Tropical Cyclones and Climate Change Assessment: Part I. Detection and Attribution

1
1 Tropical Cyclones and Climate Change Assessment: Part I. Detection

2 and Attribution
3
4 Thomas Knutson1, Suzana J. Camargo2, Johnny C. L. Chan3, Kerry Emanuel4, Chang-Hoi Ho5,
5 James Kossin6, Mrutyunjay Mohapatra7, Masaki Satoh8, Masato Sugi9, Kevin Walsh10, Liguang
6 Wu11
1
7 Geophysical Fluid Dynamics Laboratory/NOAA, Princeton, NJ USA.
2
8 Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY USA
3
9 Guy Carpenter Asia-Pacific Climate Impact Centre, City University of Hong Kong, Kowloon,
10 China
4
11 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of
12 Technology, Cambridge, MA, USA
5
13 School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea
6
14 National Centers for Environmental Information/NOAA, Center for Weather and Climate,
15 Madison, WI USA
7
16 India Meteorological Department, New Delhi, India
8
17 Atmosphere and Ocean Research Institute, University of Tokyo, Chiba, Japan
9
18 Meteorological Research Institute, Tsukuba, Japan
10
19 School of Earth Sciences, University of Melbourne, Victoria, Australia
11
20 Nanjing University of Information Science and Technology, Nanjing, Jiangsu Province, China
21
22 Submitted to BAMS
23 Revised May 20, 2019
24 Corresponding Author:
25 Thomas R. Knutson
26 GFDL/NOAA
27 201 Forrestal Road
28 Princeton, NJ 08540
29 Ph: 609-452-6509
30 Email: [email protected]
31
32
33 Capsule summary:
34 We assess whether detectable changes in tropical cyclone activity have been identified in
35 observations and whether any changes can be attributed to anthropogenic climate change.
36
Early Online Release: This preliminary version has been accepted for
publication in Bulletin of the American Meteorological Society, may be fully cited,
and has been assigned DOI 10.1175/BAMS-D-18-0189.1. The final typeset
copyedited article will replace the EOR at the above DOI when it is published.
© 2019 American Meteorological Society

2
37 Abstract.
38 An assessment was made of whether detectable changes in tropical cyclone (TC) activity are
39 identifiable in observations and whether any changes can be attributed to anthropogenic climate
40 change. Overall, historical data suggest detectable TC activity changes in some regions
41 associated with TC track changes, while data quality and quantity issues create greater
42 challenges for analyses based on TC intensity and frequency.
43 A number of specific published conclusions (case studies) about possible detectable
44 anthropogenic influence on TCs were assessed using the conventional approach of preferentially
45 avoiding Type I errors (i.e., overstating anthropogenic influence or detection). We conclude there
46 is at least low-to-medium confidence that the observed poleward migration of the latitude of
47 maximum intensity in the western North Pacific is detectable, or highly unusual compared to
48 expected natural variability. Opinion on the author team was divided on whether any observed
49 TC changes demonstrate discernible anthropogenic influence, or whether any other observed
50 changes represent detectable changes.
51 The issue was then reframed by assessing evidence for detectable anthropogenic influence while
52 seeking to reduce the chance of Type II errors (i.e., missing or understating anthropogenic
53 influence or detection). For this purpose, we used a much weaker “balance of evidence” criterion
54 for assessment. This leads to a number of more speculative TC detection and/or attribution
55 statements, which we recognize have substantial potential for being false alarms (i.e., overstating
56 anthropogenic influence or detection) but which may be useful for risk assessment. Several
57 examples of these alternative statements, derived using this approach, are presented in the report.
Accepted for publication in Bulletin of the American Meteorological Society. DOI 10.1175/BAMS-D-18-0189.1.
3
58 1. Introduction
59 The question of whether anthropogenic influence on tropical cyclone (TC) activity is detectable
60 in observations is important, particularly owing to the large societal impacts from TCs.
61 Detection and attribution studies can also inform our confidence in future TC projections
62 associated with anthropogenic warming. This report updates TC climate change assessments
63 by a World Meteorological Organization (WMO) expert team (Knutson et al. 2010) and by the
64 Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC AR5) (Bindoff et
65 al. 2013). Part II of the assessment (Knutson et al., in preparation), focuses on future TC
66 projections. Walsh et al. (2016) provides a recent extensive literature review of TCs and climate
67 change studies.
68 The authors of this report include some former members of the expert team for the WMO 2010
69 assessment (Knutson et al. 2010) along with current membership of a WMO Task Team on
70 Tropical Cyclones and Climate Change. The Task Team members were invited to become
71 members by the WMO World Weather Research Program’s Working Group on Tropical
72 Meteorology Research.
73 To conduct this assessment, we identified studies where claims were made about detection of
74 long-term (multidecadal to century scale) changes in TC activity, or about attribution of past TC
75 changes or events to anthropogenic forcing. We next developed potential detection/attribution
76 statements for the author team to evaluate. In a few cases--for completeness--we developed and
77 evaluated a potential detection and attribution statement for which no published claim has yet
78 been made. We used an elicitation procedure whereby each author provided confidence levels or
79 agree/disagree opinion on each potential statement. The summary assessment statements
4
80 highlighted in Section 6 represent cases where a majority of authors either support the statement
81 or support an even stronger version of the statement. A full list of opinions for each individual
82 author and each potential statement is contained in Supplemental Material.
83 Some related topics such as long-term trends in TC damages (Pielke et al. 2008; Mendelsohn et
84 al. 2012; Estrada et al. 2015; Klotzbach et al. 2018) or changes in mortality risk (Peduzzi et al.
85 2012), are beyond the scope of our assessment. While total storm damages have increased over
86 time, these are strongly influenced by socio-economic trends (Pielke et al. 2008; Hsiang 2010;
87 Hsiang and Narita 2012; Camargo and Hsiang, 2015).
88
89 2. Background on Detection and Attribution of TC Changes
90
91 Detection and attribution studies explore the causes of observed climate changes or events,
92 typically by comparing models with observations. TC activity has been a very challenging topic
93 for detection/attribution, owing partly to the comparative rarity of TCs and the difficulty
94 obtaining accurate and temporally consistent measures of TC properties for climate studies (e.g.,
95 Landsea et al. 2006; Kossin et al. 2013), particularly in the pre-satellite era (Vecchi and Knutson
96 2008) . Observed TC timeseries are often limited by data quality issues and by the length of
97 available records from various monitoring techniques (e.g., satellites). TC historical timeseries
98 show far less compelling evidence for long-term increases than does global mean temperature
99 (e.g., Bindoff et al. 2013). Several long TC-related observed timeseries are shown in Fig. 1
100 (discussed below). Longer TC records are potentially available from historical documents (e.g.,
5
101 Chenoweth and Divine 2008). None of these observed TC timeseries demonstrate clear evidence
102 for a century-scale increase similar to that observed for global mean temperature.
103 Within this report, we use the term “detectable anthropogenic change” to refer to a change (e.g.,
104 trend) that is both clearly distinguishable from natural variability, and is consistent--at least in
105 sign--with a modeled (statistically significant) expected change due to anthropogenic influence.
106 More narrowly, a detectable change refers to one that is highly unlikely (typically p < 0.05) to
107 occur from natural variability alone. Natural variability here refers to changes (either forced or
108 from internal variability) that occur due to natural processes alone, without human influence.
109 Attribution involves evaluating the relative contributions of different causal factors to an
110 observed change, along with an assignment of statistical confidence (Hegerl et al. 2010; Bindoff
111 et al. 2013). This can involve multivariate methods comparing spatial and temporal patterns of
112 modeled forced signals to observations (Hegerl et al. 2010), or univariate methods comparing
113 individual observed regional or global timeseries to models (Knutson et al. 2013b). While
114 attribution is normally assessed once a climate change has been established as detectable, it has
115 become more common practice for investigators to make attribution claims even in the absence
116 of establishing a detectable change, particularly in the case of extreme event attribution. We
117 refer to such attribution claims as “attribution without detection.” An example case of this is
118 Murakami et al. (2015) who infer, using model simulations, that anthropogenic forcing has
119 increased the occurrence rate of TCs near Hawaii and thus contributed to the active TC season in
120 that region in 2014, despite the lack of a significant observed long-term increasing trend in TCs
121 in the region. While such statements will typically have relatively low confidence, they can help
122 identify possible anthropogenic climate change signals that might later become detectable.
6
123 Natural variability contributions can be assessed via long preindustrial climate simulations with
124 constant forcing (which simulate internal variability) or by historical simulations forced by
125 natural forcing agents only (e.g., volcanic and solar forcing). Paleoclimate proxies can aid in
126 evaluating how unusual an observed climate variability/change is compared to preindustrial
127 behavior, but have so far been limited to a relatively small number of regions/locations. In some
128 TC-climate studies, a statistically significant linear trend or secular change is identified based on
129 conventional statistical tests (e.g., t-test). This type of statistical test does not necessarily
130 demonstrate that the observed change is highly unusual compared to natural variability. If the
131 timeseries is long enough, an estimated contribution from natural variability can be statistically
132 removed (e.g., Kossin et al. 2016; Kossin 2018) to assess whether a significant trend remains,
133 which would increase confidence in a climate change detection. Confidence in anthropogenic
134 climate change detection depends on a number of factors: confidence in the estimates of
135 anthropogenic signal and natural variability contributions; whether the observed series being
136 considered is free from significant spurious trends due to data inhomogeneity; and whether the
137 observed timeseries spans a long enough period to reliably distinguish an anthropogenic signal
138 from natural variability.
139 Lloyd and Oreskes (2016) discuss the issue of Type I and Type II errors for detection and
140 attribution. Here, we interpret a Type I attribution error as concluding that anthropogenic forcing
141 had contributed nontrivially in a certain direction to an observed change when it had not done so,
142 while a Type II error means not concluding that anthropogenic forcing had contributed to some
143 observed change or event in a certain direction when it had done so to a nontrivial extent.
144 In this report, we place an emphasis on avoiding Type I detection errors by testing against a null
145 hypothesis that an observed change was caused by natural variability. We reject the null
7
146 hypothesis (i.e., detection) when there is sufficiently strong evidence against it using a relatively
147 high statistical threshold (e.g. a 0.05 significance level), and preferably based on multiple studies
148 using well-accepted methods.
149 As discussed by Lloyd and Oreskes, whether a Type I or Type II error is more important to avoid
150 is context- and audience-dependent. If the goal is to advance scientific understanding, an
151 emphasis on avoiding Type I errors seems logical. However, for future planning and risk
152 assessment, one may want to reduce Type II errors in particular. For example, planners for
153 infrastructure development in coastal regions may want to consider emerging
154 detection/attribution findings--even if not at the 0.05 significance level-- in their planning and
155 decision-making.
156 We are motivated in this report both by the desire to improve scientific understanding and to
157 provide useful guidance to those who need to deal with future risk. Adopting only the Type I
158 perspective has substantial potential for missing anthropogenic influences that are present but
159 have not yet emerged or been identified to a high level of confidence. Therefore, we also make
160 some assessment statements considering the Type II error perspective. For this, we consider
161 whether there is a balance of evidence (more than 50% chance) that anthropogenic climate
162 change contributed nontrivially in a specified direction to an observed change. Further, for
163 climate change detection, we adopt weaker criteria (e.g., p=0.1) for statistical significance and
164 also assess whether the balance of evidence suggests that a detectable change has been
165 identified—a lower burden of evidence than used for Type I error avoidance. We recognize that
166 the alternative Type II-error framing will lead to more speculative TC detection and/or
167 attribution statements that have substantial potential for being false alarms (i.e., overstating
168 anthropogenic influence), or for being overturned in future assessments. We note that there is
8
169 also potential for confidence in TC detection/attribution statements to be revised upward in
170 future assessments.
171
172 3. Expected detectability of human influence on TC activity
173
174 A timescale of expected emergence of a detectable signal in observations can sometimes be
175 estimated (e.g., Mahlstein et al. 2012). Bender et al. (2010) estimated that the detection
176 timescale for a 10% per decade increase in Atlantic Category 4-5 hurricanes frequency would be
177 about six decades. (The term “Category” refers to the Saffir-Simpson intensity scale (Simpson
178 and Riehl 1981; https://en.wikipedia.org/wiki/Saffir-Simpson_scale). Kossin et al. (2013) noted
179 limitations of using just three decades of TC intensity data for detection, based on a heuristic
180 analysis of potential intensity (PI) changes. (PI is a theoretical estimate of the maximum
181 intensity a TC can attain for a given thermodynamic environment (Emanuel 1988).)
182
183 The expected detectability of changes in a TC power dissipation index (PDI), and PI have also
184 been assessed by Villarini and Vecchi (2013) and Sobel et al. (2016). (PDI measures the
185 accumulated cube of maximum wind speeds across all TCs and storm lifetimes.) They suggest
186 that a detectable change in TC intensity or PDI might not be expected to date, owing to a
187 significant offset of greenhouse warming influences by aerosol cooling. Sobel et al. (2016)
188 note the large multi-decadal variability, compared to the expected signal, in an observation-based
189 PDI metric. Shifting TC track/occurrence can influence storm-experienced PI, and thus signal
190 detectability (Kossin and Camargo (2009), Kossin et al. (2014), Kossin (2015)).
9
191
192 In summary, it is perhaps not surprising that no clearly detectable anthropogenic influence has
193 yet been identified for several TC metrics since a lack of detection appears consistent with
194 historical simulations of expected PI changes to date as well as some heuristically determined
195 and modeled emergence time scales (e.g., Bender et al. 2010; Kossin et al. 2013).
196
197 4. Case Studies – Assessment of detection and attribution
198
199 In this section, we review and assess several published case studies of possible detectable
200 anthropogenic influence on TCs.
201
202 a) Case Study: Aerosol forcing influence on Atlantic TC activity.
203 IPCC AR5 (Bindoff et al. 2013) concluded that there was medium confidence that increased TC
204 activity in the North Atlantic since the 1970s resulted partly from reduced aerosol forcing but
205 that it was uncertain whether any past changes in TC activity were outside the range of natural
206 variability (see Supplemental Material). These conclusions were based principally on Booth et
207 al. (2012) and Villarini and Vecchi (2012; 2013), although Booth et al.’s conclusions have been
208 challenged (Zhang et al. 2013). Dunstone et al. (2013) modeled Atlantic TC frequency response
209 to aerosol forcing, suggesting that anthropogenic aerosols could have been the dominant cause of
210 20th century low-frequency North Atlantic TC variability (e.g, Mann and Emanuel 2006).
211 Villarini and Vecchi’s (2012; 2013) statistical downscaling of Coupled Model Intercomparison
10
212 Project Phase 5 (CMIP5) models suggested that decreasing aerosols caused about 25% of the
213 observed increase in Atlantic TC activity since the 1970s. An update of Emanuel’s (2007) PDI
214 (Fig. 1e) illustrates the strong increase in Atlantic TC activity from the early 1990s to 2005,
215 though with a decreasing tendency after 2005. Note that for Fig. 1e, no adjustments for
216 incomplete sampling of TCs (such as possible missed storms in the pre-satellite era, following
217 Vecchi and Knutson 2008) have been incorporated, which may result in a low bias in PDI in the
218 earlier parts of the record shown. Yan et al. (2017b) conclude that a recent (2005 to 2015)
219 decline in Atlantic major hurricane counts was not likely due to aerosol forcing increases, but
220 was more likely associated with increased vertical wind shear due to a weakening Atlantic Ocean
221 meridional overturning circulation (AMOC).
222 Ting et al. (2015) concluded that aerosols are more effective than greenhouse gases in
223 modulating Atlantic PI changes and that the PI increases over the last 30 years have been
224 dominated by multidecadal natural variability. Zhang et al. (2017) used statistical methods to
225 infer an impact of aerosol forcing on Atlantic TC activity via its influence on African outgoing
226 longwave radiation gradients. Balaguru et al. (2018) tentatively attributed increases in Atlantic
227 hurricane rapid intensification during 1986-2015 to natural variability, though not ruling out
228 anthropogenic forcing. Malavelle et al. (2017) concluded that the large indirect aerosol effect in
229 some CMIP5 models is greatly overestimated. Murphy et al. (2017), find that 38 of 41 CMIP5
230 historical runs simulated an Atlantic Multidecadal Oscillation (AMO) that is significantly
231 correlated with the observed AMO, suggesting a possible influence of historical forcings in
232 driving AMO phases.
11
233 Vecchi and Delworth (2017), Booth (2017), and Sutton et al. (2018) provide perspectives on the
234 competing processes thought to be contributing to Atlantic multidecadal variability (AMV) and
235 thus to TC activity. While earlier studies held that this variability is primarily due to internal
236 climate mechanisms, they review a number of more recent studies that conclude that external
237 forcing of the climate system has played some role--and perhaps a crucial role--in observed
238 AMV. The North Atlantic Oscillation has been identified as an important driver of AMOC
239 variability (Delworth et al. 2017). Saharan dust changes are another mechanism that may have
240 contributed substantially to late 20th century Atlantic hurricane variability (Dunion and Velden
241 2004; Evan et al. 2006; Strong et al. 2018). Direct effects of aerosol pollution on TCs (Y. Wang
242 et al., 2014) are another means by which TC climate could be affected by anthropogenic
243 activities.
244 Considering this evidence, the following summary statement was evaluated from a Type I error-
245 avoidance perspective by the author team: “The estimated contribution of decreased
246 anthropogenic aerosol forcing to the increased Atlantic TC frequency since the 1970s is large
247 and positive and is highly unusual (e.g., p<0.05) compared to natural variability.” A majority (7
248 of 11) authors expressed only low confidence in the above statement; the remaining 4 authors
249 ranked it between low-to-medium and medium-to-high confidence.
250 From Type II error-avoidance perspective, only one1 of 11 authors agreed with the statement that
251 “the balance of evidence suggests that there has been a detectable increase in North Atlantic TC
252 activity since the 1970s”. Author opinion was divided on whether the balance of evidence
253 suggested that anthropogenic forcing (e.g., greenhouse gases and aerosols) had contributed to
254 increased North Atlantic TC activity since the 1970s, with only 5 of 11 authors concurring.
1
That author has unpublished work in progress that influenced his minority response.
12
255
256 b) Case Study: Poleward migration of latitude of maximum intensity (western North
257 Pacific)
258 The latitude of lifetime-maximum intensity (LMI) of TCs has migrated poleward globally since
259 1980 (Kossin et al. 2014)—a change which can influence TC hazard exposure and mortality risk
260 (Kossin et al. 2016a). The migration is consistent with the independently-observed expansion of
261 the tropics (Lucas et al. 2014), and has been linked to changes in areas of large-scale rising and
262 sinking motion in the tropics (Sharmila and Walsh 2018). The migration is apparent in the
263 average latitude location of TC eyes (1982-2015) as obtained from satellite imagery (Knapp et
264 al. 2018). Part of the northern hemisphere poleward migration is due to interbasin changes in TC
265 frequency (Kossin et al. 2014; Moon et al. 2015; Kossin et al. 2016b; Moon et al. 2016), and the
266 trends can be sensitive to the time period chosen (Song and Klotzbach 2018; Tennille and Ellis
267 2017; Kossin 2018b) and to subsetting of the data by intensity (Zhan and Wang 2017).
268 Significant poleward migration rates are observed in the western North Pacific, Southern Indian,
269 and South Pacific basins and significant equatorward rates in the North Atlantic (Kossin et al.
270 2016a; Kossin 2018).
271 The migration rate is particularly well-observed and robust in the western North Pacific-- the
272 focus of a number of studies (Kossin et al 2016a; Choi et al. 2016; Liang et al 2017; Oey and
273 Chou 2016; Song and Klotzbach 2018; Zhan and Wang 2017; He et al. 2015; Tennile and Ellis
274 2017; Kossin 2018; Studholme and Gulev 2018). Therefore our case study focuses on this basin.
275 Pacific LMI changes (1980-2013) are associated with a significant poleward migration of TC
276 genesis latitude (Daloz and Camargo 2018). Song and Klotzbach (2018) found that during 1961-
13
277 2016 the latitudinal distance from genesis to latitude of maximum intensity increased, whereas
278 the genesis latitude only increased in the latter part of the period. Park et al. (2014) found that
279 the location of maximum intensity in the western North Pacific moved closer to East Asia during
280 19772010, leading to increased TC landfall intensity over east China, Korea and Japan. A
281 possibly related change observed in recent decades is a TC track shift from the South China Sea
282 region toward the East China Sea (Wang et al. 2011; Kossin et al. 2016a). This shift in recent
283 decades may be due to a weakening of easterly steering flows in the western tropical North
284 Pacific (Oey and Chou 2016) or to a weakening and eastward shift of the subtropical high
285 (Liang et al. 2017)—changes which these authors interpret as linked to global warming.
286 However, past observed shifts in western North Pacific TC occurrence/tracks and circulation
287 features have not yet been demonstrated to be detectable compared to natural variability (e.g.,
288 Lee et al. 2012).
289 According to Kossin et al. (2016a), the LMI poleward migration in the western North Pacific is
290 consistent among TC datasets, and the trend remains significant over the past 60–70 years after
291 statistically removing the influence of El Niño/Southern Oscillation and the Pacific Decadal
292 Oscillation (Fig. 1h). Recent supplemental calculations by J. Kossin regress out influence of the
293 AMO on west Pacific TC activity (e.g., Zhang et al. 2018), and replace the PDO index with an
294 Interdecadal Pacific Oscillation (IPO) index for comparison to Song and Klotzbach (2018).
295 These sensitivity tests show little effect on the trend in LMI. The residual trends were also found
296 to be robust to the use of annual or July-November averaged climate indices.
297 Kossin et al. (2016a) find a nominally positive trend in western North Pacific LMI (1980-2005)
298 in CMIP5 historical runs, but it is weaker than observed, and not statistically significant in the
299 model ensemble. The LMI poleward migration is statistically significant in late 21st century
14
300 CMIP5 projections (Representative Concentration Pathway 8.5 scenario), with a similar spatial
301 pattern and magnitude to the past observed changes there, supporting a possible anthropogenic
302 contribution to the observed trends. A caveat is that the CMIP5 TC simulations use global
303 climate models that simulate only about 20-25% of the observed number of TCs in the basin
304 (Camargo 2013). The upward trend in observed LMI over the past 60-70 years has not been
305 compared with internal variability simulated by climate models.
306 Most authors (eight of 11) concluded (Type I error perspective), that there is low-to-medium
307 confidence that the observed poleward migration of the western North Pacific basin LMI is
308 detectable compared with expected natural variability, while only three authors had either
309 medium (one) or medium-to-high (two)) confidence. Author opinion was divided on whether
310 anthropogenic forcing had contributed to this change, with six of 11 having only low confidence,
311 two having low-to-medium confidence, and three having medium confidence. Alternatively,
312 from a Type II error perspective, all authors concluded that the balance of evidence suggests that
313 the observed poleward migration of the LMI in the western North Pacific basin is detectable; and
314 nine of 11 authors concluded that the balance of evidence suggests that anthropogenic forcing
315 contributed to the observed migration.
316
317 c) Case Study: Multidecadal to century-scale landfalling TC records
318 Callaghan and Power (2011) find a statistically significant century-scale decreasing trend in
319 severe landfalling TCs in eastern Australia (Fig. 1a). However, they did not assess whether the
320 change was caused by anthropogenic forcing, nor compared the trend to climate models’ internal
321 variability. Haig et al (2014) analyzed a 1500-year TC paleoclimate proxy reconstruction for the
15
322 mid-west and northeast coasts of Australia and concluded that recent levels of TC activity in this
323 region are the lowest in the past 550 to 1500 years.
324 Weinkle et al. (2012) analyzed TC landfalls globally (Fig. 1i). They make no claim of
325 significant trends. While their global index for major TCs (Fig. 1i, dark bars) has a positive trend
326 (1970-2010) with a p-value of 0.06 using a conventional linear trend test, their trend tests are not
327 meant to be robust for trend detection in the presence of substantial multidecadal internal
328 variability. The timeseries of TC landfalls for Japan since 1901 (Fig. 1b from Kumazawa et al.
329 2016) and global TC and hurricane frequency since 1970 (Fig. 1g from Maue et al. 2011) also
330 show no strong evidence for trends.
331 U.S. landfalling hurricane counts (1878-2017; Fig. 1c) show a nominally negative decline,
332 although the trend over 1900-2017 is not statistically significant (Klotzbach et al. 2018).
333 Hurricane Harvey’s U.S. landfall in 2017 ended an 11-year “drought” of major hurricane U.S.
334 landfalls (Kossin et al. 2017), which had been tentatively attributed to natural variability (Hall
335 and Hereid 2015). Hart et al. (2016) discussed several limitations of focusing on such an index
336 of major U.S. hurricane landfalls, noting the arbitrariness of several aspects of the index. An
337 alternate index, based on accumulated cyclone energy, was analyzed by Truchelet and Staehling
338 (2017). Wang et al. (2011) and Kossin (2017) found that large-scale conditions that produce
339 enhanced Atlantic hurricane activity in the deep tropics are statistically associated with
340 conditions related to track changes (Wang et al.) or storm weakening (e.g., higher shear) near the
341 U.S. coast (Kossin 2017), that would decrease the chances of a hurricane making an intense U.S.
342 landfall.
16
343 In summary, no detectable anthropogenic influence has been identified to date in observed TC
344 landfalling data, using Type I error avoidance criteria. From the viewpoint of Type II error
345 avoidance, one of the above changes (decrease in severe landfalling TCs in eastern Australia),
346 was rated as detectable, though not attributable to anthropogenic forcing (nine of 11 authors),
347 with one dissenting author expressing reservations about the historical data quality in this case.
348
349 d) Case Study: Trends in global TC intensity
350
351 Early studies of global TC intensity data found increasing trends in numbers of very intense TCs,
352 and in their proportion of overall TCs (e.g., Webster et al. 2005; Elsner et al. 2008), but these
353 results were questioned due to concerns about data quality and changes in observational
354 capabilities over time (Landsea et al. 2006; Hagen and Landsea 2012; Klotzbach and Landsea
355 2015). A global TC intensity dataset was subsequently developed from satellite data in an effort
356 to control for temporal inhomogeneities (Kossin et al. 2007). In the latest version of this dataset
357 (Kossin et al., 2013; Fig. 1f), a slight increasing trend in global intensity for the strongest TCs (at
358 least hurricane intensity) was identified (p-value of 0.1), which is broadly consistent with
359 expectations based on detectability studies (Kossin et al. 2013, Sobel et al. 2016). Kang and
360 Elsner (2015; 2016) infer a slight increase in global TC intensity over 1984-2012, but do not
361 claim to identify a detectable anthropogenic influence.
362
363 North Atlantic LMIs have a highly significant increasing trend over 1982-2009 using
364 conventional trend tests (Fig. 1f), but the Atlantic basin has marked multidecadal variability that
365 can confound climate change detection. Park et al. (2013) found that intensity trends in the
17
366 western North Pacific have varied at the sub-basin scale, with increases south of Japan offsetting
367 decreases in the Philippine Sea. Liu and Chan (2017) found an increase in PDI in the northern
368 part of the basin and a decrease in the south, consistent with the observed poleward LMI
369 migration. Mei and Xie (2016) found that over the last 37 years landfalling typhoon intensities
370 have increased by 12-15%, with the proportion of landfalling Category 4-5 storms doubling
371 during that period, though they did not compare these changes to expected natural variability or
372 anthropogenic forcing responses.
373
374 Despite this lack of robust climate change detection, the small observed increasing trend in
375 global intensity (e.g., Kossin et al. 2013) is generally consistent with expectations of increasing
376 TC intensity with global warming from potential intensity theory (e.g.., Sobel et a. 2016) and
377 high resolution models (Knutson et al. 2010), pointing to a possibly emerging anthropogenic
378 signal. For example, in Sobel et al. (their Fig. 3) the positive influence of increasing greenhouse
379 gases on model-simulated potential intensity is beginning to dominate over the negative
380 influence from aerosols by the early 21st century in both the northern and southern hemisphere
381 means.
382
383 From the perspective of avoiding Type I errors, we conclude that there is only low confidence in
384 detection and attribution of any anthropogenic influence on historical TC intensity in any basin
385 or globally. However, from the perspective of reducing Type II errors, ten of 11 authors
386 concluded that the balance of evidence suggests that there is a detectable increase in the global
387 average intensity of the strongest (hurricane-strength) TCs since the early 1980s, while one
388 author believes that at least half of half of the observed increase is due to improving observations
18
389 during the period. Eight of 11 authors concluded that the balance of evidence suggests
390 anthropogenic forcing has contributed to the increase in global average TC intensity, while the
391 other three authors were not convinced that an anthropogenic contribution has been demonstrated
392 to a balance of evidence level using appropriate attribution methodologies.
393
394
395 e) Case Study: Atlantic hurricane surge record since 1923
396
397 Grinsted et al. (2013) developed indices of U.S. surge events using six long-term tide gauge
398 datasets. Their indices minimize the influence of long-term sea level rise. They detected a
399 significant increasing trend since 1923 in the frequency of moderately large U.S. surge events (~
400 tropical storm size surge events and higher, Fig. 1d). Warm global temperature years were
401 statistically associated with occurrence of large surge events--a statistical linkage suggesting
402 anthropogenic influence. Lee et al. (2017) used alternative statistical models on the Grinsted et
403 al. surge index, and estimated that if storm surges depend on global mean temperature, a signal
404 should be confidently identified by 2030.
405
406 Our expectation is that some worsening of total inundation levels during storms is occurring
407 globally associated with the anthropogenic contributions to global and regional sea level rise
408 (Bindoff et al. 2013, Horton et al. 2018), assuming all other factors equal (Woodruff et al. 2013;
409 Sweet et al. 2013). However, a TC climate change signal has not yet been convincingly
410 identified as a residual in sea level extremes data, after accounting for background sea level rise.
411 Marcos and Woodworth (2018) conclude that extreme sea levels have increased in several North
19
412 Atlantic basin locations since 1960, but not significantly if the mean sea level rise influence is
413 removed. Wahl and Chambers (2015) removed mean sea level change influences and found
414 some significant trends in residual extremes, but mostly along the southeast coast in the winter
415 season (not the TC season). Other related metrics of U.S. landfalling TC activity, such as the
416 U.S. landfalling hurricane timeseries since 1878 (Fig. 1c), do not show evidence for a long-term
417 increase. Therefore, the evidence for detectable increases in U.S storm total inundation levels,
418 apart from changes expected from sea level rise influence, is mixed.
419
420 From the perspective of avoiding Type II errors, only two of 11 authors concluded that the
421 balance of evidence suggests that there has been a detectable increase in the frequency of
422 moderately large U.S. surge events since 1923 as documented by the index of Grinsted et al., and
423 that anthropogenic forcing had contributed to the increase. From a Type I error perspective,
424 there is only low confidence for detection or attribution of an anthropogenic influence on the
425 surge events described by the Grinsted et al. index.
426
427
428 f) Case Study: Event attribution for tropical cyclones
429 A relatively new area for detection and attribution studies is the attribution of extreme events
430 (U.S. National Academy of Sciences (NAS) 2016). One approach is “ingredients-based”, re-
431 simulating an event using changes in large-scale environmental conditions (e.g., sea surface
432 temperatures (SSTs) and atmospheric temperatures) based on model simulations of pre-
433 industrial-to-current anthropogenic climate change. Using this approach, Lackmann (2015) finds
434 no statistically significant impact of anthropogenic climate change on Hurricane Sandy’s (2012)
20
435 intensity but projected that continued 21st century warming would lead to significantly increased
436 intensity. Takayabu et al.’s (2015) simulations suggest that anthropogenic forcing increased
437 supertyphoon Haiyan’s (Philippines, 2013) intensity, using an estimated anthropogenic SST
438 change signal with relatively strong warming near the Philippines. Wehner et al. (2018)
439 simulated a decreasing anthropogenic influence on Haiyan’s intensity, though using a lower
440 resolution model. Because these approaches assume the existence of a particular storm and
441 synoptic situation, they do not address climate-induced changes in factors such as storm
442 frequency or steering flows (e.g., Barnes et al. 2013).
443 TC event attribution studies have focused on other types of TC events, such as unusually active
444 TC seasons and a 20-yr regional decline in TC activity. Some of these events have been tested
445 for anthropogenic influence using model simulations. For example, Murakami et al. (2015;
446 2017) infer a substantial anthropogenic contribution to unusually high TC frequency near Hawaii
447 in 2014 and 2015 and in the eastern Pacific basin in 2015. Zhang et al. (2016) infer an
448 anthropogenic contribution to high Accumulated Cyclone Energy (ACE) in the western North
449 Pacific in 2015. Exploring potential causes of recent decadal-scale changes, Takahashi et al.
450 (2017) infer that changes in sulfate aerosol emissions caused more than half of the observed
451 decline in TC frequency over the southeastern part of the western North Pacific during 1992-
452 2011, while Zhao et al. (2018) conclude that internal variability (the Interdecadal Pacific
453 Oscillation) contributed to the lower TC frequency observed in the western North Pacific basin
454 after 1998. Using a statistical analysis, Yang et al. (2018) infer a contribution of global warming
455 to record-setting (1984-2015) TC intensity in the western North Pacific in 2015. Murakami et al.
456 (2018) conclude that the active 2017 Atlantic hurricane season was mainly caused by high SSTs
21
457 in the tropical Atlantic relative to the global tropics, but did not attribute this seasonal event to
458 anthropogenic forcing.
459 None of the above event attribution studies provides convincing evidence for an observed change
460 in TC activity being detectable (i.e., an observed climate change signal that is highly unusual
461 compared to natural variability alone). Therefore, in the above studies, any cases of inferred
462 anthropogenic attribution are examples of attribution without detection.
463 We assessed several of the above studies in our elicitation, taking the perspective of avoiding
464 Type II errors. A slight majority of authors concluded that the balance of evidence suggests that
465 anthropogenic forcing contributed to the active 2014 TC season near Hawaii (six of 11 authors).
466 Author opinion was divided on whether the balance of evidence suggested that anthropogenic
467 forcing had contributed to the intensity of supertyphoon Haiyan in the western North Pacific
468 (with only five of 11 authors agreeing with the statement). A clear majority (eight of 11 authors)
469 concluded that the balance of evidence suggest an anthropogenic contribution to the 2015
470 western North Pacific TC season. In none of the above cases did a majority of authors assess the
471 observed changes as detectable, either from a Type I or Type II error perspective.
472
473 g) Case Study: Hurricane Harvey’s (2017) extreme rainfall
474
475 Hurricane Harvey (2017) produced record-breaking three-day rainfall totals over the Houston
476 region, mainly caused by the storm’s slow movement over eastern Texas. Trenberth et al. (2018)
477 concluded that Harvey’s rainfall totals were linked to the large ocean heat content lost from the
478 unusually warm Gulf of Mexico during the storm. Emanuel (2017) simulated large numbers of
22
479 TCs in the region under various climate conditions and found that the relative occurrence rate of
480 Harvey-like TC rainfall in the region was substantially enhanced by anthropogenic changes to
481 the large-scale environment--an example of model-based attribution without detection. Wang et
482 al. (2018) modeled the influence of observed trends since 1980 in atmospheric variables and SST
483 on Harvey’s evolution, but did not separate out an anthropogenic influence. Zhang et al. (2018)
484 concluded that urbanization contributed to the flood response to rainfall during Harvey as well as
485 to the storm total rainfall.
486
487 Van Oldenborgh et al. (2017) and Risser and Wehner (2017) examined long records of extreme
488 rainfall in the region using extreme value analysis. Van Oldenborgh et al. estimated that
489 observed extreme precipitation had increased there by 12-22% since 1880. Examining
490 simulations from three models, they concluded that global warming and associated
491 environmental changes--which were primarily caused by greenhouse gas emissions from
492 anthropogenic activities-- made the Harvey precipitation event 8-19% more intense (or 1.5 to 5
493 times more likely). Risser and Wehner inferred a large positive influence of human-induced
494 climate change on extreme rainfall in the region, which they estimated as a likely increase in
495 Harvey’s total rain of at least 19%, and the chances of the observed rainfall event (threshold
496 exceedance) by at least 3.5 times. Both studies suggest the observed precipitation increases
497 exceeded the increase expected due to increased water vapor content alone associated with a
498 warmer atmosphere. These studies are examples of climate change detections for extreme
499 rainfall in general in the east Texas region rather than specifically for hurricane-related rainfall
500 rates. We have adopted a relatively conservative approach to the assessment of extreme
23
501 precipitation trends in general in this assessment (e.g., van Oldenborgh et al 2017; Risser and
502 Wehner 2017) because this is a topic that extends beyond the scope of our assessment.
503
504 Previous assessments focusing on extreme rainfall in general have concluded that there is
505 medium confidence that anthropogenic forcing has contributed to intensification of heavy
506 precipitation over land at the global scale for regions with sufficient data coverage (Bindoff et al.
507 2013) and over the United States (Easterling et al. 2017). However, a thorough assessment of
508 long-term changes in extreme rainfall in general (including both TC-related and non-TC-related)
509 and from urbanization influences is beyond the scope of our TC-climate assessment. A
510 detectable anthropogenic influence on near-surface water vapor has also been identified at the
511 global scale with medium confidence (Bindoff et al. 2013). At the regional scale, some evidence
512 for a detectable anthropogenic contribution to centennial-scale sea surface warming has been
513 reported for the western Gulf of Mexico (Vose et al. 2017), and upper-ocean heat content in the
514 Gulf of Mexico has also increased since the 1960s (Trenberth et al. 2018), associated with the
515 surface warming.
516
517 Modeling studies suggest that hurricane-related precipitation rates (for a given storm) on average
518 increase with anthropogenic warming (e.g., Knutson et al. 2010; 2015). Patricola and Wehner
519 (2018) conclude, based on a modeling study, that anthropogenic climate change has contributed
520 to both the average and extreme rainfall characteristics of hurricanes Katrina, Irma and Maria;
521 Harvey was not included in their study. However, no observational studies have provided
522 convincing evidence of a detectable anthropogenic influence specifically on hurricane-related
523 precipitation. Kunkel et al. (2010) found a pronounced increase in frequency of TC-associated
24
524 extreme precipitation events in the U.S. comparing 1994-2008 to 1895-2008 climatology, but did
525 not claim to have detected an anthropogenic signal. Lau and Zhou’s (2012) analysis of the
526 relatively short (two decades) available satellite-based TC rainfall data does not find conclusive
527 evidence for detectable anthropogenic influence on TC rainfall.
528
529 In summary, from a Type I error perspective, the author team had low confidence that
530 anthropogenic influence specifically on hurricane precipitation rates has been detected.
531 Alternatively, from the perspective of reducing Type II errors, all authors concluded that the
532 balance of evidence suggests that there has been a detectable long-term increase in occurrence of
533 Hurricane Harvey-like extreme precipitation events in the eastern Texas region, and that
534 anthropogenic forcing has contributed to this increase.
535
536
537
538 h) Case Studies: Changes in Arabian Sea TCs
539
540 An observed increase in Arabian Sea pre-monsoon season TC intensity between 1979-1996 and
541 1997-2010 was hypothesized to be due to decreasing storm-ambient vertical wind shear
542 associated with changing meridional SST gradients due to a six-fold increase in regional
543 anthropogenic aerosol emissions (Evan et al. 2011). This interpretation has been debated (Wang
544 et al. 2012; Evan et al. 2012). Rajeevan et al. (2013) document multidecadal variations in
545 Arabian Sea pre-monsoon season TC activity, wind shear and upper ocean temperatures since
546 1955 and interpret these as arising from natural variability. Evan et al. (2011) is not an explicit
25
547 TC simulation study and no clear demonstration was made that the TC intensity changes were
548 highly unusual compared to natural variability. Consequently, from a Type II error perspective,
549 our author team did not conclude that the balance of evidence supports either a detectable change
550 or an anthropogenic contribution to the observed TC intensity increase.
551
552 In a study of recent Arabian Sea post-monsoon season extreme severe cyclonic storms (defined
553 as TCs with maximum winds exceeding 46 m s-1) Murakami et al. (2017b) reported the first
554 documented occurrence of such post-monsoon season TCs during 2014 and 2015, based on
555 records extending only back to 1998. They conducted long experiments with a 25 km-grid
556 global coupled model using idealized fixed 1860, 1990, and 2015 radiative conditions and
557 estimated more than a doubling in the probability of occurrence of such storms for 1990 and
558 2015 climate forcings, relative to 1860.
559
560 Among the caveats to this study, the 1998-2013 period of analysis is relatively short for
561 identifying a detectable anthropogenic signal. The modeling approach of using long fixed-
562 forcing runs (1990 and 2015), as opposed to transient climate change simulations, overestimates
563 the anthropogenic climate change signal, since the constant-forcing model has more time than
564 the real world to equilibrate to given forcing levels. Their model simulated a general increase in
565 TC activity with CO2 warming across other tropical regions (as seen for example in their Bay of
566 Bengal results), implying that apparent agreement between the observed and simulated increase
567 of post monsoon TC activity in the Arabian Sea region could be coincidental. However, in the
568 North Indian Ocean only the post-monsoon strong TCs show an increase with historical
569 anthropogenic forcing, in agreement with seasonal timing for the observed Arabian Sea TC
26
570 increases.
571
572 In summary, from a Type I error perspective, there is low confidence in a detectable
573 anthropogenic influence on increasing TC activity in the Indian Ocean region. Alternatively,
574 from the Type II error perspective, all authors concluded that the balance of evidence suggests
575 there has been some detectable increase in the frequency of post-monsoon season extremely
576 severe cyclonic storms over the Arabian Sea, and most authors (eight of 11) concluded that
577 anthropogenic forcing has contributed to the increase.
578
579
580 i) Case Study: Global increase in proportion of Category 4-5 TCs
581
582 Holland and Bruyère (2014) analyzed changes in TC frequency for various storm categories,
583 assessing IBTrACS/JTWC intensity data and a shorter homogenized satellite-based intensity data
584 (Kossin et al. 2013). From the satellite-based data, they conclude that the proportion of
585 hurricanes reaching Category 4 or 5 intensity has increased by 25-30% per degree Celsius of
586 global warming in recent decades. A similar statistically significant signal was found in most
587 individual TC basins.
588 They tested for anthropogenic influence during 1975-2010 because the CMIP3 historical
589 simulations they analyzed suggested little net anthropogenic influence on global mean
590 temperature before 1975. Their analysis did not compare the observed changes over 1975-2010
591 to expected internal climate variability on 35-year timescales from climate model control runs.
592 Their linkage to anthropogenic forcing as a mechanism is statistical in nature as there is no
27
593 explicit comparison between observed storm metrics and those derived from simulations using
594 historical forcings. They inferred that the observed increase in proportion of Category 4-5
595 storms (which by definition has an upper limit of 100%) may be reaching a saturation point soon
596 and may not continue increasing over the coming century which could hinder its detectability
597 using this particular metric.
598 From the Type I error perspective, the author team did not conclude that there is confidence in
599 detection of an anthropogenic climate change signal in historical proportion of very intense TCs.
600 Alternatively, from the perspective of reducing Type II errors (where we, again, require less
601 convincing levels of evidence), all authors concluded that the findings of Holland and Bruyère
602 (2014), combined with other studies linking climate warming to increased TC intensity, provides
603 a balance of evidence suggesting that the observed increase in Category 4-5 proportion in recent
604 decades represents a detectable change. Most authors (eight of 11) concluded that the balance of
605 evidence suggests that the increase in proportion of Category 4-5 storms resulted in part from
606 anthropogenic forcing.
607
608 j) Case Study: Global slowdown of TC translation speeds
609 Kossin (2018a) reported a significant decreasing trend in global TC translation speed--about 10%
610 over 1949-2016 (Fig. 1j). This decrease has been particularly strong for TCs over land areas
611 near the western North Pacific (-21%) and North Atlantic (-16%) basins, and around Australia (–
612 18%). Chu et al. (2012) had previously reported a statistically significant decrease in steering
613 flows and in storm translation speeds over 1958-2009 in the western North Pacific and South
614 China Sea regions.
28
615 Kossin (2018a) interprets the observed global slowdown as consistent with expected changes in
616 atmospheric circulation due to anthropogenic forcing. A few studies report future projections of
617 TC propagation speeds under climate warming, and these contain examples of projected
618 significant decreases in speeds (e.g., Knutson et al. 2013a; Gutmann et al. 2018) and other
619 examples where there is an increase or no significant change (Knutson et al. 2013a; Kim et al.
620 2014). Thus, there is no consensus in the studies on the sign or significance of projected late 21st
621 century change, nor do they provide direct guidance on expected historical forced changes. The
622 observed trends in Kossin (2018a) are thus examples of significant linear trends that are not yet
623 quantitatively linked to past anthropogenic influence based on direct model simulations.
624 One possible implication of a TC propagation speed decrease would be an increase in TC-related
625 rainfall amounts at fixed locations along the storms’ paths. Altman et al. (2013) reported very
626 strong century-scale increases in typhoon-related rainfall rates over Korea during 1904-2008,
627 although their study does not present enough methodology details for a careful assessment. Kim
628 et al. (2006) had previously reported large increases in TC-related rainfall rates in Korea
629 beginning around 1980, based on a shorter record extending back to 1954.
630 In summary, from a Type I error avoidance perspective, a slight majority of the authors (6 of 11)
631 had only low confidence that there has been a detectable decrease in global or western North
632 Pacific TC translation speeds since 1949. For the other five authors, four had low-to-medium
633 confidence, and one had medium confidence). Most authors (eight of 11) concluded that there
634 was only low confidence that anthropogenic forcing had contributed to the observed decrease.
635 Alternatively, from the perspective of reducing Type II errors, most authors (eight of 11)
636 concluded that the balance of evidence suggests there has been a detectable decrease in global
29
637 TC translation speeds since 1949; only one author concluded that the balance of evidence
638 suggests that anthropogenic forcing has contributed to the observed decrease.
639
640 5. Paleoclimate Perspectives
641
642 Paleoclimate studies of TC activity can be used to explore the expected background level of
643 natural variability based on datasets that are much longer than available instrumental records.
644 Such natural variability has the potential to be a confounding factor for detection and future
645 climate projections. Paleoclimate records can also be used for better assessing TC risk analysis
646 even in the absence of changing climate.
647 Paleoclimate TC records have been extracted at specific locations for periods up to about 8,000
648 years ago. Analysis of these data suggests large centennial-scale variations in TC activity over
649 the past few thousand years (Liu, 2001; Woodruff et al. 2012; Donnelly et al. 2015; Muller et al.
650 2017) at these locations. Reasons for these variations are not clear. Since there are clear
651 relationships in the current climate between El Niño/Southern Oscillation (ENSO) variations and
652 TCs, it is possible that such a relationship could also be observed in paleoclimate data. However,
653 Muller et al. (2017) conclude that there does not appear to be a clear association in the various
654 TC basins between periods of enhanced El Niño activity and TC activity that is consistent with
655 current climate relationships (Nott and Forsyth 2012; Haig and Nott 2016). Yan et al. (2017a)
656 identified a potential indicator of centennial-scale variations in TC numbers in the western North
657 Pacific in the past millennium, but these variations tend to be inconsistent with written typhoon
658 records from China during this time. Denniston et al. (2015) find more consistent relationships
30
659 between ENSO variations and TC incidence in the Australian region over the past 2,000 years.
660 Donnelly and Woodruff (2007) find a strong statistical relationship between a 5,000 year proxy
661 record of intense Caribbean hurricane landfalls and a proxy reconstruction of El Niño variability.
662 Several studies have suggested a link between TC incidence, patterns of SST and the position of
663 the Intertropical Convergence Zone (ITCZ), with a more northerly mean position favoring more
664 TC incidence, a result supported by climate models (Merlis et al. 2013; van Hengstum et al.
665 2016). Note that projections of future climate typically include a more poleward ITCZ in the
666 Northern Hemisphere summer (Ceppi et al. 2013). In the Atlantic, a number of paleo-TC studies
667 find relationships between higher SSTs in the main TC development region and greater numbers
668 of TCs (Donnelly et al. 2015; Trouet et al. 2016).
669 A cautionary note on these studies is that past climates (with different climate forcings) may not
670 provide good analogues for expected natural climate variability over the past century or for 21st
671 century variability and change. Also paleoclimate studies typically cover one or a limited
672 number of specific locations, and thus can only be used with caution to infer basin- or regional-
673 scale TC activity.
674 In summary, clear centennial-to-millennial scale variations in TC activity have been documented
675 in paleoclimate records, which provide some long-term context for observed TC variability on a
676 centennial scale. The paleo proxy studies to date do not contradict our conclusions about
677 detection and attribution in this report that are based on the historical-era TC data.
678
679 6. Summary
680
31
681 In this assessment, we have focused on the question: Can an anthropogenic influence on TC
682 activity be detected in past data? We explore this question from two perspectives:
683 avoiding/reducing either Type I or Type II errors, since we presume that different audiences will
684 have different preferences on which type of error should be avoided to a greater extent. A
685 summary of the distribution of author opinion on the various potential detection and attribution
686 statements discussed in Section 4 is provided in Table 1.
687 Using the conventional perspective of avoiding Type I error, the strongest case for a detectable
688 change in TC activity is the observed poleward migration of the latitude of maximum intensity in
689 the northwest Pacific basin, with eight of 11 authors rating the observed change as low-to-
690 medium confidence for detection (with one other author having medium and two other authors
691 having medium-to-high confidence). A slight majority of authors (six of 11) had only low
692 confidence that anthropogenic forcing had contributed to the poleward shift. The majority of the
693 author team also had only low confidence that any other observed TC changes represented either
694 detectable changes or attributable anthropogenic changes.
695 From the perspective of reducing Type II errors, a majority of the author team agreed on a
696 number of more speculative TC detection and/or attribution statements, which we recognize have
697 substantial potential for being false alarms (i.e., overstating anthropogenic influence) but which
698 may be indicators of emerging anthropogenic signals in the data. Most authors agreed that the
699 balance of evidence suggests detectable anthropogenic contributions to: i) the poleward
700 migration of the latitude of maximum intensity in the western North Pacific; ii) increased
701 occurrence of extremely severe (post-monsoon season) cyclonic storms in the Arabian Sea; iii)
702 increased global average intensity of the strongest TCs since early 1980s; iv) increase in global
703 proportion of TCs reaching Category 4 or 5 intensity in recent decades; and v) increased
32
704 frequency of Hurricane Harvey-like extreme precipitation events in the Texas (U.S.) region. In
705 addition, a majority of authors concluded that the balance of evidence suggested an
706 anthropogenic influence (without detection) on: vi) the unusually active TC season in the
707 western North Pacific in 2015. Author opinion was divided but a slight majority concluded that:
708 vii) unusually high TC frequency near Hawaii in 2014 was a case where the balance of evidence
709 suggested an anthropogenic influence (without detection). Finally, most authors concluded that
710 the balance of evidence suggests: viii) detectable (but not attributable) decreases in severe
711 landfalling TC frequency in eastern Australia since the late 1800s; and ix) detectable (but not
712 attributable) decreased global TC translation speeds since 1949.
713 Regarding storm surge, our expectation is that a widespread worsening of total inundation levels
714 during storms is occurring due to the global mean sea level rise associated with anthropogenic
715 warming, assuming all other factors equal, although we note that no TC climate change signal
716 has been convincingly detected in sea level extremes data. To date, there is not convincing
717 evidence of a detectable anthropogenic influence on hurricane precipitation rates, in contrast to
718 the case for extreme precipitation in general, where some anthropogenic influence has been
719 detected.
720 The relatively low confidence in TC change detection results from several factors, including:
721 observational limitations, the smallness of the expected human-caused change (signal) relative to
722 the expected natural variability (noise), or the lack of confident estimates of the expected signal
723 and noise levels. Going forward, continued development/maintenance of climate-quality TC-
724 related observed datasets, paleo-storm proxies, and TC statistics from appropriately designed
725 modeling studies will all be important for further progress. Monitoring and analysis of various
33
726 TC indices, and development of improved climate models with TC simulation capabilities, are
727 strongly recommended to help identify emerging anthropogenic TC climate change signals.
728
729
730 Acknowledgements: We thank I. Held, H. Murakami, E. Lloyd, editor C. Landsea , and three
731 reviewers for helpful comments on drafts of the manuscript. The first author acknowledges
732 particularly helpful comments and advice from A. Sobel regarding the Type I vs. Type II error
733 perspective which led to its adoption into the assessment. We thank S. Power, A. Grinsted, R.
734 Maue, R. Pielke, Jr., H. Fudeyasu, and R. Kumazawa for providing data for our figure panels.
34
735 References:
736
737 Altman, J., J. Dolezal, T. Cerny, and J.-S. Song, 2013: Forest response to increasing typhoon
738 activity on the Korean peninsula: evidence from oak tree-rings. Global Change Biol., 19, 498–
739 504, doi: 10.1111/gcb.12067
740 Balaguru, K., G.R. Foltz, and L.R. Leung, 2018: Increasing magnitude of hurricane rapid
741 intensification in the central and eastern tropical Atlantic. Geophys. Res. Lett., 45, 4238–4247.
742 doi: 10.1029/2018GL077597.
743 Barnes, E.A., L.M. Polvani, and A.H. Sobel, 2013: Model projections of atmospheric steering of
744 Sandy-like superstorms. Proc. Nat. Acad. Sci., 110, 15211–15215, doi:
745 10.1073/pnas.1308732110.
746 Bender, M., T. Knutson, R. Tuleya, J. Sirutis, G. Vecchi, S.T. Garner, and I. Held, 2010:
747 Modeled impact of anthropogenic warming on the frequency of intense Atlantic hurricanes.
748 Science, 327, 454–458.
749 Bindoff, N.L., P.A. Stott, K.M. AchutaRao, M.R. Allen, N. Gillett, D. Gutzler, K. Hansingo,
750 G. Hegerl, Y. Hu, S. Jain, I.I. Mokhov, J. Overland, J. Perlwitz, R. Sebbari, and X. Zhang, 2013:
751 Detection and Attribution of Climate Change: from Global to Regional. In: Climate Change
752 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment
753 Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner,
35
754 M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, P. M. Midgley.], Cambridge
755 University Press, 1535 pp.
756 Booth, B., 2017: Origins of Atlantic decadal swings: External modulation. Nature, 548, 284-
757 285.
758 Booth B.B.B., N.J. Dunstone, P.R. Halloran, T. Andrews, and N. Bellouin, 2012: Aerosols
759 implicated as a prime driver of twentieth-century North Atlantic climate variability. Nature, 484,
760 228-232.
761 Callaghan J, Power S.B., 2011: Variability and decline in the number of severe tropical cyclones
762 making land-fall over eastern Australia since the late nineteenth century. Clim. Dyn., 37. 647-
763 662.
764 Camargo, S.J., 2013: Global and regional aspects of tropical cyclone activity in the CMIP5
765 models. J. Climate, 26, 9880-9902, doi: 10.1175/JCLI-D-12-00549.1.
766 Camargo, S.J., and S.M. Hsiang, 2015: Tropical Cyclones: From the influence of climate to their
767 socio-economic impacts, Chapter 18, pp. 303-342, in Extreme Events: Observations, Modeling
768 and Economics, M. Chaves, M. Ghil and J. Urrutia-Fucugauchi, editors, AGU Monograph,
769 Willey-Blackwell, ISBN: 978-1-119-15701-4, doi: 10.1002/9781119157052.ch18.
770 Ceppi, P., Y. T. Hwang, X. Liu, D. M. Frierson, and D. L. Hartmann, 2013: The relationship
771 between the ITCZ and the Southern Hemispheric eddy-driven jet. J. Geophys. Res. Atmos.,
772 118(11), 5136-5146.
36
773 Chenoweth, M. and D. Divine, 2008: A document-based 318-year record of tropical cyclones in
774 the Lesser Antilles, 1690-2007. Geochem. Geophys. Geosyst., 9, Q0008013,
775 doi:10.1029/2008GC002066.
776 Choi, J.-W., Y. Cha, H.-D. Kim, and S.-D. Kang, 2016: Latitudinal change of tropical cyclone
777 maximum intensity in the western North Pacific. Adv. Meteorol., 2016, 5829162, doi:
778 10.1155/2016/5829162.
779 Chu, P.–S., J.-H. Kim, and Y. R. Chen, 2012: Have steering flows in the western North Pacific
780 and the South China Sea change over the last 50 years? Geophys. Res. Lett., 39, L10704,
781 doi:10.1029/2012GL051709.
782 Daloz, A.S. and S.J. Camargo, 2018: Is the poleward migration of tropical cyclone maximum
783 intensity associated with a poleward migration of tropical cyclone genesis? Clim. Dyn., 50, 705–
784 715, doi:10.1007/s00382-017-3636-7
785 Dean, L., K.A. Emanuel, and D.R. Chavas, 2009: On the size distribution of Atlantic tropical
786 cyclones. Geophys. Res. Lett., 36, L14803, doi:10.1029/2009GL039051.
787 Delworth, T.L., F. Zeng, L. Zhang, R. Zhang, G.A. Vecchi, and X. Yang, 2017: The central role
788 of ocean dynamics in connecting the North Atlantic Oscillation to the extratropical component of
789 the Atlantic Multidecadal Oscillation. J. Climate, 30, 3789-3805, doi:10.1175/JCLI-D-16-0358.1
790 Denniston, R.F., G. Villarini, A.N. Gonzales, K.H. Wyrwoll, V.J. Polyak, C.C. Ummenhofer,
791 M.S. Lachniet, A.D. Wanamaker, W.F. Humphreys, D. Woods, and J. Cugley, 2015: Extreme
792 rainfall activity in the Australian tropics reflects changes in the El Niño/Southern Oscillation
793 over the last two millennia. Proc. Nat. Acad. Sci., 112, 4576-4581.
37
794 Donnelly J.P., A.D. Hawkes, P. Lane, D. MacDonald, B.N. Shuman, M.R. Toomey, P. van
795 Hengstum, and J.D. Woodruff, 2015: Climate forcing of unprecedented intense-hurricane
796 activity in the last 2,000 years. Earth Future, 3, 49–65. doi:10.1002/2014EF000274
797 Donnelly, J. P., and J. D. Woodruff, 2007: Intense hurricane activity over the past 5,000 years
798 controlled by El Nino and the west African monsoon. Nature, 447, 465-468.
799 Dunion, J.P., and C.S. Velden, 2004: The impact of the Saharan air layer on Atlantic tropical
800 cyclone activity. Bull. Amer. Meteor. Soc., 85, 353–365.
801 Dunstone N.J., D.M. Smith, B.B.B. Booth, L. Hermanson, and R. Eade, 2013. Anthropogenic
802 aerosol forcing of Atlantic tropical storms. Nat. Geosci., 6,534-539
803 Easterling, D.R., K.E. Kunkel, J.R. Arnold, T. Knutson, A.N. LeGrande, L.R. Leung, R.S. Vose,
804 D.E. Waliser, and M.F. Wehner, 2017: Precipitation change in the United States. In: Climate
805 Science Special Report: Fourth National Climate Assessment, Volume I [Wuebbles, D.J., D.W.
806 Fahey, K.A. Hibbard, D.J. Dokken, B.C. Stewart, and T.K. Maycock (eds.)]. U.S. Global Change
807 Research Program, Washington, DC, USA, pp. 207-230, doi: 10.7930/J0H993CC.
808 Elsner, J. B., J.P. Kossin, and T.H. Jagger, 2008: The increasing intensity of the strongest
809 tropical cyclones. Nature, 455, 92-95. doi:10.1038/nature07234.
810 Emanuel K.A., 1988: The maximum intensity of hurricanes. J. Atmos. Sci., 45, 1143–1155.
811 Emanuel, K., 2017: Assessing the present and future probability of Hurricane Harvey’s rainfall.
812 Proc. Nat. Acad. Sci., 114, 12681-12684, doi:10.1073/pnas.1716222114.
38
813 Estrada, F., W.J.W. Botzen, and R.S.J. Tol, 2015: Economic losses from US hurricanes
814 consistent with an influence from climate change. Nat. Geosci., 8, 880-884,
815 doi:10.1038/NGEO2560.
816 Evan, A.T., J.P. Dunion, J.A. Foley, A.K. Heidinger, and C.S. Velden, 2006: New evidence for
817 a relationship between Atlantic tropical cyclone activity and African dust outbreaks. Geophys.
818 Res. Lett., 33, L19813, doi:10.1029/2006GL026408.
819 Evan A.T., J.P. Kossin, C.E. Chung, and V. Ramanathan, 2011: Arabian Sea tropical cyclones
820 intensified by emissions of black carbon and other aerosols. Nature, 479, 94–97.
821 Evan A.T., J.P. Kossin, C. Chung, and V. Ramanathan, 2012: Evan et al. reply to Wang et al.
822 (2012), “Intensified Arabian Sea tropical storms”. Nature, 489, E2–E3.
823 Grinsted, A., J.C. Moore, and S. Jevrejeva, 2012: Homogeneous record of Atlantic hurricane
824 surge threat since 1923. Proc. Nat. Acad. Sci., 109, 19601-19605, doi:10.1073/pnas.120942109.
825 Gutmann, E. D., R. M. Rasmussen, C. Liu, K. Ikeda, C. L. Bruyere, J. M. Done, L. Garre, P.
826 Friis-Hansen, V. Veldore, 2018: Changes in hurricanes from a 13-year convection-permitting
827 pseudo-global warming simulation. J. Climate , 31, 3643–3657, https://doi.org/10.1175/JCLI-D-
828 17-0391.1.
829 Hagen, A. B., and C. W. Landsea, 2012: On the classification of extreme Atlantic hurricanes
830 utilizing mid-twentieth-century monitoring capabilities. J. Climate, 25, 4461-4475.
831 Haig, J.E.A. and J. Nott, 2016: Solar forcing over the last 1500 years and Australian tropical
832 cyclone activity. Geophys. Res. Lett., 43, 2843-2850.
39
833 Haig, J., J. Nott, and G.-J. Reichart, 2014: Australian tropical cyclone activity lower than at any
834 time over the past 500-1500 years. Nature, 505, 667-671.
835 Hall, T., and K. Hereid, 2015: The frequency and duration of U.S. hurricane droughts. Geophys.
836 Res. Lett., 42, 3482–3485, doi: 10.1002/2015GL063652.
837 Hart, R.E., D.R. Chavas, and M.P. Guishard, 2016: The Arbitrary Definition of the current
838 Atlantic major hurricane landfall drought. Bull. Amer. Meteor. Soc., 97, 713–722, doi:
839 10.1175/BAMS-D-15-00185.1.
840 He, H., J. Yang, D. Gong, R. Mao, Y. Wang, and M. Gao, 2015: Decadal changes in tropical
841 cyclone activity over the western North Pacific in the late 1990s. Clim. Dyn., 45, 3317-3329, doi:
842 10.1007/s00382-015-2541-1.
843 Hegerl, G.C., O. Hoegh-Guldberg, G. Casassa, M.P. Hoerling, R.S. Kovats, C. Parmesan, D.W.
844 Pierce, and P.A. Stott, 2010: Good practice guidance paper on detection and attribution related to
845 anthropogenic climate change. Meeting Report of the Intergovernmental Panel on Climate
846 Change Expert Meeting on Detection and Attribution of Anthropogenic Climate Change.
847 Stocker, T.F., C.B. Field, D. Qin, V. Barros, G.-K. Plattner, M. Tignor, P.M. Midgley, and K.L.
848 Ebi, Eds. IPCC Working Group I Technical Support Unit, University of Bern, Bern, Switzerland,
849 1-8. http://www.ipcc.ch/pdf/supporting-
850 material/ipcc_good_practice_guidance_paper_anthropogenic.pdf
851 Holland G.J., and C. Bruyère, 2014: Recent intense hurricane response to global climate change.
852 Clim. Dyn., 42, 617-627, doi: 10.1007/s00382-013-1713-0.
40
853 Horton, B.P., R.E. Kopp, A.J. Garner, C.C. Hay, N.S. Khan, K. Roy, and T.A. Shaw, 2018:
854 Mapping sea-level change in time, space and probability. Annu. Rev. Env. Resour., 43, 13.1-
855 13.41, doi: 10.1146/annurev-environ-102017-025826.
856 Hsiang, S.M., 2010: Temperature and cyclones strongly associated with economic production in
857 the Caribbean and Central America. Proc. Nat. Acad. Sci., 107, 15367-15372, doi:
858 10.1073/pnas.1009510107.
859 Hsiang, S.M., and D. Narita, 2012: Adaptation to cyclone risk: Evidence from the global cross-
860 section. Climate Change Economics, 3, 1250011, doi: 10.1142/S20100781250011X.
861 IPCC, 2013: Climate Change 2013: The Physical Science Basis. Contribution of Working Group
862 I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge
863 University Press, Cambridge, UK and New York, NY, 1535 pp.
864 http://www.climatechange2013.org/report/
865 Kang, N.-Y., and J. B. Elsner, 2015: Trade-off between intensity and frequency of global tropical
866 cyclones. Nat. Climate Change, 5, 661–664, doi:10.1038/nclimate2646.
867 Kang, N. and J.B. Elsner, 2016: Climate mechanism for stronger typhoons in a warmer world. J.
868 Climate, 29, 1051–1057, doi:10.1175/JCLI-D-15-0585.1
869 Kim, J.-H., C.-H. Ho, M.-H. Lee, J.-H. Jeong, and D. Chen, 2006: Large increase in heavy
870 rainfall associated with tropical cyclone landfalls in Korea after the late 1970s. Geophys. Res.
871 Lett., 33, L18706, doi:10.1029/2006GL027430.
41
872 Kim, H-S, G. A. Vecchi, T. R. Knutson, W. G. Anderson, T. L. Delworth, A. Rosati, F. Zeng,
873 and M. Zhao, 2014: Tropical cyclone simulation and response to CO2 doubling in the GFDL
874 CM2.5 high-resolution coupled climate model. J. Climate, 27, DOI:10.1175/JCLI-D-13-
875 00475.1.
876 Klotzbach, P., S. Bowen, R. Pielke, and M. Bell, 2018: Continental United States hurricane
877 landfall frequency and associated damage: observations and future risks. Bull. Amer. Meteor.
878 Soc., 99, 1359-1376, doi:10.1175/BAMS-D-17-0184.1.
879 Klotzbach, P. and C. Landsea, 2015: Extremely intense hurricanes: revisiting Webster et al.
880 (2005) after 10 years. J. Climate, 28, 7621-7629.
881 Knapp, K. R., C.S. Velden, and A.J. Wimmers, 2018: A global climatology of tropical cyclone
882 eyes. Mon. Wea. Rev., 146, 2089-2101.
883 Knutson, T.R., J.L. McBride, J. Chan, K. Emanuel, G. Holland, C. Landsea, I. Held, J.P. Kossin,
884 A.K. Srivastava, and M. Sugi, 2010: Tropical cyclones and climate change. Nat. Geosci., 3, 157-
885 163, doi:10.1038/ngeo0779.
886 Knutson T.R., J. J. Sirutis, G. A. Vecchi, S. Garner, M. Zhao, H.-S. Kim, M. Bender, R. E.
887 Tuleya, I. M. Held, and G. Villarini, 2013a: Dynamical downscaling projections of late 21st
888 century Atlantic hurricane activity: CMIP3 and CMIP5 model-based scenarios. J. Climate, 26:
889 6591-6617. DOI: 10.1175/JCLI-D-12-00539.1
890 Knutson, T.R., J.J. Sirutis, M. Zhao, R.E. Tuleya, M. Bende, G.A. Vecchi, G. Villarini, and D.
891 Chavas, 2015: Global projections of intense tropical cyclone activity for the late 21st century
42
892 from dynamical downscaling of CMIP5/RCP4.5 scenarios. J. Climate, 28, doi:10.1175/JCLI-D-
893 15-0129.1
894 Knutson, T.R., F. Zeng, and A.T. Wittenberg, 2013b: Multimodel assessment of regional surface
895 temperature trends: CMIP3 and CMIP5 twentieth-century simulations. J. Climate, 26, 8709–
896 8743, doi:10.1175/JCLI-D-12-00567.1
897 Korty, R.L., S.J. Camargo, and J. Galewsky, 2012: Tropical cyclone genesis factors in
898 simulations of the Last Glacial Maximum. J. Climate, 25, 4348-4365.
899 Kossin, J.P., 2015: Validating atmospheric reanalysis data using tropical cyclones as
900 thermometers. Bull. Amer. Meteor. Soc., 96, 1089–1096.
901 Kossin, J.P., 2017. Hurricane intensification along United States coast suppressed during active
902 hurricane periods. Nature, 541, 390-394. Doi:10.1038/nature20783.
903 Kossin, J.P., 2018a: A global slowdown of tropical cyclone translation speed. Nature, 558, 104-
904 108.
905 Kossin, J.P., 2018b: Comment on “Spatial and temporal trends in the location of the lifetime
906 maximum intensity of tropical cyclones”. Atmosphere, 9, 241-244.
907 Kossin, J.P., and S.J. Camargo, 2009: Hurricane track variability and secular potential intensity
908 trends. Climatic Change, 9, 329-337.
43
909 Kossin, J.P., K.A. Emanuel, and S.J. Camargo, 2016a: Past and projected changes in western
910 North Pacific tropical cyclone exposure. J. Climate, 29, 5725–5739, doi:10.1175/JCLI-D-16-
911 0076.1.
912 Kossin J.P., K.A. Emanuel, and G.A. Vecchi, 2014. The poleward migration of the location of
913 tropical cyclone maximum intensity. Nature, 509. 349-352.
914 Kossin J.P., K.A. Emanuel, and G.A. Vecchi, 2016b: Comment on ‘Roles of interbasin
915 frequency changes in the poleward shifts of the maximum intensity location of tropical
916 cyclones’. Environ. Res. Lett., 11, 068001.
917 Kossin, J.P., T. Hall, T. Knutson, K.E. Kunkel, R.J. Trapp, D.E. Waliser, and M.F. Wehner,
918 2017: Extreme storms. In: Climate Science Special Report: Fourth National Climate Assessment,
919 Volume I [Wuebbles, D.J., D.W. Fahey, K.A. Hibbard, D.J. Dokken, B.C. Stewart, and T.K.
920 Maycock (eds.)]. U.S. Global Change Research Program, Washington, DC, USA, pp. 257-276,
921 doi: 10.7930/J07S7KXX.
922 Kossin, J.P., K.R. Knapp, D.J. Vimont, R.J. Murnane, and B.A. Harper, 2007: A globally
923 consistent reanalysis of hurricane variability and trends. Geophys. Res. Lett. 34, L04815.
924 Kossin, J.P., T.L. Olander, and K.R. Knapp, 2013: Trend analysis with a new global record of
925 tropical cyclone intensity. J. Climate, 26, 9960–9976.
926 Kumazawa, R., H. Fudeyasu, and H. Kubota, 2016: Tropical cyclone landfall in Japan during
927 1900–2014. Tenki, 63, 855–861 (in Japanese).
44
928 Kunkel, K.E., D.R. Easterling, D.A.R. Kristovich, B. Gleason, L. Stoecker, and R. Smith, 2010:
929 Recent increases in U.S. heavy precipitation associated with tropical cyclones. Geophys. Res.
930 Lett., 37, L24706, doi: 10.1029/2010GL045164.
931 Lackmann, G.M., 2015: Hurricane Sandy before 1900 and after 2100. Bull Amer. Meteor. Soc.,
932 96, 547-560. doi:10.1175/BAMS-D-14-00123.1
933 Landsea, C. W., Harper, B. A., Hoarau, K. & Knaff, J. A., 2006: Can we detect trends in
934 extreme tropical cyclones? Science, 313, 452–454.
935 Lau, W.K.M., and Y.P. Zhou, 2012: Observed recent trends in tropical cyclone rainfall over the
936 North Atlantic and North Pacific. J. Geophys. Res., 117, D03104, doi:10.1029/2011JD016510.
937 Lee, B.S., M. Haran, K. Keller, 2017: Multidecadal scale detection time for potentially
938 increasing Atlantic storm surges in a warming climate. Geophys. Res. Lett., 44, 10,617–10,623,
939 doi: 10.1002/2017GL074606.
940 Lee, T.-C., T.R. Knutson, H. Kamahori, and M. Ying, 2012: Impacts of climate change on
941 tropical cyclones in the western North Pacific Basin. Part I: Past observations. Tropical Cyclone
942 Res. Rev., 1, 213-230. DOI: 10.6057/2012TCRR02.08
943 Liang, A., L. Oey, S. Huang, S. Chou, 2017: Long-term trends of typhoon-induced rainfall over
944 Taiwan: In situ evidence of poleward shift of typhoons in western North Pacific in recent
945 decades. J. Geophys. Res. 122, 2750 – 2765, doi: 10.1002/2017JD26466.
946 Liu, K.S., and J.C.L. Chan, 2017: Variations in the power dissipation index in the East Asia
947 region. Clim. Dyn. 48, 1963-1985, doi: 10.1007/s00382-016-3185-5.
45
948 Lloyd, E.A., N. Oreskes, 2018: Climate change attribution: When is it appropriate to accept new
949 methods? Earth's Future, 6, 311–325, doi:10.1002/2017EF000665.
950 Lucas, C., B. Timbal, and H. Nguyen, 2014: The expanding tropics: a critical assessment of the
951 observational and modeling studies. WIREs Clim. Change, 5, 89-112, doi:10.1002/wcc.251
952 Mahlstein, I., R.W. Portmann, J.S. Daniel, S. Solomon, and R. Knutti, 2012: Perceptible
953 changes in regional precipitation in a future climate. Geophys. Res. Lett., 39, L05701,
954 doi:10.1029/2011GL050738.
955 Malavelle, F.F, and Coauthors, 2017: Strong constraints on aerosol–cloud interactions from
956 volcanic eruptions. Nature, 546, 485-491, doi:10.1038/nature22974
957 Mann M.E., and K.A. Emanuel, 2006: Atlantic hurricane trends linked to climate
958 change. EOS, 87. 233-244
959 Marcos, M., and P. L. Woodworth, 2018: Changes in extreme sea levels. CLIVAR
960 Variations/Exchanges, 16(1), 20-24.
961 Maue, R. N., 2011: Recent historically low global tropical cyclone activity. Geophys. Res. Lett.,
962 38, L14803. doi:10.1029/2011GL047711.
963 Mei, W., and S.-P. Xie, 2016: Intensification of landfalling typhoons over the northwest Pacific
964 since the last 1970s. Nature Geosci. 9, 753–757, doi: 10.1038/ngeo2792.
46
965 Mendelsohn, R., K. Emanuel, S. Chonabayashi, and L. Bakkensen, 2012: The impact of climate
966 change on global tropical cyclone damage. Nat. Climate Change, 2, 205–209,
967 doi:10.1038/nclimate1357.
968 Merlis, T.M., M. Zhao, and I.M. Held, 2013: The sensitivity of hurricane frequency to ITCZ
969 changes and radiatively forced warming in aquaplanet simulations. Geophys. Res. Letters, 40,
970 4109-4114.
971 Moon, I.-J., S.-H. Kim, P. Klotzbach, and J.C.L. Chan, 2015: Roles of interbasin frequency
972 changes in the poleward shifts of the maximum intensity location of tropical cyclones. Environ.
973 Res. Lett., 10, 104004.
974 Moon, I.-J., S.-H. Kim, P. Klotzbach, and J.C.L. Chan, 2016: Reply to Comment on “Roles of
975 interbasin frequency changes in the poleward shifts of the maximum intensity location of tropical
976 cyclones”. Env. Res. Lett., 11, 068002.
977 Muller, J., J.M. Collins, S. Gibson, and L. Paxton, 2017: Recent advances in the emerging field
978 of paleotempestology. In: J.M. Collins, K. Walsh (eds) Hurricanes and Climate Change Vol 3,
979 pp. 1-33.
980 Murakami, H., E. Levin, T. L. Delworth, R. Gudgel, and P-C Hsu, 2018: Dominant effect of
981 relative tropical Atlantic warming on major hurricane occurrence. Science, in press.
982 Murakami, H., G. A. Vecchi, T. L. Delworth, K. Paffendorf, L. Jia, R. G. Gudgel, and F, Zeng,
983 2015: Investigating the Influence of Anthropogenic Forcing and Natural Variability on the 2014
47
984 Hawaiian Hurricane Season. Bull. Amer. Meteor. Soc., 96(12), DOI:10.1175/BAMS-
985 EEE_2014_ch23.1 .
986 Murakami, H., G. A. Vecchi, T. L. Delworth, A. T. Wittenberg, S. Underwood, R. Gudgel, X.
987 Yang, L. Jia, F. Zeng, K. Paffendorf, and W. Zhang, 2017a: Dominant role of subtropical Pacific
988 warming in extreme eastern Pacific hurricane seasons: 2015 and the future. J. Climate, 30, 243-
989 264.
990 Murakami, H., G.A. Vecchi, and S. Underwood, 2017b: Increasing frequency of extremely
991 severe cyclonic storms over the Arabian Sea. Nat. Climate Change, 7, 885–889,
992 doi:10.1038/s41558-017-0008-6.
993 Murphy, L. N., K. Bellomo, M. Cane, and A. Clement, 2017: The role of historical forcings in
994 simulating the observed Atlantic multidecadal oscillation, Geophys. Res. Lett., 44, 2472–2480,
995 doi: 10.1002/2016GL071337.
996 NAS, 2016: Attribution of Extreme Weather Events in the Context of Climate Change. The
997 National Academies Press, Washington, DC, 186 pp. http://dx.doi.org/10.17226/21852
998 Nott J., and A. Forsyth, 2012: Punctuated global tropical cyclone activity over the past 5,000
999 years. Geophys. Res. Letters, 39, L14703, doi:10.1029/2012GL052236
1000 Oey, L.Y, and S. Chou, 2016: Evidence of rising and poleward shift of storm surge in western
1001 North Pacific in recent decades. J. Geophys. Res., 121, 5181-5192, doi: 10.1002/2016JC011777.
48
1002 Park, D. S. R., C. H. Ho, J. H. Kim, and H. S. Kim, 2013: Spatially inhomogeneous trends of
1003 tropical cyclone intensity over the western North Pacific for 1977‒2010. J. Climate, 26, 5088–
1004 5101.
1005 Park, D. S. R., C. H. Ho, J. H. Kim, 2014: Growing threat of intense tropical cyclones to East
1006 Asia over the period 1977–2010. Envir. Res. Lett., 9, 014008.
1007 Patricola, C. M., and M. F. Wehner, 2018: Anthropogenic influences on major tropical cyclone
1008 events. Nature, 563, 339–346.
1009 Peduzzi P., B. Chatenoux, H. Dao, A. De Bono, C. Herold, and co-authors, 2012: Global trends
1010 in tropical cyclone risk. Nat. Climate Change, 2, 289-294.
1011 Pielke, R. A., and co-authors, 2008: Normalized hurricane damage in the United States: 1900–
1012 2005. Nat. Hazards Rev. 9, 29. doi:10.1061/(ASCE)1527-6988(2008)9:1(29)
1013 Rajeevan, M., J. Srinivasan, K. Niranjan Kumar, C. Gnanaseelan, and M. M. Ali, 2013: On the
1014 epochal variation of intensity of tropical cyclones in the Arabian Sea. Atmos. Sci. Let., 14, 249-
1015 255, doi:10.1002/as12.447.
1016 Risser, M. D., and M. F. Wehner, 2017: Attributable human-induced changes in the likelihood
1017 and magnitude of the observed extreme precipitation during Hurricane Harvey. Geophys. Res.
1018 Lett., 44, 12,457–12,464. doi:10.1002/2017GL075888.
1019 Sharmila , S. and K. J. E. Walsh, 2018: Recent poleward shift of tropical cyclone formation
1020 linked to Hadley cell expansion. Nat. Climate Change, https://doi.org/10.1038/s41558-018-
1021 0227-5.
49
1022 Simpson, R.H. and H. Riehl, 1981: The Hurricane and Its Impact. Louisiana State Univ. Press,
1023 Baton Rouge (IBSN 0-8071-0688-7), 398 pp.
1024 Sobel, A. H., S. J. Camargo, T. M. Hall, C.-Y. Lee, M. K. Tippett, and A. A. Wing, 2016:
1025 Human influence on tropical cyclone intensity. Science, 353, 6296, 242-246.
1026 Song, J.-J., P. J. Klotzbach, 2018: What has controlled the poleward migration of annual
1027 averaged location of tropical cyclone lifetime maximum intensity over the western North Pacific
1028 since 1961? Geophys. Res. Lett., 45, 1148 – 1156, doi: 10.1002/2017GL076883
1029 Strong, J. D. O., G. A. Vecchi, and P. Ginoux, 2018: The climatological effect of Saharan dust
1030 on global tropical cyclones in a fully coupled GCM. J. Geophys. Res.: Atmos., 123, 5538–5559.
1031 https://doi.org/10.1029/2017JD027808
1032 Studholme, J. and S. Gulev, 2018: Concurrent changes to Hadley Circulation and the meridional
1033 distribution of tropical cyclones. J. Climate, 31, 4367-4389, doi:10.1175/JCLI-D-17-0852.1.
1034 Sutton, R. T., G. D. McCarthy, J. Robson, B. Sinha, A. T. Archibald, and L. J. Gray, 2018:
1035 Atlantic multidecadal variability and the U.K. ACSIS Program. Bull. Amer. Meteor. Soc., 99,
1036 415-425.
1037 Sweet, W., C. Zervas, S. Gill, and J. Park, 2013: Hurricane Sandy innundation probabilities
1038 today and tomorrow. Bull. Amer. Meteor. Soc., 94. S17-S20.
1039 Takahashi, C., M. Watanabe, and M.Mori, 2017: Significant aerosol influence on the recent
1040 decadal decrease in tropical cyclone activity over the western North Pacific. Geophys. Res. Lett.,
1041 44, 9496–9504, https://doi.org/10.1002/2017GL075369
50
1042 Takayabu, I., et al., 2015: Climate change effects on the worst-case storm surge: a case study of
1043 Typhoon Haiyan. Environ. Res. Lett., 10, 064011
1044 Tennille, S. A., and K. N. Ellis, 2017: Spatial and temporal trends in the location of the lifetime
1045 maximum intensity of tropical cyclones. Atmosphere, 8, 198, doi:10.3390/atmos8100198
1046 Ting, M., S.J. Camargo, C. Li, and Y. Kushnir, 2015. Natural and forced North Atlantic
1047 hurricane potential intensity change in CMIP5 models. J. Climate, 28, 3926-3942, doi:
1048 10.1175/JCLI-D-14-00520.1.
1049 Trenberth, K. E., L. Cheng, P. Jacobs, Y. Zhang, and J. Fasullo, 2018: Hurricane Harvey links to
1050 ocean heat content and climate change adaptation. Earth’s Future, 2018EF000825.
1051 Trouet, V., G.L. Harley, and M. Domínguez-Delmás, 2016: Shipwreck rates reveal Caribbean
1052 tropical cyclone response to past radiative forcing. Proc. Natl. Acad. Sci, 113, 3169–3174.
1053 doi:10.1073/pnas.1519566113
1054 Truchelut, R. E., and E. M. Staehling, 2017: An energetic perspective on United States tropical
1055 cyclone landfall droughts. Geophys. Res. Lett., 44, 12,013–12,019.
1056 https://doi.org/10.1002/2017GL076071
1057 Van Oldenborgh, G. J., K. van der Wiel, A. Sebastian, R. Singh, J. Arrighi, F. Otto, K. Haustein,
1058 S. Li, G. Vecchi, and H. Cullen, 2017: Attribution of extreme rainfall from Hurricane Harvey,
1059 August 2017. Environ. Res. Lett., 12, 124009.
1060 Vecchi, G. A., and T. L. Delworth, 2017: Origins of Atlantic decadal swings: Integrate the
1061 whole system. Nature, 548, 284-285.
51
1062 Vecchi, G. A., and T. R. Knutson, 2008: On estimates of historical North Atlantic tropical
1063 cyclone activity. J. Climate, 21(14), DOI:10.1175/2008JCLI2178.1, 3580-3600.
1064 Villarini G., and G. A. Vecchi,2012: Twenty-first-century projections of North Atlantic tropical
1065 storms from CMIP5 models. Nat. Climate Change, 2: 604-607.
1066 Villarini G, and G. A. Vecchi, 2013: Projected increases in North Atlantic tropical cyclone
1067 intensity from CMIP5 models. J. Climate, 26, 3231-3240.
1068 Wahl, T., and D. P. Chambers, 2015: Evidence for multidecadal variability in U.S. extreme sea
1069 level records. J. Geophys. Res. Oceans, 120, 1527-1544.
1070 Walsh, K. J. E., J. L. McBride, P. J. Klotzbach, S. Balachandran, S. J. Camargo, G. Holland, T.
1071 R. Knutson, J. P. Kossin, T-C Lee, A. Sobel and M. Sugi, 2016: Tropical cyclones and climate
1072 change. Wiley Interdisciplinary Reviews: Climate Change, 7(1), DOI:10.1002/wcc.371 .
1073 Wang, B., S. Xu, and L. Wu, 2012: Intensified Arabian Sea tropical storms. Nature, 489: E1-E2,
1074 doi:10.1038/nature11470.
1075 Wang C. C., B. X. Lin, C. T. Chen, and S. H. Lo, 2015: Quantifying the effects of long-term
1076 climate change on tropical cyclone rainfall using a cloud-resolving model: Examples of two
1077 landfall typhoons in Taiwan. J. Climate, 28, 66–85, https://doi.org/10.1175/JCLI-D-14-00044.1
1078 Wang, C., H. Liu, S.K. Lee and R. Atlas, 2011: Impact of the Atlantic warm pool on United
1079 States landfalling hurricanes. Geophys. Res. Lett., 38, L19702, doi:10.1029/2011GL049265
52
1080 Wang, S. S.-Y., L. Zhao, J.-H. Yoon, P. Klotzbach, and R. R. Gillies, 2018: Quantitative
1081 attribution of climate effects on Hurricane Harvey’s extreme rainfall in Texas. Environ. Res.
1082 Lett., 13, 054014
1083 Wang Y, K. H. Lee, Y. Lin, M. Levy, and R. Zhang, 2014: Distinct effects of anthropogenic
1084 aerosols on tropical cyclones. Nat. Climate Change, 4, 368–373, doi:10.1038/nclimate2144
1085 Webster, P. J., G. J. Holland, J. A. Curry, and H.-R. Chang, 2005: Changes in tropical cyclone
1086 number, duration, and intensity in a warming environment. Science 309, 1844–1846.
1087 Wehner, M.F., Colin Zarzycki, C. Patricola, 2018: “Estimating the human influence on tropical
1088 cyclone intensity as the climate changes” in Hurricane Risk, Collins, J.M. and K. Walsh, Eds.,
1089 Vol. 3. Springer, in press.
1090 Weinkle, J., R. Maue, and R. Pielke, 2012: Historical global tropical cyclone landfalls. J.
1091 Climate, 25, 4729–4735, https://doi.org/10.1175/JCLI-D-11-00719.1
1092 Woodruff, J.D., J. L. Irish, and S. J. Camargo, 2013: Coastal flooding by tropical cyclones and
1093 sea level rise. Nature, 504, 44-52. doi: 10.1038/nature12855
1094 Woodruff, J.D., S.L. Sriver, and D.C. Lund, 2012: Tropical cyclone activity and western North
1095 Atlantic stratification over the last millennium: a comparative review with viable connections. J.
1096 Quat. Sci., 27, 337–343.
1097 Yan, Q., T. Wei, and Z. Zhang, 2017a: Variations in large-scale tropical cyclone genesis factors
1098 over the western North Pacific in the PMIP3 last millennium simulations. Clim. Dyn., 48, 957-
1099 970.
53
1100 Yan, X., R. Zhang, and T. R. Knutson, 2017b: The role of Atlantic overturning circulation in the
1101 recent decline of Atlantic major hurricane frequency. Nat. Commun., 8, 1695,
1102 doi:10.1038/s41467-017-01377-8.
1103 Yang, S., N. Kang, J. B. Elsner, and Y. Chun, 2018: Influence of global warming on western
1104 North Pacific tropical cyclone intensities during 2015. J. Climate, 31, 919–925,
1105 https://doi.org/10.1175/JCLI-D-17-0143.1
1106 Zhan, R., and Y. Wang, 2017: Weak tropical cyclones dominate the poleward migration of the
1107 annual mean location of lifetime maximum intensity of northwest Pacific tropical cyclones since
1108 1980. J. Climate, 30, 6873–6882.
1109 Zhang R., T. L. Delworth, R. Sutton, D. Hodson, K. W. Dixon,I. M. Held, Y. Kushnir, D.
1110 Marshall, Y. Ming, R. Msadek, J. Robson, A. Rosati, M.Ting, G. A. Vecchi, 2013: Have aerosols
1111 caused the observed Atlantic multidecadal variability? J. Atmos. Sci. 70, 1135–1144.
1112 Zhang, L., T. Retchman, K. K. Karnauskas, L. Li, J. P. Donnely, and J. P. Kossin, 2017:
1113 Longwave emission trends over Africa and implications for Atlantic hurricanes. Geophys. Res.
1114 Lett., 44, 9075-9083, doi: 10.1002/2017GL073869.
1115 Zhang, W., G. A. Vecchi, H. Murakami, G. Villarini, T. L. Delworth, X. Yang, and L. Jia, 2018:
1116 Dominant role of Atlantic Multi-decadal Oscillation in the recent decadal changes in western
1117 North Pacific tropical cyclone activity. Geophys. Res. Lett., 45(1), DOI:10.1002/2017GL076397
1118 Zhang, W., G. A. Vecchi, H. Murakami, T. L. Delworth, K. Paffendorf, L. Jia, G. Villarini, R. G.
1119 Gudgel, F. Zeng, and X. Yang, 2016: Influences of natural variability and anthropogenic forcing
54
1120 on the extreme 2015 accumulated cyclone energy in the western North Pacific [in “Explaining
1121 Extremes of 2015 from a Climate Perspective”]. Bull. Amer. Meteor. Soc., 97(12),
1122 DOI:10.1175/BAMS-D-16-0146.1 S131-S135.
1123 Zhang, W., G. Villarini, G. A. Vecchi, and J. A. Smith, 2018: Urbanization exacerbated the
1124 rainfall and flooding caused by hurricane Harvey in Houston. Nature, 563, 384–388.
1125 Zhao, J., R. Zhan, Y. Wang, and H. Xu, 2018: Contribution of the Interdecadal Pacific
1126 Oscillation to the recent abrupt decrease in tropical cyclone genesis frequency over the western
1127 North Pacific since 1998. J. Climate, 31, 8211–8224, https://doi.org/10.1175/JCLI-D-18-0202.1
55
1128 Figure Captions
1129 Fig. 1. Sample long observed TC timeseries. a) Annual number of severe landfalling TCs in
1130 eastern Australia (1873-2011), with the linear trend shown as a black line (S. Power, personal
1131 communication 2018, updated from Callaghan and Power 2011); b) annual number (bars) and
1132 five-year running mean (line) of TC landfalls on Japan for 1900-2014 (H. Fudeyasu, personal
1133 communication 2018, adapted from Kumazawa et al. 2016; c) five-year running mean (thick
1134 line) and annual (thin line) count of U.S. landfalling hurricanes (1878-2017) from
1135 www.amol.noaa.gov/hrd/hurdat/comparison_table.html; d) index of moderately large U.S. surge
1136 events (tropical storm size and higher) for 1923-2016 (A. Grinsted, personal communication
1137 2018, updated from Grinsted et al. 2012); e) Atlantic basin TC power dissipation index and
1138 tropical Atlantic sea surface temperature index (1949-2017), low-pass filtered (updated from
1139 Emanuel 2007); f) lifetime maximum TC intensities from homogenized ADT-HURSAT dataset
1140 (1982-2009) displayed as quantiles from 0.5 (median) to 0.9 as a function of year, with linear
1141 trends superimposed (Kossin et al. 2013); g) global annual occurrence frequency of all TCs (top
1142 curve) and hurricane-intensity TCs (bottom curve) as 12-month running sums for 1970-May
1143 2018 (R. Maue, personal communication 2018, updated from Maue 2011); h) annual average
1144 latitude of maximum TC intensity in the western North Pacific, with El Niño and Pacific Decadal
1145 Oscillation influences removed by linear regression, straight line depicting the linear trend
1146 excluding the final year, and gray shading the 95% confidence bounds (Kossin et al. 2018b); i)
1147 global frequency of landfalling TCs of hurricane-strength (blue) or major hurricane strength
1148 (red) for 1970-2016 (R. Pielke, Jr., and R. Maue, personal communication 2018; updated from
1149 Weinkle et al. 2012); j) global average propagation speed of tropical cyclones (1949-2016) and
1150 its linear trend, with gray shading depicting 95% confidence bounds on the trend (Kossin 2018a).
56
1151 Panels c, g, h, i, and j use approximately the same data source: different revisions of version 3 of
1152 IBTrACS (International Best Track Archive for Climate Stewardship) without adjustment.
57
1153
Fig. 1. Sample long observed TC timeseries. a) Annual number of severe landfalling TCs in eastern Australia
1154
(1873-2011), with the linear trend shown as a black line (S. Power, personal communication 2018, updated
from Callaghan and Power 2011); b) annual number (bars) and five-year running mean (line) of TC landfalls on
1155 Japan for 1900-2014 (H. Fudeyasu, personal communication 2018, adapted from Kumazawa et al. 2016; c) five-
year running mean (thick line) and annual (thin line) count of U.S. landfalling hurricanes (1878-2017) from
www.amol.noaa.gov/hrd/hurdat/comparison_table.html; d) index of moderately large U.S. surge events
(tropical storm size and higher) for 1923-2016 (A. Grinsted, personal communication 2018, updated from
Grinsted et al. 2012); e) Atlantic basin TC power dissipation index and tropical Atlantic sea surface temperature
index (1949-2017), low-pass filtered (updated from Emanuel 2007); f) lifetime maximum TC intensities from
homogenized ADT-HURSAT dataset (1982-2009) displayed as quantiles from 0.5 (median) to 0.9 as a function of
year, with linear trends superimposed (Kossin et al. 2013); g) global annual occurrence frequency of all TCs (top
curve) and hurricane-intensity TCs (bottom curve) as 12-month running sums for 1970-May 2018 (R. Maue,
personal communication 2018, updated from Maue 2011); h) annual average latitude of maximum TC intensity
in the western North Pacific, with El Niño and Pacific Decadal Oscillation influences removed by linear
regression, straight line depicting the linear trend excluding the final year, and gray shading the 95% confidence
bounds (Kossin et al. 2018b); i) global frequency of landfalling TCs of hurricane-strength (blue) or major
hurricane strength (red) for 1970-2016 (R. Pielke, Jr., and R. Maue, personal communication 2018; updated
from Weinkle et al. 2012); j) global average propagation speed of tropical cyclones (1949-2016) and its linear
trend, with gray shading depicting 95% confidence bounds on the trend (Kossin 2018a). Panels c, g, h, i, and j
use approximately the same data source: different revisions of version 3 of IBTrACS (International Best Track
Archive for Climate Stewardship) without adjustment.
58
1156
Figure 1. (contd.)
1157
59
Table 1. Distribution of author opinion on potential tropical cyclone detection & attribution statements
.elicitation.
PERSPECTIVE: TYPE I error avoidance:
1. The estimated contribution of decreased anthropogenic aerosol forcing to the increased Atlantic TC frequency
since the 1970s is large and positive and is highly unusual (e.g., p<0.05) compared to natural variability.
Confidence: Low (7); Low-to-medium (2); Medium (1); Medium-to-high (1)
2. Observed poleward migration of latitude of max intensity in NWPac basin is highly unusual (e.g., p<0.05;
statistically distinguishable) compared with expected natural variability. Confidence: Low-to-Medium (8);
Medium (1); Medium-to-high (2)
3. Anthropogenic forcing has contributed to the observed poleward migration of the latitude of maximum
intensity in the NW Pacific basin. Confidence: Low (6); low-to-medium (2); Medium (3).
4. There has been a detectable decrease (highly unusual compared to natural variability; e.g., p<0.05) in the global
scale propagation speed of TCs since 1949. Confidence: Low (6); Low-to-medium (4); Medium (1)
5. Anthropogenic forcing has contributed to the observed decrease in the global scale propagation speed of TCs
since 1949. Confidence: Low (8); Low-to-medium (3)
6. List any other observed multidecadal to century-scale change in TC activity that is highly unusual (e.g., p<0.05;
statistically distinguishable) compared with expected natural variability (from a Type I error avoidance
perspective), and provide confidence level. None identified
PERSPECTIVE: TYPE II error avoidance. Both detection and attribution sub-statements are prefaced by "The
balance of evidence suggests …"; and "Detectable" refers to "(unusual compared to natural variability, e.g.,
p<0.1)"
7. Detectable increase in N. Atlantic TC activity since the 1970s (9% agree); and anthropogenic forcing (reduced
aerosol forcing) has contributed to this increase (45% agree)
8. Observed poleward migration of latitude of max intensity in NWPac basin is detectable (all agree); and
anthropogenic forcing has contributed to the observed poleward migration of the latitude of maximum intensity
in the NW Pacific basin (82% agree)
9. Detectable increase in TC intensity over the Arabian Sea (pre-monsoon season) 1979-2010 (none agree); and
anthropogenic forcing has contributed to this increase (none agree).
10. Detectable increase in the frequency of extremely severe cyclonic storms over the Arabian Sea (post-monsoon
season) over 1998-2015 (all agree); and anthropogenic forcing has contributed to this increase (73% agree)
11. Detectable increase in the global proportion of TCs reaching Category 4 or 5 intensity in recent decades (all
agree); and anthropogenic forcing has contributed to this increase (73% agree).
12. Detectable increase in the global average intensity of strongest (hurricane intensity) TCs since the early 1980s
(91% agree); and anthropogenic forcing has contributed to this increase of global average intensity of strongest
(hurricane intensity) TCs (73% agree).
13. Detectable multidecadal increase in TC occurrence near Hawaii (none agree); and anthropogenic forcing
contributed to the recent unusually active TC season near Hawaii in 2014 (55% agree).
14. Detectable increase in TC occurrence activity in the western North Pacific in recent decades (none agree); and
anthropogenic forcing contributed to the recent unusually active TC season, including the record-setting (1984-
2015) TC intensity, in the western North Pacific in 2015 (73% agree)
15. Detectable increase in the intensity of Hurricane Sandy-like storms in the Atlantic in recent decades (none
agree); and anthropogenic forcing contributed to the intensity of Hurricane (Superstorm) Sandy in 2012 (none
agree)
16. Detectable increase in the intensity of Haiyan-like supertyphoons in the western North Pacific in recent
decades (18% agree); and anthropogenic forcing contributed to the intensity of supertyphoon Haiyan in 2013 (45%
agree)
17. Detectable long-term increase in the occurrence of Hurricane Harvey-like extreme precipitation events in the
Texas region (U.S.) (all agree); and anthropogenic forcing has contributed to increased frequency of Hurricane
Harvey-like precipitation events in the Texas (U.S.) region (all agree);
18. Detectable increase in the frequency of moderately large U.S. surge events since 1923 as documented by the
index of Grinsted et al. (which strongly filters out sea level rise influences) (18% agree); and anthropogenic forcing
has contributed to this increase (18% agree)
60
Table 1. (contd.)
19. Detectable decrease in the global scale propagation speed of TCs since 1949 (73% agree); and anthropogenic
forcing has contributed to this decrease (9% agree).
20. Detectable decrease in severe landfalling TCs in eastern Australia since the late 1800 (82% agree); and balance
of evidence suggests anthropogenic forcing has contributed to this decrease (none agree)
21. Detectable decrease in US landfalling hurricance frequency since the late 1800s (none agree); and
anthropogenic forcing has contributed to this decrease (none agree)
22. Detectable Increase in global major hurricane landfall frequency in recent decades (none agree); and
anthropogenic forcing has contributed to this increase (none agree)
23. Detectable decrease in TC frequency in the southeastern part of the western North Pacific (1992-2011) (none
agree); and anthropogenic forcing (changes in aerosol emissions) has contributed to this decrease (50% agree).
1158
61
1159 Page left intentionally blank.
1160
1161
1162
1163
1164
1165

Tropical Cyclones and Climate Change Assessment: Part I. Detection and Attribution

Uploaded by

Copyright:

Available Formats

Tropical Cyclones and Climate Change Assessment: Part I. Detection and Attribution

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

Tropical Cyclones and Climate Change Assessment: Part I. Detection and Attribution

Uploaded by

Copyright:

Available Formats

1

1 Tropical Cyclones and Climate Change Assessment: Part I. Detection

© 2019 American Meteorological Society

42 challenges for analyses based on TC intensity and frequency.

43 A number of specific published conclusions (case studies) about possible detectable

49 TC changes demonstrate discernible anthropogenic influence, or whether any other observed

50 changes represent detectable changes.

74 long-term (multidecadal to century scale) changes in TC activity, or about attribution of past TC

75 changes or events to anthropogenic forcing. We next developed potential detection/attribution

79 agree/disagree opinion on each potential statement. The summary assessment statements

82 author and each potential statement is contained in Supplemental Material.

87 Hsiang and Narita 2012; Camargo and Hsiang, 2015).

89 2. Background on Detection and Attribution of TC Changes

126 evaluating how unusual an observed climate variability/change is compared to preindustrial

138 from natural variability.

148 using well-accepted methods.

150 is context- and audience-dependent. If the goal is to advance scientific understanding, an

153 infrastructure development in coastal regions may want to consider emerging

169 also potential for confidence in TC detection/attribution statements to be revised upward in

170 future assessments.

172 3. Expected detectability of human influence on TC activity

174 A timescale of expected emergence of a detectable signal in observations can sometimes be

178 and Riehl 1981; https://en.wikipedia.org/wiki/Saffir-Simpson_scale). Kossin et al. (2013) noted

197 4. Case Studies – Assessment of detection and attribution

200 anthropogenic influence on TCs.

202 a) Case Study: Aerosol forcing influence on Atlantic TC activity.

221 meridional overturning circulation (AMOC).

232 driving AMO phases.

249 ranked it between low-to-medium and medium-to-high confidence.

270 2016a; Kossin 2018).

288 Lee et al. 2012).

296 to be robust to the use of annual or July-November averaged climate indices.

305 compared with internal variability simulated by climate models.

315 contributed to the observed migration.

317 c) Case Study: Multidecadal to century-scale landfalling TC records

330 show no strong evidence for trends.

349 d) Case Study: Trends in global TC intensity

361 claim to identify a detectable anthropogenic influence.

372 anthropogenic forcing responses.

392 to a balance of evidence level using appropriate attribution methodologies.

395 e) Case Study: Atlantic hurricane surge record since 1923

404 should be confidently identified by 2030.

425 surge events described by the Grinsted et al. index.

428 f) Case Study: Event attribution for tropical cyclones

442 frequency or steering flows (e.g., Barnes et al. 2013).

458 anthropogenic forcing.

462 anthropogenic attribution are examples of attribution without detection.

473 g) Case Study: Hurricane Harvey’s (2017) extreme rainfall

485 to the storm total rainfall.

515 surface warming.

522 convincing evidence of a detectable anthropogenic influence specifically on hurricane-related

527 evidence for detectable anthropogenic influence on TC rainfall.

534 anthropogenic forcing has contributed to this increase.

538 h) Case Studies: Changes in Arabian Sea TCs

550 or an anthropogenic contribution to the observed TC intensity increase.

558 2015 climate forcings, relative to 1860.

577 anthropogenic forcing has contributed to the increase.

580 i) Case Study: Global increase in proportion of Category 4-5 TCs

587 individual TC basins.