Sunday, February 28, 2010

Abilene Effect and Anomalies

This post is a continuation, in some respects, of the temperature investigations I've made over the past few weeks for the U.S. land surface records, using the data placed on-line in late 2009 by the Hadley Research Center (England) and their Climatic Research Unit (CRU). The data consists of just over 1000 records, one each for a city around the world, giving monthly average temperatures from roughly October 2009 back to, in some cases, the early 1800s. Hadley states they released this data voluntarily, after the Climate Gate scandal happened in November 2009, with the purpose to show there was nothing to hide in their data.

I posted earlier on the Abilene Effect (my phrase), showing that the small town of Abilene, Texas, has had no significant warming for the past 120 years. The Abilene Effect shows that a warming could be found if one carefully selects the last 25 years of the 20th century, from 1975 to 2000, because a series of cold winters from 1975 to 1980 skewed the results. In effect, the Abilene Effect shows that global warming (or the appearance thereof) was caused by a few cold winters. (see this link, where I wrote a bit about anomalies)

This post carries the subject just a bit further, and delves into the anomalies for Abilene using the hadCRUT3 dataset. As a caveat, I have no idea what massaging went into the numbers posted by Hadley; these may or may not be the actual monthly average temperatures measured in Abilene over all those years. I suspect, though, that the monthly averages are pretty good. As I wrote earlier, it would not be very clever (or smart) for scientists (even agenda-driven ones) to tamper with the data that is rather easily verified - or refuted. Abilene has newspaper records for daily temperatures going back many decades. If it was a cold winter, one can be sure the newspaper reported on it. Similarly for a hot summer. There also is a US Air Force Base just outside of town, Dyess AFB. One can be sure that the Air Force measured the temperature every day and got it right, plus kept the records. There are also other temperature records for that city in various other databases. So, I'm guessing that the basic data shown in my charts for Abilene are not too far from reality. (as a note, computing a monthly average from daily temperatures is at least one step in the data massaging. There is very likely an earlier step, that is, computing the daily average temperature from the high and low temperature measured for that day. The number of steps in the calculations becomes important, as shown in a moment).

Where the shenanigans begin, more than likely, is in the calculations applied to the basic data - in effect, starting with the anomalies. The anomalies, to start off with the basics, are simply the differences between one day's temperature and a long-term average of the temperature for that date. Anomalies can also be made for monthly averages, annual averages, or any other time frame. The hadCRUT3 data is given as monthly averages for each city, so the anomalies shown below are for monthly, and later for annual time frames.

One key to calculating anomalies is deciding what is to be the base period, or the long-term average for that date. HadCRUT3 uses 1961 to and including 1990, a 30-year period. This is a very interesting choice, actually. The late '60s and '70's were rather cold, as will be shown in a moment, relative to the entire 120 - year record for Abilene. I am not quite sure why that 30-year period was chosen, perhaps a reader can explain this in a comment. The choice of a cold base period makes anomalies higher when the temperatures revert back to the long-term normal, and much higher when temperatures periodically exceed the long-term normal. Conversely, it is more difficult (or more rare) to find a negative anomaly when a cold base period is used.

The Hadley data includes a value for the average monthly temperature for each month, calculated as the arithmetic average for the 30 values of that month from 1961 to 1990, inclusive. I calculated these separately to verify if Hadley can do a simple average. They were actually pretty close on this, having rounded the results to one decimal place. This is good, as a mistake here would sound several alarms.

For the charts below, I computed the anomalies as (current temperature minus average base temperature); thus a positive anomaly shows hotter than the base period, and a negative anomaly shows colder than the base period. (warm is up, cold is down; this makes sense to me). For those keeping score, this is step 2 (or maybe 3) in the process.

Once the monthly anomalies are computed, then an annual anomaly can be found by simply averaging the twelve monthly anomalies for each year. (this is at least step 3 in the process, or 4 if we count averaging the daily high and low to create the average daily temperature). The chart of annual anomalies for Abilene is shown below as Figure 1:

Figure 1
Abilene TX Monthly Temperature Anomalies
(click for larger view in new window)

Next, I calculated the simple arithmetic average of all the annual temperature anomalies, which gave a value of +0.36 deg C. Inspection of Figure 1 shows that most of the data lie above the zero line, confirming that 0.36 is a reasonable value. This also shows that the base period chosen by Hadley is indeed colder than the long-term average.

Several interesting observations can be made from Figure 1. The blue connected dots are the annual anomaly data, the black line through the center is a 10-year moving average to show trends, and the red line shows the linear trend from start to finish. The linear trend has a slope of negative 0.27 degrees C per century (-0.0027 deg C per year). The linear trend for anomalies is negative, but is at a steeper slope than that shown from my earlier chart for Abilene that uses each monthly average temperature (-.0027 compared to -.0019). This is something that happens with anomalies.

More interesting, though, is the 10-year moving average line. This shows a slightly more pronounced increase from 1890 to 1940 than the increase from 1975 to 2000, both at roughly 2.4 degrees C per century. This initial period of increasing temperature anomalies would have created havoc for the world, had scientists known about it and politicians acted on it, as some propose they do now.

The moving average also shows a decline from 1940 to 1975 of 1.6 degrees C in only 35 years - a rate of cooling of almost 5 degrees C per century. No wonder the scientists in the late 1970s were all in a panic over an impending ice age. Again, compared to the monthly average data trend for the same period, the anomalies show much more change (anomaly slope negative 4.6 degrees C per century, monthly average slope negative 2.2 degrees C per century).

The 10-year moving average also shows that CO2 cannot be the cause of any temperature changes, as the AGW proponents insist that temperatures go up as CO2 increases. Clearly, in Abilene, temperatures decreased from 1940 to 1975, yet CO2 increased as world population increased and fossil fuel consumption increased. From fundamentals of process control, one cannot control any system unless changes in the manipulated variable (CO2) have a measurable and consistent effect on the system; stated another way, if CO2 makes things warmer, then it must make things warmer at all times. CO2 does not do this.

A final point or two about Figure 1. What the AGW proponents continually harp on about is that the 2000s and the 1990s were the hottest decades on record. One would have a hard time proving that from examining the Abilene data, especially the anomalies. The black moving average trend line shows that Abilene's decades were hotter than the 1990s in the 1910s, 1920s, 1930s, 1940s, 1950s, and 1960s. Also, the 2000s were exceeded by the 1930s, 1940s, and 1950s. Finally, hottest years did not occur in the 1990s and 2000s. These temperatures were exceeded nine times, in 1910, 1921, 1927, 1933, 1934, 1938, 1939, 1946, and 1954.

To summarize the overall anomalies chart of Figure 1, the temperature trends are consistent in direction with the monthly averages, but much greater in the rate of change (slope).

The next chart, Figure 2, shows the annual anomalies for the period 1975 to 2000, that period heralded by the AGW proponents as proof that CO2 is warming the earth and that the rate of heating is unprecedented. They claim that the entire world must curb fossil fuel consumption to stop the runaway warming that is certain to result. And if all I had to look at was Figure 2, I would be alarmed, also. That is one scary chart. The black linear trend line shows a rate of increase of 2.9 degrees C per century. If that rate were to continue, the world would indeed be in trouble. But let's examine this chart.

Figure 2
Abilene TX Monthly Temperature Anomalies 1975-2000
(click for larger view in new window)

As I mentioned earlier, in the discussion on statistics in the Abilene post, one can achieve an upward slope in a line by having lower values at the beginning of the data, and higher values at the end. This is the see-saw analogy. Referring again to Figure 2, there are very low values at the beginning (lower left portion of the data), plus higher values at the end (upper right portion of the chart). Yet, the data in the middle, between those extremes, shows a downward trend! This is a prime example of the Abilene Effect, where careful selection of start and end dates show a warming. In this case, not only cold winters in the late 1970s but anomalies that show a warm trio of years for 1998, 1999, and 2000 contribute to the warming trend.

If all this is revealed in a simple analysis of one little town in the USA, one can only wonder how much additional BS (bad science) is there to be uncovered from full analysis of all the temperature records. Unfortunately (for me, at least), I do not have time to do a similar analysis on all the 86 or 87 cities in the USA's lower 48 states (see this link for charts of all 86 hadCRUT3 USA cities showing monthly average temperatures).

The conclusions are clear. Temperature anomalies create alarm where none exists; they exaggerate trends; and they are used by the alarmists to show a global warming where none exists.

UPDATE: March 2, 2010 A point about manipulating the data. Note that the hadCRU data that was released has only the monthly averages. These can easily be manipulated by not including selected months temperature, for example if one wanted to show a slight warming in recent years (say since 1975), one could leave out the data for a cold month. The resulting average of 11 months will be slightly higher than if all 12 months were included. Similarly, one could make the earlier years data a bit colder, by leaving off a hot month here and there. This is not an accusation that this type of thing was done, but it could have been done. That is but one reason that the full data set must be released for all to see, if climate scientists hope to regain their lost credibility. [end update]

Roger E. Sowell, Esq.
Marina del Rey, California

No comments: