atmo- cyclones. Climate Dyn., doi:10.1007/s00382-015-2544-y, in press.
spheric model: Frequency and wind intensity analyses. Yablonsky, R., and I. Ginis, 2008: Improving the ocean initiali-
J. Meteor. Soc. Japan, 84, 259–276, doi:10.2151/jmsj.84.259. zation of coupled hurricane–ocean models using feature-
Pielke, R. A., J. Gratz, C. W. Landsea, D. Collins, M. A. Saunders, based data assimilation. Mon. Wea. Rev., 136, 2592–2607,
and R. Musulin, 2008: Normalized hurricane damages in the doi:10.1175/2007MWR2166.1.
United States: 1900–2005. Nat. Hazards Rev., 9, 29–42, Yoshimura, J., and M. Sugi, 2005: Tropical cyclone climatology in a
doi:10.1061/(ASCE)1527-6988(2008)9:1(29). high-resolution AGCM—Impacts of SST warming and CO2
Song, J.-J., Y. Wang, and L. Wu, 2010: Trend discrepancies among increase. SOLA, 1, 133–136, doi:10.2151/sola.2005-035.
three best track data sets of western North Pacific tropi- Zhai, A. R., and J. H. Jiang, 2014: Dependence of US hurricane
cal cyclones. J. Geophys. Res., 115, D12128, doi:10.1029/ economic loss on maximum wind speed and storm size. En-
2009JD013058. viron. Res. Lett., 9, 064019, doi:10.1088/1748-9326/9/6/064019.
Sugi, M., A. Noda, and N. Sato, 2002: Influence of global warming Zhao, M., and I. M. Held, 2010: An analysis of the effect of global
on tropical cyclone climatology: An experiment with the JMA warming on the intensity of Atlantic hurricanes using a GCM
global model. J. Meteor. Soc. Japan, 80, 249–272, doi:10.2151/ with statistical refinement. J. Climate, 23, 6382–6393, doi:10.1175/
jmsj.80.249. 2010JCLI3837.1.
——, H. Murakami, and J. Yoshimura, 2009: A reduction of global ——, ——, S.-J. Lin, and G. A. Vecchi, 2009: Simulations of global
tropical cyclone frequency due to global warming. SOLA, 5, hurricane climatology, interannual variability, and response to
164–167, doi:10.2151/sola.2009-042. global warming using a 50-km resolution GCM. J. Climate, 22,
——, ——, and ——, 2012: On the mechanism of tropical cyclone 6653–6678, doi:10.1175/2009JCLI3049.1.
frequency changes due to global warming. J. Meteor. Soc. Ja- ——, and Coauthors, 2013: Robust direct effect of increasing at-
pan, 90A, 397–408, doi:10.2151/jmsj.2012-A24. mospheric CO2 concentration on global tropical cyclone
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of frequency: A multi-model inter-comparison. U.S. CLIVAR
CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., Variations, 11 (3), U.S. CLIVAR Office, Washington, D.C.,
93, 485–498, doi:10.1175/BAMS-D-11-00094.1. 17–23.

Clim Jcli D 15 0129.1

Uploaded by

Copyright:

Available Formats

Clim Jcli D 15 0129.1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Clim Jcli D 15 0129.1

Uploaded by

Copyright:

Available Formats

15 SEPTEMBER 2015 KNUTSON ET AL.

Global Projections of Intense Tropical Cyclone Activity for the Late

THOMAS R. KNUTSON, JOSEPH J. SIRUTIS, AND MING ZHAO

MORRIS BENDER AND GABRIEL A. VECCHI

(Manuscript received 11 February 2015, in final form 30 June 2015)

Ó 2015 American Meteorological Society

1. Introduction climate change despite the coarse-resolution limitations

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Modeling center (or group) Model name

2. Model description based on the time average (1982–2005) of the HadISST

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

North Northeast Northwest North South Southwest

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

model overestimating rainfall—and the TRMM under-

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

North Northeast Northwest North South Southwest

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

frequency projected by Emanuel (2013) using a statistical–

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

FIG. 12. Percent change (warm climate minus con-

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

5. Discussion and conclusions In terms of regional distribution, the increase in pro-

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

Unauthenticated | Downloaded 04/11/24 03:28 AM UTC

You might also like