Tuesday, December 03, 2013

Hurricane forecasts are worthless

I don't know the methods that folks in NOAA are using to forecast the number of Atlantic hurricanes and I am not interested in these methods because they clearly don't work. Pages such as this one (with different years in the URL) show you tables with the predicted and actual number of named storms, hurricanes, and major hurricanes for the relevant year.

Let us look at years 2001-2013 and focus on the actual numbers from NOAA in May. Note that the hurricane season includes months between June and November and the 2013 season has officially ended last Saturday.

Let's start with major (=Category 3,4,5) hurricanes. I am ready to bet that the qualitative message we could extract from all hurricanes or all named storms would be similarly discouraging and the same would be true for other teams that try to forecast the hurricane season before it starts.

The May NOAA forecasts vs actual numbers after the season are:

2001: 2-3, 4
2002: 2-3, 2
2003: 2-4, 3
2004: 2-4, 6
2005: 3-5, 7
2006: 4-6, 2
2007: 3-5, 2
2008: 2-5, 5
2009: 1-3, 2
2010: 3-7, 5
2011: 3-6, 4
2012: 1-3, 2
2013: 3-6, 0

Some of the numbers are OK within the interval, some of them are underestimates, some of them are overestimates. Concerning the latter, 2013 is the greatest blunder in this list (the years 2001, 2004, 2005, 2006 will contribute the same terms to the "badness score" below but that's only because these forecasts had smaller i.e. more ambitious values of the accuracy \(\Delta N\)). The prediction expected the 3rd or 4th most intense year (after 2006, 2010, tied with 2011). Instead, we got zero major hurricanes (both 2013 hurricanes were the minimum-size Category 1 hurricanes) – the lowest number in many decades.

Let's analyze the success rate in a slightly sophisticated way, using the \(\chi\)-squared distribution techniques. Let's define "chi squared" as\[

\chi_{13}^2 = \sum_{y=2001}^{2013} \frac{(N^{\rm predicted}_y-N^{\rm actual}_y)^2}{(\Delta N_y)^2}.

\] The subscript \(13\) means that we are dealing with a system of \(13\) variables (hurricane counts for 13 different years).

What are we doing? We are calculating the differences of the actual number of major hurricanes from the mean value of the predicted range, square the result (because the second power is more mathematically natural than the absolute value), and divide it by the square of the expected error margin which is, to mention an example, \(1.5\) for the year \(y=2013\). In other words, if you know a certain semitechnical jargon, \(\chi_{13}^2\) is the sum of thirteen squared "numbers of sigmas" encoding the error of each estimate.

To continue with this example, the contribution to \(\chi_{13}^2\) from \(y=2013\) is \((4.5-0)^2/1.5^2=9\). All the \(13\) terms are\[

\chi_{13}^2 &= 9+1+0+9+9+9+4+\\
&+1+0+0+\frac 19+0+9 = \frac{460}{9}\approx 51.1.

\] Great. This is NOAA's score; the higher value of \(\chi_{13}^2\) we get, the poorer the estimates were. How do we decide whether \(51.1\) is a good score or a bad one? My recipe is: permute the actual 13 annual numbers of the hurricanes by random permutations and calculate the score \(\chi_{13}^2\) for each random permutation. Then compare the actual unpermuted score with the permuted ones to see how many permutations give you worse results than the unpermuted score. That's a measure whether NOAA's estimates are better than guesswork – especially whether it makes sense to prepare different forecasts every year, pretending or assuming that we can distinguish individual years in advance.

So what do the scores \(\chi_{13}^2\) from the random permutations look like? The histogram of the scores for 150,000 random permutations looks as follows:

The maximum of the distribution is around \(\chi_{13}^2\approx 60\) (that's the value you read on the \(x\)-axis) and the secondary minor bump near \(\chi_{13}^2\approx 110\) is no fluke – it really follows from the character of the data and the forecasts. The average \(\chi_{13}^2\) over the permutations was \(75.5\) or so but a very significant percentage of the random permutations, namely 23 percent, gave a better (i.e. lower) score than the actual \(\chi_{13}^2\approx 51.1\) from the unpermuted numbers of hurricanes (it's the percentage of the area in the graph above on the left side from the \(\chi_{13}^2=51.1\) vertical line).

You are invited to repeat the calculation for the total number of hurricanes and the total number of named storms.

At any rate, the calculated probability 23 percent means that if a dog belonging to the NOAA hurricane experts actually eats the homework and randomly permutes the predictions for the years, the likelihood is 23 percent that the dog will surprisingly improve the predictions for the ensemble of the 13 years relatively to his or her master or lord (I mean the owner).

In other words, the empirical evidence suggesting that the forecasts are better than guesswork (that the people are better than their dogs) is as weak as the observation that "holy crap, a random number between 0 and 100 happened to be smaller than 23, what a miracle, it must mean something!". Well, it often happens that a random number between 0 and 100 is smaller than 23. The NOAA forecasts are pretty much as successful as guesswork. To claim that there is at least 2-sigma evidence that NOAA is better than guesswork, the number 23 percent above would actually have to be smaller than 5 percent! So one more description of the apparent failure says that there's no statistically significant evidence that their forecasts are better than random guesswork.

We could also try to replace their variable and seemingly arbitrary forecasts for individual years by a different model. In the years 2001-2013, the actual number of major hurricanes was 44 (the sum of the "actual" thirteen numbers at the top) which is 3.38 major hurricanes per year.

To make our "universal perdiction" as bold as NOAA's prediction, we have to have our \(\sum(\Delta N_y)^2\) equal to theirs. That's equivalent to choosing \(\Delta N=1.185\) or so, the average (root mean square) \(\Delta N_y\) used in the NOAA forecasts. So imagine that we would compete with NOAA and we would say that the number of hurricanes is from \(3.38-1.185\) to \(3.38+1.185\) i.e. between \(2.195\) and \(4.565\). Our prediction would be the same for every year.

How well would our simple prediction – which is as accurate in average as NOAA's – fare in comparison with NOAA? Well, our score would be simply\[

\sum_{y=2001}^{2013} \frac{ (N^{\rm actual}_y - 3.38)^2 }{ 1.185^2 } \approx 33.5.

\] I kid you not: our score is \(33.5\), vastly smaller (i.e. better) than NOAA's \(51.1\). How many permutations (using the original NOAA predictions) give a better result than \(33.5\)? It's just 4.7 percent of the permutations. (Of course, permutations don't improve anything with our prediction because it doesn't depend on the year. We would have to use another measure of "relative success" of such a forecast.)

In this sense, there does exist some slightly significant, 2-sigma evidence that our simple forecast is better than NOAA's forecasts randomly permuted by their dog. That's a result that NOAA could only dream about. The actual numbers suggest that NOAA actually made the predictions less accurate by deviating from the universal prediction – that their deviations were anticorrelated with the actual annual fluctuations! Too bad.

So the recipe for NOAA to save one-half of the budget for the major hurricane forecasting and to improve the forecasting of major hurricanes is straightforward: fire all the people who are currently employed to do such things, send one-half of the budget every year to your humble correspondent, and read the e-mail from me that will predict \(2.2-4.6\) major hurricanes every year.

I can similarly save one-half of the money for all other forecasts related to the tropical cyclones, too. And improve the forecasts at the same moment.

Feynman's 1974 Caltech commencement speech begins in the following way:
During the Middle Ages there were all kinds of crazy ideas, such as that a piece of of rhinoceros horn would increase potency. Then a method was discovered for separating the ideas--which was to try one to see if it worked, and if it didn't work, to eliminate it. This method became organized, of course, into science. And it developed very well, so that we are now in the scientific age. It is such a scientific age, in fact, that we have difficulty in understanding how witch doctors could ever have existed, when nothing that they proposed ever really worked--or very little of it did.
Well, it is not hard to imagine how our ancestors could fail to care whether their methods were working: the people in the Earth sciences clearly don't care, either. It's straightforward and sort of trivial to prove that their forecasting methods don't work but no one seems to care and the same bogus forecasts are generated every year again and again. In fact, hundreds of billions of dollars are being wasted every year as "scapegoats" for numerous similar predictions that demonstrably fail to work as well.

So I have concluded it's not a scientific world (well, of course, Feynman made the same conclusion in 1974, too).


  1. NOAA's forecasts typically have enormous ranges, especially their May forecast. Of course, basically all forecasters have egg on their face this year.

    I think (but I'm not sure) that CSU has a better track record. Predicting weather sure is hard, though!

  2. The enormous ranges still fail to increase the success rate in any way. Clearly, in many years, they were outside even these "enormous ranges".

    Predicting weather is hard, it's so hard that it's really impossible beyond a certain point and a sensible person shouldn't pretend that he's producing scientific predictions is the methodology demonstrably doesn't work.

    The word "forecast" or "prediction" is a pretty strong one and it shouldn't be used in cases that are accurately described as "guesswork".

  3. But inquiring minds still want to know, can an Amazon drone flying on
    one side of the world cause a hurricane on the other side of the world
    by way of the Butterfly Effect? ;~)


  4. Hm, good point.

    Anyway, I am sorry to say it looks like CSU's *preseason* numbers are not better than the mean (they are actually worse) over the period 2001-2013.

    My understanding is that it is easier to accurately predict the aggregate activity (ACE) than it is the breakdown into named storms, hurricanes, and major hurricanes. CSU is pretty candid that the best they generally do on numbers long term is, will it be above average or below average. Which is, IMAO, better than NOAA.

    At any rate, this presentation their team gave for their forecast at the Florida Governor's Hurricane Conference is interesting:


    Their predictions by August usually improve-see also their statistical model for "NTC" (a measure of aggregate activity) Interestingly, both their statistical prediction for named storms (15.6) and their average of analog years (10.4) are much closer to the actual value than their final august number (18). I'm not sure how they derive their final numbers, it seems to be a subjective judgement. This perhaps adds error rather than useful information.

  5. choosing the in-sample average of 3.38 as a static prediction and noting that it did better is silly. of course it did better. you chose it with the benefit of perfect hindsight. their, admittedly crappy, chi-squared is at least computed by comparing out-of-sample results to forecasts. better to run some experiments. experiment#1: draw many length-13 sequences of an iid N(0,1) variable and compute each sequence's chi-square (just the sum of squares, since mean=0 and sd=1). what fraction of the sequences have a value as large as 50? answer: basically 0 (less than 1e-5). experiment#2: since the chi-square depends both upon knowledge of both the mean and the standard deviation, let's make life easy by assuming we really do know the standard deviation and let's denominate our uncertainty about the mean in units of the sdev. in other words, let's say that for each of the 13 "reads" in a sequence we draw the mean from a distribution centered on the expected mean with uncertainty alpha * stdev. question: how big does alpha have to be before chi-square has probability >= 1/3 of being at least as large as 50? running the experiment tells you that alpha has to be at least 1.5, i.e. the uncertainty in the mean is much larger than the claimed fluctuation size about the mean. so, i agree that their results are crap. but i don't agree that you can compare to an in-sample result to draw the conclusion. who cares, really... by the way, i know you're a mathematica fan-boy and i like it,too, but R is free. here are a couple of R lines to show the result:




    for (i in 1:length(v)) { v[i]=rnorm(mean=rnorm(mean=0,sd=alpha,n=1),sd=1,n=1);






  6. Well, I am not sure why you think you don't agree with me since I agree with almost everything you wrote above, except perhaps some history. That is: Western Ukraine was a part of the Russian Empire only when Poland was a part of the same empire. As you know very well, it was part of Poland between the wars. Contrary to what many Westerners beleive, that was not what the Poles wanted, or more accurately, not what the creator of pre-war Poland, Jozef Pilsudski wanted - he started the war with the Bolsheviks in 1920 by advancing on Kiev in alliance with Simon Petliura (http://en.wikipedia.org/wiki/1920_Kiev_Offensive ) in order to try to an independent Ukrainian state that he saw as a necessary buffer between Poland and Bloshevik Russia and a future ally. It was only the failure to achieve decisive victory in that war that resulted in Poland dividing Ukraine with the Soviets. Pisludski, already out of power, apologised to the Ukrainians.

    The situation of Western Ukraine thus differs from the situation of Poland only during the post war period, during which Western Ukraine was a part of the Soviet Union while Poland was a satellite state. This is a difference but not as significant as you seem to suggest.

    As for Eastern Ukraine, again pure history does not provide sufficient explanation for the difference between the countries since Kiev was Polish until 1667 and Poland lost its independence in 1795, about 130 years later. Again, while 130 years is not insignificant it is also not also not “many centuries”.

    On the other hand, there is also a great deal of truth in what you write. In fact, in Poland alone there is a large difference in economic and cultural development between the western part and the eastern parts, even if you discount the territories acquired from Germany. Even in old, pre-partition Poland, the East was always more backward economically and culturally. These differences persist even in spite of the mixing of the populations that took place after the war. (The rapid economic and social decay of the former German lands after Poland took them over after the second world war was largely due to the fact that Poles expelled by the Soviets from the former Polish lands in the east were resettled there).

    In fact, this phenomenon of the West being more developed than the East is not just a Polish but a European phenomenon.

    So yes, I do not expect independent Ukraine, even in the EU to reach the level of development of Poland just as I don’t expect Poland to fully catch up with Czechia - although that is more of a possibility if your economy gets badly mismanaged. But I have no doubt that Ukraine will be vastly less well off if left to itself and even worse if absorbed by a revitalised Russian empire (although actually it is much more likely to drag this empire down like a heavy weight drags down a leaking balloon).

    The other thing I mentioned is the geopolitical aspect. A loss of independence by Ukraine, whether de facto or (very unlikely) formalised will actually create a tremendous destabilisation in Europe - much more threatening than the break up of Yugoslavia ever did.

    As usual, people who suggest that its best to keep out of this matter are ignoring the fact that all that are doing is making sure that they will get involved when the situation is much worse in the future. This has, in fact been a repeated theme in the foreign policy of the West since the end of the first world war.

  7. The forecasting failure starts very deep down, judging by the inability of anyone to forecast the overall ACE value anywhere near accurately. There has been quite obvious liberalization of the criteria for recording 'named' storms, to the point that even minimal storms in mid ocean now get included. The result is that even if the forecast were correct on the number of storms, it does not relate to the damage potential, which eliminates whatever residual utility it might have had. That is of course the point of the memo.
    Sadly, it could be applied with equal force to any longer term weather forecast that I know of.

  8. It matters not whether they are right or wildly off. Atlantic hurricanes that strike land are dispersed over 3,000 miles of coast line. If NOAA predicted 4 severe hurricanes in a season, and it was known to be right, no one anywhere would do anything different. It's a useless fact. Except, of course, it's not even fact.

  9. "There has been quite obvious liberalization of the criteria for recording 'named' storms"

    Not this bullshit again.

    What you are seeing is *improved analysis and observations*, not any redefinition of criteria. Those marginal storms should be counted. It's the *failure* to count such systems-because of inadequate technology or methodology-that was wrong, not that we count them now.

    TL;DR Counts in the past were too low, not too high now.

  10. dude, even if you don't publish my comment you should at least correct your error. your conclusion is right only by accident right now...

  11. Don't be silly. You know very well that the distribution of hurricane numbers per year *is* static plus minus undetectable noise. I could have easily made the same prediction (plus minus epsilon) before 2001, too. Just calculate the number of hurricanes in any other decade.

    Nothing is silly about what I write. If one can show - and I can show - that one doesn't gain (in fact, does loses) accuracy by doing different predictions every year, it is very important. A sensible person shouldn't try to do year-specific predictions - at least until the moment when something remarkably improves about the forecasting techniques.

  12. I don't understand why you think that we essentially agree. First, you apparently began to talk about "Western" Ukraine. Well, if we split Ukraine and talk about the regions separately, the results may be different. But I think that even Western Ukraine won't catch up with Poland and I disagree that the difference from Poland is just in the Soviet post-war years. The difference from Poland goes back 1,000 years due to the Catholic-Orthodox boundary which you seem to completely neglect. It is not negligible and it's really just the "Catholic" countries that underwent the Western revolution since the renaissance etc. Those things didn't automatically spread to the Orthodox world.

  13. I believe etudiant's point was that because the criteria have changed and/or the observations have improved, inferring a long-term trend line regarding the frequency of such storms is difficult, as the sign of the changed criteria automatically produces a (misleading, probably false) increase with time.

  14. Dear Lucretius, none of your "subtle refinements" matters. There may have been Orthodox folks in Poland but their number is negligible today. Almost all the Poles are "heirs" to the Catholic tradition and this is what determines whether the Western habits are being naturally adopted. Similarly, Ukraine had parts of tendencies to be Catholic in the past but those tendencies or people went nearly extinct. Much of the current Ukrainian population are heirs to the Orthodox tradition which also means that they have more similar attitude to certain social and economic questions as Russia or Greece than Poland or Germany.

  15. “Similarly, Ukraine had parts of tendencies to be Catholic in the past but those tendencies or people went nearly extinct.”

    Well, I am not sure what you are basing your information on, but a trip to Lviv should convince you that it is simply wrong.

    Every major church in this city is catholic. They belong to the Ukrainian Greek Catholic Church


    (known also as the Union Church in the Slavic languages,.e.g http://pl.wikipedia.org/wiki/Kościół_unicki )

    Visually the churches look “orthodox” and the liturgy, in particular very beautiful singing, is also orthodox but in all important respects the church is catholic, and in particular it recognizes the Pope as its head. About 10% of Ukrainians are members of this church and they form a large majority in the city of Lviv and the surrounding areas.

    Morevoer, the Greek catholics have always represented the elite in Western Ukraine (which is one reason why the Soviets try to extinguish the church completely) and most nationalist leaders come from this church (Stepan Bandera was the son of a Greek Catholic priest).

    The reason why the Roman Catholic Church is virtually non-existent in Ukraine is that under the terms of the Union of Brest, the Ukrainian Greek Catholic Church has a monopoly on catholicism in Ukraine and the Roman Catholic church only caters to the Polish population.

    All of this has nothing at all to do with Greece, as contacts with that country, which were restricted only to religious matters in the Middle Ages, have been minimal since then. Please note that the Churches in Armenia, Georgia, Rumania, Bulgaria, Serbia also call themselves “orthodox” although this means very little. For example, the Armenian church (which usually calls itself the Armenian Apostolic Church) is older than both the Catholic Church and the Orthodox Church. There is nothing sensible that can be deduced from the fact that it sometimes called “orthodox” (which really just means the same as catholic - the one that upholds the true doctrine).

    The idea that there exists anything called “orthodox civilisation” (and that in particular it has any relation to development) was a creation of Russian propaganda in the tsarist days and although it was adopted by some western thinkers (e.g. Samuel Huntington) it remains just one of many myths of this kind.

  16. OK, I don't understand what's the goal of this silly and obvious cherry-picking of yours.

    We were talking about Ukraine, not Lvov. You can't weaken my statement that Ukraine is predominantly Orthodox and its Catholicism is nearly extinct by counting churches in one cherry-picked town.

    The Lednice castle has a minaret but that doesn't mean that Czechia is partly a Muslim country especially because it's the only minaret we have in Czechia. ;-)

  17. 10% of a any population is not "nothing". Obviously you know nothing about the differences and the lack of thereof between the Catholic and the Orthodox churches. Think about it: the Catholic Church considers the Othodox "schismatics" and not "heretics" because there is no serious difference between the churches except in relation to the role of the Pope. Any orthodox church that recognises the position of the Pope automatically can become a part of the Catholic church, without any other change. This is what happened in the Union of Brest.
    There is some significance in the role of the Pope as opposed to the Patriarchs of the Orthodox churches, because the position of the Pope tended to counter the role of the state, but this difference applies also to that between the Catholic Church and the Protestant ones. In fact, the difference between the Catholic Church and Orthodox churches (there is no single orthodox church - which is one of many things you seem to have no inkling about) is about the same as that between the Catholic Church and the High Church part of the Anglican Church. To draw any far reaching conclusion about such matters, without knowing even the most basic facts is absurd.

    You are as ignorant in this area as I am of String Theory, but unlike you I do not pretend to know more than I do.

  18. 10% is not nothing but it's a small minority that affects the life and evolution of the society less than those 90% do. Moreover, the Catholics in Ukraine are below 10% (8% or so) - and Orthodox folks in Poland make just 1%. This is a huge difference in demographics that is guaranteed to have consequences in the political life of the countries, too.

    It doesn't matter if you present the differences between the churches as "just as the Pope". The differences are significant, see e.g. these basic 12 examples,


    and if you're ignorant about them, it's just too bad. They're differences about who is divine and who isn't, who may be trusted and who shouldn't, and so on, and all these things markedly affected the evolution of the countries in the recent 2,000 years.

    Believe ten times that your knowledge of irrelevant factoids may replace the gaping holes you have about the basic separation of the world's civilizations but the people who prefer observations and facts will still see that your belief is invalid.

  19. First, Greek Catholics in Poland are counted as catholics, therefore it is difficult to know their true number. They are mostly in the areas neighbouring Belarus and Ukraine. The Orhtodox are counted as a separate religious community.

    The supposedly important differences between Roman Catholicism and Greek Orthodoxy that you listed are completely trivial, about as trivial as the excuse for the original split:


    The best proof that they are trivial is that the Roman Catholic Church did not require the orthodox to change any of their practices (none of the ones you have listed) in order to become a part of the Catholic church. As I wrote above: Orthodox Churches are not considered heretical, unlike all the protestant churches (except the Anglicans). That means, in particular, that a dying catholic can receive last sacraments and absolution from an orthodox priest in no catholic priest is available, but not from any protestant.
    You can claim that the things you have listed in you blog are culturally very significant but actually they are trivialities that have no significant affect on society. What does have a much larger effect is the attitude to the state and the ruler - and in that there have been significant differences. But that is not something you seem to know anything about.

    Anyway, it's better to stop this argument before we get carried away too far. I really do not like these quarrels with you because basically I consider that on all important matters we are allies and it's stupid too quarrel with allies over inessential things. So I restrain on many occasions (including your various provoking comments on the nature of mathematics vs physics) from writing anything - exactly for that reason. But if we quarrel over these kind of things there will eventually be nothing left to discuss.

  20. Dear Lucretius, the previous discussion about the minority religions in Poland was about Polish Orthodox folks, not Greek Catholics, so you have changed the topic once again - you're doing it all the time apparently because it's more comfortable for you to start to talk about yet another, completely different thing - Lvov or Greek Catholics in Poland - than to admit that you're wrong about the essential points.

    The Catholic-Orthodox difference could have looked minor in the past but the differences became the seed of most of the differences between the West and the East (of Europe) we see today.

    For example, the Roman Catholic trust in the Pope is special and could look like a "license to a dictator who will be a hardliner and stop progress". But the actual correlation is exactly the opposite. The Pope's authority is exactly what allowed the Catholic countries to evolve. Take a recent example. The Pope apologized to Galileo, fine. Such changes are possible *exactly* because an individual can do them - perhaps under the influence of the evidence "around". If the authority is collectively given to all the mullahs in a nation, such radical transformations will never materialize because all potentially growing new opinions will be minorities that will be treated as heretics and pushed back to zero influence. With a single Pope, the evolution could occur.

    The influence of this "detail" on the history might fail to be self-evident at the beginning - and it is incomprehensible e.g. to you even today - but it was essential in shaping the history of subcontinents for long centuries. The situation is analogous to genetics. Differences between genes, pieces of DNA, may look harmless and negligible but they're what decides about the ultimate difference between birds and mammals and other things. Differences have implications and when the peaceful exchange of ideas or copulation is suppressed, the differences increase in time.

  21. Well, this time you got to the real point with which I completely agree and I have been suggesting this all along. The orthodox churches in various countries lacking external authority and support from it all fell under the sway of the state and became its tool. The final and extreme step was achieved by Peter the Great, who made all the orthodox clergy employees of the State, who received salaries from it and could be fired at any time. The ruler became not only the ultimate decision maker in secular matters but also in religious ones. In particular, all priests in Russia were required to report anything that heard during confession that was of interest to the authorities and generally did so. The could be no secrets from the Tsar.

    The Bolsheviks obviously try to continue this policy. While persecuting the church, they also controlled appointments to the hierarchy. That's why practically every senior figure in the Russian Orthotox church has been accused of having been a KGB informer, and in most cases this was probably true. It was also one of several reasons for the break-up of the Ukrainian Orhtodox Church (and this time I do not mean the Greek Catholic Church). The Orthodox Church is also split



    On the other hand, in the West, the struggle between the Popes and the kings lead eventually to the separation of Church and state and the acceptance that rulers should not interfere with what the subjects think. In Poland it happened very early, the words that King Sisigmund Augustus addressed in the 16th century to the Polish Parliament "I am not the ruler of human consciences" - are often quoted as the example of the difference between Poland and Russia (which was then ruled by Ivan the Terrible) but can also be viewed as describing the difference between the direction in which catholicism was evolving and the path taken by the Russian Orthodox.

  22. Nice to see some agreement. I decided not to reduce our harmony by further detailed comments, thanks for yours.

  23. If that was his point he should have made that point instead of making the ill informed claim that storms that are named today shouldn't be.

    But yes, the improved technology does create false trends. But nobody says "before national Doppler radar, that short lived tornado out in the middle of nowhere wouldn't have been counted, *therefore it isn't a tornado and shouldn't be counted*"! That's a non sequitur. This is an analogous situation.

    Anyway, if you want to see what I think is the definitive publication on whether there is really an long term trend in tropical storms, see this:


    which confirms that two effects bias the total count and give an illusory long term trend:

    first, undercounts of storms generally due to plain missing them in the ship era (pre satellite) and undercounts of short lived storms likely to have not been considered in the past. Add estimated missed storms back in and take the short ones out, trend disappears.

    But they are all still tropical storms.

  24. One should be careful inferring the properties of counts from the HURDAT data, since we can be pretty confident it underestimates past counts.

    As I mentioned above, this:


    is a good paper on the subject. It shows that there aren't likely real long term trends in counts, thought there remains of course interannual and interdecadal variability.

  25. Interesting take on the situation in the Ukraine from a close and yet distant observer.

    Czechia and Germany/Austria seem astoundingly (to the uninformed visitor) similar in so many aspects: e.g. historical architecture, genetics, food and even some parts of the mindset.
    The language, however, preserved the wonderful special identity of Czechia.

  26. Yes, Barbara, we're sort of brothers across the language groups (it's been also said that Czechs are the Germans among Slavs) but after the last 50 years, Czechs are still the visibly poorer brothers. ;-)

    The language hasn't been threatened for something like 200 years - two centuries ago, all the Czech elite would tend to switch to the German language and only peasants would be the safe carriers of the language. However, that was countered by the Czech National Revival, a pro-Czech movement of reasonably intelligent folks - arguably overcompensated during much of the 19th century. We're elsewhere today.

    Just before I saw your comment, I was reading some news about a $200 fine for a Slovak TV for having aired things in foreign languages which happened to be Czech (some Czech dubbing of a movie was aired). ;-) I think that this silly law about the fines was mainly anti-Hungarian but when it's used against Czech and English, it looks a bit comical, especially if the fines are so ludicrously low. Well, they threatened that they could be higher next time.

    We've agreed with the convention that Slovak is an independent language but some reality is still independent of conventions, including the fact that it's stupid for a Slovak to voluntarily reduce his "sphere of influence" by declaring Czech borderline incomprehensible to him even though he can clearly understand almost all of it. If I could speak German, I would surely love to be viewed as a potential German, too. My German stands at the funny level only, however. But the Slovaks still don't have this pragmatic attitude - the Czech revivalists probably didn't have it, either.

  27. Well, in Hungary, some time ago, they passed a law that says that 60% of all TV content must be European and 50% Hungarian. It sounds to me like is puts the non-Hungarian Europeans at a great disadvantage vs. the Americans.

    Also 35% of all music broadcast on the radio has to be Hungarian. Again, I am not sure if this refers to the number of pieces or (more likely) their duration, but they they must be using a clever computer program to match the lengths of the pieces available, particularly if, as seems natural, repetitions are not allowed.

  28. A useful new dictionary for anyone following events in Kiev. Some things I have known for some time but many are new. I notice they RFERL do not give an accurate translation of ""Януковича в жопу, Україну в Європу!" (they admit it) so if there is a demand for it I (or Lubos) can provide it.


    The East-Europeans still retain a rater rough sense of humour. I am reminded of this popular anti-Putin video on youtube, which was a great hit in Russia a while ago. It is entitled: "Our madhouse is voting for Putin":


  29. But imagine what would happen if Ukraine joined the EU and bulk immigration to Europe was allowed. Brussels would need to embark on a program to build massive numbers of new mental asylums.

  30. The Poles now in the UK, Germany, Ireland etc, and the Bulgarians and Rumanians who will soon follow, would get some serious competition.

    I suppose you must be referring to more mental hospitals for the locals? Don't worry, under Sharia mentally ill people are considered to be possessed by a jin and are treated by exorcism (ruqya). Overall, a great saving for the Health Service.

  31. There's a Turkish-German author, name of Akif Pirincci, who writes the most biting and cutting anti-Islam pieces. He came to Germany as a child, quickly learned the language, and became a successful and prosperous author of novels and screenplays. His vision of Europe's future is hellish, a carbon copy of the dysfunctional muslim societies that keep pumping immigrants into the continent. It will be especially bad for the fools who thought that by importing hostile, barely literate people from backward societies, they could make up for their demographic shortfall. In Pirincci's vision of the future they are reduced to begging on the streets for handouts from their new masters.

  32. How curious! I have known him for many years but only as the author of strange and very good detective novels about ... cats:


    I have always thought of him as a far better writer than this Rawlings woman (although I have never read anything of what she has written as I cannot get through more than a few pages. It's much better as television). The psychological tension of his books reminds me of the magic books by Janusz Korczak that had a big impact on me as a child:


    Well, I see that Pirincci has turned in the direction of more realistic nightmares. I wish they don't come true for him as they did for Korczak.

  33. I once took the annual hurricane counts for the Atlantic back to about 1875 and plotted up the distribution of hurricanes per year. It was nearly a perfect Poisson distribution with a mean of 5+. I then plotted the distribution of time intervals between hurricanes. That was a very good approximation to an exponential as it should be for a Poisson process. These results were presented on McIntyre’s website about five years ago.

    The Pacific hurricanes were harder to pin down as a Poisson process since there are many more per year, about 25, but with only 100 years the distribution looks terrible. However, the distribution of the time intervals between hurricanes also was a very good exponential.

    Since a Poisson process with mean of 5+ describes the Atlantic hurricanes so well, I think that all the fluctuations in hurricane numbers are just a random walk and the mechanism for creating them has been stable for at least the length of time that records have been kept.

  34. You might enjoy this rant from a Canadian on American's sense of geography: