Monday, May 18, 2009

Electra, California Garbage in; Garbage Out

Lets look at one station in the US Historical Climate network. I downloaded the raw data from, plotting it via their interface. It is shown below

An examination of the above chart shows several things. There is a step-function shift in 1911 or there abouts. That was probably a station move. Another shift occurred about 1960. You can see that the temperature level changes at the years I cite. Someone who is versed in dealing with data streams can see this instantly.
Then of course, immediately after 1984 or so, something very very bizarre happens to the raw data. Don't ask me what it is, but it is there and it is clearly erroneous.
No one can assert with a straight face that Electra, California suddenly heated by 17 degrees Fahrenheit between 1984 and 1986. If it had, it would have made huge headlines. "City wilts beneath the crush of global warming" would have been one of many headlines.

Well, I went to here to download the edited data. It is shown below.

Two of the step function changes are more evident on this graph and the post 1984 squirliness is gone from the data. One might think that this is well edited data, and everything is hunky dory. It isn't.

The first question one must ask is, how do they know PRECISELY how much temperature to subtract from the raw data to arrive at the final temperature? They compare Electra with surrounding cities and average it, geographically. But the nearest city to Electra in the US Historical Climate Network, is Lodi, California, 43 miles away. All others are further. And Lodi is at 49 feet elevation while Electra is at 755, hardly an equivalent site with which to use to replace bad values in Electra.

I subtracted the Edited data from the Raw data. I did it this way to show the difference as a positive number. That is the plot below.

Now, what is utterly amazing to anyone who understands statistics, even at a freshman level, if the error in the raw data is over 17 degrees, there is no way to decide to correct a given year, say 1985, by EXACTLY 17.31 deg F. There is absolutely NO way that they could reconstruct the temperature in Electra to the accuracy of .31 degrees by looking at Lodi's temperature 43 miles away or by looking at any other temperature dataset even further away. This is statistical lunacy, yet that is what your government climatological hysteriacs are saying that they can accomplish. They are saying that they can do mathematical magic to one-one hundredth of a degree. What utter crap.

Let's all give a hip-hip hooray for the magical abilities of the weather service.


  1. Once again: those giant 17 degree jumps? I haven't checked for this particular data set, but for the last 2 times you posted on similar ones I went back to the _real_ raw data from USHCN. In every case I checked, the reason for the jump was that some of the daily data was missing either in the summer or winter. The graphs you are posting take the average absolute data for the year. The "edited" data presumably takes the anomaly average for the year and uses that.

    This is a fairly trivial correction that even someone who only understands statistics at a freshman level should be able to figure out.

    (if you look at my comments on previous posts I went into more detail)

    Correcting changes due to station moves, time of observation, or such changes are admittedly more complicated, and there is probably legitimate room for disagreement on how much to trust the USHCN adjustments.

  2. Dave, I don't recall you mentioning this before, but I have been very busy. I will look into this.

  3. Just checked and I didn't see any missing data of that magnitude. One month was missing in 1984 and 85. The data I downloaded was from CO2 and they have an early 2007 download of the raw data. I do know that the USHCN has altered the raw data since that time. I don't know how one changes raw data, but that is what they did.