Nature Unleashes a Flood … of Bad Science.

Guest Post by Willis Eschenbach

Recently, Nature Magazine published a paywalled paper called “Human contribution to more-intense precipitation extremes” by Seung-Ki Min, Xuebin Zhang, Francis W. Zwiers & Gabriele C. Hegerl (hereinafter MZZH11) was published in Nature Magazine. The Supplementary Information is available here. The study makes a very strong claim to have shown that CO2 and other greenhouse gases are responsible for increasing extreme rainfall events, viz:

Here we show that human-induced increases in greenhouse gases have contributed to the observed intensification of heavy precipitation events found over approximately two-thirds of data-covered parts of Northern Hemisphere land areas.

Figure 1. Extreme 1-day rainfall. New Orleans, Katrina. Photo Source

There are two rainfall indices which are used in their analysis, called the RX1day and RX5day indices. The RX1day and RX5day indices give the maximum one-day precipitation and five-day precipitation for a given station for a given month. These individual station datasets (available here, free registration required) have been combined into a gridded dataset, called HADEX (Hadley Climate Extremes Dataset) . It is this gridded dataset that was used in the MZZH11 study.

So what’s wrong with the study? Just about everything. Let me peel the layers off it for you, one by one.

Other people have commented on a variety of problems with the study, including Roger Pielke Jr., Andy Revkin,  Judith Curry . But to begin with, I didn’t read them, I did what I always do. I went for the facts. I thrive on facts. I went to get the original data. For me, this is not the HADEX data, as that data has already been gridded. I went to the actual underlying data used to create the HADEX dataset, as cited above. Since they don’t provide a single datablock file with all of the areas (grrrr … pet peeve), I started by looking at the USA data.

And as is my habit, the first thing I do is just to look at the individual records. There are 2,661 stations in the USA database, of which some 731 contain some RX1day maximum one day rainfall data. However, as is usual with weather records of all kinds, many of these have missing data.  In addition, only 9% of the stations contain a significant trend at the 95% confidence level. Since with a 95% confidence interval (CI) we would expect 5% of the stations to exceed that in any random dataset, we’re only slightly above what would be expected in a random dataset. In addition, the number of stations available varies over time..

Now, let me repeat part of that, because it is important.

91% of the rainfall stations in the US do not show a significant trend in precipitation extremes, either up or down.

So overwhelmingly in the US there has been

No significant change in the extreme rainfall.

And as if that wasn’t enough …

Of the remaining 9% that have significant trends, 5% of the trends are probably from pure random variation.

So this means that

Only about 5% of the stations in the US show any significant change in rainfall extremes.

So when you see claims about changes in US precipitation extremes, bear in mind that they are talking about a situation where only ~ 5% of the US rainfall stations show a significant trend in extreme rainfall. The rest of the nation is not doing anything.

Now, having seen that, let’s compare that to the results shown in the study:

Figure 2. The main figure of the MZZH11 study, along with the original caption. This claims to show that the odds of extreme events have increased in the US.

Hmmmm …. so how did they get that result, when the trends of the individual station extreme precipitation show that some 95% of the stations aren’t doing anything out of the ordinary? Let me go over the stages step by step as they are laid out in the study. Then I’ll return to discuss the implications of each step.

1. The HADEX folks start with the individual records. Then, using a complex formula based on the distance and the angle from the center of the enclosing gridcell, they take a weighted station average of each month’s extreme 1-day rain values from all stations inside the gridcell. This converts the raw station data into the HADEX gridded station data.

2. Then in this study they convert each HADEX gridcell time series to a “Probability-Based Index” (PI) as follows:

Observed and simulated  annual extremes are converted to PI by fitting a separate generalized extreme value (GEV) distribution to each 49-year time series of annual extremes and replacing values with their corresponding percentiles on the fitted distribution. Model PI values are interpolated onto the HadEX grid to facilitate comparison with observations (see Methods Summary and Supplementary Information for details).

In other words, they separately fit a generalized three-parameter probability function each to gridcell time series, to get a probability distribution. The fitting is done iteratively, by repeatedly adjusting each parameter to find the best fit. Then they replace that extreme rainfall value (in millimetres per day) with the corresponding probability distribution value, which is between zero and 1.

They explain this curious transformation as follows:

Owing to the high spatial variability of precipitation and the sparseness of the observing network in many regions, estimates of area means of extreme precipitation may be uncertain; for example, for regions where the distribution of individual stations does not adequately sample the spatial variability of extreme values across the region. In order to reduce the effects of this source of uncertainty on area means, and to improve representativeness and inter-comparability, we standardized values at each grid-point before estimating large area averages by mapping extreme precipitation amounts onto a zero-to-one scale. The resulting ‘probability-based index’ (PI) equalizes the weighting given to grid-points in different locations and climatic regions in large area averages and facilitates comparison between observations and model simulations.

Hmmm … moving right along …

3. Next, they average the individual gridcells into “Northern Hemisphere”, “Northern Tropics”, etc.

4. Then the results from the models are obtained. Of course, models don’t have point observations, they already have gridcell averages. However, the model gridcells are not the same as the HADEX gridcells. So the model values have to be area-averaged onto the HADEX gridcells, and then the models averaged together.

5. Finally, they use a technique optimistically called “optimal fingerprinting”. As near as I can tell this method is unique to climate science. Here’s their description:

In this method, observed patterns are regressed onto multi-model simulated responses to external forcing (fingerprint patterns). The resulting best estimates and uncertainty ranges of the regression coefficients (or scaling factors) are analysed to determine whether the fingerprints are present in the observations. For detection, the estimated scaling factors should be positive and uncertainty ranges should exclude zero. If the uncertainty ranges also include unity, the model patterns are considered to be consistent with observations.

In other words, the “optimal fingerprint” method looks at the two distributions H0 and H1 (observational data and model results) and sees how far the distributions overlap. Here’s a graphical view of the process, from Bell, one of the developers of the technique.

Figure 2a. A graphical view of the “optimal fingerprint” technique.

As you can see, if the distributions are anything other than Gaussian (bell shaped), the method gives incorrect results. Or as Bell says (op. cit.) the optimal footprint model involves several crucial assumptions, viz:

•  It assumes the probability distribution of the model dataset and the actual dataset are Gaussian

•  It assumes the probability distribution of the model dataset and the actual have approximately the same width

While it is possible that the extreme rainfall datasets fit these criteria, until we are shown that they do fit them we don’t know if the analysis is valid. However, it seems extremely doubtful that the hemispheric averages of the probability based indexes will be normal. The MZZH11 folks haven’t thought through all of the consequences of their actions. They have fitted an extreme value distribution to standardize the gridcell time series.

This wouldn’t matter a bit, if they hadn’t then tried to use optimal fingerprinting. The problem is that the average of a PI of a number of extreme value distributions will be an extreme value distribution, not a Gaussian distribution. As you can see in Figure 2a above, for the “optimal fingerprint” method to work, the distributions have to be Gaussian. It’s not as though the method will work with other distributions but just give poorer results. Unless the data is Gaussian, the “optimal fingerprint” method is worse than useless … it is actively misleading.

It also seems doubtful that the two datasets have the same width. While I do not have access to their model dataset, you can see from Figure 1 that the distribution of the observations is wider, both regarding increases and decreases, than the distribution of the model results.

This seems extremely likely to disqualify the use of optimal fingerprinting in this particular case even by their own criteria. In either case, they need to show that the “optimal fingerprint” model is actually appropriate for this study. Or in the words of Bell, the normal distribution “should be verified for the particular choice of variables”. If they have done so there is no indication of that in the study.

I think that whole concept of using a selected group of GCMs for “optimal fingerprinting” is very shaky. While I have seen theoretical justifications for the procedure, I have not seen any indication that it has been tested against real data (not used on real data, but tested against a selected set of real data where the answer is known). The models are tuned to match the past. Because of that, if you remove any of the forcings, it’s almost a certainty that the model will not perform as well … duh, it’s a tuned model. And without knowing how or why the models are chosen, how can they say their results are solid?

OK, I said above that I would first describe the steps of their analysis. Those are the steps. Now let’s look at the implications of each step individually.

STEP ONE: We start with what underlies the very first step, which is the data. I didn’t have to look far to find that the data used to make the HADEX gridded dataset contains some really ugly errors. One station shows 48 years of August rains with a one-day maximum of 25 to 50 mm (one to two inches), and then has one August (1983) with one day when it is claimed to have rained 1016 mm (40 inches) … color me crazy, but I think that once again, as we have seen time after time, the very basic steps have been skipped. Quality doesn’t seem to be getting controlled. So … we have an unknown amount of uncertainty in the data simply due to bad individual data points. I haven’t done an analysis of how much, but a quick look revealed a dozen stations with that egregious an error in the 731 US datasets … no telling about the rest of the world.

The next data issue is “inhomogeneities” (sudden changes in volume or variability) in the data. In a Finnish study, 70% of the rainfall stations had inhomogeneities. While there are various mathematical methods used by the HADEX folks to “correct” for this, it introduces additional uncertainty into the data. I think it would be preferable to split the data at the point of the inhomogeneous change, and analyze each part as a separate station. Either way, we have an uncertainty of at least the difference in results of the two methods. In addition, the Norwegian study found that on average, the inhomogeneities tended to increase the apparent rainfall over time, introducing a spurious trend into the data.

In addition, extreme rainfall data is much harder to quality control than mean temperature data. For example, it doesn’t ever happen that the January temperature at a given station averages 40 degrees every January but one, when it averages 140 degrees. But extreme daily rainfall could easily change from 40 mm one January to an unusual rain of 140 mm. This makes for very difficult judgements as to whether a large daily reading is erroneous.

In addition, an extreme value is one single value, so if that value is incorrectly large it is not averaged out by valid data. It carries through, and is wrong for the day, the month, the year, and the decade.

Rainfall extreme data also suffers in the recording itself. If I have a weather station and I go away for the weekend, my maximum thermometer will record the maximum temperature of the two days I missed. But the rainfall gauge can only give me the average of the two days I missed … or I could record the two days as one with no rain on the other day. Either way … uncertainties.

Finally, up to somewhere around the seventies, the old rain gauges were not self emptying. This means that if the gauge were not manually emptied, it could not record an extreme rain. All of these problems with the collection of the extreme rainfall data means it is inherently less accurate than either mean or extreme temperature data.

So that’s the uncertainties in the data itself. Next we come to the first actual mathematical step, the averaging of the station data to make the HADEX gridcells. HADEX, curiously, uses the averaging method rejected by the MZZH11 folks. HADEX averages the actual rainfall extreme values, and did not create a probability-based index (PI) as in the MZZH11 study. I can make a cogent argument for either one, PI or raw data, for the average. But using a PI based average of a raw data average seems like an odd choice, which would result in unknown uncertainties. But I’m getting ahead of myself. Let me return to the gridding of the HADEX data.

Another problem increasing the uncertainty of the gridding is the extreme spatial and temporal variability of rainfall data. They are not well correlated, and as the underlying study for HADEX says (emphasis mine):

[56] The angular distance weighting (ADW) method of calculating grid point values from station data requires knowledge of the spatial correlation structure of the station data, i.e., a function that relates the magnitude of correlation to the distance between the stations. To obtain this we correlate time series for each station pairing within defined latitude bands and then average the correlations falling within each 100 km bin. To optimize computation only pairs of stations within 2000 km of each other are considered. We assume that at zero distance the correlation function is equal to one. This may not necessarily be the best assumption for the precipitation indices because of their noisy nature but it does provide a good compromise to give better gridded coverage.

Like most AGW claims, this seems reasonable on the surface. It means that stations closer to the gridbox center get weighted more than distant stations. It is based on the early observation by Hansen and Lebedeff in 1987 that year-to-year temperature changes were well correlated between nearby stations, and that correlation fell off with distance. In other words, if this year is hotter than last year in my town, it’s likely hotter than last year in a town 100 km. away. Here is their figure showing that relationship:

Figure 3. Correlation versus Inter-station Distance. Original caption says “Correlation coefficients between annual mean temperature changes for pairs of randomly selected stations having at least 50 common years in their records.”

Note that at close distances there is good correlation between annual temperature changes, and that at the latitude of the US (mostly the bottom graph in Figure 3) the correlation is greater than 50% out to around 1200 kilometres.

Being a generally suspicious type fellow, I wondered about their claim that changes in rainfall extremes could be calculated by assuming they follow the same distribution used for temperature changes. So I calculated the actual relationship between correlation and inter-station distance for the annual change in maximum one-day rainfall. Figure 4 shows that result. It is very different from temperature data, which has good correlation between nearby stations and drops off slowly with increasing distance. Extreme rainfall does not follow that pattern in the slightest.

Figure 4. Correlation of annual change in 1-day maximum rainfall versus the distance between the stations. Scatterplot shows all station pairs between all 340 mainland US stations which have at least 40 years of data per station. Red line is a 501 point Gaussian average of the data.

As you can see, there is only a slight relationship at small distances between extreme rainfall event correlation and distance between stations. There is an increase in correlation with decreasing distance as we saw with temperature, but it drops to zero very quickly. In addition, there are a significant number of negative correlations at all distances. In the temperature data shown in Figure 3, the decorrelation distance (the distance where the average correlation drops to 0.50) is on the order of 1200 km. The corresponding decorrelation distance for one-day extreme precipitation is only 40 km …

Thinking that the actual extreme values might correlate better than the annual change in the extreme values, I plotted that as well … it is almost indistinguishable from Figure 4. Either way, there is only a very short-range (less than 40 km) relation between distance and correlation for the RX1day data.

In summary, the method of weighting averages by angular distances used for gridding temperature records is supported by the Hansen/Lebedeff temperature data in Figure 3. On the other hand, the observations of extreme rainfall events in Figure 4 means that we cannot use same method for gridding of extreme rainfall data. It makes no sense, and reduces accuracy, to average data weighted by distance when the correlation doesn’t vary with anything but the shortest distances, and the standard deviation for the correlation is so large at all distances.

STEP 2: Next, they fit a generalized extreme value (GEV) probability distribution to each individual gridcell. I object very strongly to this procedure. The GEV distribution has three different parameters. Depending on how you set the three GEV dials, it will give you distributions ranging from a normal to an exponential to a Weibull distribution. Setting the dials differently for each gridcell introduces an astronomical amount of uncertainty into the results. If one gridcell is treated as a normal distribution, and the next gridcell is treated as an exponential distribution, how on earth are we supposed to compare them? I would throw out the paper based on this one problem alone.

If I decided to use their method, I would use a Zipf distribution rather than a GEV. The Zipf distribution is found in a wide range of this type of natural phenomena. One advantage of the Zipf distribution is that it only has one parameter, sigma. Well, two, but one is the size of the dataset N. Keeps you from overfitting. In addition, the idea of fitting a probability distribution to the angular-distance weighted average of raw extreme event data is … well … nuts. If you’re going to use a PI, you need to use it on the individual station records, not on some arbitrary average somewhere down the line.

STEP 3: Hemispheric and zonal averages. In addition to the easily calculable statistical error propagation in such averaging, we have the fact that in addition to statistical error each individual gridpoint has its own individual error. I don’t see any indication that they have dealt with this source of uncertainty.

STEP 4: Each model needs to have its results converted from the model grid to the HADEX grid. This, of course, gives a different amount of uncertainty to each of the HADEX gridboxes for each of the models. In addition, this uncertainty is different from the uncertainty of the corresponding observational gridbox …

There are some other model issues. The most important one is that they have not given any ex-ante criteria for selecting the models used. There are 24 models in the CMIP database that they could have used. Why did they pick those particular models? Why not divide the 24 models into 3 groups of 8 and see what difference it makes? How much uncertainty is introduced here? We don’t know … but it may be substantial.

STEP 5: Here we have the question of the uncertainties in the optimal fingerprinting. These uncertainties are said to have been established by Monte Carlo procedures … which makes me nervous. The generation of proper data for a Monte Carlo analysis is a very subtle and sophisticated art. As a result, the unsupported claim of a Monte Carlo analysis doesn’t mean much to me without a careful analysis of their “random” proxy data.

More importantly, the data does not appear to be suitable for “optimal fingerprinting” by their own criteria.

End result of the five steps?

While they have calculated the uncertainty of their final result and shown it in their graphs, they have not included most of the uncertainties I listed above. As a result, they have greatly underestimated the real uncertainty, and their results are highly questionable on that issue alone.

OVERALL CONCLUSIONS

1. They have neglected the uncertainties from:

•  the bad individual records in the original data

•  the homogenization of the original data

•  the averaging into gridcells

•  the incorrect assumption of increasing correlation with decreasing distance

•  the use of a 3 parameter fitted different probability function for each gridcell

•  the use of a PI average on top of a weighted raw data average

•  the use of non-Gaussian data for an “optimal fingerprint” analysis

•  the conversion of the model results to the HADEX grid

•  the selection of the models

As a result, we do not know if their findings are significant or not … but given the number of sources of uncertainty and the fact that their results were marginal to begin with, I would say no way. In any case, until those questions are addressed, the paper should not have been published, and the results cannot be relied upon.

2.  There are a number of major issues with the paper:

•  Someone needs to do some serious quality control on the data.

•  The use of the HADEX RX1day dataset should be suspended until the data is fixed.

•  The HADEX RX1day dataset also should not be used until gridcell averages can be properly recalculated without distance-weighting.

•  The use of a subset of models which are selected without any ex-ante criteria damages the credibility of the analysis

•  If a probability-based index is going to be used, it should be used on the raw data rather than on averaged data. Using it on grid-cell averages of raw data introduces spurious uncertainties.

•  If a probability-based index is going to be used, it needs to be applied uniformly across all gridcells rather than using different distributions a gridcell by gridcell basis.

•  No analysis is given to justify the use of “optimal fingerprinting” with non-Gaussian data.

3. Out of the 731 US stations with rainfall data, including Alaska, Hawaii and Puerto Rico, 91% showed no significant change in the extreme rainfall events, either up or down.

4. Of the 340 mainland US stations with 40 years or more of records, 92% showed no significant change in extreme rainfall in either direction.

As a result, I maintain that their results are contrary to the station records, that they have used inappropriate methods, and that they have greatly underestimated the total uncertainties of their results. Thus the conclusions of their paper are not supported by their arguments and methods, and are contradicted by the lack of any visible trend in the overwhelming majority of the station datasets. To date, they have not established their case.

My best regards to all, please use your indoor voices in discussions …

w.

[UPDATE] I’ve put the widely-cited paper by Allen and Tett about “optimal fingerprinting” online here.

About these ads
This entry was posted in Bad science, flooding and tagged , . Bookmark the permalink.

185 Responses to Nature Unleashes a Flood … of Bad Science.

  1. Harry the Hacker says:

    Seems a lot like Garbage in, Garbage Out.

  2. Mark T says:

    Of course their conclusions are not supported… it was published in Nature.

    Mark

  3. tokyoboy says:

    This is one of their (semi)final death rattles?

  4. Charlie A says:

    Have you offered to Nature your services as a one man editorial review board.

    It would be interesting to see the peer review comments for this paper.

  5. ferd berple says:

    A study that said there was no statistically significant change in rainfall would not have been published by Nature. Look at any newspaper. An increase in the rate of plane crashes is a story. A decrease is not.

  6. Colonel Sun says:

    The initial precipitation data has been subjected to such extreme torture that it would appear appropriate for Amnesty International to take up the case.

    /humour

    For each of the points you listed, reproduced below,

    • the bad individual records in the original data

    • the homogenization of the original data

    • the averaging into gridcells

    • the incorrect assumption of increasing correlation with decreasing distance

    • the use of a 3 parameter fitted different probability function for each gridcell

    • the use of a PI average on top of a weighted raw data average

    • the use of non-Gaussian data for an “optimal fingerprint” analysis

    • the conversion of the model results to the HADEX grid

    • the selection of the models

    one question that is not clear to me. Do the authors employ a sensitivity analysis to yield an estimate of the systematic errors associated with performing each of the above? Do they report such errors?

    A general observation: climatology seems to be the only field of science that does not think that error estimates and error bars on data are necessary.

  7. J says:

    I know it is unlikely to happen, but you should actually send this analysis as a comment to Nature and see what happens …

  8. Nylo says:

    Great analysis as usual, Willis. It’s a pleasure to read your guest posts, which are IMHO always among WUWT’s greatest contents.

  9. michel says:

    Willis, very nice piece, knocks the paper into a cocked hat.

    Do you have time to turn your attention on the other paper in this issue, the one about the flooding in the UK in 2000?

    The claim there was that a week or so of autumn flooding was provably connected to rising CO2 levels….. At least, that’s what I think it was saying.

  10. Willis Eschenbach says:

    Colonel Sun says:
    February 20, 2011 at 11:43 pm

    … one question that is not clear to me. Do the authors employ a sensitivity analysis to yield an estimate of the systematic errors associated with performing each of the above? Do they report such errors?

    Basically, no. They do a couple of sensitivity analyses, but they don’t touch the important issues like data quality, model selection, or improper averaging. They also do not do any analysis of whether this data is suitable for the “optimal fingerprint” analysis.

    w.

  11. steven mosher says:

    I knew you’d get a kick out of it.

  12. Willis Eschenbach says:

    Charlie A says:
    February 20, 2011 at 11:28 pm

    Have you offered to Nature your services as a one man editorial review board.

    It would be interesting to see the peer review comments for this paper.

    Yes, it would be very interesting to see the comments. That’s why I have argued elsewhere that peer reviews should be signed, and published electronically when the paper is published.

    w.

  13. jorgekafkazar says:

    Bottom line: Their method doesn’t permit the inclusion of error bars. The result isn’t just garbage, we can’t even tell how bad the garbage might be. Alarmists are so certain that a minuscule increase in CO2 produces warming, rain, dead frogs, etc., that they have to resort to increasingly bizarre statistical methods in an attempt to tease out some sort of signal from chaotic data. Rain is more chaotic than temperature, it would seem, so the validity of this paper is well below that of temperature-based papers. How did this ever get published? Oh, yes. Nature.

  14. Malaga View says:

    A beautifully sharp analysis that cuts right through the mumbo jumbo to reveal witch doctors publishing more voodoo in their house magazine.

    First Witch: When shall we three meet again
    In thunder, lightning, or in rain?

    Second Witch: When the hurlyburly’s done,
    When the battle’s lost and won.

    William Shakespeare, Macbeth, 1.1

  15. SSam says:

    I’m not a stats sort of person… but that read really, really well.

    Thank you for putting it into common language.

  16. Martin Brumby says:

    After torturing the data to that extent, it will obviously confess to whatever the “scientists” want to hear.

    It would have at least been more honest to have said:-
    “It’s all the fault of Man’s CO2. A little Polar Bear told me.”

  17. Manfred says:

    It is annoying, that such poor science slipped again through Nature’s peer review process.

    I associate with co-author Gabriele Hegerl questionable practices within the IPCC, disproven Hockey team “science”, climategate and her few rooms apart neighbour Geoffrey Boulton running the failed Muir/Russell “inquiry”.

  18. Mike Haseler says:

    I was wondering whether there was any basis to your concerns … and then I saw the scatterplot!

    If there is so little correlation between stations – how can anyone create a gridded pattern by averaging local stations?

  19. Willis Eschenbach says:

    steven mosher says:
    February 20, 2011 at 11:52 pm

    I knew you’d get a kick out of it.

    Thanks for the heads-up on the paper, mosh. It was the first time I’ve explored deeply into the terrain surrounding the analysis of data extremes as opposed to the data itself. My previous foray involved peak river flows, shown as Update 12 here. Interesting stuff.

    It also brought me back to “optimal fingerprinting”, which I think I understand, but which I don’t understand the implications of. Does anyone know a clear explication and discussion of the subject? I’ve posted up the Allen and Tett paper on OF here.

    I generally distrust multiple linear regressions, of which optimal fingerprinting is one flavor, because so little of nature is linear and because of the problem of overfitting … yes, multiple linear regressions have their uses, but they are a blunt and often misleading tool.

    w.

  20. Peter Plail says:

    Thanks for your hard work Willis.

    They have done a lot of calculations to get their results. I am appalled that they assume Gaussian distribution of actual and model datasets without checking. Surely it is not a difficult task to verify this pretty fundamental assumption, at least using the actual dataset?

    Another question I keep asking myself is where is the science in this – this is a statistical analysis (and a pretty amateur one, judging from Willis’ analysis).

  21. John Trigge says:

    Peer review, that much lauded and oft quoted as the arbiter of quality, fails again.

  22. I remember reading about this article before, and the SI had this pearl:

    “…There is a sudden drop in observational coverage after 2000. For 2001-2003, fewer than 60% of grid-points have data compared to the 1961-1990 mean…”

    Kinda hard to tell if there’s been an increase in rain, when there’s been a decrease in data.

  23. Roy says:

    I think that the general point that a change in climate will produce changes in the (frequency of particular types of) weather is logically sound!

  24. Ed Zuiderwijk says:

    That “fingerprint method” looks to me like the “methods” used to arrive at the infamous hockey stick. If you introduce a clearly “new statistical technique”, if that’s what it is, you would want to have a reputable statistician have a look at it (and any publisher worth his salt would insist on that).

    If I want to find out if two distributions are statistically significantly different, I use the good old non-parametric Kolmogorov-Smirnov test. I’m sure I know what comes out of that if applied to these data.

  25. Merovign says:

    It is, frankly, difficult for the layman to understand how so many errors can be made in a row, accidentally, by professionals.

    I can only describe it as a statistical horror show. It’s actually kind of depressing.

    Yes, seeing the peer review comments would be interesting. I doubt they will be forthcoming, however. Science as an open process seems to be anachronistic.

  26. Christopher Hanley says:

    They seem to be sidestepping an essential link in their chain of reasoning viz. the global mean temperature.

  27. R.S.Brown says:

    Willis,

    In your top paragraph you use the phrase,

    “extreme rainfall events”

    to describe what MZZH11 is looking at, drawn from the
    HADEX data set from 1950 up to and including 1999.

    However, the authors of the study describe the data with
    the more inclusive phrase,

    “heavy precipitation events”

    to describe the target of their investigation.

    Going through the commentaries, the press releases, and
    particularly the Supplementary Information at:

    http://www.nature.com/nature/journal/v470/n7334/extref/nature09763-s1.pdf

    I don’t see any differentiation between rainfall and snowfall
    as forms of precipitation. Rainfall is immediate in it’s “wetting”
    and can be measured in real time via rain gauges.

    Snowfall, on the other hand, can be reduced down to it’s
    equivalent in rainfall at the time it’s precipitated, but the
    “wetting” effects on the environment don’t occur until
    the actual snow melts.

    The rain that falls on Tuesday is credited to Tuesday. The
    snow that falls on Wednesday may be credited to Wednesday,
    but may nor melt until Saturday, or two Saturdays later.

    2.2 inches of rain over a weekend is a good, strong, but not
    all that heavy a rain event in many U.S. locations. The 22 inch
    dry snow equivalent occuring in most U.S. locations generally
    is considered a heavy precipitation event.

    My point is, this study may be “off” a bit on the basic assumption
    of what’s way down inside the HADEX data set.

  28. Puckster says:

    “Being a generally suspicious type fellow, I wondered about their claim that changes in rainfall extremes could be calculating by assuming they follow the same distribution used for temperature changes.”

    Before submitting this to Nature……….maybe change “calculating” to “calculated”?

    ……..I’m just saying…….just a little proof reading.

    Anyway, just how “overlooked” can submissions get. Anthony, doubtless your submission will live up to the requisite standards.

  29. John Marshall says:

    Looks like a fiddle to me. You could compare extreme rainfall to shoe size, instead of model output and get the same result.
    Models are not science!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

  30. 2kevin says:

    …and in this corner, fighting for for humanity and sa nity as whole; Willis ‘The Eviscerator’ Eschenbach!

  31. kadaka (KD Knoebel) says:

    Editorial Annoyances:

    STEP ONE: [...] I haven’t done an analysis of how much, but a quick look revealed a dozen station…
    stations

    In addition, extreme rainfall data is much harder to quality control than mean temperature data. For example, it doesn’t ever happen that the January temperature at station averages 40 degrees every January but one, when it averages 140 degrees.
    Needs clarity and/or correction. Delete the “at station” and change to “the January temperatures average”?

    Being a generally suspicious type fellow, I wondered about their claim that changes in rainfall extremes could be calculating by assuming they follow the same distribution used for temperature changes.
    Should be calculated

    If I decided to use their method, I would use a Zipf distribution rather than a GEV. The Ziph distribution is found in a wide range of this type of natural phenomena. One advantage of the Zipf distribution is that it only has one parameter, sigma.
    H or F? (Pronunciation-based error, “pf” in German can be the same as “ph”, ex: Pfizer)

    Otherwise, good reading, good analysis.

    I shall now wait for the regulars of Tami’s Troupe to complain “That’s just the US which is less than 2% of the land area,” “If you actually knew statistics you’d see they are perfectly justified…” Etc, ad nauseam.

  32. Mac says:

    Poor data, bad methodologies, flawed science, misleading headlines.

    It would seem that is all you need do to publish in Nature.

  33. Anoneumouse says:

    In Climate Science, to observe something, you have to create it. Now this sounds scarily close to bullshit. But if it is bullshit, then at least it’s bullshit with equations.

    Where A= a quantity and B= a quantity and C=a quantity and BS =Bullshit

    here is the formula that defines bullshit:

    A+B+C = D
    A+B+C+BS=BS
    and BS is not equal to D

    The significance of this formula is that even when you solve the variables A. B and C once you add BS to it your answer is also BS. Simply adding the bullshit factor completely destroys your ability to solve the equation that would otherwise be represented by the value D.

  34. fFreddy says:

    Seems a lot like Garbage in, Garbage Out.

    No, to quote a commenter elsewhere, in climatology it is “Garbage in, Gospel Out”

  35. Richard Telford says:

    only ~ 5% of the US rainfall stations show a significant trend in extreme rainfall. The rest of the nation is not doing anything.

    ——————
    This is a very weak analysis, of the type beloved by climate “sceptics”.

    If time is a weak predictor of extreme rainfall, then only a few individual stations will have a statistically significant trend, perhaps few more that expected from the Type I error rate. But there may still be a highly significant relationship taking the data en-mass.

    Climate “sceptics” like this, because they can pick a record and show that there is no *statistically significant* change, ignoring the aggregate data which may show a highly statistically significant change.

    Here is some R code that illustrates this:

    set.seed(123)
    x=rnorm(100)
    y=as.data.frame(replicate(1000,rnorm(100,x*.05)))#1000 replicate y variables with weak relationship to x
    plot(x,y[[1]])#plot of first y replicate against x
    sum(sapply(y,function(Y)cor.test(x,Y)$p.value)<.05)#number of replicates with a statistically significant trend == 68
    cor.test(rowMeans(y),x)$p.value # p-value of the aggregate data set highly statistically significant

  36. Don V says:

    Willis, starting on page 424 of the Allen and Tett paper they discuss “Consistency checks to detect model inadequacy”. They even give a ” simple test of the null hypotheis” (I suck at statistics and it didn’t look that simple to me). Were you able to determine if MZZH11 ever conducted this “simple test” for using optimal fingerprinting to prove modeled rainfall extremes retrospectively correlated with the
    actual data and therefore could be used to prognosticate? I kept falling asleep at this late hour while wading through the paper. . . . z z z but I couldn’t find any such tests.

  37. Steeptown says:

    Are the authors all like Steig, i.e. not statistical experts? Were statisticians involved at all, including in the peer review process?

  38. Brian H says:

    Heh. “Not even wrong” seems like a suitable ‘peer review’ comment.

  39. Edwin Cottey says:

    An excellent analysis, as always, by Mr Eschenbach. Reading the explanation of the authors’ ‘curious transformation’ of the data, I thought I was reading something written by Michael Mann!!
    Malaga View says it beautifully with a quotation from Shakespeare. The US navy during the last war said it no less elegantly – ‘Bullshit baffles brains’.

  40. tallbloke says:

    “Thinking that the actual extreme values might correlate better than the annual change in the extreme values, I plotted that as well … it is almost indistinguishable from Figure 4. Either way, there is only a very short-range (less than 40 km) relation between distance and correlation for the RX1day data.”

    I had to laugh when I got to your figure 4 Willis. A comprehensive FAIL analysis, thank you.

    This reminds me of the tropospheric hotspot paper (who was it?) which splurged red all over the graph, by including zero among the numbers to be painted red.

  41. On Katrina, here’s a clue: don’t build a city below sea level.

  42. Charlie A says:
    February 20, 2011 at 11:28 pm

    Have you offered to Nature….

    Why did you pick Nature in particular? It is because of the controversy in some of their global warming papers publicized by them? You chose Nature because of their global warming bias?

  43. Roger Longstaff says:

    A very impressive analysis. The cumulative application of approximations that you describe surely leads to a meaningless result.

    I also would be interested in your thoughts on the other Nature paper, where a single SINGLE EXTREME RAINFALL EVENT, IN A single (“cherry picked”?) region of England, 11 years ago (“also “cherry picked”?), was “shown” to have been 50% more likely due to anthropogenic carbon dioxide. Massive parallel processing was required to run the models – presumably a “Monte Carlo” simulation?

    This was widely reported by the UK MSM as further proof of harmful AGW, and by implication to justify the terrible cost that we are paying to mitigate against its effects. Somewhere else on this blog (in a different thread) someone published the raw rainfall data for England, which, to the untrained eye, showed no discerable pattern for the last hundred years.

    All of this sent my BS detector off the scale, but the average response down at the pub was – if all of these clever people used every computer in the world to prove this, it must be right. And Nature used to be a fine publication!

  44. Cementafriend says:

    Willis, I have not checked the actual statistical distribution of rainfall but I can say absolutely that it is not a Gaussian normal distribution. For example a station near where I live (now closed but I have been measuring daily rainfall for a couple of years and have infilled a few missing monthly records to make a complete record back to 1893). The rain is seasonal with three summer months having about five times the three winter months while the other months vary between the high and low months. The average in one of the wet months is about 230mm and the individual months (117 years) range from 5mm to 1380mm (std dev 205mm). For one of the drier months the average is 59mm and the range from 0mm to 158mm (std dev 52mm). Any one can see that this is a skewed distribution and both average and SD do not make much sense. My look at this rainfall data including a check of some daily rainfalls (max of 665mm one day in 1898) shows that there has been no increase in daily or monthly maximum rainfall but possibly a very small increasing trend in the annual total of less than 5 mm in about 1800mm total over the 117years but the record length is not sufficient to really judge.

  45. Guam says:

    @ Michael

    Yes That is EXACTLY what was being claimed, I saw the interview on Sky News regarding this University of Oxford claim (whilst there were caveats) it was quite clear they claimed a risk increase by a factor of 2 due to AGW of the 2000 events in the UK

  46. Juraj V. says:

    I will concentrate on UK data available since ~1750:

    Monthly precipitation for England&Wales
    http://climexp.knmi.nl/data/pHadEWP_monthly_qc.png
    Eyeballing it, nothing special but some statistic tools should be applied.

    Annual precipitation for England&Wales
    http://climexp.knmi.nl/data/pHadEWP_monthly_qc_mean1.png
    Some 60-years long wave patter has emerged; now I remind you that those scientists concentrated on post-1950 trend and “after dozens of climate runs they found out, that only with programmed GH forcing they get the rainfall increase”. Those scientist of course cherry-picked the rising rainfall trend after 1950. Hey, run that junk model of yours since 1750, whether we will see the wave pattern or not! [lotsa profanities self-snipped]

    Of course, maybe the annual totals are not much changing, but the precip is spread more unevenly during the year? Let’s calculate annual standard deviation of monthly values!
    http://climexp.knmi.nl/data/pHadEWP_monthly_qc_sd1a.png
    No bloody increase in precipitation spread, or instability, or irregularities, nothing.. [more self-snippage]

    Of course, maybe the rise in extremes is hidden in the daily data? Let’s see annual standard deviations of daily data:
    England&Wales, only since 1930
    http://climexp.knmi.nl/data/pHadEWP_daily_qc_sd1a.png
    Year 2000 was rather remarkable and there is some trend;
    Scotland: no trend
    http://climexp.knmi.nl/data/pHadSP_daily_qc_sd1a.png
    N. Ireland: trend to more stability
    http://climexp.knmi.nl/daily2longer.cgi
    SE England: no obvious trend
    http://climexp.knmi.nl/data/pHadSEEP_daily_qc_sd1a.png

  47. danbo says:

    Figure 1. Extreme 1-day rainfall. New Orleans, Katrina. Photo Source

    I’m not sure where this came from. But see that large body of water at the top, of the source photo. http://www.cces.ethz.ch/projects/hazri/EXTREMES/img/KatrinaNewOrleansFlooded.jpg?hires It’s called lake Pontchartain. It’s the second largest inland body of saltwater in the US. That’s where the water came from. Although it rained, this was a tidal event. Not a rain event.

  48. Viv Evans says:

    Thank you, Willis, for this entertaining ‘look-under-the-hood’. It is like showing that there is only a lawn mower motor under the hood of the latest, hugely lauded model of a Ferrari – and it is stuttering as well.

    One wonders who did the pal review on this paper …

  49. fenbeagle says:

    Looks like a heavy flood to me. Meanwhile, wading through the mire on the Lincolnshire Fens…

    http://fenbeagleblog.wordpress.com/

  50. ImranCan says:

    Thanks Willis
    I downloaded the paper and also was extremely skeptical that there had even been a statistically significant increase in extreme precipitation. But I am glad you have really gone into it – thanks for that !

  51. John A says:

    I’d be intrigued to know what Steve McIntyre or Ross McKitrick thinks about this analysis…

  52. Solomon Green says:

    Mr. Eschenbach,

    As always a great piece of work. A couple of questions though. You say “In addition, only 9% of the stations contain a significant trend at the 95% confidence level. Since with a 95% confidence interval (CI)…”.

    How do you know that the trend is significant? And at the 5% level? Is the trend significant only by observation or have you assumed a Gaussian distribution? Or have you used your own preferred Zipf distribution? And, if the latter, have you also used it to obtain your 95% confidence limit?

  53. Chris H says:

    Nature used to be the premier scientific journal with rigorous peer review. To have a paper published in Nature was the passport to fame, tenure and grants. Sadly, the journal seems to have become a mere propaganda vehicle.

    The misuse of statistical processes you highlight in this paper shows that either the reviewers are so fixated on the conclusions which fit their own preconceived notions or they simply do not understand the methodology. If they didn’t, did they tell the editor? Did the editor take specialist statistical advice? One suspects strongly the answer to both those questions is “no”.

    I am sure that peer review retains it’s integrity in some areas of scientific endeavour but “climate science” is clearly not one of them.

  54. Cassandra King says:

    If I understand their methodology correctly the team has looked for an alternate CAGW signal because global temperatures are not doing what they should and the scope for adjustments has reached its end and in finding this alternate round peg they have tried to smash it into the square hole of reality by starting out with the determination and then trying to find and fit and adjusts methods and data to support their case. Observation versus modelling and data extrapolation?

    The CAGW theorists believe they need to show the ordinary people an easy to understand easy to digest and simple visual cue, its easy to see rain and rainfall is easy to understand. We see the attempt to find an anthropogenic signal, any signal will do, is taking ever more desperately bizarre forms, the data and causal links ever more tenuous.

  55. Roy says:
    February 21, 2011 at 12:58 am
    I think that the general point that a change in climate will produce changes in the (frequency of particular types of) weather is logically sound!

    Agreed, but the problem here (and with the Oxford study about flooding) is that they both take the conclusion that their studies show the increased risk of precipitation/flooding is as a result of co2 forcings, where as all they actualy show (if we accept them as correct, which is a big if), is an increased risk due to climate change. They do not prove correlation with co2 and CC.

    This strikes me as a major problem with a lot of alarmist papers/studies in that they always assume that co2 is the only possible cause. Thus any changes in weather/flood/draught patterns are proof of AGW.

  56. Mycroft says:

    Afraid to say the horse has bolted on this one,the paper was aired on the BBC’s
    6 o’clock news Friday 18/2 evening.think it was Shukman frothing at the mouth over it.
    So much for the IPCC stance of not be able to say that any extreme weather event should/could be attributed to AGW!!
    i seem to remember it being put down to the jet stream again??

  57. Peter Plail says:

    Richard Telford says:
    February 21, 2011 at 1:55 am

    Gosh, Richard. I wish I was clever like you. I tried to get cleverer when I first started getting concerned about global warming, and I went to real Climate because they claimed to be the fount of all climate knowledge. Their condescension rapidly turned me off.

    What I found useful here (on WUWT) was the way this community helped people who didn’t understand. They didn’t talk down, and where the author of a post was too technical for me, I usually found illumination from the comments of bloggers.

    I take it that you are not persuaded by Willis’ analysis of this paper, but making vague assertions and flaunting a bit of code leaves me thinking that you are simply trying to show off. Climate thickies like me would value some explanation – for all I know you may have a valid point – but if you can’t learn to communicate in normal language then pray, keep your thoughts to yourself.

  58. Geoff Sherrington says:

    Australia’s new Commissioner for Climate, Tim Flannery, was on National TV a few minutes ago quoting both this paper and the one about the UK year 2000 floods. He said they supported man-made climate change as linked to the Queensland floods of recent months.

    Opposition politician Senator Barnaby Joyce (who would be Rupublican in US terminology) asked Prof Flannery about his main doctoral discipline, which was palaeontology. The green-leaning host Tony Jones, then asked Sen Joyce if he would draw Prof Flannery into discussions in his new role. Joyce replied that he would call him in if palaeontology was to be discussed……

    From Willis’ analysis of this paper, it seems that once again, statisticians should be employed more often by climate workers.

  59. Beesaman says:

    Well done Mr.Eschenbach, a pity you were not on the review panel!
    Is it me or does there seem to be great haste to publish doom and gloom reports, papers and articles recently. There must be someone, somewhere tracking this. Has this output of doom gone exponential yet? Where will the tipping point be?

  60. Gre McLaughlin says:

    Please do not use a moving average when trying to fit temperature data. As a statistican and meteorologist , there is far too much variation in the data (Monthly temperatures) to support the use of a moving average. In fact, the variation in the data does not permit adequate forecasting abilities (although it does permit trending applications). This is why climate predictions of temperature deviations (changes) are somewhat meaningless. It would be better to select a time period, when variations demonstrate randomness, to then use an average or predict a trend. Also, using a single number or even a small set of numbers to predict a trend is meaningless if the variation (deviation month to month, season to season, etc.) is non-random.

  61. Allanj says:

    The heaviest rains tend to occur in unstable air masses that tend to be local, e.g. thunderstorms. Steady light rains tend to be from wide area stable air masses. It should be clear that extreme rain events will not correlate well with distance.

    Also errors in reporting will always tend high. You may report a 4mm rain as 40mm but very unlikely as negative 40mm.

    As always, your contribution is greatly appreciated.

  62. Juraj V. says:

    There is one hook inside: now it is not the “warmer climate causing extremes” claim, but directly the pure anthropogenic forcing itself! This flip in argumentation is not surprising, since Northern hemisphere cools down big time, being bellow 1990 at the moment:
    http://climexp.knmi.nl/data/icrutem3_hadsst2_0-360E_30-90N_n_su_1990:2011a.png

    So first it was “CO2 warms planet and thus extremes will increase” but now expect reasoning “CO2 directly causes extremes even the temperature does not rise, because temperature was never that important, but the forcing, the forcing”.

  63. Wade says:

    I’m confused. NOAA just had a study that said global warming was going to cause more desertification.
    http://wattsupwiththat.com/2011/02/19/noaas-compendium-of-climate-catastrophe/

    More atmospheric dust from global desertification could lead to increases of harmful bacteria in oceans, seafood.

    And now this study in Nature says global warming is causing more rain. Which is it?

  64. Alberta Slim says:

    Nature Mag’s article is a good example of the old saying:
    Liars go “figure” [MZZH11]
    Figures don’t lie [Willis's].
    IMHO

  65. David says:

    I need help finding PAST weather related extreme events world wide. Wik is not a good resource. I am looking for detailed reports of drought heatwaves storms , cyclones hurricanes, extreme cold, etc. Here is a list of what I have found so far for the 1930 to 1936 or so period I am researching. China, Russia US and Canada are covered, but little else. Thanks in advance.

    1930 May 13th Farmer killed by hail in Lubbock, Texas, USA; this is the only US known fatality due to hail.
    1930 June 13th 22 people killed by hailstones in Siatista, Greece.

    1930 Sept 3rd Hurricane kills 2,000, injures 4,000 (Dominican Republic).
    1930s Sweden The warmest decade was the 1930s, after which a strong cooling trend occurred until the
    1970s INTERNATIONAL JOURNAL OF CLIMATOLOGY http://onlinelibrary.wiley.com/doi/10.1002/joc.946/abstract
    1930 Russian heat wave in the 1930′s, for the decade was 0.2 degrees below 2000 to 2010 heat wave.
    1930 set 3 all time HIGHEST state temperatures, Delaware, 110F Jul. 21, Kentucky, 114 Jul. 28, Tennessee 113 Aug. 9, and one all time LOWEST state record, Oklahoma -27 Jan. 18. About 400% more then a statistical average.
    1931 set two highest State temp ever, FL, 109 Jun. 29, and HI, 109 Jun. 29
    1931 Europe LOWEST temp ever in all of Europe −58.1 °C (−72.6°F)
    1931 The 20th centuries worst water related disaster was the Central China flooding of 1931, inundating 70,000 square miles and killing 3.5-4 million people.
    1931 July Western Russia heat wave 6 degrees F monthly anomaly above normal, 2nd warmest on 130 year record. Decade of 1930 to 1940 within 0.2 degrees of 2000 to 2010 western Russia July

    1931 Sept 10th The worst hurricane in Belize Central America history kills 1,500 people.

    1932 TORNADO OUTBREAK SEVERE 1932, March 21 Alabama 268 DEAD
    1932 November 9th Santa Cruz Del Sur Cuba category 5 hurricane 2,500 dead.
    1932 Madagascar cyclone crosses Reunion Island 35,000 homeless 45 dead.
    1932 June 19th Hailstones kill 200 in Hunan Province, China
    1932 / 33 Soviet famine. 7 to 14 million. Mostly human caused, but drought and low crop yields in 1931 and 32 contributed.
    1933 Sept Cat 3 Florida landfall.
    1933 4 LOWEST state temp ever were recorded in Oregon -54 Feb. 10, Texas -23 Feb. 8,
    Vermont -50 Dec. 30, Wyoming -66 Feb. 9
    1933 February 6 Highest recorded sea wave (not tsunami), 34 metres (112 feet), in Pacific hurricane
    1933 Highest temp ever in SWEDEN 38.0 °C (100.4 °F) tied in 2009
    1933 Lowest temp ever recorded in ASIA −68 °C (−90 °F) tied in 02 and 06
    1933 NORTH KOREA LOWEST temp ever North Korea −43.6 °C ( -46.48°F)
    1933 August 11th Highest World Temperature ever reaches 136 degrees F (58 degrees C) at San Luis Potosí, Mexico (world record).
    1933 Nov 11th Great Black Blizzard” first great dust storm in the Plains of the USA.

    1934 May 11th Over two days, the most severe dust storm to date in the USA sweeps an estimated 350 million tons of topsoil from the Great Plains across to the eastern seaboard.
    1934 Fastest recorded with an anemometer outside of a tropical cyclone: 372 km/h (231 mph) sustained 1-minute average; Mount Washington, New Hampshire,
    Michigan -two states recorded their highest ever temperature both 118 degrees Idaho and Iowa, and two states recorded their lowest ever temperatur Michigan -51 and New Hampshire -47
    1934 LOWEST temp ever Singapore 19.4 °C (66.9 °F)
    1934 Typhoon strikes Honshu Island, Japan, kills 4,000

    1935 Ifrane Morocco, LOWEST temperature continent of Africa ever recorded, minus 11
    1935 Florida, A CAT ONE HURICANE AT LANDFALL.
    1935 Nepisiguit Falls, New Brunswick 39.4 °C 12th highest temp ever in Canada.
    1935 Collegeville, Nova Scotia 38.3 °C 15th highest temp ever in Canada.
    1935 Iroquois Falls, Ontario −58.3 °C 5th lowest temp ever in Canada.
    1935 Western Russia, 9th coldest July in 130 years.
    1935 145,000 dead 1935 Yangtze river flood China
    1935 August 1935 and 36 two typhoons hit Fukien province in China, hundreds dead.
    1935 Labor Day hurricane one of the most intense hurricanes to make landfall in U.S. in recorded history. More than 400 people were killed. 185 MPH sustained winds
    1935 Hati 21 October: hurricane in Sud and Sud-Est départements. 2,000 people perished.

    1936 HIGHEST state temperature ever recorded in Nebraska 118 Jul. 24, New Jersey 110 Jul. 10, North Dakota 121 Jul. 6, Oklahoma 120 Jun. 27, Pennsylvania 111 Jul. 10, South Dakota 120 Jul. 5, Virginia 112 Jul. 10, Wisconsin 114 Jul. 13, Arkansas 120 Aug. 10, 1936, Indiana 116 Jul. 14, ever recorded Kansas 121 Jul. 24, Louisiana 114 Aug. 10, Maryland 109 Jul. 10

    1936 TORNADO outbreak April 5-6 Mississippi and Georgia 436 dead

    1930 to 1936 20 Twenty state record all time HIGHEST in 6 year period plus 7 were tied ONLY in the same 6 year period. 9 record Lowest in same period. Contrast that to 5 highs set in 1990 – 2000 all 5 in 1994. And 5 lows in the same period ten year period.
    Six of Canada’s highest ever records were set in the same period.
    1936 Bay of Bengal Myanmar May 1st cyclone 72,000 homes lost 360 dead
    1936 Drought related famine in China, five million dead. (
    NOAA’S TOP GLOBAL WEATHER, WATER AND CLIMATE EVENTS OF THE 20 TH CENTURY)
    1936 July 11th St. Albans, Manitoba 2nd highest temp ever in Canada 44.4 C

    1936 Northeast Flood – Spring 1936
    Rain concurrent with snowmelt set the stage for this flood. It affected the entire state of New Hampshire.[17] … In all, damage totaled US$113 million (1936 dollars), and 24 people were killed.
    1937 record state lowest temp California -45 Jan. 20
    1937 state LOWEST record Nevada -50 Jan. 8
    1937 January Ohio/Great Miami River Flood
    Two days later, the Ohio River crested in Cincinnati at a record 24.381 m (79.99 ft). Flooding in the city lasted 19 days…… Damages totaled US$20 million (1937 dollars).[23]
    1937 Highest recorded temp in Canada 45 °C (113 °F) Midale

  66. Jit says:

    Willis:

    9% of the raw data were trended. As you say, we expect 5% to be trended at 95% level, but we expect 2.5% to be + trended and 2.5% – trended. What percentage of the series showed +ve and -ve trends?

    Also, don’t quite follow how they do the grid averages. How big are the gridcells? Are they really using data from 2000km away in the calculation of cell values in gridcells that are much smaller (when spatial correlation dies off so fast)?

    Why have gridcells at all – why not just interpolate all available series and calculate standard errors based on some sort of cross-correlation? (And limit contributing series to those within the range of +ve spatial correlation)

  67. David says:

    Help, looking for worldwide historic records of extreme weather, storms, typhoons, hurricanes tornados, floods, droughts etc. North America China, and Russiaa are somewhat covered, but none but the US in detail. The period is 1930 to 1936 or there about. This is what I have so far…Thanks in advance

    1930 May 13th Farmer killed by hail in Lubbock, Texas, USA; this is the only US known fatality due to hail.
    1930 June 13th 22 people killed by hailstones in Siatista, Greece.

    1930 Sept 3rd Hurricane kills 2,000, injures 4,000 (Dominican Republic).
    1930s Sweden The warmest decade was the 1930s, after which a strong cooling trend occurred until the
    1970s INTERNATIONAL JOURNAL OF CLIMATOLOGY http://onlinelibrary.wiley.com/doi/10.1002/joc.946/abstract
    1930 Russian heat wave in the 1930′s, for the decade was 0.2 degrees below 2000 to 2010 heat wave.
    1930 set 3 all time HIGHEST state temperatures, Delaware, 110F Jul. 21, Kentucky, 114 Jul. 28, Tennessee 113 Aug. 9, and one all time LOWEST state record, Oklahoma -27 Jan. 18. About 400% more then a statistical average.
    1931 set two highest State temp ever, FL, 109 Jun. 29, and HI, 109 Jun. 29
    1931 Europe LOWEST temp ever in all of Europe −58.1 °C (−72.6°F)
    1931 The 20th centuries worst water related disaster was the Central China flooding of 1931, inundating 70,000 square miles and killing 3.5-4 million people.
    1931 July Western Russia heat wave 6 degrees F monthly anomaly above normal, 2nd warmest on 130 year record. Decade of 1930 to 1940 within 0.2 degrees of 2000 to 2010 western Russia July

    1931 Sept 10th The worst hurricane in Belize Central America history kills 1,500 people.

    1932 TORNADO OUTBREAK SEVERE 1932, March 21 Alabama 268 DEAD
    1932 November 9th Santa Cruz Del Sur Cuba category 5 hurricane 2,500 dead.
    1932 Madagascar cyclone crosses Reunion Island 35,000 homeless 45 dead.
    1932 June 19th Hailstones kill 200 in Hunan Province, China
    1932 / 33 Soviet famine. 7 to 14 million. Mostly human caused, but drought and low crop yields in 1931 and 32 contributed.
    1933 Sept Cat 3 Florida landfall.
    1933 4 LOWEST state temp ever were recorded in Oregon -54 Feb. 10, Texas -23 Feb. 8, Vermont -50 Dec. 30, Wyoming -66 Feb. 9
    1933 February 6 Highest recorded sea wave (not tsunami), 34 metres (112 feet), in Pacific hurricane
    1933 Highest temp ever in SWEDEN 38.0 °C (100.4 °F) tied in 2009
    1933 Lowest temp ever recorded in ASIA −68 °C (−90 °F) tied in 02 and 06
    1933 NORTH KOREA LOWEST temp ever North Korea −43.6 °C ( -46.48°F)
    1933 August 11th Highest World Temperature ever reaches 136 degrees F (58 degrees C) at San Luis Potosí, Mexico (world record).
    1933 Nov 11th Great Black Blizzard” first great dust storm in the Plains of the USA.

    1934 May 11th Over two days, the most severe dust storm to date in the USA sweeps an estimated 350 million tons of topsoil from the Great Plains across to the eastern seaboard.
    1934 Fastest wind speed recorded with an anemometer outside of a tropical cyclone: 372 km/h (231 mph) sustained 1-minute average; Mount Washington, New Hampshire,
    Michigan -two states recorded their highest ever temperature both 118 degrees Idaho and Iowa, and two states recorded their lowest ever temperatur Michigan -51 and New Hampshire -47
    1934 LOWEST temp ever Singapore 19.4 °C (66.9 °F)
    1934 Typhoon strikes Honshu Island, Japan, kills 4,000

    1935 Ifrane Morocco, LOWEST temperature continent of Africa ever recorded, minus 11
    1935 Florida, A CAT ONE HURICANE AT LANDFALL.
    1935 Nepisiguit Falls, New Brunswick 39.4 °C 12th highest temp ever in Canada.
    1935 Collegeville, Nova Scotia 38.3 °C 15th highest temp ever in Canada.
    1935 Iroquois Falls, Ontario −58.3 °C 5th lowest temp ever in Canada.
    1935 Western Russia, 9th coldest July in 130 years.
    1935 145,000 dead 1935 Yangtze river flood China
    1935 August 1935 and 36 two typhoons hit Fukien province in China, hundreds dead.
    1935 Labor Day hurricane one of the most intense hurricanes to make landfall in U.S. in recorded history. More than 400 people were killed. 185 MPH sustained winds
    1935 Hati 21 October: hurricane in Sud and Sud-Est départements. 2,000 people perished.

    1936 HIGHEST state temperature ever recorded in Nebraska 118 Jul. 24, New Jersey 110 Jul. 10, North Dakota 121 Jul. 6, Oklahoma 120 Jun. 27, Pennsylvania 111 Jul. 10, South Dakota 120 Jul. 5, Virginia 112 Jul. 10, Wisconsin 114 Jul. 13, Arkansas 120 Aug. 10, 1936, Indiana 116 Jul. 14, ever recorded Kansas 121 Jul. 24, Louisiana 114 Aug. 10, Maryland 109 Jul. 10

    1936 TORNADO outbreak April 5-6 Mississippi and Georgia 436 dead

    1930 to 1936 20 Twenty state record all time HIGHEST in 6 year period plus 7 were tied ONLY in the same 6 year period. 9 record Lowest in same period. Contrast that to 5 highs set in 1990 – 2000 all 5 in 1994. And 5 lows in the same period ten year period.
    Six of Canada’s highest ever records were set in the same period.
    1936 Bay of Bengal Myanmar May 1st cyclone 72,000 homes lost 360 dead
    1936 Drought related famine in China, five million dead. (
    NOAA’S TOP GLOBAL WEATHER, WATER AND CLIMATE EVENTS OF THE 20 TH CENTURY)
    1936 July 11th St. Albans, Manitoba 2nd highest temp ever in Canada 44.4 C

    1936 Northeast Flood – Spring 1936
    Rain concurrent with snowmelt set the stage for this flood. It affected the entire state of New Hampshire.[17] … In all, damage totaled US$113 million (1936 dollars), and 24 people were killed.
    1937 record state lowest temp California -45 Jan. 20
    1937 state LOWEST record Nevada -50 Jan. 8
    1937 January Ohio/Great Miami River Flood
    Two days later, the Ohio River crested in Cincinnati at a record 24.381 m (79.99 ft). Flooding in the city lasted 19 days…… Damages totaled US$20 million (1937 dollars).[23]
    1937 Highest recorded temp in Canada 45 °C (113 °F) Midale

  68. Geoff Sherrington says:

    Further to the above on Australia’s new Commissioner for climate, the statement was made that 2 recent papers had shown a link between flooding and man-made climate change. This was amid the Queensland flood context. We can presume these papers to be
    Pardeep Pall, Tolu Aina, Dáithí A. Stone, Peter A. Stott Toru Nozawa Arno G. J. Hilberts, Dag Lohmann & Myles R. Allen, 2011: Anthropogenic greenhouse gas contribution to flood risk in England and Wales in autumn 2000. Nature vol 470, pp 382–385 DOI:doi:10.1038/nature09762

    Seung-Ki Min et al 2011, Xuebin Zhang, Francis W. Zwiers, Gabriele C. Hegerl 2011: Human contribution to more-intense precipitation extremes, Nature, vol 470 , pp 378–381

    There were heavy Spring rains in Queensland in late Dec 2010 to Jan 2011. Several people including the above are spreading the story that high sea surface temperatures in the few months before the rains prepared the way.

    Here is a graph of SST, Spring, the Coral Sea, where the hot rains were supposed to come from.
    http://www.geoffstuff.com/BOMsst_cor_0911_11969.png

    Here is a graph of rainfall over Northen Australia land. I cannot find a map to match rain over the sea. If you can see a correlation between high SST in Spring and heavy northern rainfall, you have a better stats pack than I do.

    Remember that Spring down under is the few months before Christmas.

    Even from NOAA global maps, the S-O-N-D-J period in recent years has has Coral Sea SSTs that can be described as

    2005 hot
    2006 cold
    2007 warmish
    2008 hot
    2009 average
    2010 3 lukewarm months (S N J) and 2 average to coolish (O, D).

    Like Willis above, one gets a clearer story by delving into the actual data. These are not a 1:1 match, but they are indicative of no significant correlation. So where is the hand of man in all this? Drawing graphs, I suspect.

  69. David says:

    Peer review that says CO2 not causing extreme events…

    Over the period of 1965–2008, the global TC activity, as measured by storm days, shows a large amplitude fluctuation regulated by the ENSO and PDO, but has no trend, suggesting that the rising temperature so far has not yet an impact on the global total number of storm days.” Wang, B., Y. Yang, Q.‐H. Ding, H. Murakami, and F. Huang, 2010. Climate control of the global tropical storm days (1965–2008). Geophysical Research Letters,

    “(1) There is no significant overall long-term trend common to all indices in cyclone activity in the North Atlantic and European region since the Dalton minimum.
    Bärring and Fortuniak, 2009 International Journal of Climatology,

    “Over the past 24 yr, the land falling tropical cyclones clearly show variability on inter-annual and inter-decadal time scales, but there is no significant trend in the landfall frequency. from Zhang et al., 2009

    Chan and Xu write “An important finding in this part of the study is that none of the time series shows a significant linear temporal trend, which suggests that global warming has not led to more landfalls in any of the regions in Asia.” from Chan and Xu, 2009 Proceedings of the Royal Society A, 465, 3011-3021.

    Phillipines 1902 – 2005 Annual TLP from 1902 to 2005 using the two definitions shows dominant periodicity of about 32 years before 1940 and of about 10–22 years after 1945; however, no trend is found.” Chan and Xu, 2009 International Journal of Climatology, 29, 1285-1293.

    The 1900–01 to 2006–07 trends in the annual percentage of high- and low-extreme snowfall years for the entire United States are not statistically significant.”
    Sorrel, P., B. Tessier, F. Demory, N. Delsinne, D. Mouaze. 2009.

    France, …no evidence is found of any increase in the frequency or intensity of storms, and in fact, the large storms of southern France seemed more frequent more than 100 years ago. Sabatier, P., L. Dezileau, M. Condomines, L. Briqueu, C. Colin, F. Bouchette, M. Le Duff, and P. Blanchemanche. 2008. Reconstruction of paleostorm events in a coastal lagoon (Hérault, South of France). Marine Geology,

    Analyses show that although economic losses from weather related hazards have increased, anthropogenic climate change so far did not have a significant impact on losses from natural disasters. The observed loss increase is caused primarily by increasing exposure and value of capital at risk. Laurens M. Bouwer Bulletin of the American Meteorological Society 2010

  70. Richard M says:

    I first thing that hit me when I first read about this paper were the dates … 1951-1999. Did the world stop measuring rainfall in 1999? Why not 2009? Why not start in 1900? Does anyone else feel like this entire study was likely another alarmist cherry picking exercise?

  71. Bernie says:

    Willis:
    Nice job. The Nature editorial board appears to have some more explaining to do.
    Do you know if there have been verifications of the original Hansen and Lebedeff (1987) results? I realize that precipitation is likely to be different from temperature but the difference in your analysis from H&L is so dramatic that it seems to me to be worth verifying – if for no other reason than the passing of 25 years. Also, I would assume that the analysis of the satellite data would verify any findings with respect to temperature.

  72. ozspeaksup says:

    one word only- is enough to know its shonky…Hadley.
    we have some fool here in aus saying that WA drought is agw, but the east coast cant be proven to be agw…I suspect hes looking for a Hadley or other handout

  73. Jose Suro says:

    Hello Willis,

    Another masterfully exposed, infinitely complex statistical house of cards. Very well done, and thanks so much for your time spent on this. A number of questions immediately came to mind:

    Forget man-made influences for a moment, lets not even go that far.

    Is there any proof that these rain indexes actually correlate to all past flood events, especially when gridded?

    I can think of a couple of problems there…..

    Are there causal factors in past flood events other than an immediate precipitation event?

    How do they handle a correlation when the precipitation event takes place a very large distance away from the flood event?

    sarc on/ Do they take into account wind caused waves slamming onto earthen levies until they collapse thus flooding the lower terrain (the Katrina event in Figure 1) ? /sarc off :).

    Best,

    Jose

  74. Latitude says:

    I hope people have finally figured out that peer review is not what they think it is….

    ..and that having something published, does not mean it’s right

  75. BFL says:

    Looks like the Ioannidis Demon (John P.) at work again. Common in th AGW crowd.

  76. Ken Harvey says:

    Gridding a temperature measurement over distance might be argued to have some validity. It would be a difficult argument to win, but doing the same thing with rainfall? Absurd. Precipitation extremes can have a span of only a couple of miles or so.

    It seems to me to be time to prohibit anyone from publishing statistical analysis that is not supported by an “audit certificate” provided by a registered and named statistician.

  77. richard verney says:

    Roy says:
    February 21, 2011 at 12:58 am
    I think that the general point that a change in climate will produce changes in the (frequency of particular types of) weather is logically sound!
    ////////////////////////////////////////////////////////////
    I would agree with the general thrust of that premise.
    The issue is whather 0.5C, or o.9C, or 1.5C or 2C etc amounts to a change in climate. It is hard to see that a change of no more than 0.8C (that appears to be the max warming of the last century even if the temp record is reliable) amounts to a significant change in climate and it is therefore extremely unlikely that the present change is culpable at this stage.

    Good post Willis.

  78. MattN says:

    It is becoming easier and easier to debunk the crapscience coming from these guys. If the study had used the exact same methodology and came to the exact opposite conclusion, what are the chances it sees publication? Zero.

  79. Vince Causey says:

    Excellent analysis by Willis again. Unfortunately, the whole point of these flawed alarmist papers is not to produce sound science, but to allow the media to scream ‘science shows extreme rainfall is caused by man.’ They know full well that whilst the sceptics will pull it apart strand by strand, the message has already been broadcast, and to achieve their ends – legally binding carbon mitigation legislation – that is all that matters.

  80. pyromancer76 says:

    Like Viv Evans (3:39), I enjoyed this entertaining look under the hood. A lot of hoods need to be lifted more often — rust and gunk and sabotage found everywhere. I guess this means public, signed peer review comments at the very least. Thanks for this exceptional effort, as usual, Willis.

    I agree with Manfred, 12:34 am:
    It is annoying, that such poor science slipped again through Nature’s peer review process.
    I associate with co-author Gabriele Hegerl questionable practices within the IPCC, disproven Hockey team “science”, climategate and her few rooms apart neighbour Geoffrey Boulton running the failed Muir/Russell “inquiry”.”

    Both Nature and Science enterprises are run now as last ditch efforts to save their (and their corporate-government-academic sponsors) cushy jobs, incomes, benefits, pensions on taxpayers’ generosity. Since there is so much lucre involved, they will fight to the last breath — like the “public servant” unions in Wisconsin and other states in the U.S.

    We must be there to take down their names and put them into a “shamed” pool of pseudo-scientists (or authoritarian unionists as compared to “democratic” unions) who should never be hired again. I am tired of payin’ for ‘em.

  81. Peter Miller says:

    I am not sure if Nature is as unscientific as Greenpeace – that would be very difficult – so it was like a breath of fresh air to hear this from one of Greenpeace’s founders:

    http://www.stockwatch.com/News/Item.aspx?bid=Z-C:*CURRENT-1810801&symbol=*CURRENT&news_region=C

  82. Craig Loehle says:

    It’s even worse than you say. It is known that the models are very bad at simulating historic rainfall at regional/local scales. There have been issues over time with rain gages and changes in methods, and rain data in many parts of the world are just awful. But it is just unbelievable that they don’t bother to check their assumptions of correlation with distance and normality etc and especially what is the null about number of stations that would show a trend by chance. Simply appalling. Furthermore, their test doesn’t show what they claim: if it is getting warmer naturally vs by GHG, rain will go up either way (as it seems to have done) and the models will simulate it going up without that proving that it is due to GHG.

  83. grayman says:

    Willis, great read. The grid cell distribution they use how big are the cells? Here in Texas, it can rain in small or large areas, as anywhere else, but most of the time i have seen it rain on one side of a fence line and not the other, so IMO i do not see this being a very good way of saying that there are more exreme rain events for any particular areas much less N.H. or global. And one more question,why do they take measured data and put in a model instead of actually using OBSERVED and measured data and using it? Sorry if the last is confusing because it is confusing me also. Models=GIGO!

  84. Bill Illis says:

    Here is a scatterplot of HadUKP England and Wales precipitation (a fully quality controlled dataset) versus HadCET temperatures back to 1766.

    Technically, there should be a trend in precipitation of 2% to 3% per 1.0C increase in temperatures.

    There is no trend in this data and no evidence that global warming is causing more precipitation in the UK.

    http://img249.imageshack.us/img249/1669/engwalesprecipvshadcetk.png

    Either the temperature data is wrong, the precipitation data is wrong or the Clausius Clapeyron relation is wrong (on a local level at least – and if that is the case, the studies are based on a false premise to start with).

    There is also no trend in daily precipitation in England and Wales back to 1931 (over 29,000 individual datapoints), a period when temperatures have supposedly increased by 0.8C in the UK.

    http://img101.imageshack.us/img101/4800/dailyrainfallenglandwal.png

  85. Rick says:

    On our farm, located on the western Canadian prairie, we have helped along with other neighbors, to record the observed rainfall for our area for the last 35 years. These records are kept for the Agriculture dept. of our province and cover the growing period of April 1 to Oct.30. Our son, with an interest in numbers, decided to plot these records, in an attempt to discern a pattern. He simply straight lined the average and plotted the monthly totals and yearly totals. Predictably the tracking line jumped above and below the average with huge month to month and year to year variation. Even a generally dry year would surprise you with a spike in monthly rainfall or likewise a wet year would produce a month with very little rain. There was less rain in the dryer 1980′s but not every year in the 1980′s. There has been more rain in the last decade but not every year of the last decade. We have been unable to discern any pattern. The only thing that we could conclude from our little exercise was the erratic and unpredictable nature of rainfall on the Canadian Prairie but then everyone around here knew that already.

  86. jrwakefield says:

    Rainfall data is all suspect to begin with. Let me give you an example why. I live just outside London Ontario. There is a weather station at London airport.

    Last summer, as an example, we had very little rain in August where I am, but the city of London got dumped on many times. I could see the rain clouds pass right over us and drop nothing, only to drop it all over London. My son’s place is a 20 minute drive from me, and he will get dumped on while we don’t and visa versa.

    Snow is worse. We get lake effect snow, sometimes 2 to 3 feet in 48 hour period. But those are fingers of deep snow drop from Lake Huron, and completely dependant on the wind pattern. We get 3 ft of snow, London airport get’s nothing.

    So what is the rainfall for our ‘grid’? Change the wind pattern and we get dumped and not London. So London’s station records too little rainfall, or too much rainfall (both due to highly localized dumps).

    Unless you have rain stations every kilometer, and observe changes in drop zones due to wind patterns, there is no way you can deduce if there is any trend happening. And there is some indication, from all those wind turbines, that wind over the past 6 years have been slowing. That will change the drop zones of the rain even though the rain volume hasn’t changed.

  87. Darkinbad the Brightdayler says:

    They could have used a Binomial test to compare the observed value in each grid square with the expected value (ie assuming an equal chance of finding activity). The expected value (I assume) varied from year to year, reflecting their observation of increased activity. Since there were potentially hundreds of tests (thus exposing the possibility of numbers of false positives accuring in the data set) they might have used used a Bonferroni correction which could be repeated to test, for example, 5 times the average number. The Bonferroni correction would have made the results more rigorous so that, where previously, there was a 1 in 20 chance of the result for any particular grid square being wrong, after applying the correction, there would be only a 1 in 20 chance of any square in the whole grid being wrong.
    However, no amount of applying otherwise appropriate statistical practices improves poor hypothesis formulation or testing

  88. David L says:

    Merovign says:
    February 21, 2011 at 1:06 am

    “It is, frankly, difficult for the layman to understand how so many errors can be made in a row, accidentally, by professionals…”

    This is the same thought that I had while reading Willis’ excellent post. It’s sobering that so many so-called “experts” can make so many mistakes. I guess it should be expected when people try to push a religion based on faith. Everything can be twisted and contorted into evidence to support their cause.

  89. Davidg says:

    ‘Not Even wrong’ , these miserable excuses for scientists at Nature make it quite clear that this is a war over truth. An attempt to manipulate mass behavior the world over.
    It’s beyond time to throw the gauntlet down; gentlemanly discussion is useless. It’s time to call a spade a spade. Nature needs to be publicly called out and humiliated over their
    new policy of scientific fictionalizing.

  90. steven mosher says:

    Thanks Willis,

    When I read it I knew you would have a field day. As I see it there are at least three angles at the piece. The first on instruments, the one you did on data and methods,
    and then a last one on the GCMs they selected.

    I havent found anything on optimal fingerprinting.

  91. steven mosher says:

    Richard M says:
    February 21, 2011 at 5:36 am

    I first thing that hit me when I first read about this paper were the dates … 1951-1999. Did the world stop measuring rainfall in 1999? Why not 2009? Why not start in 1900? Does anyone else feel like this entire study was likely another alarmist cherry picking exercise?

    #########
    there was a “problem” with data after 1999. Its covered in the SI. But I didnt have access to the SI to tell you what the problem was

    The cutoff date doesnt really effect the main conclusions. BUT it does show that models underestimate extreme events. Go figure 1998 and 1999 had a lot of extreme events. Since models dont represent the timing of el ninos the models under estimate extreme events in that period.

  92. Ben of Houston says:

    How does such sophomoric analysis get into a “prestigious” journal? The tacking of calculation upon assumption upon calculation upon assumption introduces such uncertainty that the data could not be used to show nuclear weapons cause cancer.

    After hearing this description, my wife, a business major with no math background, said “it sounds like that plugged stuff into their calculator until they got the answer they wanted”.

  93. vigilantfish says:

    Peter Miller says:
    February 21, 2011 at 7:31 am

    I am not sure if Nature is as unscientific as Greenpeace – that would be very difficult – so it was like a breath of fresh air to hear this from one of Greenpeace’s founders:

    http://www.stockwatch.com/News/Item.aspx?bid=Z-C:*CURRENT-1810801&symbol=*CURRENT&news_region=C

    ————

    Peter, I wish this were something new, but it isn’t. Patrick Moore has increasingly distanced himself from Greenpeace since he left in 1986. What would be exciting would be if other Greenpeace zealots also saw the light.

    ————–

    Willis, I am in awe, as always. You have restored my faith in Divine Providence. That being said, I had to look up Weibull distribution and have to confess I am none the wiser, although apparently its very important for engineering.

  94. jrwakefield says:

    There are just too many locallized surface variables that affect where and when rain will fall. The US suffered the dust bowl years in the 1930s, but was that offset with more rain in another region of the continent? Was that drought caused by changes in wind patterns that just took the rain to another location? Thus there is no way of knowing if the current data is showing any actual change in rain fall, or just changes in where the rain falls.

    The only way to measure rain fall properly is to get the actual downfall a storm system drops and add them all up across the globe. Since that has not appened in the past, and would be most difficult today (radar might be one method), then there is no way we can know if there is any change in rain over time.

  95. Michael Moon says:

    Is it not absurd to assume that max rainfall data correlates to stations by distance? We have all seen the thunderstorm raining only on the front yard, not the back yard, so their choice of “gridding” seems contrived. Why doesn’t anyone in climate science just average ALL the stations together and say what the yearly trend of the average is? No, they would rather create data they don’t have…

  96. Robuk says:

    Unfortunately the brain dead UK politicians believe in this study.

    http://s446.photobucket.com/albums/qq187/bobclive/?action=view&current=Pakistanfloods.mp4

  97. James Evans says:

    This seems pretty much par for the course. Take some data that obviously doesn’t support your idea, and then do a huge amount of complicated statistical jiggery-pokery with it. Then, hey presto, the data says what you want it to say. You get some nice headlines, and nobody in the media is interested when a while later the shenanigans are uncovered. It’s what the Team does best.

  98. Steve Keohane says:

    Thanks Willis for another clear analysis. There is a new precipitation database, measuring system in the US. It has been getting organized for the past few years, all volunteers. I don’t know whether this will be a separate database or is amended onto an existing one. It is the Community Cooperative Rain Hail & Snow network, or CoCoRAHS. Here is the home page http://www.cocorahs.org/
    Looks like if you go to ‘Data Analysis’ under Resources, then the bottom of that page lets one download data either as Excel or KMZ files.

  99. JJB MKI says:

    @Richard Telford
    February 21, 2011 at 1:55 am

    Could this be the birth of a new logical fallacy – Argumentum ad R? Adding more tea leaves to the cup doesn’t make divination any more accurate, and if you need to use code rather than plain English to make an argument, I suspect you don’t really have one to make. All you have proven is that you most likely found the article too painful to read beyond the fifth paragraph, you find something disdainful in the word ‘sceptic’, and that commenter kadaka (1.29am) was resoundingly correct in his prediction (and I’ll bet he didn’t even have to resort to inappropriately applied statistical methods used to deliberately blur the line between observation and model results to make it).

  100. Robuk says:

    David says:
    February 21, 2011 at 5:15 am

    “Here is a list of what I have found so far for the 1930 to 1936 or so period I am researching. China, Russia US and Canada are covered, but little else. Thanks in advance.”

    I bet there is not much UHI effect in those temps.

  101. Willis Eschenbach says:

    Peter Plail says:
    February 21, 2011 at 12:37 am

    Thanks for your hard work Willis.

    They have done a lot of calculations to get their results. I am appalled that they assume Gaussian distribution of actual and model datasets without checking.

    Just to be clear, Peter, we don’t know if they checked or not. If they did, what they did not do is report which test for gaussian normality they used, and the results.

    w.

  102. philincalifornia says:

    http://www.nature.com/siteindex/index.html

    Has a list of all the Nature publications, for example:

    Nature
    Nature Biotechnology
    Nature Cell Biology
    Nature Chemical Biology
    Nature Chemistry, etc.

    Perhaps the Editors should consider adding Nature Lysenkoism to their list ?

  103. danj says:

    e-based papers. How did this ever get published? Oh, yes. Nature.

    Malaga View says:
    February 21, 2011 at 12:11 am
    A beautifully sharp analysis that cuts right through the mumbo jumbo to reveal witch doctors publishing more voodoo in their house magazine.

    First Witch: When shall we three meet again
    In thunder, lightning, or in rain?

    Second Witch: When the hurlyburly’s done,
    When the battle’s lost and won.

    William Shakespeare, Macbeth, 1.1
    —————————————————————————-

    I think Birnam Wood us getting very close to Dunsinane…

  104. Willis Eschenbach says:

    Puckster says:
    February 21, 2011 at 1:24 am

    “Being a generally suspicious type fellow, I wondered about their claim that changes in rainfall extremes could be calculating by assuming they follow the same distribution used for temperature changes.”

    Before submitting this to Nature……….maybe change “calculating” to “calculated”?

    ……..I’m just saying…….just a little proof reading.

    Thanks, fixed. You might want to work a bit on your delivery, however. I’m not sure what a submission to Nature has to do with a typo, or why the tone of your post.

    Or as you put it,

    ……..I’m just saying…….just a little politeness.

  105. Willis Eschenbach says:

    kadaka (KD Knoebel) says:
    February 21, 2011 at 1:29 am

    Editorial Annoyances: …

    Thanks, kadaka, fixed.

    w.

  106. Theo Goodwin says:

    Brilliant work, Willis. You nailed them totally.

  107. Tim Folkerts says:

    I haven’t had time to look through it all, but the first conclusion (which was repeated “because it is important”) it was is invalid.

    Now, let me repeat part of that, because it is important.

    91% of the rainfall stations in the US do not show a significant trend in precipitation extremes, either up or down.

    So overwhelmingly in the US there has been

    No significant change in the extreme rainfall.

    The aggregate of a lot of data sets that are not individually significant can quite easily be significant. For example, very few if any years by themselves would show statistically significant increase in temperature from the previous year. But a century of such years can show a statistically significant trend.

    Or bet red/black on a roulette table 50 times. Rarely would this be statistically different from 50-50. But if you believe that overwhelmingly the odds are no different from 50-50, I would love to go to Las Vegas with you. :-)

    The rest of the article might address this, but it is not encouraging when the first, key conclusion does not follow from the data.

  108. David Smith says:

    Hi, Willis. Another data problem to be considered has to do with the time of observation.

    If an observer takes a 24-hour reading in the afternoon and resets the rain gauge then there is a chance that an afternoon heavy rain (afternoon thunderstorm) will get split between two days in the records. If an observer takes the reading at say midnight or dawn then the risk of splitting a thunderstorm into two days is diminished.

    Afternoons, with their heat-driven precipitation tendency, tend to be the time of heavy rain events moreso than the cooler hours.

    Over the decades there was, I believe, a move in the US to switch observation times to dawn or midnight. I think that this was the basis for the well-known time-of-observation bias adjustment to the temperature records. Less well known, or not at all known, is the possible impact of the observation time shift on the precipitation records.

    At some point in time it’d be good to investigate the data to see if a precipitation TOB truly exists.

  109. Oliver Ramsay says:

    Puckster says:
    ” ……..I’m just saying…….just a little proof reading.”

    ————————-
    Is reading a proof equivalent to proofreading?

    Just asking.

  110. Malaga View says:

    Nature Magazine published a paywalled paper…..

    Surely they will have to increase the subscription to keep the riff raff out… you know Willis… the ones that ask the awkward questions… the ones that lower the tone of the debate with their off the wall reality checks… the ones that aren’t prepared to wear a collar and Team tie… the ones that aren’t prepared to Hide the Decline in Climate Science… after all it really doesn’t pay to wash your dirty linen in public.

  111. Willis Eschenbach says:

    Richard Telford says:
    February 21, 2011 at 1:55 am

    only ~ 5% of the US rainfall stations show a significant trend in extreme rainfall. The rest of the nation is not doing anything.

    ——————
    This is a very weak analysis, of the type beloved by climate “sceptics”.

    If time is a weak predictor of extreme rainfall, then only a few individual stations will have a statistically significant trend, perhaps few more that expected from the Type I error rate. But there may still be a highly significant relationship taking the data en-mass.

    Climate “sceptics” like this, because they can pick a record and show that there is no *statistically significant* change, ignoring the aggregate data which may show a highly statistically significant change.

    A couple of comments on that claim:

    1. If the Nature article used good data and reasonable methods, you’d be correct, because then we might have a chance of finding a “highly statistically significant change”. I note that rather than acknowledge any of the dozen or so problems with the claim, instead you say that my analysis is weak … might be, Richard, but their analysis is weaker.

    2. The MZZH11 study didn’t do the type of analysis you discuss, an average of a large mass of data. Instead, they averaged a large number of unknown transformations (PIs with each using a unique and different distribution) of improperly calculated area-angular weighted averages of non-quality controlled data using unknown coefficients … and since we don’t know the unknowns in that, we cannot say if their results are significant or not.

    3. As a result, I was unable to do any sophisticated analysis of their results because they are not replicable as they stand … and even if they were, why would I want to convert data to PIs using a different distribution for each datapoint? That way lies madness.

    Here is some R code that illustrates this:

    set.seed(123)
    x=rnorm(100)
    y=as.data.frame(replicate(1000,rnorm(100,x*.05)))#1000 replicate y variables with weak relationship to x
    plot(x,y[[1]])#plot of first y replicate against x
    sum(sapply(y,function(Y)cor.test(x,Y)$p.value)<.05)#number of replicates with a statistically significant trend == 68
    cor.test(rowMeans(y),x)$p.value # p-value of the aggregate data set highly statistically significant

    Thanks for the code, Richard, it’s always useful. You are correct that the average of data of which only a few individual trends are statistically significant may be statistically significant as a whole. There are some problems with your example, however. And curiously, one of them is the same trap the MZZH11 authors fell into. That problem is the requirement that the data be Gaussian for the results to be valid, or as the authors of the package put it:

    If method is “pearson”, the test statistic is based on Pearson’s product moment correlation coefficient cor(x, y) and follows a t distribution with length(x)-2 degrees of freedom if the samples follow independent normal distributions.

    Since what was averaged in the MZZH11 are PIs, we can be pretty sure that their averages are of samples that do NOT follow independent normal distributions … which means that if the authors had used the exact method that you recommend, their results would still be wrong …

    In addition, to do any analysis of the type that you are proposing, we need to look at both the distribution of the data and the distribution of the errors in the data. Your method will only work when (as in your code) both the error and the data are independent and Gaussian (normal). We know the data that they are using is not normal … and we don’t know the nature of the errors, but it is unlikely that they are normal, since the maximum possible negative error is the size of the underlying data point (e.g. rainfall was 20 mm, record says 0 mm), while there is no maximum positive error (e.g. rainfall was 20 mm, record says 1067 mm) … but in either case, if either the data or the errors are non-normal, your method gives incorrect answers.

    I use a simple analysis of trends to get an idea of what we are looking at. And in climate science, because of the data quality issues that I have listed above, that may be the best that we can do. Yes, as you point out we can do increasingly selective and specific analyses to dig out tiny signals, we have a whole arsenal of methods to do that (although “optimal fingerprinting” isn’t usually one of them) … but we cannot do that without a very careful analysis of the data that we are using. The problem is that when we look for big differences in the results, it doesn’t matter if there are small problems with the data.

    But when there are big problems with the data and we are looking for small differences …

    Thanks for your thoughts,

    w.

  112. Darkinbad the Brightdayler says:

    .05 level means that 5 times out of a hundred, you will be accepting a hypothesis you should be rejecting.
    When you have several hundreds or thousands of similar tests, you need to check whether the results you think you are interpreting are drawn from the 5% group or the 95% group.
    Selecting the 5% group isn’t a valid scientific option

  113. Oliver Ramsay says:

    Tim Folkerts says
    “The aggregate of a lot of data sets that are not individually significant can quite easily be significant. For example, very few if any years by themselves would show statistically significant increase in temperature from the previous year. But a century of such years can show a statistically significant trend.”
    ————————————
    What made you think this was a one-year trend?

  114. Willis Eschenbach says:

    Don V says:
    February 21, 2011 at 2:03 am

    Willis, starting on page 424 of the Allen and Tett paper they discuss “Consistency checks to detect model inadequacy”. They even give a ” simple test of the null hypotheis” (I suck at statistics and it didn’t look that simple to me). Were you able to determine if MZZH11 ever conducted this “simple test” for using optimal fingerprinting to prove modeled rainfall extremes retrospectively correlated with the
    actual data and therefore could be used to prognosticate? I kept falling asleep at this late hour while wading through the paper. . . . z z z but I couldn’t find any such tests.

    Don, the problem is not the “simple test of the null hypothesis”. They’ve done that, and reported it. The problem is the huge uncertainties in the data, the grid-cell averaging, the conversion to PIs, and the huge problem of the non-normality of the dataset. With those uncertainties, the “simple test of the null hypothesis” (and indeed the “optimal fingerprinting” method itself) gives meaningless results.

    w.

  115. Sam Parsons says:

    danbo says:
    February 21, 2011 at 3:29 am
    Figure 1. Extreme 1-day rainfall. New Orleans, Katrina. Photo Source

    “I’m not sure where this came from. But see that large body of water at the top, of the source photo. http://www.cces.ethz.ch/projects/hazri/EXTREMES/img/KatrinaNewOrleansFlooded.jpg?hires It’s called lake Pontchartain. It’s the second largest inland body of saltwater in the US. That’s where the water came from. Although it rained, this was a tidal event. Not a rain event.”

    It was a broken levee event. A levee at the south extremity of Lake Pontchartrain failed. Neither rain nor tide caused the flooding in New Orleans’ Ninth Ward during Katrina.

  116. Willis Eschenbach says:

    Amino Acids in Meteorites says:
    February 21, 2011 at 2:43 am

    Charlie A says:
    February 20, 2011 at 11:28 pm

    Have you offered to Nature….

    Why did you pick Nature in particular? It is because of the controversy in some of their global warming papers publicized by them? You chose Nature because of their global warming bias?

    Ummm … because the original paper was published in Nature. It’s also why “Nature” is in the title of the piece. Your speculation is baseless.

    w.

  117. Sam Parsons says:

    Amino Acids in Meteorites says:
    February 21, 2011 at 2:38 am
    “On Katrina, here’s a clue: don’t build a city below sea level.”

    The Old City, aka The Crescent City, is well above sea level and was not touched by flooding during Katrina. The 20th Century city known as the Ninth Ward is what was built below sea level and was flooded because of a failed levee during Katrina. The Old City is called “The Crescent City” because it sits on a crescent shaped ridge that borders the Mississippi.

  118. Sam Parsons says:

    Richard Telford says:
    February 21, 2011 at 1:55 am
    only ~ 5% of the US rainfall stations show a significant trend in extreme rainfall. The rest of the nation is not doing anything.

    ——————
    This is a very weak analysis, of the type beloved by climate “sceptics”.

    “If time is a weak predictor of extreme rainfall, then only a few individual stations will have a statistically significant trend, perhaps few more that expected from the Type I error rate. But there may still be a highly significant relationship taking the data en-mass.”

    “Climate “sceptics” like this, because they can pick a record and show that there is no *statistically significant* change, ignoring the aggregate data which may show a highly statistically significant change.”

    This is remarkably thoughtless criticism of the kind that you must expect from Warmista. Having found that individual stations show no significant trend, he states his preference for aggregated data which does show a trend. That is, he expresses his preference for the analysis that shows a warming signal. DUH! Is that reasoning? Is that analysis? No, it is cheating!

    If he is serious what he must do is explain the methodological reasons for using the aggregated data. Of course, this never occurs to him. Getting the “Warmista answer” is all that matters.

  119. Willis Eschenbach says:

    danbo says:
    February 21, 2011 at 3:29 am

    Figure 1. Extreme 1-day rainfall. New Orleans, Katrina. Photo Source

    I’m not sure where this came from. But see that large body of water at the top, of the source photo. It’s called lake Pontchartain. It’s the second largest inland body of saltwater in the US. That’s where the water came from. Although it rained, this was a tidal event. Not a rain event.

    Let’s compare NOAA, emphasis mine:

    Gulf Coast

    Rainfall from Katrina’s outer bands began affecting the Gulf coast well before landfall. As Katrina came ashore on August 29th, rainfall exceeded rates of 1 inch/hour across a large area of the coast. NOAA’s Climate Reference Network Station in Newton, MS (60 miles east of Jackson, MS) measured rainfall rates of over an inch an hour for 3 consecutive hours, with rates of over 0.5 in/hr for 5 hours during August 29th. Precipitation analysis from NOAA’s Climate Prediction Center show that rainfall accumulations exceeded 8-10 inches along much of the hurricane’s path and to the east of the track.

    with your claim:

    Although it rained, this was a tidal event. Not a rain event.

    Your explanation (tides) would only affect that part of New Orleans that is below sea level … but there were huge areas flooded that were above sea level or storm surge, they were flooded by the rain …

    So your claim is that extreme rain had nothing to do with Katrina, it wasn’t a rainfall event, and you’re seriously going to bust me because I used a picture of Katrina, not to prove anything scientific, but simply as an example of what can happen during times of extreme rains?

    Dude, you seriously have too much time on your hands … sounds like you’re a smart guy, how about applying that smarts to something further down the page than the picture at the top of an analysis?

    w.

  120. Al Gored says:

    Another brilliant dissection Willis. It should be an autopsy but…

  121. Willis Eschenbach says:

    Wade says:
    February 21, 2011 at 4:56 am

    I’m confused. NOAA just had a study that said global warming was going to cause more desertification.
    http://wattsupwiththat.com/2011/02/19/noaas-compendium-of-climate-catastrophe/

    More atmospheric dust from global desertification could lead to increases of harmful bacteria in oceans, seafood.

    And now this study in Nature says global warming is causing more rain. Which is it?

    (sarc) Clearly you are new to climate science … the answer is that global warming increases both floods and droughts. And heat waves and cold waves. They’re all a predictable result of global warming. What could be simpler? (/sarc)

    w.

  122. Willis Eschenbach says:

    Jit says:
    February 21, 2011 at 5:16 am

    Willis:

    9% of the raw data were trended. As you say, we expect 5% to be trended at 95% level, but we expect 2.5% to be + trended and 2.5% – trended. What percentage of the series showed +ve and -ve trends?

    About three quarters positive, one quarter negative … but remember, there are problems in the raw data, big problems, and we only have about 25 datasets with a significant trend. Given those issues, finding that distribution (one quarter/three quarters) is not surprising.

    Also, don’t quite follow how they do the grid averages. How big are the gridcells? Are they really using data from 2000km away in the calculation of cell values in gridcells that are much smaller (when spatial correlation dies off so fast)?

    HADEX use an odd sized gridcell, hang on … OK, their reference says:

    Grids are 3.75 longitude by 2.5 latitude.

    At the latitude of the mid-US this would be about 195 miles wide by 150 tall (320 km by 240 km).

    Unfortunately, we don’t know how far out they are averaging precipitation. It appears from their data that they may be using the temperature decorrelation data, which hits 50% at about 2000 km (12oo miles), but it’s not clear. Although they give decorrelation lengths (L, in km) for a variety of temperature and rainfall indices (here, Appendix A), they don’t give the decorrelation length for either RX1day or RX5day data. In fact, there is so little correlation between adjacent stations for RX1day that I don’t think a distance/angle calculation is appropriate or accurate … you’d end up with lots of gridcells with no data because there is no rainfall station within 40 km of .

    Why have gridcells at all – why not just interpolate all available series and calculate standard errors based on some sort of cross-correlation? (And limit contributing series to those within the range of +ve spatial correlation)

    I like the question, don’t have an answer. I think they want to average the data into a single value (e.g. northern hemisphere extreme rainfall PI) so that they can use the “optimal fingerprinting” technique. However, I dislike gridbox averaging for the purpose.

    w.

  123. Willis Eschenbach says:

    Richard M says:
    February 21, 2011 at 5:36 am

    I first thing that hit me when I first read about this paper were the dates … 1951-1999. Did the world stop measuring rainfall in 1999? Why not 2009? Why not start in 1900? Does anyone else feel like this entire study was likely another alarmist cherry picking exercise?

    The answer is that they wanted to compare observational data to model results. The model results in question are the 20th century model runs archived at CMIP3, the “20c3m” runs. With a few exceptions, these runs all end in 1999 … so that’s why the study ends in 1999. Regarding the start date, that’s the start of the observational data on rainfall extremes.

    I make no accusations of “cherry picking” of the data, they’ve used what they have available.

    w.

  124. Willis Eschenbach says:

    Richard M says:
    February 21, 2011 at 5:36 am

    I first thing that hit me when I first read about this paper were the dates … 1951-1999. Did the world stop measuring rainfall in 1999? Why not 2009? Why not start in 1900? Does anyone else feel like this entire study was likely another alarmist cherry picking exercise?

    Start date is the start of the collated observational data, end date is the end of the CMIP3 twentieth century hindcast model runs archived at CMIP3. They haven’t cherry picked the data.

    w.

  125. Willis Eschenbach says:

    Bernie says:
    February 21, 2011 at 5:48 am

    Willis:…
    Do you know if there have been verifications of the original Hansen and Lebedeff (1987) results? I realize that precipitation is likely to be different from temperature but the difference in your analysis from H&L is so dramatic that it seems to me to be worth verifying – if for no other reason than the passing of 25 years. Also, I would assume that the analysis of the satellite data would verify any findings with respect to temperature.

    I’ve run the results myself and find results similar to Hansen/Lebedeff. In addition, the HADEX paper cited above (Appendix A) recalculates the values for many of the variables, and finds similar answers.

    That reference also says:

    Statistical tests were not generally applied to precipitation data analyzed at the workshops but any obvious outliers, identified by careful examination of graphs, were checked manually. Careful post workshop analysis was employed and data processed outside of the workshops were similarly tested for outliers but methods varied from source to source. Statistical tests, local knowledge, an investigation of station histories or comparison with neighboring stations can all be applied to determine whether an outlying precipitation value is erroneous. It is particularly important to identify multiday precipitation accumulations that can appear erroneously in records of daily precipitation [Viney and Bates, 2004]. These occur when accumulated rainfall values are reported as daily totals. For example, data extracted from GHCN-Daily for Brazil were rejected if a rainfall value greater than 1 mm fell after a missing observation [Haylock et al., 2006]. Even after data were processed and collated for this study, annual time series of total precipitation and diurnal temperature range for each station were assessed again to identify outliers that may have been missed in the initial quality control procedure.

    While that sounds fine, how is it that the ETCCDI data still contains crazy outliers? … this is again the problem with the lack of transparency. There should be an audit trail so that we can compare the pre-QC data with the post-QC data, or at least there should be a record of what changes were made. Without that, we simply cannot trust the HADEX data … which should sound familiar, it’s no different than the situation with other Hadley Center products.

    w.

  126. Willis Eschenbach says:

    Bill Illis says:
    February 21, 2011 at 7:40 am

    Here is a scatterplot of HadUKP England and Wales precipitation (a fully quality controlled dataset) versus HadCET temperatures back to 1766.

    Technically, there should be a trend in precipitation of 2% to 3% per 1.0C increase in temperatures.

    There is no trend in this data and no evidence that global warming is causing more precipitation in the UK.

    http://img249.imageshack.us/img249/1669/engwalesprecipvshadcetk.png

    Either the temperature data is wrong, the precipitation data is wrong or the Clausius Clapeyron relation is wrong (on a local level at least – and if that is the case, the studies are based on a false premise to start with).

    There is also no trend in daily precipitation in England and Wales back to 1931 (over 29,000 individual datapoints), a period when temperatures have supposedly increased by 0.8C in the UK.

    http://img101.imageshack.us/img101/4800/dailyrainfallenglandwal.png

    Excellent work as usual, Bill. Richard Telford is right that looking for a trend in individual stations is a weak test, but you are showing the national averages.

    In addition, teasing out some kind of very small relationship from climate data, while possible, suffers from a couple problems. The first is that the rainfall data is short, spotty, and contains a host of missing observations. In recent years, of course, the data for any given station is more complete than in early years … which will not have much of an influence on averages, but could easily affect extremes.

    In addition, if the relationship is so weak that we need layers and layers of sophisticated analysis to find it in fifty years of data … then I doubt that it is big enough to make any difference at all.

    On the other hand, if the difference is as big as they claim it is (an increase in extreme events of 0.3-0.5% per year over much of the US), then we should have seen a 15-25% increase in the number of extreme events in the record … but that hasn’t shown up, and despite Richard’s claims, an increase of 25% in any trend over 50 years will definitely show up in the simple trend data. But it hasn’t.

    w.

  127. Tim Folkerts says:

    Willis Eschenbach says: February 21, 2011 at 11:52 am

    “About three quarters positive, one quarter negative … but remember, there are problems in the raw data, big problems, and we only have about 25 datasets with a significant trend. Given those issues, finding that distribution (one quarter/three quarters) is not surprising.”

    But isn’t that like saying “I flipped a coin 5 times, and repeated the experiment 731 times. Of those experiments, 75% of the time I got more heads than tails. But given the small number of trials, it is not surprising to get more heads.” When in fact, getting more heads 75% of the time is incredibly unlikely in 731. For 731 trials where there is no trend, then getting 75% going one way would only happen about 1 out of 10^43. (If you had less than 731 actual stations to use, then the odds of a 3:1 split would not be as extreme, but even with as few as 20 stations, getting a 15:5 split would be a statistically significant difference from 50:50.)

    So what you have shown is that there is indeed a VERY statistically significant difference from the null hypothesis (“there is no trend”).

    Could you provide one more statistic? How many times were there statistically significant DECREASES compared to INCREASES? 9% of the time time you say there was a trend. If there is indeed no trend, then about half of these should trend up and half trend downward. If instead, close to 0% of stations show a decrease while close to 9% show an increase, then once again that is evidence that there is a statistically significant change.

    NOTE: There is a big difference between “statistically significant” and “meteorologically significant”. The increase could well have no impact on people or nature.

  128. Willis Eschenbach says:

    Tim Folkerts says:
    February 21, 2011 at 10:52 am

    I haven’t had time to look through it all, but the first conclusion (which was repeated “because it is important”) it was is invalid.

    Now, let me repeat part of that, because it is important.

    91% of the rainfall stations in the US do not show a significant trend in precipitation extremes, either up or down.

    So overwhelmingly in the US there has been

    No significant change in the extreme rainfall.

    The aggregate of a lot of data sets that are not individually significant can quite easily be significant. For example, very few if any years by themselves would show statistically significant increase in temperature from the previous year. But a century of such years can show a statistically significant trend.

    Thanks, Tim. You make the same point as Richard, that testing individual trends is a weak test. While you are correct in principle, that claim is often wrong in practice. See my answer to Richard here, along with my comments here.

    I put all of those trend numbers in because they are important as an indication of the size of the change that we are looking at. It is a very tiny change if it exists at all. We know that because of the amount of the underlying data which contains no trend at all. This is a different result from e.g. finding a tiny trend that results from the averaging of a large number of significant trends.

    In addition, it indicates that the size of their claimed effect is doubtful. They claim that the number of extreme events in much of the US has increased by 30-50% over the last 50 years … are you seriously claiming that if that happened a simple trend test would not be able to detect a 30-50% change, and that such a 30-50% change would be statistically insignificant in over 90% of the stations?

    Because that’s what your claim means, and if I’m to believe it, you’ll have to do better than say “the aggregate of a lot of data sets that are not individually significant can quite easily be significant”. Yes, they can … but you to to show in this case that they are significant, not that they “can quite easily be” significant. You are saying that a 30-50% change in a dataset average is only visible in 9% of the underlying data making up the average … citation? Explanation? Supporting data?

    It might be explained by very large changes in a small subset of the data … but then you’re left trying to explain why rainfall only changes in a small percentage of the data, and the rest is unchanged … while “global warming” is said to cause droughts and floods, I’ve not heard anyone claim that it will only affect less than 10% of recording stations.

    Finally, my analysis of the individual trends is far from the only evidence that their analysis is deeply and seriously flawed … I don’t want anyone to come away thinking that your point would make any difference if it were true. We still have the other huge problems with the dataset and the the analysis, which your points don’t touch at all.

    w.

  129. TonyK says:

    I must admit that a lot (most?) of this analysis went right over my head! The whole affair reminds me of those join-the-dot pictures. It seems that the warmists simply add a whole lot more dots between the measuring stations that actually exist until the picture comes out like they wanted in the first place.
    Personally, I view ANY processing of the raw data with a variable amount of scepticism. If there is a trend in temperature or rainfall, surely it would show in the raw data. Simply look at a good long temperature record. Is it going up or down? If up, is it a linear rise or accelerating? Or is it levelling out?
    You know the old chestnut, ‘If a tree falls in the forest when no-one is there to hear it, does it still make a noise?’ (Duh! Of course it does!) Perhaps the climate equivalent is ‘If it rains exceptionally hard where no-one is there to see it, does it count? A warmist would say ‘Yes, and we have to allow for that in the models.’ A heretic would say ‘If there was no-one there to see it, how do you know it rained?’

    [Question: Why do our UK cousins hyphenate "no one"? ~dbs]

  130. Tim Folkerts says:

    I have no specific knowledge of the “fingerprint” analysis they did, but ….

    1 ************************************
    “The problem is that the average of a PI of a number of extreme value distributions will be an extreme value distribution, not a Gaussian distribution.”

    The Central Limit Theorem states

    Let X1, X2, X3, …, Xn be a sequence of n independent and identically distributed (iid) random variables each having finite values of expectation µ and variance σ2 > 0. The central limit theorem states that as the sample size n increases, the distribution of the sample average of these random variables approaches the normal distribution with a mean µ and variance σ2/n irrespective of the shape of the common distribution of the individual terms Xi.
    from Wikipedia

    The average of numbers from ANY distribution WOULD approach a normal (Gaussian) distribution when data is combined regardless of the initial distribution.

    2 ************************************
    “In other words, the “optimal fingerprint” method looks at the two distributions H0 and H1 (observational data and model results) and sees how far the distributions overlap. ”

    You CLAIM that they must use the technique you mention, but there are plenty of legitimate ways to compare non-normal (non-Gaussian) distributions. Can you cite any evidence that the analysis in the current paper actually uses the method you suggest, or that the method would fail if they did use the method you suggest (especially in light of my fisrt point that the data will indeed ?

  131. Tim Folkerts says:

    Willis,

    I agree that there are plenty of opportunities to critique the paper. Any time there is such substantial statistical analysis, there is lost of room for errors. And then there is lots of room for critiques of the critiques. :-)

    You state in your reply “They claim that the number of extreme events in much of the US has increased by 30-50% over the last 50 years”. This is the first I had seen that claim, so I was not thinking about that in my response. Even so, I could argue that if a given station went from 2 extreme events per decade to 3 per decade, that would be an increase of 50% increase in the number of extreme events. Finding a statistically significant trend in such data would be difficult unless you averaged a lot of stations to see a change of 200 to 300 per decade. But again, I don’t have the data, nor do I have the original paper, so this is simply speculation about how such a “big” 50% change could be difficult to spot.

  132. kadaka (KD Knoebel) says:

    From Tim Folkerts on February 21, 2011 at 1:09 pm:

    “In other words, the “optimal fingerprint” method looks at the two distributions H0 and H1 (observational data and model results) and sees how far the distributions overlap. ”

    You CLAIM that they must use the technique you mention, but there are plenty of legitimate ways to compare non-normal (non-Gaussian) distributions. Can you cite any evidence that the analysis in the current paper actually uses the method you suggest…

    http://localgov.nccarf.edu.au/resources/human-contribution-more-intense-precipitation-extremes
    Emphasis added:

    In this Nature article the authors state that “human-induced increases in greenhouse gases have contributed to the observed intensification of heavy precipitation events found over approximately two-thirds of data-covered parts of Northern Hemisphere land areas. These results are based on a comparison of observed and multi-model simulated changes in extreme precipitation over the latter half of the twentieth century analysed with an optimal fingerprinting technique.”

    Cute site. The Aussies have a National Climate Change Adaptation Research Facility (NCCARF) to help them prepare for the Challenging Turbulent Times To Come. Who knew? I didn’t. I’m sure our Aussie brethren are all very happy to be paying for something so useful.

  133. John from CA says:

    Great Post Willis — this is a perfect example of a great content object exercise for the classroom. Would you consider allowing us to convert some of your posts to free SCOs for educators?

  134. HFC says:

    Biased Brainwashing Corporation has been selling the AGW = more rain story today.

    Listen or simply read the summary to understand the BBC angle.

    http://www.bbc.co.uk/programmes/b00yjs49

  135. Don Horne says:

    Willis,

    Shouldn’t “temperatures” be “precipitation” in the sentence below which is just after Fig. 2?

    Hmmmm …. so how did they get that result, when the trends of the individual station extreme temperatures show that some 95% of the stations aren’t doing anything out of the ordinary?

  136. Don Horne says:

    OOPS,

    That should be just after Fig. 1 not Fig. 2. Sorry ’bout that.

    Previous post by me…
    Willis,

    Shouldn’t “temperatures” be “precipitation” in the sentence below which is just after Fig. 2?

    Hmmmm …. so how did they get that result, when the trends of the individual station extreme temperatures show that some 95% of the stations aren’t doing anything out of the ordinary?

  137. Willis Eschenbach says:

    Tim Folkerts says:
    February 21, 2011 at 12:31 pm

    Willis Eschenbach says: February 21, 2011 at 11:52 am

    “About three quarters positive, one quarter negative … but remember, there are problems in the raw data, big problems, and we only have about 25 datasets with a significant trend. Given those issues, finding that distribution (one quarter/three quarters) is not surprising.”


    Could you provide one more statistic? How many times were there statistically significant DECREASES compared to INCREASES? 9% of the time time you say there was a trend. If there is indeed no trend, then about half of these should trend up and half trend downward. If instead, close to 0% of stations show a decrease while close to 9% show an increase, then once again that is evidence that there is a statistically significant change.

    I’m sorry for my lack of clarity. The figures I gave were for statistically significant stations only. The numbers are slightly different for all of the stations, and for the mainland long stations. Of the 340 mainland US stations that have 40 years of data or more, there were 27 stations with statistically significant trends. Of these, 5 were negative and 22 were positive, with a mean of +3 ± 2 mm (2σ) per decade increase.

    But before doing analyses on those numbers, remember that there is some very bad data in the mix, as I pointed out in the head post. This includes data that’s out by an order of magnitude or more. Until that data is fixed, you can’t make any definitive statements about anything.

    And again, this is very peripheral to my main point, which was the underlying weaknesses of their analysis.

    w.

  138. danbo says:

    Sam agreed. The levees broke. Not just in the 9th ward. I’ve seen photos of the water splashing over the levees. One would argue if it had not been for the tide. Would the levees have broken.

    http://en.wikipedia.org/wiki/File:Hurricane_Katrina_winds_1200utc29Aug05_landfall_LA_1hr.gif
    Here it’s pushing water in the lake.

    http://en.wikipedia.org/wiki/File:Hurricane_Katrina_winds_1500utc29Aug05_landfall_MS.gif
    Here the winds shift and push the water at the levees.

    Either way. It ain’t rain.

  139. Michael Barnes says:

    Was the flood picture of New Orleans from the original article? If so, it is irrelevant to their argument.

    The flooding in New Orleans was not from the rain. New Orleans flooded because a levee failed.

  140. danbo says:

    Sam. When you drive across the parish line from Jefferson. The next I-10 interchange (in the photo) as I recall isn’t part of the 9th ward I believe it’s the 4th. It also flooded around the Superdome. Pretty much in the CBD. There were many breaches and collapses.

  141. Willis Eschenbach says:

    Tim Folkerts says:
    February 21, 2011 at 1:09 pm

    You CLAIM that they must use the technique you mention, but there are plenty of legitimate ways to compare non-normal (non-Gaussian) distributions.

    Tim, I don’t understand this. Where have I claimed that they have used what techniques?

    Can you cite any evidence that the analysis in the current paper actually uses the method you suggest, or that the method would fail if they did use the method you suggest (especially in light of my fisrt point that the data will indeed ?

    Again, confusion. If by “the method [I] suggest” you mean “optimal fingerprinting”, that’s what they say they use. I gave evidence that Bell says the data must be random Gaussian.

    Finally, yes, I am well aware of the Central Limit Theorem. However, what you may not realize is the range that this encompasses. For example, here is the violinplot of the time-series of the average of 10,000 datasets of the length of the MZZH11 data (49 years).

    As you can see, despite the numbers being large (10,000 sets of 49-year pseudo-data), the distribution is nowhere near normal.

    And indeed, the question is not whether averages in general are normal or not. The Law of Large Numbers is meaningless here. The question in this case is, how far from normal is the actual data used by MZZH11, and what effect does that departure have on the accuracy of their results?

    Since they give no indication that they have even considered this question (although they may have done so), and since we cannot repeat their analysis because of their use of a different and unspecified probability distribution for each gridcell on the map, at this point their claims are both unsupported and incapable of replication.

    This again highlights the need for scientists to publish their data and their code. At present, nobody on the planet can confidently replicate their results. And to make things worse, the same is true about the HADEX gridcell averages. There is no audit trail, there’s not even sufficient description to establish the decorrelation length used with the RX1day and RX5day data.

    I’m not saying that their analysis is wrong because of the issue of whether the average is Gaussian, Tim. I’m saying that MZZH11 and HADEX have not given us anywhere near enough data to determine if their claims even make it above the noise level, much less if they are significant.

    w.

  142. danbo says:

    Sorry Willis I wasn’t sure if you grabbed it. Of if it came from someone who was using it pretend it was rain. If it was a quick grab by you no problem.

    Forgive me. But Katrina isn’t something theoretical to me. I lived in New Orleans. And lived on the Mississippi coast saw the lead up to Katrina. Had to run. (from Mississippi.) And return home. And have listened to BS about Katrina and AGWing till I’m sick.

    No offense intended guy.

  143. Willis Eschenbach says:

    John from CA says:
    February 21, 2011 at 3:06 pm

    Great Post Willis — this is a perfect example of a great content object exercise for the classroom. Would you consider allowing us to convert some of your posts to free SCOs for educators?

    Do as you wish, as long as you point out to the students that I can be wrong just like anyone …

    w.

  144. danbo says:

    Willis. I know the area in the photo. Granted it floods with a heavy dew. But the food in the picture. That was tidal or as others prefer, The levee caving in. And of course every hurricane has heavy rain. Supposedly we had a 30 ft high tide. I know where the high tide marks are. I believe it.

  145. Willis Eschenbach says:

    Tim Folkerts says:
    February 21, 2011 at 1:26 pm

    …You state in your reply “They claim that the number of extreme events in much of the US has increased by 30-50% over the last 50 years”. This is the first I had seen that claim, so I was not thinking about that in my response.

    It was a slight simplification, and on re-reading, not at all clear. Plus it contains a typo, it should have been half of that, 15% to 25% … let’s start again.

    Note the legend in the figure at the top. This is the change in the PI, the probability index which goes from 0 to 1. They say that in much of the US the PI has increased at the rate of 0.3% to 0.5% for fifty years, which is a change of 15% to 30% in fifty years.

    In reality, as thePI increases, the occurrences are becoming larger and larger. An extreme rainfall with a PI of 0.80 is much more than twice as large as the rainfall with a PI of 0.40. So their 15% to 25% increase PI will be reflected in a much larger increase in 1-day maximum rainfall.

    Sorry for the confusion,

    w.

  146. Look at this flood — 1927 – - makes Hurricane Katrina flood look like a puddle. Regions underwater from Ponchartrain up to Kentucky and Arkansas.
    http://en.wikipedia.org/wiki/Great_Mississippi_Flood_of_1927
    Good Gauss! Put that in your distribution and try to make it look normal.

  147. banjo says:

    [Question: Why do our UK cousins hyphenate "no one"? ~dbs]

    http://www.grammar-monster.com/easily_confused/no-one_no_one.htm

    Coz we is cool an up to date, innit?
    Unfortunately my grammar is not as healthy as my grandpa.

  148. john reeves says:

    Hey excellent analysis there good fellows. I saw the abstract for that and it looked very bogus, as in plenty of “suggests” “could be” and partly caused” kind of statemnets..didn’t sound convincing at all.

    Yr research states why in very clear terms and i just wish this attitude of jumping to conclusions and trying to proven them via ‘science’ is really causing plenty of problems in real terms. The older school emprical method of observing data then seeking to understand what it says seem to be a better way of approaching the science of CC.

    Keep up the good work..

  149. kadaka (KD Knoebel) says:

    Questions:

    I found a University of Victoria press release about the paper:
    http://communications.uvic.ca/releases/release.php?display=release&id=1205

    A new study co-authored by Francis Zwiers, the director of UVic’s Pacific Climate Impacts Consortium, suggests that human-induced global warming may be responsible for the increases in heavy precipitation that have been observed over much of the Northern Hemisphere including North America and Eurasia over the past several decades. (…)

    1. It says at the bottom:

    To receive a copy of the study email a request to press.nature@gmail.com

    Does that mean that Anthony Watts, publisher of the “new media” publication Watt’s Up With That?, can request a copy?

    2. We were repeatedly assaulted with complaints that the Medieval Warming Period was not a global event, it was only in the Northern Hemisphere (when they allowed that much), doesn’t detract from Michael Mann’s hockey stick graph, etc.

    Now they are looking at Northern Hemisphere data, and it is cited as showing global warming. Specifically the effects of human-induced global warming. [And don't bother to argue the exact semantics, it is being cited as proof of (C)AGW.]

    It’s not global, it’s only Northern Hemisphere. It’s Northern Hemisphere, it’s global. WUWT?

  150. Greg Cavanagh says:

    I design storm water networks for a local council. Let me point out that a 1 day event will be a localised rainfall. A 5 day event is expected to be a slightly larger area event, but not guaranteed to be.

    I can’t see any point in grid celling what is essentially point data.

  151. johanna says:

    [Question: Why do our UK cousins hyphenate "no one"? ~dbs]
    —————————————————
    We do it in Australia too. It is because “no one factor is proven to be more important” has a different meaning to “no-one knows which factor” etc. :)

    Thanks for this Willis. I am no statistician, but there are enough red flags for even the ordinary reader who has had a passing exposure to stats to grasp that there is a problem. The way that data is aggregated and averaged and extrapolated bedevils climate science (quite apart from the poor quality of the data in the first place).

    Rainfall data is potentially very misleading, even if it is robust. There can be wild variations within small areas, and large variations over periods of many years in the same spot (unless you are in the equatorial zone, perhaps).

    I think that the methodology has to be just about bulletproof before you could make even the most cautious findings. This one doesn’t even come close.

  152. Smokey says:

    banjo,

    Thanx for that link. And from my handy desktop dictionary widget:

    hyphen n.]: the sign (-) used to join words to indicate that they have a combined meaning or that they are linked in the grammar of a sentence (as in pick-me-up, rock-forming), to indicate the division of a word at the end of a line, or to indicate a missing or implied element (as in short- and long-term).

    As your link states, “no one” without a hyphen is always grammatically correct.

  153. Jeff Alberts says:

    Woohoo! Let’s hear it for Peer Review!

    What?

    This got through?

    Oh.

    Never mind!

  154. Tim Folkerts says:

    “One station shows 48 years of August rains with a one-day maximum of 25 to 50 mm (one to two inches), and then has one August (1983) with one day when it is claimed to have rained 1016 mm (40 inches) … color me crazy, but I think that once again, as we have seen time after time, the very basic steps have been skipped. Quality doesn’t seem to be getting controlled. “

    To me this is the single biggest and most obvious problem. However correct the analysis may be, they can’t get around “garbage in, garbage out”.

  155. Cementafriend says:

    Willis, I am no expert on statistics but one little book I have on statistics, M Moroney “Facts from figures” has a chapter called “Goals, Floods, and Horse-kicks -The Poisson Distribution”. It would appear that the rainfall in my area has a distribution close to a Poisson distribution but I have not yet done a Chi square test to check the goodness of fit ( the author says “mean and variance for a Poisson are identical” – see my figures in the post above). I have not read the paper but did they look at a particularly area and compare the probablity of an extreme event (say one in fifty years) to actual events (say over two consecutive fifty year periods)?. However, as I said I believe that records are not long enough to determine accurately one in fifty year events let alone one in hundred year events. Further, rainfall is very localised as shown by events in Australia over the last year (people were killed by local flash floods- not helped by poor stormwater design)

  156. Willis Eschenbach says:

    danbo says:
    February 21, 2011 at 4:59 pm

    Willis. I know the area in the photo. Granted it floods with a heavy dew. But the f[l]ood in the picture. That was tidal or as others prefer, The levee caving in. And of course every hurricane has heavy rain. Supposedly we had a 30 ft high tide. I know where the high tide marks are. I believe it.

    My point was that whatever else was going on in New Orleans or the surrounding area with levees and sea level, Katrina was certainly an example of a 1-day extreme rain … see the caption under figure 1. I meant nothing more than that. Plus I liked the picture …

    w.

  157. Willis Eschenbach says:

    Greg Cavanagh says:
    February 21, 2011 at 6:42 pm

    I design storm water networks for a local council. Let me point out that a 1 day event will be a localised rainfall. A 5 day event is expected to be a slightly larger area event, but not guaranteed to be.

    I can’t see any point in grid celling what is essentially point data.

    I couldn’t agree more. To me, a technique like “kridging” would be better. Geologists use it because they know kriging makes more dollars, and to me it makes much more sense.

    w.

  158. AusieDan says:

    good work Willis.

  159. danbo says:

    Willis. I guess rain is relative. An inch an hour isn’t normal but not that unusual for us. (http://en.wikipedia.org/wiki/May_8th_1995_Louisiana_Flood) is an extreme. Normal is about 60+ inches a year. With about 70 thunderstorm days a year. http://coolweather.net/staterainfall/louisiana.htm

    I understand what you were doing. I brought up the issue after hearing about and seeing so much disinformation about Katrina. Trying to blame the least to blame person. Trying to claim it was AGWing. I know you weren’t doing that. But others have. And I didn’t know who used the photo with that heading. And if they claimed that was AGWing induced rain. They were less than honest.

    It was an impressive flood. You might like these photos during and after. Paris ave at the MRGO bridge. http://www.mgcollins.com/Katrina/MRGOPage.html

    Take care Willis. And keep up the good work. I appreciate your posts.

  160. Brian H says:

    Solomon, Dirkinbad, Willis —
    and all others commenting as though the 5% standard was usable..

    It is not a gold standard. Nor silver, nickle, copper, or iron. Barely makes it as a lead standard. Soft, pliable, poisonous when eaten. Suitable only for ultra-soft pseudo-sciences like Psychology and Sociology (it’s the best they can do).

    Physics likes 5-sigma stuff. How ’bout we try for that?

  161. Richard Telford says:

    Tim Folkerts says:
    February 21, 2011 at 7:35 pm

    “One station shows 48 years of August rains with a one-day maximum of 25 to 50 mm (one to two inches), and then has one August (1983) with one day when it is claimed to have rained 1016 mm (40 inches) … color me crazy, but I think that once again, as we have seen time after time, the very basic steps have been skipped. Quality doesn’t seem to be getting controlled. “

    To me this is the single biggest and most obvious problem. However correct the analysis may be, they can’t get around “garbage in, garbage out”.
    ————————-
    But would it not be useful to check if data are correct before declaring them to be false. A metre of rain, perhaps dumped by a hurricane, should have caused massive flooding and should be easy to verify from local sources. Of course it is much better for the narrative to skip this step.

  162. Willis Eschenbach says:

    Richard Telford says:
    February 22, 2011 at 12:43 am

    Tim Folkerts says:
    February 21, 2011 at 7:35 pm

    “One station shows 48 years of August rains with a one-day maximum of 25 to 50 mm (one to two inches), and then has one August (1983) with one day when it is claimed to have rained 1016 mm (40 inches) … color me crazy, but I think that once again, as we have seen time after time, the very basic steps have been skipped. Quality doesn’t seem to be getting controlled. “

    To me this is the single biggest and most obvious problem. However correct the analysis may be, they can’t get around “garbage in, garbage out”.
    ————————-

    But would it not be useful to check if data are correct before declaring them to be false. A metre of rain, perhaps dumped by a hurricane, should have caused massive flooding and should be easy to verify from local sources. Of course it is much better for the narrative to skip this step.

    A metre in a day? Yes, that’s possible, but just barely, and only in some places. However, in the record I referenced above, the monthly one-day maximum rainfall was about 2″ throughout the record … so it’s not in a rainy zone. Hang on, let me look it up …

    OK, this is good. The weather record is from Cascabel, Arizona, not far from Tucson. The date of the supposed 1016 mm rain in one day was July, 1982. Here’s the rainfall record for Cascabel. While there’s lots of Augusts with total rainfall of 3 to 5 inches, the total rain for the month of July 1982 was … wait for it … 0.8 inches.

    You are right to insist on checking, Richard. However, I’ve looked at so many weather records that I can spot a bogus one at ten paces … and in this case, I was 100% correct, the data is in fact in error. My gut is not wrong very often on this kind of question, the human eye/brain combo is a marvelous tool for finding inconsistencies. I didn’t skip the step to improve the narrative. I skipped it because I knew what I would find … and I was right.

    w.

  163. Willis Eschenbach says:

    Brian H says:
    February 22, 2011 at 12:04 am (Edit)

    Solomon, Dirkinbad, Willis —
    and all others commenting as though the 5% standard was usable..

    It is not a gold standard. Nor silver, nickle, copper, or iron. Barely makes it as a lead standard. Soft, pliable, poisonous when eaten. Suitable only for ultra-soft pseudo-sciences like Psychology and Sociology (it’s the best they can do).

    Physics likes 5-sigma stuff. How ’bout we try for that?

    I agree entirely, but I’ve given up fighting that fight. Instead I do what I did here, point out that one time in twenty a 95%CI is exceeded by random chance …

    w.

  164. Willis Eschenbach says:

    Cementafriend says:
    February 21, 2011 at 8:37 pm

    Willis, I am no expert on statistics but one little book I have on statistics, M Moroney “Facts from figures” has a chapter called “Goals, Floods, and Horse-kicks -The Poisson Distribution”. It would appear that the rainfall in my area has a distribution close to a Poisson distribution but I have not yet done a Chi square test to check the goodness of fit ( the author says “mean and variance for a Poisson are identical” – see my figures in the post above). I have not read the paper but did they look at a particularly area and compare the probablity of an extreme event (say one in fifty years) to actual events (say over two consecutive fifty year periods)?. However, as I said I believe that records are not long enough to determine accurately one in fifty year events let alone one in hundred year events. Further, rainfall is very localised as shown by events in Australia over the last year (people were killed by local flash floods- not helped by poor stormwater design)

    Good thinking, keep exploring the questions. You’ll get a better fit by using a Zipf distribution, the poisson distribution doesn’t give good values for extreme events.

    w.

  165. jgc says:

    Thanks for that very interesting post, which leaves Nature a rather dubious position regarding the quality of reviews.

    I was curious about the trends, so I did a little test in R with the Met Office Hadley Centre observations datasets (HadEX_RX1day_1951-2003)

    There are many grid cells with very little data, only 5 years in 53 years time series, so the attached plot is only for complete time series.
    The interesting values are
    Mean trend for USA: 0.08
    Mean global trend: 0.06
    Mean Global trend including incomplete series: -0.0046

    Grid cells are 300 miles across, which might be far to large an area to be meaningful

    I am a very beginner in R, so please do not take my word and check the code and data, which is here: http://tinyurl.com/4kkxg44 , together with the figure for USA

  166. jgc says:

    Adding to my previous comment, I think we should not forget that using GCMs to estimate any change in global precipitation is rather pointless.

    This figure shows how bad the performance of the main GCMs is when compared with global observations, though the discrepancy between global observations themselves is not any smaller: http://a.imageshack.us/img14/6786/gcmmodobspcp.png
    Note specially the flat lines and how models fail to capture inter-annual variability.
    data are from the climate explorer:
    http://climexp.knmi.nl

    Not to mention the very large regional biases in specific humidity of GCMs, a variable that should be related to precipitation (See: GEOPHYSICAL RESEARCH LETTERS, VOL. 34, L18704, doi:10.1029/2007GL030429, 2007)

  167. donkeygod says:

    Really excellent analysis, and thanks for it. Seems to me that much grief would be avoided if someone passed a law requiring all climate science papers to be reviewed by impartial statisticians. Unlikely, of course, but it’s pretty clear that many climate scientists are astonishingly weak in this area. Misuse, abuse, and plain ignorance of statistics is a common thread across way too many papers. Magazines with the status of Nature have no excuse for repeatedly feeding oxygen to claims which demonstrably abuse statistics, or misuse them in jejune ways. Would that some investigative reporter would take the time and trouble to learn enough statistics to take on the worst perps. It’d be a gold mine for anyone who made the effort. These concepts don’t translate easily into plain language, but it can be done. I’m still waiting to see it.

  168. Alexander K says:

    Willis, Terrific article, as is usual from you. Even though I struggled to understand the stats, I got there in the end.
    Like you, sometimes I am still an eighteen year old farmhand in my head, looking at the blue-black thunderheads piling up behind the mountains and knowing that bad weather is coming and that the sheep I will shear tomorrow, if I can get them into the shed while they are dry, will be cranky and fidgety because of the low barometer readings and will try to kick my head off.
    That long-ago experience has made me look at the seen and the unseen with weather and its by-products.
    In the case of the English floods in question , my reading tells me that the trend in precipitation levels in the UK have no particular upticks for the past several generations, but the volume of new buildings on the flood plains increased considerably at the same time as the care of the rural environment was scaled back considerably by regional councils in the thrall of a central government which had little knowledge of the requirements of the countryside for ongoing flood protection works and maintainence of river banks, clearing obstructions such as weed from rivers, streams and drains. In the view of many rural people I have discussed this with, the general consensus is that the old sensible and cautious countrymen’s opinions and practices were ignored while the Met Office preached a gospel of coming droughts during the everlasting Summer of Global Warming, but reality eventually intervened in the form of normal rainfall patterns, hence the flooding where flooding has occurred for generations. But the lack of maintainence plus the surge of building on flood plains caused the floods to seem worse than they ever had been.
    In a sense, those floods were a result of the continuous promotion of Global Warming rather than Global warming itself; the paper in question seems to be not merely in error, but a look in the wrong direction for the causes of those floods.

  169. Willis Eschenbach says:

    Alexander K says:
    February 22, 2011 at 4:49 am


    In the case of the English floods in question , my reading tells me that the trend in precipitation levels in the UK have no particular upticks for the past several generations, but the volume of new buildings on the flood plains increased considerably at the same time as the care of the rural environment was scaled back considerably by regional councils in the thrall of a central government which had little knowledge of the requirements of the countryside for ongoing flood protection works and maintainence of river banks, clearing obstructions such as weed from rivers, streams and drains.

    In the view of many rural people I have discussed this with, the general consensus is that the old sensible and cautious countrymen’s opinions and practices were ignored while the Met Office preached a gospel of coming droughts during the everlasting Summer of Global Warming, but reality eventually intervened in the form of normal rainfall patterns, hence the flooding where flooding has occurred for generations. But the lack of maintainence plus the surge of building on flood plains caused the floods to seem worse than they ever had been.

    In a sense, those floods were a result of the continuous promotion of Global Warming rather than Global warming itself; the paper in question seems to be not merely in error, but a look in the wrong direction for the causes of those floods.

    You have raised an interesting issue, Alexander. The authors assume a linear relationship between extreme rainfall events and floods.

    But as you point out, the changes in the UK landscape have altered that relationship. Here’s a good description of the issues, from the UK source for the data used in the other and equally poor Nature flood analysis (emphasis mine)

    Appraisal of Long Hydrometric Series

    Some lengthy historical data series are excellently documented and reflect close attention to the highest standards of hydrometric data acquisition appropriate to the period when the data were collected. For most older series, however, considerable curatorial skill is required to prepare the datasets for dissemination and further analysis. Validation of these early datasets is an ongoing task.

    In order to capitalise fully on important historical data series it is essential to critically review the likely data accuracy and appraise, at least qualitatively, temporal changes in artificial influences and their impact on the flow regimes and aquifer recharge patterns. Data precision and consistency can be a major problem with many early hydrometric records. Over the twentieth century instrumentation and data acquisition facilities improved but these improvements can themselves introduce inhomogeneities into the time series - which may be compounded by changes (sometimes undocumented) in the location of the monitoring station or methods of data processing employed. In addition, man’s influence on river flow regimes and aquifer recharge patterns has become increasingly pervasive, over the last 50 years especially. The resulting changes to natural river flow regimes and groundwater level behaviour may be further affected by the less perceptible impacts of land use change; although these have been quantified in a number of important experimental catchments generally they defy easy quantification.

    What he said is what you said. In general, the increase in impervious surfaces (roads, roofs, parking lots) increases the flood potential. I have seen no recognition of these land use changes in either of the Nature flood papers. The authors of the above quote conclude by saying (emphasis mine)

    It will be appreciated therefore that the recognition and interpretation of trends relies heavily on the availability of reference and spatial information to help distinguish the effects of climate variability from the impact of a range of other factors; seldom is it safe to allow the data series to speak for themselves.

    And from what I have seen, seldom is it safe to allow the climate scientists to speak for the climate …

  170. Mark T says:

    The floods along the Missouri and Mississippi Rivers (particularly early 1990s) have always been attributed to the Army Corp of Engineers making the channels along cities unmovable. Preventing long-term river bed changes comes at the cost of short term extremes in which the rain water has nowhere to go. Whole towns were destroyed as a result.

    Mark

  171. Willis Eschenbach says:

    Mark T says:
    February 22, 2011 at 1:21 pm

    The floods along the Missouri and Mississippi Rivers (particularly early 1990s) have always been attributed to the Army Corp of Engineers making the channels along cities unmovable. Preventing long-term river bed changes comes at the cost of short term extremes in which the rain water has nowhere to go. Whole towns were destroyed as a result.

    Mark

    While this is true, this study (in principle) avoids that question by using repeated simulations of the same year. Because of that, long term changes in the landscape that affect flood probability shouldn’t be a factor.

    However, in the real world it is a huge factor in real world river flow and flooding. As a result, the underlying river flow data used to calibrate the P-R model contains that trend (generally of increasing flooding for a given amount of rain). This introduces a spurious trend of unknown size into the calculations.

    w.

  172. Brian H says:

    Ah, yess. Culverts vs. streams. Here in Vancouver, Can., some effort has been made to rehab once prolific salmon streams now running beneath the ground on the mountainous North Shore in urban pipes and culverts. To the astonishment of virtually all, it sometimes works and fish come to spawn.

  173. pete says:

    Think, before you apply blindly some basic statistical method.

    You cut away the ~10% of extremest events:
    “91% of the rainfall stations in the US do not show a significant trend in precipitation extremes, either up or down.”

    … and then wonder, that there are no extreme events left.
    “No significant change in the extreme rainfall.”

    What did you expect? The extreme events are in the tails (by definition). If you want to analyse them you better not cut them away. On the contrary, you rather cut away the bulk of the events which for the analysis of extreme cases is not really interesting.

    Let’s construct a little example for you from an arbitrary –non climate related– field, such that your judgement is not blurred by your beliefs.

    A researcher measures the lengths of earthworms. He does this for 30 years now. Last year he says he noticed, that he collected more larger earthworms and more shorter earthworms than in the years before. The average length stayed the the same though.
    Now you arrive, look at his collection of earthworms. The first thing YOU do is, you remove all the largest and the shortest earthworms. Then you say: “hey, you are wrong, there is no extreme case left! And look, the average is the same as before”.

    What you little toy analysis shows is, that the average precipitation does not change. Something which is in agreement with what climate scientists say. You’ve done well this part, ..

  174. Willis Eschenbach says:

    Brian H says:
    February 23, 2011 at 11:26 pm

    Ah, yess. Culverts vs. streams. Here in Vancouver, Can., some effort has been made to rehab once prolific salmon streams now running beneath the ground on the mountainous North Shore in urban pipes and culverts. To the astonishment of virtually all, it sometimes works and fish come to spawn.

    As a man who has fished commercially for the wily and noble salmon in the oceans off the left coast of North America, from Bristol Bay Alaska on the Bering Sea to the Monterrey Bay in California’s Pacific, I can only applaud these efforts. Because they reproduce inland, salmon have been hit hard by humans changing the landscape. Life is incredibly tenacious, however, and given half a chance, nay, a tenth of a chance, salmon will charge through the opening, leaping waterfalls along the way …

    And yes, the more impervious the land surface, the more runoff, the more floods.

    w.

  175. Willis Eschenbach says:

    pete says:
    February 24, 2011 at 2:36 pm

    Think, before you apply blindly some basic statistical method.

    You cut away the ~10% of extremest events:
    “91% of the rainfall stations in the US do not show a significant trend in precipitation extremes, either up or down.”

    … and then wonder, that there are no extreme events left.
    “No significant change in the extreme rainfall.”

    What did you expect? The extreme events are in the tails (by definition). If you want to analyse them you better not cut them away. On the contrary, you rather cut away the bulk of the events which for the analysis of extreme cases is not really interesting.

    pete, these kinds of things are better posed as questions than imperatives. You have misunderstood what I have done. I have not cut away the tails and then proclaimed there is no trend in the tails, I don’t do that.

    When you see anyone, including me, saying something that seems impossible (e.g. cutting off the tails and then proclaiming no trend in the tails), the first thing to ask yourself is “have I mis-understood what they are saying”?

    A long thread like this one is a great resource in that regard. You can read through it and see if anyone else has noticed what you think you see going on. If no one else has noticed what would be a very elementary (and very important) error after 170+ comments, the odds that you have misunderstood the situation go up sharply. There’s plenty of very bright folks following the discussion, elementary errors get noticed, you can’t fool the Argus-eyed Intarwebs …

    In this case, no, I haven’t made that elementary mistake.

    w.

  176. Brian H says:

    Willis;
    The astonishment is partly from the fact that there can be no surviving salmon from those previous streams’ spawning grounds, so no “scent-impressed” returnees. Speculation abounds.

  177. Brian H says:

    Addendum;
    I think in a few cases the re-newed streams were “seeded” with eggs or hatchlings, but I’m not sure.

  178. Willis Eschenbach says:

    Brian H says:
    February 24, 2011 at 4:48 pm

    Willis;
    The astonishment is partly from the fact that there can be no surviving salmon from those previous streams’ spawning grounds, so no “scent-impressed” returnees. Speculation abounds.

    Addendum;
    I think in a few cases the re-newed streams were “seeded” with eggs or hatchlings, but I’m not sure.

    Perhaps, although somewhere in my long and intermittent career as a commercial fisherman I read that for every thousand salmon (or some such number) coming in from the ocean, one of them ends up in the wrong stream … which would actually seem to me to confer a big survival advantage, allowing your home stream’s genes to end up in some other watershed, and allowing the repopulation of streams empty of salmon for whatever reason.

    w.

  179. kadaka (KD Knoebel) says:

    If anyone’s interested, the paper has been posted online:

    http://www.warwickhughes.com/agri/Min%20Zwiers%20et%20al%20Nature%202011.pdf

    Showed up on a Google search.

  180. David Socrates says:

    Willis,

    Congratulations – you have produced a gem of a blog article. I am not a professional statistician so I can’t easily verify or refute the details of much of what you say but the logic and common sense commentary looks spot on. I had not previously looked at the HadEX extreme events database. As you have already commented here, it runs only from 1951 to 2003, so the authors of the paper cannot be criticised for starting their analysis at 1951. However I think Hadley Centre should be challenged. Rainfall data has been collected around the world for at least as long as instrumental temperature data and Hadley’s own UK rainfall statistics start in 1766. So why did they choose to start the HadEX series from 1951? And for that matter why did it end at 2003? Did it stop raining then?

    A more profound question occurs to me. What motivated Hadley in the first place to develop the HadEX series? I cannot see how mining original rainfall data and extracting extreme events from it adds any value at all. In fact, it subtracts value because it opens up a debate about how best to define an extreme event. It seems an odd statistical exercise – a bit like developing a map of a territory consisting of features above a certain level, and features below another level, with all the features in between thrown away.

    To investigate this whole issue further, I downloaded the standard Hadley rainfall series for England and Wales and plotted it from 1770 to 2010 (see http://www.thetruthaboutclimatechange.org – 1st diagram). As I suspected, it reveals that the linear trend in average annual rainfall across England and Wales over that period was a distinctly un-alarming +20mm per century!

    Call me suspicious, but I confess it did cross my mind at that stage that Hadley might have chosen to start their HadEXseries in 1951 (and terminate it in 2003) because that happened to correspond to a unusually large upward linear trend. Shame on me! Anyway, when I calculated the trend over that period it turned out to be 78mm per century – a whopping factor of 4 higher that the long term trend.

    So I decided to investigate the first one from 1770 to 1822. This turned out to have a negative trend of -59mm/century. Wow! Was I perhaps onto a cherry picking scandal? Well the answer is no. After trying a few more 53 year trends in between, I discovered that they were all over the place – some steeply positive, some steeply negative and others flat. It seemed to be just an artefact of the ‘jumpy’ nature of rainfall data. Hadley was off the hook.

    But in the end (of course) I got hooked and generated the lot – that is all 193 trend slopes, each of 53 years, between 1818 and 2010. In the resulting chart (see http://www.thetruthaboutclimatechange.org – 2nd diagram) each vertical bar is a measure of the trend in rainfall, positive or negative, in the 53 years up to and including the date where the bar is positioned. So the first bar at year 1822 shows the trend between 1770 and 1822, the second bar at 1823 shows the trend between 1771 and 1823, …and so on up to the last bar at 2010 which shows the trend between 1958 and 2010.

    I was astonished by the results. I had been expecting a fairly random scattering of slopes, some negative, some positive, which collectively would be consistent with the modest long term trend of 20mm/century. Instead I got a remarkably clear sign of a natural cyclic oscillation of about 50 year periodicity. I then generated a 20 year symmetrical running mean (the red plot) which shows this underlying cycle even more distinctly.

    All this is very interesting and unexpected. But the final irony for our MZZH11 friends is in the trend slope bar chart. Running one’s eye from left to right, what does one see? Steep slopes from 1867-1887 and from 1918-1940 indicating rapidly increasing rainfall over the upswings of those two cyclic periods – but a significantly shallower late 20th century upswing from 1970 to 2000. So it would seem that, based on Hadley’s own data, late 20th century rainfall trends in England and Wales were significantly less extreme than at any other time in the last 200 years. Inadvertently, no doubt, their chosen span for their extreme precipitation data set covers almost exactly the period when the rainfall was exhibiting its least dramatic behavior.

    Of course, the above analysis relates only to England and Wales. It would be interesting to see if this result, based on real rainfall data over a decent period (250 years) and requiring no voodo statistical techniques, scales to the rest of the world.

  181. Willis Eschenbach says:

    David Socrates says:
    February 28, 2011 at 11:06 am (Edit)

    … To investigate this whole issue further, I downloaded the standard Hadley rainfall series for England and Wales and plotted it from 1770 to 2010 (see http://www.thetruthaboutclimatechange.org – 1st diagram). As I suspected, it reveals that the linear trend in average annual rainfall across England and Wales over that period was a distinctly un-alarming +20mm per century! …

    Fascinating stuff, David. You should write it up. Where did you get the data?

    w.

  182. David Socrates says:

    Willis,

    Thanks for your encouraging response. The data is freely available from http://hadobs.metoffice.com/hadukp/ .

    I would also welcome direct contact with you, if you wish, via davidsocrates2010 at gmail.com concerning my ongoing work in this area.

    David

  183. Willis Eschenbach says:

    David Socrates says:
    March 1, 2011 at 5:57 am

    Willis,

    Thanks for your encouraging response. The data is freely available from http://hadobs.metoffice.com/hadukp/ .

    I would also welcome direct contact with you, if you wish, via davidsocrates2010 at gmail.com concerning my ongoing work in this area.

    David

    Oh, OK, same place I got my data.

    Thanks,

    w.

  184. diogenes says:

    “TheLastDemocrat says:
    February 21, 2011 at 5:35 pm

    Look at this flood — 1927 – – makes Hurricane Katrina flood look like a puddle. Regions underwater from Ponchartrain up to Kentucky and Arkansas.
    http://en.wikipedia.org/wiki/Great_Mississippi_Flood_of_1927
    Good Gauss! Put that in your distribution and try to make it look normal.”

    Bessie Smith sang Back Water Blues about poor folks sitting on the roofs of their houses during this flood…you can look at statistical data and you can look at other sources…just like the MWP

Comments are closed.