Sunday, July 12, 2009

When Hell Froze in July

Those of us who grew up in the southern part of the US know how hellishly hot it can be down here in July. Indeed, it can be hellishly hot all from May to October. During these times we all wish for cool weather all the time.

So, imagine my surprise when I was comparing the temperature records of two southern towns, Hawkinsville, GA and Eastman, GA, two towns separated by only 20 miles, when I noticed that the residents of Hawkinsville, GA got their wish, for 6 days in July 2001. Below is the raw data record for Hawkinsville, GA for July 2001. Notice that all the temperatures dropped into the low 20s from the 6th to the 11th of July. Surely this was one of the more memorable events, weatherwise, ever experienced by anyone in the south.

Sadly the surrounding towns didn't get to experience this cold weather in July as they show that they were not affected. I guess God didn't grant them a week's reprieve from the hellish temperatures which prevail at that time of the year. Maybe only atheists live in Eastman. Below is the chart for the same time for Eastman, GA just 20 miles away.

But winter in July isn't the only thing that one finds in the temperature records of used by the global warming hysteriacs (most of whom have never actually seen the raw data). The following winter in that part of Georgia was also fascinating.

I plotted the average temperature measured by each station for the winter of 2001, starting in Dec. 2001 and ending March 31, 2002. That chart can be seen below.

While the two curves are similar in the low frequency components of the temperature curve, they are not at all similar in the high frequency components. By this, I mean that on a weekly scale, the two temperatures move together, but on a daily basis, the temperature is, well, bizarre.

Subtracting the temperature of Eastman from that of Hawkinsville gives the following curve for that time frame.

Note that a 10 degree difference between the two towns during this period is not at all unusual. I count 9 times when the temperature difference was greater than 10 deg F. As noted before on this blog, such a temperature gradient is almost unheard of in meteorology, yet there it is in the data the hysteriacs use to determine global warming.

In reality what this means is that the data is so bad, and the error bar in measuring temperature is so large that one can't know what the climate is actually doing.

I next ploted the max and min temperatures for the two towns. We see the same kind of nonsense going on. Large differences between the two towns just a few miles apart.

And when the differences between the two maxes and two mins are plotted you can see the same wide error in the measurment of the local temperature.

Indeed, the averaging of the max and min make the temperature stream seem tame by comparison. I count in the above 16 times that the difference of the maximum daily temperature between these two towns exceeded 10 deg. Difference. and 19 times when the difference between the daily minimum temperatures exceeded 10 degrees. I also calculated the standard deviations of the difference of the average temperature, standard deviation of the difference of the maximum temperatures and the standard deviation of the difference of the minimum temperatures.

Let me explain. These two towns are only 20 miles apart and in some sense, each daily temperatures should be very similar to the measurement of the town only 20 miles away. One can view each day's measurments as two attempst to measure the climate at that location. The difference in daily max, min and average, tell you how accurate the temperature measuring system is. Using this time period, one can see that the measurment of the average daily temperature is only accurate to within 5.4 deg F of the true value. For the Max and Min the measuring system is only accurate to within 6.4 and 7.0 Deg. F. That means that if a global warming hysteriac wants to claim that the earth has warmed by 1.1 deg F over the past century, he has a problem. His data set is only accurate to about 5 deg, and these errors, not being statistical in nature, can't be said to be amenable to normal statistical treatment of random variation. The errors are not random.

Don't let anyone tell you that the measured temperature data is solid evidence for anything. It is utter crap.

The data again comes from Dave's favorite site--Dave who seems to still be absent after chiding me for not using this source for raw data. A source he thought which was solid and truly useful data. data source


  1. A couple points:

    1. It seems that you have yet to actually deal with the temperature record in the way the climatologists use it. Namely in gridded averages on a continental scale.

    Surely you are aware there is error in data. Be it systemic or random. It is possible to deal with this error.

    Now, I took the two stations you mentioned and I, after hopefully parsing out all the missed months etc. I found the median difference to be 0.00 between the two stations.

    That brings me to my second point.

    2. You keep talking about these data as if they defy any statistical analysis. In fact there are non-parametric methods for assessing the quality of the data.

    I will admit I'm no statistician, but it is possible to get a "confidence interval" of sorts on the MEDIAN of this data. It is skewed in that there is a tail on it. It isn't a "Normal" distribution, but that doesn't mean it isn't possible to get an idea of how close these two stations are.

    You can use the information from a binomial distribution equation to get an idea of the confidence interval on a given quantile ( It is, admittedly, imprecise and there may be other methods available.

    Now when I took the difference data between the two Georgia stations and ran these numbers I came out with the median at 0 (about point 9845 in the ranked data set of all valid points) and found the upper 95% ci at about point 9983 (0, again) and the lower 95% ci at about point 9707 (0, again).

    Again, I will defer to any real statisticians, but I really don't see this temperature record as horribly flawed. These two stations seem to track along pretty well with each other.

    Again, no one is denying that there is bad data in the record. It's nearly 60 years of data there, taken daily. That's about 19,690 data points (when you remove the points that were dropped or missed). That's a lot of data.

  2. I think I am beginning to see your point about the non-normality of the distributions. Because of the "limit theorems" in statistics, small, random, independent errors very often result in more "normal" distributions. However I would be interested to know if it is necessary that normality be retained for _all_ errors of a random nature.

    In the present case there is a highly liklihood of some "gross" errors due to failure to read the instrument correctly or instrumental problems.

    But the distribution is still quite tightly centered around 0 difference but with exceptionally long tails which deviate it from true normality.

    Perhaps you should post the histograms of the differences between the stations in the future along with a discussion of the mean, median and confidence intervals.

  3. This comment has been removed by the author.

  4. I am glad you are finally starting to understand that this data is crap and not subject to the normal gridded statistics. Data errors must be GAUSSIAN to be fixable by the methods you suggested in your first not. If they are not Gaussian, then grids will give you a number, but not the correct number.

    You mention the exceptionally long tails. Another guy and I got an article published in the Journal of Theoretical Biology because the distribution of gene orientations on eukaryotic genomes was not Gaussian (random) and one of the signs of its non-Gaussianarity (if that is a word) was that there were long long tails on the distribution.

    see Gordon Simons and Glenn Morton "The gene-orientation structure of eukaryotes" Journal of Theoretical Biology
    Volume 222, Issue 4, 21 June 2003, Pages 471-475

  5. One additional comment. You say I haven't looked at the data in a gridded form. I have. I did it for Missouri in the post A Mosey Across Missouri?