Figures: Top: Eastern Pacific Warming (EPW) Atlantic storm tracks for seasons indicated on the plot. EPW is the “traditional El Nino”. Bottom left: Central Pacific Warming (CPW) or the “new” El Nino described in the paper and right: Eastern Pacific Cooling or La Nina. Track legend: [TS,Hurr,Major,Cat5]:[blue,green,red,pink]
A new paper by Kim, Webster, and Curry in Science (2009) has arrived. The paper is accessible to the general reader without training in tropical meteorology or an exhaustive knowledge of the rapidly evolving state of El Nino science.
Abstract:
Two distinctly different forms of tropical Pacific Ocean warming are shown to have substantially different impacts on the frequency and tracks of North Atlantic tropical cyclones. The eastern Pacific warming (EPW) is identical to that of the conventional El Niño, whereas the central Pacific warming (CPW) has maximum temperature anomalies located near the dateline. In contrast to EPW events, CPW episodes are associated with a greater-than-average frequency and increasing landfall potential along the Gulf of Mexico coast and Central America. Differences are shown to be associated with the modulation of vertical wind shear in the main development region forced by differential teleconnection patterns emanating from the Pacific. The CPW is more predictable than the EPW, potentially increasing the predictability of cyclones on seasonal time scales.
Press release:
New type of El Nino could mean more hurricanes make landfall
El Niño years typically result in fewer hurricanes forming in the Atlantic Ocean. But a new study suggests that the form of El Niño may be changing potentially causing not only a greater number of hurricanes than in average years, but also a greater chance of hurricanes making landfall, according to climatologists at the Georgia Institute of Technology. The study appears in the July 3, 2009, edition of the journal Science.
“Normally, El Niño results in diminished hurricanes in the Atlantic, but this new type is resulting in a greater number of hurricanes with greater frequency and more potential to make landfall,” said Peter Webster, professor at Georgia Tech’s School of Earth and Atmospheric Sciences.
That’s because this new type of El Niño, known as El Niño Modoki (from the Japanese meaning “similar, but different”), forms in the Central Pacific, rather than the Eastern Pacific as the typical El Niño event does. Warming in the Central Pacific is associated with a higher storm frequency and a greater potential for making landfall along the Gulf coast and the coast of Central America.
Even though the oceanic circulation pattern of warm water known as El Niño forms in the Pacific, it affects the circulation patterns across the globe, changing the number of hurricanes in the Atlantic. This regular type of El Niño (from the Spanish meaning “little boy” or “Christ child”) is more difficult to forecast, with predictions of the December circulation pattern not coming until May. At first glance, that may seem like plenty of time. However, the summer before El Niño occurs, the storm patterns change, meaning that predictions of El Niño come only one month before the start of hurricane season in June. But El Niño Modoki follows a different prediction pattern.
“This new type of El Niño is more predictable,” said Webster. “We’re not sure why, but this could mean that we get greater warning of hurricanes, probably by a number of months.”
As to why the form of El Niño is changing to El Niño Modoki, that’s not entirely clear yet, said Webster.
“This could be part of a natural oscillation of El Niño,” he said. “Or it could be El Niño’s response to a warming atmosphere. There are hints that the trade winds of the Pacific have become weaker with time and this may lead to the warming occurring further to the west. We need more data before we know for sure.”
In the study, Webster, along with Earth and Atmospheric Sciences Chair Judy Curry and research scientist Hye-Mi Kim used satellite data along with historical tropical storm records and climate models.
The research team is currently looking at La Niña, the cooling of the surface waters in the Eastern and Central Pacific.
“In the past, La Nina has been associated with a greater than average number of North Atlantic hurricanes and La Nina seems to be changing its structure as well,” said Webster. “We’re vitally interested in understanding why El Niño-La Niña has changed. To determine this we need to run a series of numerical experiments with climate models.”
Climate Audit 
112 Comments
“In the past, La Nina has been associated with a greater than average number of North Atlantic hurricanes and La Nina seems to be changing its structure as well,” said Webster. “We’re vitally interested in understanding why El Niño-La Niña has changed. To determine this we need to run a series of numerical experiments with climate models.”
Average number (counts) of NA hurricanes, I don’t need no stinkin numbers, I need Cat45 percentages.
“This new type of El Niño is more predictable,” said Webster. “We’re not sure why, but this could mean that we get greater warning of hurricanes, probably by a number of months.”
Maybe I’m just dense today, but I find this series of statements completely baffling.
With regards to this season’s hurricane forecast, there is a 50% chance of the activity being significantly higher than average, based on what is projected to happen in the pacific. for EMCWF’s coupled climate model prediction of Nino 3 and 4 (note most ENSO sites only show Nino 3.4) can be found at:
Nino 3 (EPW): http://www.ecmwf.int/products/forecasts/d/charts/seasonal/forecast/seasonal_range_forecast/nino_plumes_public_s3/
Nino 4 (CPW): http://www.ecmwf.int/products/forecasts/d/charts/seasonal/forecast/seasonal_range_forecast/nino_plumes_public_s3!4!200906!/
The model ensembles suggest a high probability of full blown El Nino in August, with Nino 4 looking warmer than Nino 3. For a Modoki event, would require Nino 4 greater than 1 degree anomaly, but Nino 3 less than 1 degree. most likely scenario is a mixed event, with both Nino 3 and Nino 4 greater than 1 degree at some point in late summer.
What this means is that we could see some significant late season hurricane activity, with elevated likelihood of gulf and central america landfalls. If it is a regular El Nino, then lower than usual activity. A highly unpredictable situation that turns on the nuance of how warm Nino 3 gets.
so did any of you make a high forecast for this season?
Re: Judith Curry (#3),
“With regards to this season’s hurricane forecast, there is a 50% chance of the activity being significantly higher than average, based on what is projected to happen in the pacific. for EMCWF’s coupled climate model prediction of Nino 3 and 4 (note most ENSO sites only show Nino 3.4)”
Is “significantly higher than average” a quantitative or qualitative standard?
And what is meant, precisely, by “activity”? Huricane days? Named storms? ACE? Integrated Radial Energy?
I think I get the “50% chance” part. 😉
Re: Judith Curry (#3),
Implying that there’s a 50% chance that it’s significantly lower than average! Flip the coin, anyone?
Re: John Baltutis (#14), Uh, are you excluding the middle? what about exactly average?
Re: John Baltutis (#14),
No, it means there is a 50% chance that the activity will be lower than “significantly higher”. I.e., the median of the prediction range is “significantly higher”.
It is an imprecise, but not vacuous, prediction.
Re: Judith Curry (#3),
I put my guess in several years ago. It is a matter of record.
-snip
snip by author
Re: ryanm (#5),
I’ve read that El Nino’s that occur in the negative phase of the PDO are milder than when in the positive phase (and the reverse, that La Nina’s are strong in negative pdo and weaker in positive pdo). Is this what you are referring to or is this something different. Also, is there any breakdown of “Modoki” El Nino’s vs normal El Nino’s? Are they more likely during negative PDO, positive PDO, or totally unrelated to PDO?
Re: Jonathan Schafer (#7), the Kim et al. (2009) Science paper does not address this issue. The August-October mean SST in the Pacific is detrended from 1950-2006 in the paper, so it is very difficult to determine the relationship between PDO and ENSO. Indeed, the El Nino Modoki idea coined largely in 2004, is somewhat controversial with some thinking it is indeed a reflection of the PDO and interactions with the North Pacific Gyre.
The El Nino Modoki years (Central Pacific warming > 1C) as described in the paper are: 1969, 1991, 1994, 2002, 2004.
Traditional El Nino years with Eastern Pacific warming: 1951, 1957, 1963, 1965, 1972, 1976, 1982, 1987, 1997
Eastern Pacific cold years (La Nina): 1954, 1955, 1964, 1970, 1971, 1973, 1975, 1985, 1988, 1995, 1998, 1999.
As you can clearly see from those 5 Modoki events out of the past 57 years, the proportion of Modoki events has been increasing during recent decades at the expense of traditional El Nino events/years.
“El Niño Modoki”-hm…any time series for this that could allow us to test the claims about it? a ENMO index or something?
Note: Modoki is Japanese: “similar but different”.
So are they saying this new type of El Nino is really ‘NEW’ or is it just that our modern technology allows us to pick up, for want of a better word, signals that were always present only we didn’t know it?
snip
Wouldn’t the El Nino warm pool tend to be further west during times when the PDO is in its cool phase, therefore preventing the suppression of hurricanes that happened eg in 1997?
It looks to my amateur eyes that El Nino Modoki is another phrase for “weak El Nino”.
the first thing i did was look for a PDO relationship, but doesnt’ seem to be there. frustrating thing is that SST in the tropical pacific for discerning spatial variability aren’t reliable prior to about 1920
The other pacific mode of variability to look at is the north pacific gyre oscillation (NPGO)
Click to access DiLorenzo_Schneider2009.pdf
Click to access DiLorenzo_GRL_2008.pdf
ryan note: Di Lorenzo, when he saw your ACE paper, thought he saw NPGO in your ACE plot
i’ll try sending again, my first version is “awaiting moderation”
The other pacific mode of variability to look at is the north pacific gyre oscillation (NPGO)
Click to access DiLorenzo_Schneider2009.pdf
Click to access DiLorenzo_GRL_2008.pdf
ryan note: Di Lorenzo, when he saw your ACE paper, thought he saw NPGO in your ACE plot
Re: Judith Curry (#13), Why does NPGO (and NPO?) drop off the face of the Earth after the 97-8 El Nino?
Good to see Dr. Curry back.
snip by author
i just checked some of the news articles coming out about our paper, doesn’t look like there is any controversy on this one, even landsea and klotzbach made explanatory rather than critical comments. we’ll see . . .
Judith, will the paper be available at your website in the near future?
David, will get it posted next monday on my website. in the mean time, if anyone is interested in a copy, send me an email.
Moderation needs to get a bit stronger.
It is both rude and pointless to reply to a paper and remarks by Dr Curry, who is a serious and objective scientist, with these silly gibes which seem witty only to their originators. This stuff is worthy of the worst aspects of the know-nothing sections of the AGW lobby. It is as bad or worse than the worst comments one meets with on the poisonous rabbet.
Skeptics have to do better, and if they cannot, should be deleted.
Steve: I don’t think that the comments in question were “poisonous” or fit the above description, other than there were some pointless comments, which I’ve thinned. I continually ask readers not to vent or pile on or to make editorial comments about the sins of models and do so again.
I don’t see anything unusual happening in the Nino 4 region or the Nino 3 vs the Nino 4 region.
Secondly, the wind shear affecting hurricanes in the Atlantic does not come from the Nino 4 region but mainly from around Mexico and the equatorial region in the Americas (sometimes from the Pacific side but only right near the coast, not out in the middle of the Pacific). It is possible that the global weather conditions that affect wind shear coming from Mexico can have an impact on the ENSO, but it is not the ENSO itself that is causing this.
It is intuitively hard to grasp that an increase of a degree or so in SST changes large scale patterns of storms like these, especially as they often start away from the hottest parts of the sea region. The favoured mechanisms seem to rely upon this higher SST (which is a most imprecise measurement in any case).
Also often seen is that statement that the storm intensity decreases after land is crossed. This is not the case with a number of the West Australian tropical cyclones that that have persisted over land for up to 1200 km of track. Obviously, SST is not part of the mechanism over land.
Would it not be prudent to explain exceptions like this, before finessing too much about Midoki? Perhaps there is a Fushigi type?
Re: Geoff Sherrington (#26),
The West Australian tropical cyclones with the “1200 km of track” over land are actually no longer tropical cyclones despite the continued usage of the term after prolonged landfall. Instead, they are warm core tropical cyclones transitioning and transitioned into cold core extratropical cyclones. As the warm core tropical cyclone moves westwards it also veers southwards into the mid-latitudes until the prevailing mid-latitude trade winds drags it eastwards and southwards. When the mid-latitude lows produce an occluded front, an extratropical cyclone may develop, which can merge with a tropical cyclone veering into the mid-latitude track of the extratropical cyclone. The extratropical cyclone obtains its energy from the temperature differentials existing in the occluded fronts of the air masses. When the low pressure region of the warm core tropical cyclone merges with the low pressure region of the cold core extratropical cyclone, the tropical cyclone transitions into a merged extratropical cyclone. As the transition takes place from a warm core tropical cyclone into a cold core extratropical cyclone, the cyclone becomes less and less dependent upon energy obtained from the SST of warm maritime seas and more dependent upon thermal differentials between the mid-latitude or extratropical air masses existing along the fronts as they occlude. Since the transitioning and transitioned cyclone obtains its energy principally from the occluded air masses, the cyclone can sustain its energy and sometimes increase its activity while over land and without energy input from a warm maritime sea. Since the tropical cyclone contributes strong differentials in temperature, moisture, and winds to the merged extratropical cyclone, the merged extratropical cyclone has an opportunity to stengthen its potential for storm activity and storm damage over land to levels subsantially greater than if the extratropical cyclone had not encountered and merged with the tropical cyclone.
Re: D. Patterson (#61),
Thank you for the reply. Some of the differences seem teminological, some seem physical. I guess that you have replied with such a complicated mechanism that I can ask a simple question. Is the additional energy that could be contributed by an SST raised by 1 deg C adequate to either nucleate more storms or to sustain them for greater distances? Are the heat transfer equations of the right magnitude?
Re: Geoff Sherrington (#92),
Geoff, I have to chuckle when you say “a simple question.”. You are wading into a topic so complex not even the experts can agree on some basic aspects of the interactions. The description I used was one very oversimplified example used only for the purpose of illustrating the basic idea that mesoscale, intermediate, and synoptic scale cyclones undergo fundamental transitions between sources and types of energy. The transitioning of a tropical cyclone into an extratropical cyclone is the basic reason why the transitioned cyclone can no longer require maritime heat and humidity as an energy source while tracking over a continent. The actual details of these transitions and properties of the cyclones are as varied and as complex as the myriad of factors which do and do not come into play in each instance. Like automotive collisions, the consequences of encounters between the tropical depression, extratropical depression, and fronts vary greatly according to the factors which come into play in each instance.
During cyclogenesis of a tropical cyclone, an additional 1C in SST may or may not affect the development of the tropical cyclone, because there are many other factors which can offset the increased SST. Wind shear can inhibit or abort the development of a tropical cyclone and result in premature cyclolysis despite increased SST. For a taste of the complexities see for example:
Pasquero and Emanuel, Tropical cyclones and transient upper ocean warming. J. Climate, submitted.
Click to access pasquero_emanuel_JC07.pdf
Note, the paper uses modeling which has been the subject of controversy in this forum and others.
For an Australian view of tropical cyclone activity and transitioning into extratropical cyclones see for example:
J.B. Courtney and B.J. Santos; The South Pacific and southeast Indian Ocean tropical cyclone season 2005-06; Regional Office, Bureau of Meteorology, Perth, Australia, (Manuscript received October 2007)
Click to access courtney.pdf
It should also be noted how baroclinic instabilities can energize subtropical and polar cyclones with enough strength to rival the barotropic energized tropical cyclones. During the Second World War, the wartime convoys in the North Atlantic were repeatedly battered by a number of extraordinarily powerful extratropical cyclones. So it should be recognized that tropical cyclones are not the only type of cyclone which is substantially involved in the transfer of the planet’s thermal energy.
huh? snip
snip – don’t rise to this.
Quick question for Judith or anyone else familiar with the paper – should 1963 in Figure S2 be a CPW year? I’ve only scanned the paper and may not be grasping the seasonal classification method.
Here’s S2 –
It looks to me like 1963’s Nino 4 is more than +1 SD while Nino 3 SD is lower than Nino 4 SD. That, I think, is the definition of a CPW year. Thanks.
Is this a new type of El Nino, or newly discovered? I doubt that there is anything happening now that hasn’t happened before during the Holocene. I hope this study isn’t shown to be ‘statistically erroneous’ like Holland and Webster (2007).
Here’s my quick-scan synopsis of the paper (which means that I may well miss or misunderstand some things at this point). It presents some interesting ideas. I look forward to the authors, and others, exploring those ideas further. Even if there are weak spots and some overreaching, the overall paper is good.
El Nino, of course, is the occasional anomalous warming of the Pacific Ocean along the equator. This warming carries extra air to the top of the troposphere (via thunderstorms) which has to go somewhere. This extra air tends to blow eastward, across the Atlantic, in the form of stronger winds.
These stronger winds can “blow the tops off” of Atlantic hurricanes and storm seedlings, weakening them.
The traditional view is that there is only one type of El Nino which has one type of weakening effect on Atlantic hurricanes. Kim et al notes that there may be two types of El Nino, which differ from each other in the portions of the tropical Pacific which are anomalously warm.
In an “EPW” El Nino the eastern Pacific has the greater anomalous warming. In a “CPW” El Nino the central Pacific has the greater warming.
The idea is that location matters, as it affects where the extra air (stronger wind) will blow across the Atlantic. The location and strength of the stronger wind may affect both how many and where storms form in the Atlantic, and maybe other factors.
Kim presents evidence that a CPW El Nino suppresses early-season activity but does not suppress hurricanes in the worst part of the season (August thru October). Also, a CPW El Nino may affect the paths of the storms, tending to move more hurricanes towards the US.
An EPW El Nino, on the other hand, tends to suppress activity across most of the Atlantic, especially during the worst part of the season. An EPW El Nino is our friend.
Kim et al presents track anomaly and wind shear anomaly evidence to support the hypothesis.
Kim also offers evidence that a CPW El Nino may be more-easily predictable months in advance than an EPW El Nino, which would be a nice social benefit.
I think the paper is a plausible and well-packaged set of thoughts and that any controversy will be minimal. Any quibbles (and I may well have some 🙂 ) will probably center around whether the historical SST data (which has warts) sufficently support the headline suggestion of “…Shifting Patterns of Pacific Warming…” and whether the five identified CPW events are really sufficient in number to draw many conclusions.
Re: David Smith (#30),
“It” in the first paragraph refers to Kim et al, not to my comments 🙂 P
Poor wording becomes obvious about a picosecond after hitting “Submit Comment”.
Re: David Smith (#30),
David, I found the paper and SI well presented and easy to understand. Since I am a layperson interested in learning more about TC developments, I do appreciate papers of this nature.
The bootstrap technique to show 90 and 95% CLs for track density appears to be the essence of statistically separating the EPW, CPW and EPC annual event classifications. Interesting that the remainder of the results are not shown as statistically significant differences.
The distributions and numbers for TC counts appear very similar for CPW and EPC with the exception of the month July. The TC counts are are visually different between CPW/EPC and EPW and differently distributed. CPW and climatology have the same ACE indexes while EPW is significantly lower and EPC significantly higher.
The S2 figure in the SI shows EPW and EPC events rather unformily distributed over the 1950-2006 period and CPW events occurring mainly in the more recent years. It will be a simple calculation to determine the probability of obtaining that distribution. I counted 31 years in this time period as neutral.
I guess the implication from this paper is that the newer classifcation of annual events, i.e the CPW could be caused by AGW and result in a new regime for hurricane numbers, ACE and track locations. I would wonder about the linkage of AGW to CPW to increased Cat45% and whether the CPW years have increased percentages of Cat45 hurricanes.
Re: Kenneth Fritsch (#35), leading question: could you explain in your own words how physically CPW and EPW events affect Atlantic hurricane tracks, frequency, and duration (ACE)? The paper mentions vertical shear and compares the monthly means of 5-events out of 58 years using reanalysis data. Intuitively, is this compositing technique telling you anything…?
Re: Ryan Maue (#36),
The short answer off the top of my head is that I could not explain in any detail physically how CPW and EPW events (differently) affect tracks, frequency and duration of TCs. I am here to learn and if that requires me to display my ignorance so be it.
I am aware of earlier studies of the effects of vertical wind shear on TC counts and the good correlations with those counts. I would suppose that duration should also be affected by wind shear, but off the top of my head I do not remember the differing effects it might have on lower versus higher intensity TCs. The location of the genesis of the TCs obviously would effect how long the TC, once formed, can build over warm water and very roughly where it might end up.
The wind shear anomalies shown in Kim (2009) for EPW and CPW in the genesis area appear to this layperson as being much more alike than both are different from EPC.
The 5 years out of 58 would make the CPW event a relatively rare event and difficult to do statistical studies on. I need to think on this part some more. Actually, since most years are neutral, we have only 9 EPW and 12 EPC years so 5 CPW years in that light is less rare, but makes the amount of data for all these events rather limited.
The Geopotential height and SST graphs in the paper, like the vertical shear comparisons, and make CPW look like EPW lite. Also the ACE index implies that CPW is EPW lite.
Here is the Nino 4, 3.4 and 3 anomalies since 1950.
Nino 4 vs 3.4
Nino 4 vs 3
Nino 4 alone
Nino 3.4 back to 1871.
Here is nice animation of clouds and winds over 1 full year.
https://www.ucar.edu/publications/nsf_review/animations/ccm3.512×256.mpg
Re: Bill Illis (#33), Nice plots (BTW, is that how 3.4 is generated from the SST, detrending?) Where did you get the data?
Also, I’m pretty sure that the movie is climate model output-though I’m perfectly okay with being corrected on this- since the label looks like a familiar name…
Re: Andrew (#44), I raise my hand and clamor for an answer as well.
Here is the SST data used: I already averaged it for ASO and put it in a netCDF file, 1950-2008. This is the ERSSTv2 data.
From the paper: Figure Supplement 3 (click to zoom):
Just a sanity check: these are raw anomalies calculated from some long-term average (most likely 1950-2006)?
Welcome back Dr. Curry you’ve been missed here.
Dr. Curry,
Your participation is most welcome.
The 1994/95 and 1997/98 El Nino events appear quite well the JPL Sea Surface Height Animation I posted on YouTube:
The east to west INITIAL formation of the 1994/95 El Nino event is easily visible. The warm anomaly forms in NINO3, then shifts west toward NINO4, before travelling east again. Note the secondary “normal” west to east propagating wave that starts in December 2004. Then there’s the west to east unleashing of the remarkable 1997/98 El Nino. Keep in mind that it’s SSH, not SST, so there’s subsurface anomalies coming into play.
I hadn’t noticed until now, but that YouTube video is the shorter version. The longer one as a 19MB mpeg is available thru JPL here:
http://sealevel.jpl.nasa.gov/gallery/tiffs/videos/tpj1global.mpeg
It also captures the east to west initial formation of the 2002/03 El Nino.
Too bad JPL hasn’t updated those videos.
Regards
David, re 1963:
The Nino3 in 1963 is larger than its standard deviation.
EPW, CPW and EPC definition is as (our paper):
“EPW, CPW, and Eastern Pacific cooling (EPC) events are defined by: Nino 3 (see Fig. S1 in supporting online material: SOM) warming greater than one standard deviation (SD) for EPW; Nino 3 or Nino 3.4 cooler than one SD for EPC; and, for CPW, Nino 4 warming exceeding one SD, while Nino 3 stays below this range. ”
p.s. thanks to hurd, geronimo, michel for supportive comments, and thanks to david for his thoughful analysis
Steve or ryanm: My earlier comment (~12:30PM) was picked up by the spam filter. Please check. If it’s gone, I kept a copy.
Thanks
Thanks, Judith, for the explanation – I missed the 1 SD cutoff.
That raises, though, the sensitivity question. The 1963 season slightly misses being a CPW while 1969 is counted as a CPW by a hair. I suspect that a shift of about 0.1C in the SST estimates could flip those classifications. SST estimate errors of 0.1C are believable to me, given the remoteness of the regions and the lack of IR satellite.
I wonder how sensitive the results are to the exclusion of 1963 and the inclusion of 1969. Normally this would not be a problem, but here the number of CPWs are so small (5) that a shift in that group of of one or two could matter.
Re #36 Ryan, I think I know where you and Ken are headed on wind shear so I’ll leave that alone. There’s one thing on shear analysis I do want to mention, though. My observation is that the shear from tropical upper-tropospheric lows are just as harmful to seedlings and storms as is the shear from straight-line upper winds. But, it seems like the standard 850-200mb zonal shear analysis would tend to underestimate the shear caused by a spinning upper low. Is my thinking offbase?
This is probably silly, but based simply on looking at RyanM’s very nice pictures, I would classify 1963, 1665, 1969, 1987, 1991, and 1994 as being centrally located; and 1951, 1957, 1987, 1991, 1994, 2002, and 2004 as being Easterly located. The differences between, say, 1965 and 2002 appear to me to be more the degree of warming than the location. With my classification, the different types of El Ninos could just as easily be occurring in groups of about three instead of changing form one form to another.
I’ve charted up the Nino 4 and Nino 3 region anomalies versus the number of US landfalling hurricanes each year (versus the average of 1.6 per year and plotted the change from average centred on September for the seasonal element) and …
I don’t see the correlation proposed in this paper (except that maybe there might be fewer landfalling hurricanes during El Nino years and more during La Nina years although that is not really clear either).
Nino 4 versus change from average landfalling hurricanes.
Nino 3 and 4 versus average landfalling hurricanes.
snip by author
If there is a difference in landfalling hurricanes between the El Ninos they classify as EPW versus CPW, wouldn’t it be just as likely be due to the strength of the El Ninos? Looking at the graphic, it appears to me that the CPW El Ninos are generally weaker than the EPW El Ninos. In fact, I’d say the CPW’s are the weakest five, with perhaps the 1963 EPW and the 2002 CPW being on the boarder line.
borderline, that is.
Re: MJW (#50), here is a mouseover/animation that may give you a better visual indicator of how these “events” are defined:
I’ve included a larger domain to give a better indication of what else is going on, instead of simply focusing on the tropical Pacific. The anomalies are calculated relative to the 1950-2008 mean.
Re: ryanm (#51), Thanks for the very interesting illustration. I think I’ll need to read Curry’s paper on Monday when she puts it on her website to fully understand the distinction between EPWs and CPWs.
Looking at the pictures, I see two different types of El Ninos, but, as I mentioned above, it doesn’t match her categorization. I differentiate between those whose warmest area is adjacent to South America and those whose warmest area is closer to the middle of the Pacific. For example, in 1963, which is called an EPW, the hotspot is far from the South American coast. I think its configuration looks similar to the 2002 CPW.
Even if the paper has reasonable definitions for EPWs and CPWs, I still may wonder whether one could come up with dozens of reasonable categorizations, some of which will, given the limited sample of El Ninos, appear to increase with time.
My two (er, make that three) cents are that CPW events
1 are arguably weaker than EPW events,
2. are farther away from the Atlantic than EPW events, so that any Atlantic effects are probably muted
3. dump their thunderstorm exhaust into a different region of the tropical upper troposphere than do EPW events, and their exhaust probably flows via different paths than does the EPW exhaust
(The arguable aspect of strength is related to the Central Pacific being warmer than the Eastern Pacific, which affects water vaporization and probably stability.)
Given the above, I think that Ken’s description of CPW events as probably having an El Nino Lite impact on the Atlantic is a good one.
Regarding storm tracks, whether the reported effects of the five CPW events are indeed representative of longer-term CPW reality is an open question to me. When there are only five seasons involved then single seasons can have a large impact on the composite. Seasons are affected by atmospheric factors other than just Pacific SST, such as AMO, dust, anomalous 500mb behavior unrelated to ENSO and so forth. And there is randomness.
I might be more comfortable with the composite results if the five seasons were close matches, but they aren’t. The reported CPW storm counts are 17,8,7,12 and 14 and their ACE values are 158, 34, 32, 65 and 225, respectively, which is a lot of variety.
The idea of a location spectrum of El Ninos, with CPE on one end and CPW on the other, is plausible to me. The ideas that different locations tend to have different impacts on the Atlantic, with father away being El Nino Lite, are also plausible and unsurprising to me. I like the idea that the topic of El Nino location and hurricanes has been raised for discussion. I’m reluctant to accept the idea that the presented composites of the 1969, 1881, 1994, 2002 and 2004 seasons necessarily reflect how CPW affect hurricanes.
In all of this my deeper curiosity is focused on Figure S2, which shows the Nino 4 and Nino 3 time series, and the question of whether it can be said that CPWs have indeed been increasing. I’ve been trying to replicate S2 but without success so far and that is a key to exploring the matter.
Hi All,
I just spent 2 hours writing a response to some of the comments and making some points on what I think of the paper. Alas, it disappeared. I will try and reconstruct later. That will teach me to answer online!
Sorry
PW
I am in the process of bootstrapping the CPW, EPW and EPC events to determine how probable are the distributions, but I would gladly defer to a “real” statistician doing the calculations and reporting the results. From my initial looks at resampling even the CPW events do not look all that improbable.
snip by author
Here is some insight into the nuts and bolts of methodology:
* To replicate the Figure S2, please check if you detrended the timeseries and normalized each timeseries.
Re: #56
*The smoothing is using 9 boxes, but still has 50% weighting on the center.
*Figure 2 doesn’t explain anything about the category/intensity of TCs at all. This paper does not focus on the intensity of hurricanes.
Re: Judith Curry (#57),
I hope that does not make this paper’s results passé.
PS: Michel, please allow for my (weak) attempts at humor even if you feel the discussion should remain serious at all times.
Ken Fritsch (unaffiliated)
Re #57 Thanks. Perhaps my problem is in the choice of detrending methods. What method was used on S2?
I did a resampling of the CPW occurrences in the Kim (2009) paper where 5 events occurred in 57 years with 4 of them occurring with a 16 year time frame. I resampled 1000 times and determine the portion of times we could expect to see 3 and 4 events occur within a 16 year time period.
The portions of 0.191 (4) and 0.772 (3) would appear not to make the grouping of CPW events a very rare event – and even if one could make a case for the CPW events occurring in the latter versus the earlier part of the time period studied. If one were to consider the criteria for selection of a CPW event on the arbitrary side and subject to further sensitivity testing then the probabilities above would have to be adjusted upward. The R code is listed below.
x=c(rep(0,52),rep(1,5))
resx=sapply(1:1000, function(i)
sample(x))
z=rep(42000,0)
for(j in 1:1000){
for(i in 1:42){
z[i+(j-1)*42]=sum((resx[,j])[i:c(i+15)])
}
}
y=matrix(data=z,nrow=42,ncol=1000)
zmax=rep(1000,0)
for(i in 1:1000){
zmax[i]=max(y[,i])
}
znum=subset(zmax,zmax>3)
length(znum)/1000
[1] 0.191
znum3= subset(zmax,zmax>2)
length(znum3)/1000
[1] 0.772
Re #58;
Detrend: we removed the best straight-line fit linear trend from the data.
trend is calculated using linear regression (least squares approach)
The actual time span for the 4 events is 14 years and not 16 as I noted in the post above. The portions for the same test using 14 years for 4 and 3 events, respectively, are 0.115 and 0.668. The 4 events are rarer, but not all that rare by normal statistical standards.
I will call the former calculation a sensitivity test.
I’ll do a resampling of the EPW event distribution and fall off in recent years next. I suspect that the EPW event fall off might have statistical significance.
Looking closer at the distribution of the three event classifications, I am beginning to wonder why the EPW to CPW change was not considered a modification of EPW events and foregoing the introduction of a new classification (CPW).
What I thought might be a significant fall off of EPW events was not borne out by my resampling, where I determined what portion of the resamples would have only a single event in a 19 year stretch (as the last EPW event in the Kim graph did) when the resampling places randomly the 9 EPW events over 57 years. The portion was 0.72.
Of course, we can obtain a better look at the fall off by including 2007 and 2008 and soon to be on record the 2009 season.
Re #62 Judith, thanks for the help in this basic exercise.
I’ve found part of the problem. I’ve been including 2007 and 2008 SST data (1950-2008) while the paper stops with 2006 data (1950-2006).
The 1960s are still a problem for me, though. Among other things, my 1969 Nino 4 is considerably weaker than the value shown in the paper.
I think I’ll restart from scratch. To confirm, I’m using Smith-Reynolds Extended Reconstruction SST data, which I believe is what the paper used.
snip by author
FYI, Damage implications:
All data from http://www.icatdamageestimator.com Updated to 2009 values using Pielke et al. 2008. Following Kim et al. designation of ENSO types as follows (1950-2006):
EPW: 1951, ’57, ’63, ’65, ’65, ’72, ’76, ’82, ’87, ’97
CPW: 1969, ’91, ’94, ’02, ’04
EPC: 1954, ’55, ’64, ’70, ’71, ’73, ’75, ’85, ’88, ’95, ’98, ’99
EPW
Average damage: $3.2 billion, std. dev. = $6.3B
Median damage: $0.4B
CPW
Average damage: $5.1 billion, std. dev. = $7.9B
Median damage: $0.8B
EPC
Average damage: $3.3 billion, std. dev. = $5.5B
Median damage: $0.8B
All Years 1950-2006
Average damage: $4.3 billion, std. dev. = $10.8B
Median damage: $0.5B
All Warm Event Years
Average damage: $3.8 billion, std. dev. = $6.4B
Median damage: $0.6B
All Non-El Nino Years
Average damage: $5.0 billion, std. dev. = $14.5B
Median damage: $0.4B
NOTE: CPW average drops to $3.0B without 2004
Bottom line = suggestive that the warm flavors make a difference on damages but more years needed. Also, the analysis based on Kim et al. seems somewhat different than Pielke and Landsea 1999, due to using a different definition of warm and cold events as well as a different time span in the analysis. Perhaps the relationship is non-stationary over time?
For further reading:
Pielke, Jr., R.A., and C.W. Landsea, 1999: La Niña, El Niño, and Atlantic Hurricane Damages in the United States. Bulletin of the American Meteorological Society, 80, 10, 2027-2033. (PDF)
http://rogerpielkejr.blogspot.com/2009/07/hurricane-damage-by-enso-flavor.html
Re: Roger Pielke, Jr. (#67),
Since EPW (earlier) and CPW (later) events occurred at somewhat different times would not the damages have to be normalized for CPI and populations changes?. Of course, those large stdevs (if the distributions are normal) would not allow determining a significant statistical difference with these small number of events.
Please 1965 must have been a very good year year.
Re: Roger Pielke, Jr. (#67),
You need to correct your 1965 year duplication for EPW.
Progress – here’s a plot of the Nino 3 values as shown in Kim Figure S2 and as calculated by me, based on Judith’s guidance
The Nino 3 values are a near-perfect match.
But, I’m still stuck on my original problem, which was, and is, replication of Kim et al Nino 4 values.
Kennet (68), Ff course, please see the site and Pielke et al. 2008 which normalizes the data to a common base year.
Looking again at Kim et al Figure S2 (see Post #29 above) and looking at the bars (Nino 4) and dots (Nino 3) as separate events, we see the Nino 4 (effecting CPW events) anomalies as occurring in a rather uniformly cyclical and predictable manner. In that regard we see nothing of significance changing over time for Nino 4.
It is only when the additional restriction is put on the criteria for a CPW event of Nino 3 anomaly being less than 1 stdev while the Nino 4 is more than 1 stdev that we see what the graph shows as CPW increasing in recent years.
That restriction would have to affect the statistics used to determine the probability of this distribution of CPW events – as would the choice of 1 stdev for a criterion.
The zero line of the Figure S2 graph would appear to be more centered on the Nino 4 events than the Nino 3 events. That could be due to my lack of a calibrated eyeball, but I would like to see the criteria applied using a zero base line for the Nino 3 and Nino 4 events that were separated and not combined as I assume is the case in the graph.
I also want to look at the statistical significance of some of the CPW, EPW, EPC related events like ACE, total TC counts and Cat45 hurricanes. The data that Pielke Jr shows for damage related to these categories obviously does not show significant differences.
Here is my replication effort for Nino 4:
The methodology is the same one as used in the successful replication of Nino 3, but the match is much poorer.
Of particular note is 1969, which is one of the listed CPW years and had the most CPW storms (17), so its characteristics probably carry considerable weight in the analysis.
The year 1991 also falls slightly below the 1 SD criteria for a CPW year, which is minor except that the precise definition was used to exclude a very-close 1963 from being a CPW year, so it matters.
I’ll continue to look for the source of the mismatch. I’d like to resolve it so that I can explore the impact on the CPW population of slight changes in the definition of CPW, changes in the period examined (inclusion/exclusion of 2007 and 2008, for example) and changes in SST set used.
My concerns are that the CPW population is small and that the definition of CPW is somewhat arbitrary, so that seemingly minor changes in the definition of CPW, period examined, SST data, etc may substantially shift which years are classified as CPW and thus shift the reported impact on Atlantic hurricanes.
Again, I like the Kim et al concept that different “flavors” of El Nino can have different impacts on Atlantic storms. I’m concerned, though, that the available data is too limited to support any conclusions or suggestions about the expected impacts.
snip by author
snip by author
I compiled the NATL ACE values for 1950-2008 for the 4 classifications of Neutral, EPW, CPW, EPC and than compared EPW to CPW and EPW to EPC for significant difference in the mean ACE values. The t value for the first comparison above was 0.45 and for the second comparison t was 1.09. Neither of these t values rises to near the level of significance.
Neutral EPW CPW EPC Climatology
ave= 100.8 70.6 102.8 125.5 101.4
stdev= 58.5 40.2 84.9 63.5 60.3
count n= 33 9 5 12 59
It would appear that it will be difficult to show statistical significance for differences in the manifestations of these event classifications outside the Kim et al claim for track densities.
Re: Kenneth Fritsch (#76),
Correction for the above post:
The p value for CPW and EPW being different indicates that the difference calculated could be 46% due to chance. The p value for the difference between EPW and EPC being different indicates that the difference calculated could be 2.6 % by chance. Therefore the results indicate that the EPW/CPW differences are not significantly different but that the EPW/EPC differences are.
These results are more in line with CPW as EPW (or EPC) lite.
Looking at CPW with the other event designations from Kim et al shows its TC manifestations as very neutral and in line with climatology. The tabled results are for average annual counts/index values for the period 1950-2008 by event designations.
Re: Kenneth Fritsch (#78),
Kenneth, just from looking at your numbers, I see no discernable difference between neutral, cpw (el nino lite), and la nina. Is that the direction you are heading?
Re: Jonathan Schafer (#79),
Jonathan, that is the direction the data are headed. I like to see what the data that are sometimes not offered in these papers show – and at the same learn something about the science of TCs.
CPW, I judge from the data (ACE, hurricanes and Cat345), is closer overall to neutral and climatology than it is to EPC (la nino), but nothing statistically significant.
Here is some interesting news back in 2004:
Bill Gray’s hurricane forecast issued on April 2, 2004:
So, the numbers were pretty good, but the intensity distribution was way off, as the long-track intense Cape Verde storms were not anticipated.
“This year did not behave like any other year we have studied.”
Re: ryanm (#80),
This comment is a pretty standard version of what a forecaster will provide for an erroneous prediction. Using empirical data for the forecast makes statements like these more appropriate.
I am wondering how climate modelers, who claim to forecast without empirical data, comment when their models provide the wrong forecasts. Maybe something like: while the ensemble average was wrong we did have model runs that covered the actual occurrence and therefore we avoid the claims that there is statistically different result between observed and modeled. We also had noted that the models predicted the actual occurrence had a 30% chance of occurring so we were not completely wrong.
From Jeff Masters blog today
I bet it plays off the Kim paper.
Still no success in duplicating the Nino 4 values.
Re: David Smith (#81), I’ve got some Kool-Aid ready.
MEI ENSO update for July 6, 2009 — a must read for the current state of tropical SST in the context of the previous 60 years.
I continue to have difficulty getting my head around Figure S2 in the Kim et al SI. We have 2 distinct regions of Nino 3 and Nino 4, yet in the S2 chart I see a common zero line and standard deviation units for selecting Nino 3 and 4 events. In my simple minded view, why would not the Nino 3 and 4 event criteria be made separately with a separate zero line for each region and separate standard deviation units.
Maybe Kim et al does use the method I note above and I do not recognize what the initial caption statement in S2 is saying by “Time series of normalized Nino 3 (circle) and Nino 4 (bar)…” What does normalize mean?
Re: Kenneth Fritsch (#84), yes, the Nino 4 region has warmed or jumped up, if I am allowed to talk about changepoints, following the record La Nina of 1975 which is the most extreme anomaly during the past 60-years in the Nino 4 time series.
Did you reconcile the fact that the first author interpolated the SST from 2×2 degrees to 2.5×2.5 to match the NCEP reanalysis grids and therefore allow the Nino 4 box (5S-5N) to be defined without partial gridpoints?
The Nino 3.4 box has not seen a trend upwards in SST during the past 60-years.
The SST anomalies do NOTHING. It is the SST itself that is important. The Nino 4 region is correctly described by an author as much warmer than the Easterly boxes, on average. Thus, a common normalization factor is not appropriate.
Ryan, I was confused about absolute anomalies and anomalies in terms of standard deviation units in addressing the S2 graph in Kim et al. Using standard deviation units separately calculated for each Nino 3 and 4 regions can be used in the same graph as Kim has apparently done in S2. I then assume that the “normalize” reference was to the calculations of the standard deviation units.
After I came to terms with my confusion on standard deviation units, I then understood your point about absolute temperature being the important factor in terms of TC physics.
What I would like to see is a plot of the absolute temperatures and/or the anomalies along with the average temperature used for the anomalies for the separate regions of Nino 3 and Nino 4. It might be of interest to see those plots broken down for each year into temperatures for August, September and October.
Bob Tisdale has an interesting article on El Nino Modoki at Anthony Watts’ site
Re: David Smith (#88),
I read that also. So many issues that are discussed on this site lead back to centering and date range selection.
Bob Tisdale created this plot of Modoki seasons stretching back to 1900
I’m particularly interested in the 1969 value, which Kim shows as a CPW season but my calculations and Bob’s show 1969 as a rather tame season.
Oops, the Modoki image didn’t make it. Here’s the link –
Re: David Smith (#91),
David, I think that Tisdale notes what I already above noted from the Kim S2 figure and that being that the Nino 4 anomalies (standard deviation units) events by themselves show nothing new or unusual in the pattern of occurrence. That is not, however, the Kim et al definition of a CPW event. A CPW event must have a Nino 4 temperature anomaly of +1 standard deviation unit or more and at the same time a Nino 3 event with a temperature anomaly less than + 1 standard deviation unit.
It is my opinion when you allow yourself these additional parameters (including an arbitrary selection of 1 standard deviation unit criteria) you could make a case for calling out multiple classifications the likes of CPW.
The grouping of the CPW events over the time period 1950-2006 cannot be shown to be significantly different than a random occurrence.
What I would like to see is a side by side comparisons of the actual Nino 3 and Nino 4 temperatures in a time series from 1950-2006 with the ASO months defined.
I also am curious why a CPC event is left out of the otherwise symetrical designations (EPW has its EPC, but CPW lacks a CPC). It would appear that by the S2 graph in Kim et al that 3 CPC events come close in 1964,1971 and 1998, but no cigar, and leaving us with CPC being a very rare event.
I am posting my attempt to replicate Kim et al. (2009) S2 graph (with units in “normalized” standard deviations) and the Nino 3 and 4 region temperatures for the 1950-2006 time period.
I got my ERSST v2 SST data from KNMI at the link below.
http://climexp.knmi.nl/daily2longer.cgi
The proper link to KNMI is below. This page has the ERSST v2 access.
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
My replication of Kim gives the following breakdown by year of designations for EPW, CPW, EPC and CPC as compared with Kim.
Rep EPW: 1951, 1957, 1963, 1965, 1972, 1976, 1982, 1987, 1997 (9)
Kim EPW: 1951, 1957, 1963, 1965, 1972, 1976, 1982, 1987, 1997 (9)
Rep CPW: 1953, 1969, 1977, 1992, 1994, 2002, 2004 (7)
Kim CPW: 1969, 1991, 1994, 2002, 2004 (5)
Rep EPC: 1954, 1955, 1964, 1970, 1971, 1973, 1975, 1988, 1995 (9)
Kim EPC: 1954, 1955, 1964, 1970, 1971, 1973, 1975, 1985, 1988, 1995, 1998, 1999 (12)
Rep CPC: 1998 (1)
Kim CPC: (0)
1964 is marginal between EPC and CPC. 1963 is marginal between EPW and CPW. Both called for EPC and EPW.
Correlation for Nino 3 versus Nino 4 (detrended ) can be seen in the graph presented below. I think this strong correlation calls into question using separate event designations for Nino 3 and Nino 4 temperature anomalies.
Bender, i didn’t do a public forecast this year (did one last year in april, pretty much hit it as did everyone else). had I been making a forecast this past april, i would have went low
I want to present some results for the designations EPW, EPC, CPW, CPC and Neutral as used ostensibly in KIM et al (2009) that came from the temperature series for ERSSTV2 as extracted by Kim, David Smith and Kenneth Fritsch (evidently from three different sources) and HadSST1, ERSSTv3b2 and COADS SST. The last 3 temperature series and the ERSSTV2 by KF were all extracted by me using the KNMI web site. The differences appear at the margins for designations, but do make a difference in the mix of designations. I’ll call it a sensitivity test for Kim et al’s results until I can determine what causes these temperature series differences.
Re #100 Somehow the word non-robust jumps to mind.
MY SST source is here . I assume that it provides an earlier version than does KNMI.
With this paper coming out, pretty much ends the debate on the hurricane PDI/Power/ACE/integrated energy being anything but a function of ENSO, as I showed in my Maue (2009) GRL paper.
Influence of the Southern Oscillation on tropospheric temperature
McClean et al. (2009) JGR
Re: Ryan Maue (#102),
I suspect that there will be a number of scientists who dispute that this paper “ends the debate.” 😉
Re: Mike B (#103), probably should get a press release ready. I think it would be a good writing exercise to put together possible press releases and issue them preemptively to combat the agenda-driven, partisan skeptics.
Re: Ryan Maue (#102),
I was going to suggest an audit of the paper, but then I noticed the following at the very end:
😉
Re: RomanM (#105), this paper is a must-read. Should have been in Science or Nature. The data is easily downloadable, the methodology well explained, and the conclusions fundamental. Unfortunately, when you do a Google News search for El Nino or Southern Oscillation, all you come up with is Kim et al. (2009), which is a direct result of the very aggressive sales pitching by the Science journal.
Re: Ryan Maue (#102)
Re: RomanM (#107)
It is quite evident that a thorough audit of the McClean et al. (2009) paper — done in the same way as the audit CA-et-al did of the Steig Antarctica paper — would be most useful in constructively critiquing the McClean paper’s conclusions and in establishing the paper’s overall defensibility, or possible lack thereof, as the case may be. Lucia has started the auditing process over on her own blog with critical comments that echo some of the ones we see here.
My comment was meant in a humourous vein.
I haven’t read the paper thoroughly, but I must admit that the first sentence in the Analysis and Results section gave me some pause: “We start by considering 12-month running averages of our data.” There seems to be widespread use of correlations later in the paper and I think that prior smoothing could make interpretation of results considerably more difficult.
Before you get too excited, in the paper they tell us: “To remove the noise, the absolute values were replaced with derivative values based on variations. Here the derivative is the 12-month running average subtracted from the same average for data 12 months later.”
The effect of this is to remove any linear trend. We know CO2 has been increasing approximately exponentially over the period of their data, which, according to the theory of climate sensitivity to greenhouse gas increases, should result in a linear increase in temperatures.
By eliminating any linear trends from the time series, all they’ve done is emphasise the effect of ENSO, already known to be the major influence on short-term variation. Their conclusion that “The findings presented here are consistent with the Southern Oscillation being a major driver of temperature anomalies,” is certainly not supported by their analysis.
They lucked out on scoring referees who’d dozed off during the first week of Calculus 101 🙂
Re: Philip Machanick (#108).
Careful. Taking the derivative doesn’t exactly “remove” a linear trend. Rather it transforms a linear function into a constant function, whose constant value shows the strength of the trend.
OTOH, unless I missed something, their study can only justify the conclusion that the derivative of the Southern Oscillation is a “dominant and consistent influence” on the derivative of mean global temperature.
Re: Philip Machanick (#108), exactly why it should have been in Nature or Science — no need to worry about those pesky assumptions. That’s what the press release is for.
(C H E [tongue] E K)
After my first read of the McClean paper and from a layperson’s point of view, I thought the paper was off to a good start with Figure 1 showing the 12 month MA for SOI and GTTA.
Unfortunately, from there I think the authors went on an ill-advised over fitting expedition in order to make their models look better and explain more of the variance with the R^2 inflation. They played the volcano effects and lag and the temperature lags with little or no a prior reasoning for their actions.
I am guessing here and I need to read the paper again, but when one puts a model together like in this case and without regard to the dangers of over fitting and one gets a good fit it is difficult for the over fitter to understand how easy it can be to manipulate data to show something that might well not hold up out-of-sample.
Again as a layperson with one read of the paper I would have thought the authors would have better proceeded by simply explaining with a principled argument the data presented in Figure 1.