## Atlantic Hurricane Season 2008 Withers on the Vine

The North Atlantic hurricane season has nearly come to an end. As November progresses, the chance of another storm developing becomes smaller. Climatology (last 60 years) tells us that roughly 4 in 10 years see a November storm formation including 4 in 2005 (Beta, Gamma, Delta, Epsilon), Hurricane Michelle (2001), Hurricane Lenny (1999), and Hurricane Kate (1985). Jeff Masters from the Weather Underground has an image of previous early-November storm tracks especially clustered in the Western Caribbean.

So, what has the 2008 season wrought in the North Atlantic and how well did the seasonal prognosticators fare?

Even with the expected post-season tinkering of the real-time storm tracks by the folks at the National Hurricane Center, we can provide fairly accurate preliminary numbers. The community at Wikipedia constantly updates many interesting facts about the ongoing 2008 hurricane season.

Total Named storms (34 knots + one-minute maximum sustained winds): 15
Total Hurricanes (64 knots +): 7
Total Major Hurricanes (96 knots +): 4
Accumulated Cyclone Energy: 132

The respective forecasts made by CSU (Klotzbach and Gray), NOAA, as well as the UK Met Office came in quite close to the actual experienced storm activity. Before handing out trophies, please keep in mind that forecast “skill” is a function of many forecasts over longer time periods. Each of the forecasting outfits prefers to use different techniques and variables to calculate their storm numbers, so we will have to wait until each completes their post-season analysis to determine if they were “right for the right reason” or got lucky.

Now, to answer the question: how active was the 2008 hurricane season, we need to define climatology. This is where the tricksters can play pranks on the public. Where is the beginning point of the analysis? How well do we trust the frequency and the estimated intensities of each storm? What metric do we use – number of tropical storms, number of hurricanes, ACE (accumulated cyclone energy), Power Dissipation, or perhaps some complicated statistical measure? All of these questions are entangled in the debate surrounding whether anthropogenic climate change is indeed a modulating influence upon current and future Atlantic hurricane activity.

A well-accepted metric which convolves storm frequency, intensity, and duration is called accumulate cyclone energy (ACE) and is calculated very simply: take the maximum sustained winds reported by the NHC every 6-hours for all storms (> 34 knots), square this value, and sum over the entire lifetime, then divide by 10,000. In 2007, even though there were also 15 storms, the ACE was only 72 compared to 132 for 2008 with the same number of named storms. This is partially because the storms in 2008 were much longer lived especially Bertha.

Here are three different views of the Atlantic hurricane climatology depending upon what period you look at. The data is from the NHC Best Tracks without any corrections to the intensity data.

Links to two other time periods:

1978-2008

1944-2008

Thus, since 1995, Atlantic hurricane activity measured by ACE is hugely variable with feast (i.e. 2005) and famine (1997). 2008 ACE is nearly equivalent to 2006 and 2007 combined, but about half as what was experienced in the record 2005 season. The choice of 30-years is a particular favorite for many researchers in the tropical cyclone community (1978-2007). The second image clearly shows the nearly stepwise increase in ACE between 1994 and 1995. In this reference frame, 2008 ranks as one of the more active years of the past 30. Now, back up to 1944, when admittedly the intensity (and detection) data is somewhat less reliable. However, since the ACE metric is the convolution of an entire year’s worth of storm lifecycle information, and is most sensitive to higher wind speeds, the track data points prior to satellite observation (~1970s) are probably sufficient for this exercise.

Final verdict: When encapsulated in the recent active period in North Atlantic activity (1995-2007), 2008 experienced normal or expected activity as measured by ACE. In terms of a long-term climatology, either the last 30 or 65 years, 2008 is clearly an above average year.

Note: for the Climate Audit seasonal forecasters, especially those that showed exemplary skill (however you wish to measure it), please fill us in on your methodology and perhaps provide guidance for 2009. Or, for those feeling shame about being “blown off track”, time to think of good excuses.

Also, a new Science perspective has been published by Vecchi et al. (2008) entitled Whither Hurricane Activity? More on that later…

1. Posted Nov 1, 2008 at 12:46 AM | Permalink

Yes, but Paul Linsay and I win, as we predicted $6 \pm 3$ named hurricanes. What’s our prize?

2. TerryS
Posted Nov 1, 2008 at 3:53 AM | Permalink

John A:
November 1st, 2008 at 12:46 am

Yes, but Paul Linsay and I win, as we predicted 6 \pm 3 named hurricanes. What’s our prize?

You get to pick the first and last letter of the first named Atlantic storm next season. You can pick any letter you want as long as it comes before ‘B’ in the alphabet.

3. Sekerob
Posted Nov 1, 2008 at 5:20 AM | Permalink

Steve, you might want to read up on Sahara sand/dust sitting in the equatorial belt over the Atlantic. The more of it the less chance of hurricanes. Not much to do but what the winds blew up there in the first place. How does that fit in with your theories?

4. Sekerob
Posted Nov 1, 2008 at 5:23 AM | Permalink

Oh, sorry, I see a new author (to me) ryanm… I don’t look too much for names as CA = SMcI synonymous. Is that a bad sign :D

5. Pierre Gosselin
Posted Nov 1, 2008 at 6:44 AM | Permalink

In the 1944 – 2008 chart, just eyeballing, ACE seems to correlate with PDO.
ACE appears to trend downward during the cool PDO phase, and upward during the warm phase.

Just coincidental?
We’ve entered another cool PDO phase, and so one could expect a downward ACE trend over the next 20 years.

6. Pierre Gosselin
Posted Nov 1, 2008 at 6:49 AM | Permalink

John A
Terry S didn’t specify which alphabet. So next year’s first may not necessarily have to start with an A.

7. Sekerob
Posted Nov 1, 2008 at 7:04 AM | Permalink

It is more stronger related to ENSO and lesser with PDO, so say current publications. More El Nino, less hurricanes.

8. Pierre Gosselin
Posted Nov 1, 2008 at 9:08 AM | Permalink

Sekerob,
I’m getting confused.
A cool PDO would mean more La Ninas, meaning more hurricanes.
But we are getting the opposite! During the cool PDO, ACE trends downward.
Perhaps I’m eyeballing wrong, or I’m totally misunderstanding something.

9. Posted Nov 1, 2008 at 9:51 AM | Permalink

Here are a few plots on ACE, for background.

First is the distribution of Atlantic ACE since 1979. Many storms fall into the little-impact category.

Here’s how that distribution shifted (1980-94 versus 1995-2007). I shifted to a line plot and smoothed the values, for ease of viewing:

It shows some reshaping of the curve, with a lower fraction of weak storms and higher strong storms. Of course, the annual count of storms is higher in 1995-2007, which also contributes to the increase in ACE.

Finally, here’s the contribution to total ACE from each category:

The top 25% of storms contribute about 75% of the total ACE.

10. Pierre Gosselin
Posted Nov 1, 2008 at 11:27 AM | Permalink

WCR website, sometimes quiet there, has recently posted some worthwhile reports on the subject.

http://www.worldclimatereport.com/

11. Paul Linsay
Posted Nov 1, 2008 at 11:41 AM | Permalink

TerryS

In Ethiopic that would be K. So let’s see, Katrina, hmm, better think about this some more.

12. Paul Linsay
Posted Nov 1, 2008 at 11:43 AM | Permalink

http://en.wikipedia.org/wiki/Ge%27ez_alphabet

13. Posted Nov 1, 2008 at 4:10 PM | Permalink

I’ve no idea what Ryan Maue is talking about. Named hurricanes were smack in the middle of the Poissonian distribution. ACE would presumeably follow a similar distribution, so I’ve no idea what “above average” means, since in any stochastic process there would be over-average periods as well and under-average.

There was one consistent result in all of this: the number of people killed was inversely proportional to the GDP of the country. Hurricanes will always affect the Carribean, but the ability of people to survive them is directly related to the wealth of the country.

• Kenneth Fritsch
Posted Nov 2, 2008 at 9:15 AM | Permalink

Re: John A (#15),

John A, the annual ACE index has a normal distribution over the historical records, or at least when one uses the annual ACE of TCs measured west of the latitude 60W for the NATL — as I did in a long ago post. The David Smith graphs above are not used to show that distribution, but something else.

Should that be GDP per capita?

I’ve no idea what Ryan Maue is talking about. Named hurricanes were smack in the middle of the Poissonian distribution. ACE would presumeably follow a similar distribution, so I’ve no idea what “above average” means, since in any stochastic process there would be over-average periods as well and under-average.

There was one consistent result in all of this: the number of people killed was inversely proportional to the GDP of the country. Hurricanes will always affect the Carribean, but the ability of people to survive them is directly related to the wealth of the country.

14. Steve McIntyre
Posted Nov 2, 2008 at 9:21 AM | Permalink

#15. John A, the ACE distribution could easily be non-Poisson, as it’s a sort of integral and not a count. Indeed, one would expect it not to be a Poisson distribution. I, for one, would be very slow to bet against Ryan’s interpretation of these things.

15. Posted Nov 2, 2008 at 10:45 AM | Permalink

David, I do like the distribution plots of ACE and the ratio that the stronger storms represent each season. There is a lot of information in the ACE metric and research has shown that higher-ACE seasons are often associated with a positive Atlantic Meridional Mode (AMM, i.e. Vimont and Kossin 2007). The mean genesis locations are shifted well eastward into the Atlantic with lower latitude formations. Thus, the storms last longer and churn over a much longer fetch of warmer SSTs and accumulate more ACE.

The Western Pacific ACE distribution operates in a similar manner with El Nino events seeing development further southeast closer to the International Dateline. The opposite during La Nina was largely seen this year with WPAC ACE barely above the North Atlantic’s levels. WPAC ACE was only lower (thru November 2) in 1998 and 1999 during the past 30 years.

16. Posted Nov 2, 2008 at 11:30 AM | Permalink

One of my favorite plots shows Atlantic ACE and August-October AMM (AMM is a measure of SST difference between two key regions):

As indicated, the r-squared for unsmoothed data is 0.64

As an exercise I think I’ll remove the single strongest storm from each season (sometimes there’s a monster outlier) and see what that does to the correlation.

17. Posted Nov 2, 2008 at 3:07 PM | Permalink

David #19 — let me add a wrinkle to your research. Consider how the AMM is calculated and the regressions of the index upon low-level wind as well as SST. The particulars as well as an image are on the Climate Bureau web page.

Now, one of the kinematic mechanisms associated with the AMM (or ENSO) in modulating Atlantic hurricane activity is a reduction of vertical shear in the Main Development Region. Some references in the Vimont and Kossin BAMS article from November 2007 suggest both AMM and ENSO affect the upper-level winds much more than the low-level winds. I would surmise that the relative contribution to a 200 hPa – 850 hPa vector wind shear calculation of each level is a changing quantity. So, if you have a lot of hurricane activity in a given season, how much of the storm circulation itself is contributing to each level’s respective wind?

I ask this because Kyle Swanson in a 2008 GRL article (“False Causality”) undertakes an exploration into the effects of storm circulations upon 925 hPa (low-level) winds to ascertain whether the methodology employed by Saunders and Lea (2008, Nature) as well as countless other studies is in fact robust. I attach the abstract:

Statistical studies suggest a link between anomalies in seasonally averaged lower atmospheric dynamical fields and Atlantic hurricane activity. Here we show that lower atmospheric seasonal wind anomalies result primarily from the presence of the hurricanes themselves. This is done by assuming a hypothetical vortex structure whose radial structure is constrained by observations derived from aircraft probing of tropical cyclones and whose vorticity magnitude is scaled to time varying, best track intensities. Seasonal vorticity anomalies associated with Atlantic hurricane activity are accumulated by summing these idealized vorticities along observed tropical cyclone tracks. Winds associated with these seasonal vorticity anomalies explain the bulk of observed hurricane activity-related fluctuations in the seasonally averaged lower tropospheric wind. Hence, seasonal wind anomalies appear to have little causal information relevant to understanding why hurricane activity in the Atlantic has fluctuated in the past, and may be of limited value in projecting future hurricane activity.

An open question to ponder then: are the coincident correlations between SST, vertical wind shear, low-level vorticity, etc. in the Atlantic sufficiently independent of direct hurricane influence to yield meaningful dynamical and causal insight? More plainly, if you have years with lots of storms like 1995, 1999, 2004, 2005 and compare your “favorite index” regression variables to years with few storms, are you a priori prescribing relationships which may not be necessarily there?

• Kenneth Fritsch
Posted Nov 2, 2008 at 6:50 PM | Permalink

Re: Ryan Maue (#20),

I ask this because Kyle Swanson in a 2008 GRL article (“False Causality”) undertakes an exploration into the effects of storm circulations upon 925 hPa (low-level) winds to ascertain whether the methodology employed by Saunders and Lea (2008, Nature) as well as countless other studies is in fact robust.

Ryan, my first read of this paper (Swanson) as a layperson did not leave me with much to take away. I know when I have looked at the 925 hPa zonal and 200 hPa and 850hPa vertical wind shear (as was studied and published by Saunders and Lea) a good correlation is obtained with NATL TC activity, and, furthermore, those correlations were more consistent over time than the correlation with SST. Of course, if it is the TC activity causing the wind fields than one should expect a correlation.

What I got for the Swanson paper was that they looked at the observed zonal wind at 850 hPa and attempted to determine theoretically how much it could be affected by TC activity. They did not look at how TC activity affected vertical wind shear and implied in the conclusion that they were suggesting the zonal wind field effect was sufficiently large that one needed to “cleanse” it from the climatology so that the effects and relationships of other to TC activity could be better determined.

Is Swanson proposing that TC activity can enhance the climate for more TC activity or does it inhibit TC activity also? How long lasting is the effect and what would that do for well spaced storms? Has he empirically tested his theoretically based calculations.

• Posted Nov 2, 2008 at 7:48 PM | Permalink

Re: Ryan Maue (#20), Thanks for the link and writeup, Ryan. It’s something to ponder. Kenneth took a close look at it a few months ago, as he notes above. I scanned Swanson at that time but did not read it closely enough to have an opinion.

For fun, I looked at the AMM average for June + November (the edges of a season) rather than August-September-October (the heart of a season). This is in hope of avoiding contamination from cyclone winds on the AMM calculation. June and November are poor storm months basin-wide. The adjacent months of July and October are also poor months in the region covered by AMM (east of 75W). So, I think/hope that the June + November average is relatively free of any cyclone contamination.

First, here’s the correlation between the June-November value and the August-September-October one:

(I choose the last thirty years because I have higher confidence in the satellite-aided data.)

The r-squared between the “suspect” ASO and June-November is pretty good. If there was heavy ASO contamination from cyclones then I’d expect to see a poorer correlation. It’s obviously not conclusive but my confidence rises.

Next is June-November AMM versus seasonal ACE:

This presumably contamination-free AMM matches up pretty well with ACE, even though the AMM values are at the edges of the season rather than the heart. Again, this is nothing conclusive but it reinforces my sense that AMM is not simply an artifact of storm winds.

My personal conjecture is that hyperactive seasons involve good co-location of seedling vorticity centers and the associated heavier precipitation across the MDR. This produces healthier seedlings. Some healthy seedlings develop into eastern and central-Atlantic storms while others bide their time until they reach the western basin, at which time they are better-able to take advantage of any favorable development conditions. I suspect that the AMM is somehow involved in the co-location aspect.

18. Posted Nov 2, 2008 at 7:17 PM | Permalink

DAVE SMITH
How did the regular bloggers on the 2008 hurricane tracks do with respect to correctly predicting the number of hurricanes and the number of named storms?

19. Posted Nov 2, 2008 at 7:17 PM | Permalink

#22, you are correct Kenneth that Swanson (2008) does not do the same calculation for vertical shear. However, if the lower level wind values are affected, I would expect similar contamination at upper-levels. I think the location and size of a given region, say the MDR between 6-18N and 60-20W, in many years would show considerable hurricane influence (e.g. 1995, 2004).

Is Swanson proposing that TC activity can enhance the climate for more TC activity or does it inhibit TC activity also? How long lasting is the effect and what would that do for well spaced storms? Has he empirically tested his theoretically based calculations.

He does not speculate on these issues. However, serial hurricane formation and tracks (3-4 or more) have been observed many times during the past active period. I would favor an explanation that non-local climatic modulation is responsible for this effect rather than well-spaced serial occurrences. In other words, I have no clue.

20. Posted Nov 2, 2008 at 8:09 PM | Permalink

At the risk of further drifting off-topic I’ll offer one more AMM plots and then that should be the end of it :)

This is the time series of AMM (A-S-O) and ACE (seasonal) since 1948. I removed the 9-yr average and then I averaged the remainders for ease of viewing:

What intrigues me is the saw-tooth pattern since the 1960s. If that pattern reflects reality (and isn’t somehow an illusion from my simple method or a chance pattern) then what caused the swing in AMM and hurricane activity? Is is continuing? What about prior to the mid-60s: is that flatness real or just poor-quality data? Dunno.

21. Kenneth Fritsch
Posted Nov 3, 2008 at 9:57 AM | Permalink

From a reread of Swanson’s “False Causality” paper, I found that he implies (see excerpt below) that the TC effects on the wind fields can last for weeks.

TCs are localized cyclonic vorticity anomalies several orders of magnitude larger than ambient relative vorticity levels in the tropics [Mallen et al., 2005]. Recently, Sobel and Camargo [2005] showed that in the Northwest Pacific, ECMWF reanalysis seasonal vorticity anomalies regressed against fluctuations in TC activity are significant in magnitude, and last on the order of several weeks. This suggests the possibility that although localized in both space and time, the accumulated impact of TC vorticity anomalies could be substantial in the seasonally averaged vorticity fields. While TCs are only crudely resolved in the various reanalyses [Maue and Hart, 2007], their remote impacts upon the wind field still remain, particularly in those regions strongly constrained by observations, i.e., areas with sounding coverage.

I would argue that either Swanson is onto something here- big time – or he has made some major error in his assumptions. I would also think that analyses such as David Smith has posted above could be used to determine the validity of the Swanson assumptions.

My forecasts (see below) for the 2008 NATL TC season were based on the phase of the AMM as forecast by Klotzbach/Gray in the early spring. Was the AMM favorable for TC activity in the NATL this season? Or was the TC activity favorable for a favorable AMM that in turn was favorable for TC activity?

PS: I do not take forecasting, in general, very seriously and that goes in spades for my own.

I present in this post my long awaited May NATL TC activity forecast for the 2008 season. As explained previously, it is based on named TCs and hurricanes belonging to 2 Poisson distributions depending on AMM being negative or positive. The ACE index is similarly based on 2 normal distributions depending on a negative or positive AMM. The TC and hurricane counts and ACE indexes were adjusted for changing detection capabilities based on their Easy to Detect counterparts.

I used AMM positive for my forecasts based the Klotzbach/Gray 2008 forecast notes a favorable (positive) AMM for the upcoming NATL TC season – kind of – and found in this link:

22. Posted Nov 3, 2008 at 1:54 PM | Permalink

The AMM peaks in boreal spring in both the Atlantic and the Pacific basins, and shows good persistence into the summer.

Or was the TC activity favorable for a favorable AMM that in turn was favorable for TC activity?

I am unsure of a mechanism with which TCs could “excite” or impart influence upon the AMM. However, I am still curious about the contamination of monthly mean fields by the TCs themselves. I will think of an alternative way to test this. I would prefer to maintain the months of August-October in the analysis.

• Kenneth Fritsch
Posted Nov 4, 2008 at 6:48 PM | Permalink

Re: Ryan Maue (#27),

If Swanson’s False Causality theory is correct whereby TC activity affects 850 hPA zonal winds, and not the reverse causality, then I would suspect that a season with high TC activity (as measured here by ACE) should increase the differences in the wind field between the start and finish of the TC season, whereas a low TC activity should have a lesser effect from start to finish.

To test whether the evidence agrees with the Swanson theory, I used the 850 hPa zonal winds and subtracted the July value from the Nov value and the did the same for the Oct and Aug values. I used the region from 12.5 – 17.5N and 85 – 40W for the zonal winds. I then regressed the seasonal ACE for all storms and the seasonal ACE for those storms west of 60W (Easy to Detect Storms) against the zonal monthly differences.

The results in the 4 graphs below show little correlation between the end-finish monthly differences and ACE. Also note that the differences trend to smaller values with higher values of ACE – which is the wrong direction for agreement with the Swanson theory.

• Kenneth Fritsch
Posted Nov 5, 2008 at 9:26 PM | Permalink

The 2 graphs on the right side should read Oct -Aug on the x axis (as is noted in the graph title) and not Nov – July.

• Kenneth Fritsch
Posted Nov 7, 2008 at 5:29 PM | Permalink

I had some time to duplicate what Swanson did for False Causality in that I used the same region that he did for the 850 hPa wind component U at 7.5-17.5N and 90-20W. My data came from NCEP/NCAR Reanalysis whereas Swanson used ECMWF/ERA-40 Reanalysis (that I, as a non-academic, do not have access to). I understand that there is not material differences in these data sets at the pressure levels used here.

I did the same comparisons as in the above linked previous post and had the same conclusions. The correlation for the ASO (Aug-Sep-Oct) 850 hPa zonal wind (component U) with ACE from 1950-2007 is reasonably good and therefore using Swanson’s reasoning that the wind component is caused by the ACE of the NATL TCs, and not the other way around, one should expect to see some effect during the TC season progression whereby the higher levels of ACE (or TC activity) will change from the beginning to the end of season and more than in those seasons with lower ACEs. That is not the case, as one can see that increasing ACE during the season over the period of ASO produces an increase in the 850 hPa zonal wind with a reasonably good correlation, while when comparing the beginning to end of season months the correlation is poor and goes in the other direction.

I think this evidence and that provided in posts above by David Smith show that Swanson’s “False Causality” hypothesis is probably wrong and was not tested properly. Does anyone care to discuss this issue or is it time to move on?

23. Posted Nov 4, 2008 at 3:37 AM | Permalink

Here is a link to the Met Offices’ June 2008 press release on its forecast for the hurricane season:

http://www.metoffice.gov.uk/corporate/pressoffice/2008/pr20080618.html

24. Mary Ann
Posted Nov 4, 2008 at 3:47 PM | Permalink

Hi, long time lurker.. first (and last) time posting. I wanted to tell Steve that I appreciate the website and only have one gripe – I go into withdrawls when he takes a day off! Other than that, I recommend this site to everyone who offers me an opinion on the state of climate science. Keep it up! Everyone here is amazing. I’m in Project Controls and it has really opened my mind to the many ways to look at data and to never take anything at face value.

25. MikeJ
Posted Nov 5, 2008 at 5:21 PM | Permalink

A bit late perhaps, but I have been away the last few days.
Atlantic Hurricane forecast by Philip Klotzbach and William Gray clearly links increased hurricane activity to global cooling. See their paper http://hurricane.atmos.colostate.edu/Forecasts/2008/aug2008/aug2008.pdf in particular their graphic, Figure 15, on page 34.

26. Nick
Posted Nov 5, 2008 at 5:45 PM | Permalink

Ryan,

I don’t know if you can contact Drudge, but he has butchered his headline linking to your website:

‘Record inactivity’ — 2008 Atlantic Hurricane Season ‘lowest in 30-years’…

27. Posted Nov 5, 2008 at 6:45 PM | Permalink

I am trying! Matt left me hanging out to dry last year as well! UGH!

28. Bob
Posted Nov 6, 2008 at 7:50 AM | Permalink

This is not supposed to be happening. We have catastrophic “global warming” and we are supposed to have hurricanes of epic numbers and strength. Is it just possible that there is no such thing as “global warming”? Is it really just a global socialist money grab?

29. Posted Nov 6, 2008 at 11:04 AM | Permalink

Drudge has kindly fixed his link — I think the development of Paloma made his headline seem silly.

30. Tim
Posted Nov 6, 2008 at 3:12 PM | Permalink

All of Al Gores Blow-harding must have been in a counter-clockwise direction, which negated the natural spin of a hurricance. Thanks Al, you saved the day.

31. Posted Nov 6, 2008 at 8:40 PM | Permalink

ACE for 2008 is currently about 131. Hurricane Palona is adding 1 to 2 points per day, perhaps higher if it strengthens. So, in a few days we may be in the 135-140 (“above-average”) range.

The CA contestants who forecasted an above-average season are:

(terry) (105)
Bill F (112)
Steven Mosher (130)
Staffan (133)
TSR (135)
ma in va (140)
Kenneth Fritsch (145 (average))
David Smith (146)
Gray/Klotzbach (150)
US NOAA
Accuweather
Weather456
raleighwx

Please post if I’ve omitted or misstated any entry. Later in November I’ll construct a table of those who made named-storm forecasts.

• Kenneth Fritsch
Posted Nov 7, 2008 at 9:41 AM | Permalink

So, Steven Mosher, please tell me, who finally wins the contest. Please email me the results as I do not want to ruin the suspense for others.

32. Posted Nov 6, 2008 at 8:40 PM | Permalink

Oops, that Paloma, not Palona.

33. Posted Nov 8, 2008 at 8:37 AM | Permalink

Re #42 Probably time for us to move on. This has chicken-or-egg aspects and, as such, we can do reasonableness checks which give suggestive results but nothing that’d lead to closure.

Paloma is contributing nicely to seasonal ACE at the moment. We may well be above 140 in a few days.

• Kenneth Fritsch
Posted Nov 8, 2008 at 6:58 PM | Permalink

David, you are probably correct, but I wanted to offer one more analysis for review here before moving on. My problem is that the results I obtain and my interpretation of them are so counter to those of Kyle Swanson’s in False Causality that I need a check by someone to determine if I am analyzing the results correctly. I am always suspicious when results (even my own) seem to so overwhelmingly confirm my hypotheses.

I looked at the NATL TC seasons with the 5 highest and 5 lowest ACE indexes and the corresponding 850 hPa zonal winds differences for the months Nov-July and Oct-Aug. for the time period 1950-2007. I also calculated the averages for all TC seasons from 1950-2007 for ACE and zonal wind differences Nov-July and Oct-Aug. The results are reported below.

TC seasons with 5 highest ACE indexes:
Average ACE index = 228; Average zonal wind difference Oct-Aug = -0.59; Average zonal wind difference Nov-July = 0.32

TC seasons with 5 lowest ACE indexes:
Average ACE index = 26; Average zonal wind difference Oct-Aug = 0.05; Average zonal wind difference Nov-July = 1.24

Averages for all seasons 1950-2007:
ACE = 101; Zonal wind difference Oct-Aug = -0.21; Zonal wind difference Nov-July = 0.54

From my previous posts on this subject, one can see that the correlation of ACE to zonal wind averages for Aug-Sep-Oct is such that, given the causality hypothesis of Swanson, we would expect the ACE to increase or make more positive the zonal wind differences from start to finish of the TC season. Remember that Swanson assumes/implies that the TC effects can last for several weeks. Therefore if I subtract the end of season month zonal wind from the beginning I should see more positive results for those seasons with the highest ACE indexes. This difference should be more pronounced when comparing the 5 highest with the 5 lowest ACE index seasons. I see just the opposite to what Swanson would predict. My point is whether I am overlooking or doing something incorrectly here. I actually would expect these differences to be rather neutral and to that end I tested the statistical significance of these differences by assuming a normal distribution of zonal wind differences over this time period. The results below show that the zonal wind differences are within the 95% CI for a 5 season sample in all cases.

95% confidence range for 5 sample average for Oct-Aug = -1.03 to 0.61
95% confidence range for 5 sample average for Nov-July = -0.39 to 1.46

Just to show that the 5 highest and 5 lowest ACE are out of the 95% CI for a five sample average, I list that range also:

95% confidence range for 5 sample average for ACE = 48.5 to 153.9

34. Posted Nov 8, 2008 at 11:03 PM | Permalink

NOAA offers a US Climate Extremes Index (“CEI”) which is to illustrate whether the US climate is becoming more extreme. Their effort includes an experimental “Hurricane Season CEI”.

This CEI is based on the squares of windspeed for all the US landfalling storms (including tropical storms) in a season. It has some resemblence to ACE.

The plot indicates a tendency for the CEI to rise over the last 100 years. That was a mild surprise to me, as I expected to see no-change.

so, I plotted the same data as is used in the CEI but with what I thought were three minor exceptions – I count storms only once (no multi-landfall storms), I exclude tropical storms (wind below 65 knots) and I used the Saffir Simpson ratings (the CEI uses a mix of sources).

My result is here:

Interestingly, the trend is downwards, the opposite of the CEI.

So, I need to adjust for double-landfalls, add tropical storms and (if possible) mimic the CEI methodology. Perhaps those adjustments will allow me to match the NOAA CEI. If they do, then I need to ponder why double-landfalls and weak systems have such a strong effect on the CEI. If they do not, then the question is why.

35. Kenneth Fritsch
Posted Nov 9, 2008 at 2:59 PM | Permalink

David Smith, I find Karl’s Climate Extreme Index somewhat arbitrary and particularly when precipitation and drought are used in three of the six categories.

Anyway, I took the data provided by NOAA in your link and looked at the regressions of the total index and that for each of the six categories that are averaged to obtain the total index. The total has no statistically significant trend and only one of the six categories had a statistically significant trend.

That trend was for precipitation (Step 4) and with an R^2 = 0.16 and an adjusted (for an auto correlation with r = 0.28) trend with 95% CI from 0.018 to 0.082 units (percent change per year).

I think it would definitely be time to move on from this one unless you want to look in greater detail at Step 4.

• David Smith
Posted Nov 9, 2008 at 3:40 PM | Permalink

Re: Kenneth Fritsch (#46) Step 4 (extreme precipitation events) is the one which may have problems with the changes in precipitation observation time. There were some posts on this earlier. Steps 1 and 2 have problems with shifts in mean temperature. The Index is not impressive, at least at this time.

I’d like to understand exactly how step 6 is constructed and hope to gather enough information to get there.

• Kenneth Fritsch
Posted Nov 10, 2008 at 12:08 PM | Permalink

David, I was surprised that Karl, who headed up the review reported in the CA thread introduced by Ryan Maue, titled, “Consensus Report on North American Climate Extremes” and linked here http://www.climateaudit.org/?p=3191 , would produce a rather arbitrary extreme climate index that really shows nothing extreme happening in the US since 1910.

There appears to be a disconnect between the index and report. Am I missing something here or is it perhaps that Karl will do some future fine tuning of the index?

36. Gerald Machnee
Posted Nov 9, 2008 at 4:34 PM | Permalink

Hurricane Paloma died a quick death as soon as it hit Cuba.

37. David Smith
Posted Nov 9, 2008 at 6:39 PM | Permalink

Re #48 Paloma went from 125 kt winds to 30 kt winds in just 24 hours, which is as rapid a decay as I recall (excluding direct hits on the high mountains of Hispaniola). Shear, dry air and land delivered 1-2-3 punches to destroy the small-sized storm.

Paloma did end up with an ACE of 10, which brings the seasonal total to 141. Kenneth Fritsch (145) and ma in va (140) are now well-positioned in our contest.

38. Posted Nov 9, 2008 at 11:40 PM | Permalink

Paloma, with a maximum intensity of 125 kts (briefly) and an ACE of 9.915 (or 10), ranks as one of the least-ACE producing “intense” storms since 1981. I use 1981 because I can then compare with the rest of the Northern Hemisphere and have some confidence in the storm data. Iris (2001) had slightly more ACE, but less than 10 as well with a maximum wind of 125 knots.

When the Northern Hemisphere is then considered, it is apparent that Paloma is one of the least-ACE producing storms for its intensity (1981-2008). On the flip side, Bertha (2008) was one of the most prolific ACE producers for its max intensity (105 knots, ACE of 28)

There is some very interesting variability in the distribution of Northern Hemisphere ACE per storm during the past three decades, but I am still trying to make sense of it. However, it is clear that the spread of ACE per storm at higher intensities is much larger than at lower intensities.

39. Posted Nov 11, 2008 at 10:14 PM | Permalink

Interesting story on Hurricane Ike, which comes to me from several reliable sources:

As Ike’s calm eye passed over a Texas community its emergency response crew left their shelter to do a quick community damage assessment. They were surprised to find several brightly lit vans amidst the darkness. The vans belonged to a US cable network which was doing a live national broadcast using, believe this or not, a very large fan and water sprayer to blow wind and “rain” onto the reporter while he did his live broadcast.

I suppose they came equipped for those moments when the wind and rain momentarily did not coincide with the broadcast time. Hollywood.

40. Posted Nov 12, 2008 at 7:03 AM | Permalink

Hmmm, could this cable network have part of its name three letters NBC? A favorite stunt pulled by Michelle Kosinski on NBC shows her rowing around in a canoe getting all dramatic about the flooding in New Jersey in October 2005, when she could have easily got out and walked about. A video link.

Also, USA Today did a short story on the Northern Hemisphere TC Inactivity that I have discussed here on CA in the past…
USA Today Tropical Cyclone Article

41. David Smith
Posted Nov 12, 2008 at 7:41 AM | Permalink

Re #53 Well, no. I’ll let you know offline.

They also had, according to the witnesses, wet leaves and twigs in buckets for use in the fan-derived wind. It was a surreal moment.

42. Posted Nov 12, 2008 at 11:37 AM | Permalink

53: (Ryan)

Congrats on having your hurricane study written about in the USA Today on page 3. I did not know we were dealing with royalty here!!

Tom

43. Posted Nov 12, 2008 at 1:47 PM | Permalink

The author of the USA Today article on Northern Hemisphere Tropical Cyclone Activity did an excellent job of concisely outlaying the set of facts at hand. While I have a GRL paper trudging through the final steps of revision, the link by Roger Pielke and subsequent pub on the Drudge Report generated a fair bit of positive reaction to the research.

I am rather bored with prognosticating the future of hurricanes in a warming world and prefer to understand current variability. This graphic is a good place to start.

44. Posted Nov 12, 2008 at 2:59 PM | Permalink

56 (ryan):

That graph. How far back do your ACE records go?? How well does ACE correlate with PDO and ENSO anomolies??

Thanks, Tom

45. Posted Nov 12, 2008 at 3:11 PM | Permalink

Tucker, #57, Since Northern Hemisphere is 80% Western Pacific + Eastern Pacific basin contribution, the relationship with ENSO/PDO is indeed very strong. The NATL is thus indirectly affected by the teleconnections associated with ENSO (modulation of upper-level Walker circulation, etc.) so some compensation is expected. The NH records can go back as far as the 1940s, but the data quality falls off faster than General Motors share prices.

46. Posted Nov 12, 2008 at 5:46 PM | Permalink

58 (Ryan):

One would think that with correlation data in hand, that one could set up a data profile from various locations to predict ACE probabilities for any given year. Somewhat simlar to SOI data that can predict ENSO probabilities into the future. Whoever can do that would earn the big bucks from the insurance industry.

As for data quality from the 1940’s and before, that is fast ….

47. Posted Nov 20, 2008 at 8:52 PM | Permalink

Our 2008 contest involves only ACE forecasts. However, several brave souls offered forecasts for the number of named storms and the number of hurricanes.

I consider the 2008 season to be over, with little chance of another storm between now and 30 November. So, here is the 2008 count:
16 named storms
8 hurricanes

Our fearless forecasters are:

John A forecast 6 hurricanes, plus or minus 3, a forecast seconded by Paul Linsay

ma in va forecast 10 hurricanes with 4 hitting the US (3 actually hit)

matt vooro forecast 15 named storms and 6 hurricanes

terry went with 16 named storms and 7 hurricanes

Mike B forecast 27 named storms and 14 hurricanes

Dan Hughes forecast “climatology” (not bad if the reference period is 1995-2007)

moshe forecast 15 named storms and 8 hurricanes

Kenneth Fritsch forecast 14 named storms and 8 hurricanes

Folks, these were some pretty good forecasts. Kudos to matt, terry, moshe and Kenneth for doing especially well.

Next, the winners of the official ACE contest.

48. Posted Nov 25, 2008 at 6:51 PM | Permalink

I’m experimenting with the Certificate of Achievement for those who win the 2008 hurricane contest. Here’s one possibility, for your consideration and comments. I am quite open to suggestions:

• Kenneth Fritsch
Posted Nov 28, 2008 at 8:22 AM | Permalink

I like it, David. The dice that inform as to the forecasting and placing them directly below Einstein was a nice touch and too clever by far.

I wanted to do a quick analysis of just how lucky my picks were (using the methods that I used). Also we need to keep in mind that selecting from a group of forecasters increases the odds for obtaining some close picks just by chance. I suppose if we wanted to be real sneaky we could do some averaging from ensemble forecasts.

My confessions would be more appropriate if I were to win the ACE contest but as of this writing my calcualtions show that the better Ma(n) won.

49. Posted Nov 26, 2008 at 11:05 PM | Permalink

Bob Tisdale explores a lot of ocean data. A recent post of his examines the North Atlantic subpolar gyre, a region associated with thermohaline activity and AMO variation.

Bob’s SST time series for the gyre looked familiar, so I smoothed the annual SST then plotted it against the smoothed hurricane-season ACE values. ACE is red and the gyre SST is blue:

The unsmoothed data has an r-squared value of 0.43, which is not bad.

Interpretation of this is not straightforward, as it might be seen as simply reflecting correlation of Atlantic-wide SST or as a reflection of wind and pressure patterns or as an indicator of ocean/thermohaline activity. Interesting nevertheless.

50. Posted Nov 27, 2008 at 6:29 AM | Permalink

58 (Ryan):

I am not sure if you are heading in this direction, but if you are looking for ENSO correlation to NATL activity, may I suggest you break out ENSO by region. You may find that regions 1+2 have a much higher correlation to Atlantic activity than the general ENSO numbers, and especially regions 3 and 4.

Tom

51. Posted Nov 27, 2008 at 7:03 PM | Permalink

DAVID SMITH

Thanks for the fine previous posts on this blog’s hurricane forecasting record. The forecasts show that we can match and even out perform the professionals. An interesting post season observation. If one takes the LINEAR trend of the number of named storms , number of hurricanes and ACE for the years 1995-2007 ,you will get the trend line projecting the actual numbers for 2008 for named storms[16] and hurricanes[8] and an ACE of about 130.

52. Posted Nov 27, 2008 at 8:14 PM | Permalink

Ryan [re #58]

Have you looked at the ACE levels in the Atlantic when the ENSO INDEX is in neutral during the hurricane season. [like 2008] It would appear on the surface that there are a higher number of hurricanes than the long term average[6] during ENSO neutral years. Taking the years like 1978, 1979, 1980,1981,1989,1990, ignoring 1992/1993 due to Pinatubo,1996,2001,2003,ignoring 2005[very high case] , the LINEAR TREND line for 2008 is 9 .[actual was 8] To me this is what may happen next year if the neutral trend holds throughout 2009

53. Posted Nov 28, 2008 at 11:22 AM | Permalink

In general, the “SST patterns” in the Pacific associated with ENSO do not correlate well with Atlantic activity. It is true that El Nino portends less hurricane development while La Nina and neutral conditions provide more favorable environments. This is partially a result of vertical shear modulation (especially at upper-levels of the troposphere) as well as upstream African jet influences — coming from the Indian Ocean side. Since most hurricanes form as a result of African Easterly Wave development or are heavily influenced by such wave progression, it makes sense to look at the large-scale climate signals which would affect them most.

Tropical storms develop when conditions permit and usually in little cocoons of favorable atmosphere. The SSTs are always sufficient for genesis in the tropics from July – October. Positive SST anomalies on the local scale easily feed-back to the overlying atmosphere and may increase the favorability for genesis or intensification of storms. So, most Atlantic development probably can be attributed to how the climate deals with the African Easterly Waves.

54. Posted Nov 29, 2008 at 11:24 AM | Permalink

Ryan

You make some interesting points.It would appear to me that the African Easterly Waves seems to be solar heat driven and whether they end up generating a high or low storm and hurricane season in the Atlantic or Caribbean depends on the ENSO conditions coming from the Pacific.You can compare the years 2005 and 2008. Both years were ENSO neutral, both had more than average number of storms and hurricanes but the solar heating as shown by the temperature data in my judgement was significantly more in 2005.The conditions were ideal for a record storm season , namely an unusually strong and steady African Easterly wave with little or no oppostion or interference from the Pacific ENSO effect. A similar condition existed in 1995 with neutral and weak La Nina conditions during the latter part of the year and warm solar heating.[11hurricanes]

55. Posted Nov 29, 2008 at 1:58 PM | Permalink

Our 2008 ACE contest officially closes November 30’th. Nature, however, has already ended the season and therefore the contest winners can now be announced.

Our contestants chose of one five ACE groupings (“much below average, below average, average, above average or much above average”), each of which corresponded to a 20% quantile of ACE (1950-2007). This year the season’s ACE was 141, which falls into the “above-average” quantile.

In recognition of their accomplishment I offer this collector’s edition :) Certificate of Achievement to the six forecasters who correctly predicted an “above average” season ACEwise:

Among the six there are three who deserve special acknowledgment. They are ma in va, whose ACE forecast was the closest (140 forecast vs 141 actual); Kenneth Fritsch who was close behind (145 vs 140) and who also forecast the exact number of hurricanes, and Steven Mosher who offered additional forecasts of other aspects of the season, most of which were close to the mark.

Congratulations to those who came close and thanks to everyone who participated. Let’s hope these forecasters will share their secrets of success here at CA.

56. bender
Posted Dec 11, 2008 at 8:24 PM | Permalink

I still contend that 2010 is the year to watch, that 2008 confirms that the 5 year cycle is intact (and currently rebounding, as predicted), that matching the observed number of storms in any one given year, such as 2008, is not nearly as indicative of “skill” as consistently getting it more-or-less right across many years. In time I WILL claim my prize.

57. jeez
Posted Dec 11, 2008 at 9:43 PM | Permalink

Just post as Team Bender and claim you have already won.

58. team bender
Posted Dec 12, 2008 at 1:04 AM | Permalink

I have won. Besides, there was no contest. Or rather, the contest was the wrong contest. Or …

59. steven mosher
Posted Dec 13, 2008 at 5:15 PM | Permalink

for 2009 I have a new muse. I interviewed an event coordinator and when I asked her for the biggest blunder she ever made she replied that she had scheduled
a beach bikini party and a Typhoon attended rather unceremoniously. I’ll simply ask her to plan a bunch of parties for 2009 in the Gulf region and then on the assumption that she is to hurricanes what Tyrone Sloproth was to V2 rockets ( see Gravity’s Rainbow) I will make my prediction accordingly.

60. Posted Dec 21, 2008 at 1:12 AM | Permalink

For those that are following Roger Pielke Jr’s discussions on the RMS elicitation or modeling of hurricane landfalls over the next 1-5 years, please have a gander at a couple links:

Can Hurricane Activity be predicted out to 5-years with skill? (No!)

But, the flashback to the 2005 RMS elicitation has the 4 expert scientists associated with the crummy forecasts. This also links back to Prometheus.

Roger nailed all of this well over 2-years ago. Thank goodness Florida didn’t accept that RMS elicitation at that time and jack up insurance rates 40%.

61. Posted Jan 27, 2009 at 11:28 AM | Permalink

The UK Metoffice issued a self-congratulatory press release describing the unmitigated success of their 2008 Atlantic hurricane forecast. From their presser: Spot-on storm forecasting

The 2008 North Atlantic hurricane season was the fourth most active since detailed records began and this was accurately forecast by the Met Office. Richard Graham, climate specialist at the Met Office, said: “For two years running our model has given accurate guidance on tropical storm numbers, and we aim to build on that success this year.

“Our capability in this emerging area shows how the science of climate change can be used to help us today, as well as in the future.”

The Met Office uses a dynamical numerical model of the climate system, known as GloSea, to create the forecast. This technique of modelling atmospheric processes to predict tropical storm activity ahead of the season is a departure from traditional methods, which use statistical data.

Here is the graphic from their page and some text that goes along with it:

Issued 18 June 2008

15 tropical storms are predicted as the most likely number to occur in the North Atlantic during the July to November period, with a 70% chance that the number will be in the range 10 to 20. This represents above-normal activity relative to the 1990-2005 long-term average of 12.4.

I need to see more evidence of “forecasting” success than just two seasons before I start applauding the utility of climate models for seasonal forecasting. Since the climate model can’t resolve the tropical cyclones themselves, an adjustment process is applied to get the forecast number. This adjustment is a function of previous season’s activity and is essentially an analogue method.