On Dr. Roger Pielke’s Weblog, Dr. Ben Herman discusses the maximum temperature trends with respect to the HO-83 thermometers, which have been shown to have a warm bias. He states that the “thermometers have… been replaced, but to the best of my knowledge, none of the station data have been corrected for this problem.” This post will take a look at one station’s maximum temperature data. In particular, I’ll be looking at the temperature record from Tucson, station 28815. The data is available here, and I believe it’s freely available.

The main problem is that I don’t know the exact dates when the HO-83 thermometer was used at this station. However, Steve McIntyre states that these thermometers were “introduced in the early 1990s”, and Ben Herman states that they were “were replaced in the mid to late 90’s”. He also states that the “error in the Tucson data was about 2-3 deg F.” From this information, it’s possible to see what effect the HO-83 thermometers may have had on the temperature trends in Tucson.

First, assume that the USHCN data distributed at the above site has not corrected for errors from the HO-83 thermometers. Now, define “early 1990s” as somewhere between 1990 and 1994, and “mid to late 1990s” as somewhere between 1995 and 1999. We can not calculate the new temperature data, and its corresponding trend for a range of years where the HO-83 thermometers were biasing the temperature data.

Above is a plot of the yearly averaged maximum temperature as provided by the USHCN in black. The red represents a correction to the maximum temperatures of 2.5 degrees F for the years 1992-1997. As can clearly be seen, the correction of the temperature data changes the linear trend in the temperature also. In fact, if we eyeball the fit, it appears that the new trend line is about zero. However, the above plot represents just one possible change. And even if the HO-83 was installed in 1992 and replaced in 1997, the trend would still be positive, although less than the uncorrected version as seen below.

For each starting year between 1990 and 1994, and each ending year between 1995 and 1999, I corrected the maximum temperatures from the Tucson station by 2.5 degree F. I then found the linear slope for the years 1980 through present. The results are displayed below. The control represents no adjustments to the data. (All values are deg F per decade.)

Control: 0.628+/-0.347

1990-1995: 0.683+/-0.433
1990-1996: 0.628+/-0.410
1990-1997: 0.555+/-0.374
1990-1998: 0.463+/-0.353
1990-1999: 0.354+/-0.408
1991-1995: 0.628+/-0.458
1991-1996: 0.573+/-0.439
1991-1997: 0.500+/-0.407
1991-1998: 0.409+/-0.389
1991-1999: 0.299+/-0.441
1992-1995: 0.592+/-0.422
1992-1996: 0.537+/-0.402
1992-1997: 0.463+/-0.370
1992-1998: 0.372+/-0.351
1992-1999: 0.262+/-0.410
1993-1995: 0.573+/-0.374
1993-1996: 0.518+/-0.355
1993-1997: 0.445+/-0.319
1993-1998: 0.354+/-0.300
1993-1999: 0.244+/-0.369
1994-1995: 0.573+/-0.350
1994-1996: 0.518+/-0.331
1994-1997: 0.445+/-0.295
1994-1998: 0.354+/-0.278
1994-1999: 0.244+/-0.353

When the correction is applied to the early part of the time series, the adjustment has little effect on the linear trend. However, if the adjustment is applied to the later part of the time series, the linear trend is shifted to a lower value. This is exactly as expected. The only way to know which of the above values represents reality would be to figure out when the HO-83 thermometer was installed at Tucson. Note that while the values to the linear trend vary from about 0.2-0.6, the values of the error are all fairly high, 0.3-0.4, which means that they are not significantly differenent. (No actual statistical significance tests were applied to this data. There was no attempt to account for the obvious red nature of temperature data.)

The above analysis was done only for the years 1980 through present because this period represents the “modern warming” period. However, as more years are taken into account, the adjustments will have a smaller impact upon the temperature trends. For the other stations, where the average adjustment would be around 0.6F, the change in the temperature is even less. If the temperature from the surface stations are not adjusted for the biased HO-83 thermometers they should definitely be changed to reflect more accurate values.

We can also assume that the temperature of the entire HCN needs to be corrected. Jones and Young estimate that the average correction due to the HO-83 thermometers is 0.625 degrees. Below, I’ve plotted the GISTEMP as downloaded with trend line, and then after I’ve applied the correction for the HO-83 thermometers assuming that they all went into service in the year 1994 and were retired in 1999. This would be the absolute worst case scenario.

As shown, the trend with the corrected version is less than the uncorrected, as expected. These values are statistically different after accounting for the red nature of the time series. However, as stated above, these thermometers are now out of service (in the US at least), and there bias is no longer contributing to the maximum temperature time series. Since the values from GISTEMP from 2000 onwards are the highest of the time series, and since this period was not affected by the HO-83 thermometers, I must conclude that either the bias introduced by these thermometers is not as great as Dr. Herman expects, or that GISS has already accounted for the biases of the HO-83 thermometers.

(No Ratings Yet)

 Loading ...

Related Posts:

Trackback URI |