Anthony Watts has an excellent post showing the calibre of the quality control carried out by Phil Jones and Jim Hansen and the quality of Phil Jones’ “proof” that the “overall urban bias …is greater than 0.05 deg in the 20th century”. Marysville CA (425745000030), GISS population 12,000, is in the USHCN network and is used in the GISS, CRU and NOAA calculations of global temperature. Here’s one of Anthony’s pictures; go to his site for more sickening pictures.
Marysville CA weather station. Cars with heated radiators park inches from the thermometer (among other things)
Here is a plot of the temperature history at Marysville as downloaded by Anthony from GISS
Anthony contrasted this with results from nearby Orland CA (425725910040), GISS rural area, where the temperature history is shown below:
One small tweak on Anthony’s interpretation. His GISS graphic shows the “raw” not the “adjusted” GISS version. But as far as I can tell, the NOAA calculation uses the “raw” version according to their the readme which says:
This data set contains gridded temperature anomalies calculated by the “anomaly method”. Gridpoint temperatures are calculated by averaging the unweighted raw data from all stations within the grid box, and then the anomaly is the difference from 1961-90 mean.
So this is worth bearing in mind when NOAA tells you that March 2007 was the warmest in a milllllll-yun years or was it a billlll-yun years. The “adjusted” Marysville data from GISS is shown below:
The UHI warming here isn’t as bad, but it’s still arbitrary and biased relative to Orland. (And they reduced “warm” 19th century Orland temperatures no doubt for a “good” reason.) They also reduced Orland temperatures in the 1930s; the adjusted Orland temperatures are warmer now than in the 1930s; the opposite is the case with unadjusted. I wonder how they adjusted the Marysville temperatures. But what to do with the adjustments to the Dawson, Yukon data which Rob Wilson rejected. How does one go about sometimes using their adjustments and dometimes not.
But let’s say that they got all their adjustments exactly right. What does that say about the quantum of UHI? Here’s a town of 12,000 which qualifies as “rural” in all the UHI studies. The non-climatic effect here is at least 3 deg C. In this case, we can estimate it because there’s a less bad station nearby. What does one do in China or Indonesia where nearly all the stations seem to be in large cities (even if they started out small, they’re large now)?
[UPDATE: Eli Rabett has claimed that I “blew” it when I said that this “town of 12,000 qualifies as ‘rural’ in all the UHI studies”, because it is not used as a GISS unlit site. The phrase “UHI studies” is not especially clear, but is intended to cover studies like Jones et al 1990, Karl et al 1998 purporting to show that there is negligible UHI effect in global composites; the GISS composite is not a “UHI study”. Jones et al 1990 used Chinese cities with populations under 100,000 as “rural” comparanda. Karl et al 1988 said:
Due to the large number of stations located in sparsely populated areas [over 85% (70% of all stations had a 1980 population of less than 25 000 (10 000)], the impact of urbanization is not large in relation to decadal changes of temperature in the United States.
Marysville falls squarely into Karl’s definition of a “sparsely populated” area. The salient point is that the GISS adjsutment estimated a substantial UHI effect at Marysville requiring adjustment; if this adjustment is justified, then GHCN and CRU (and IPCC relying on CRU) should have used it. Eli can’t have it both ways. [End UPDATE Jul 30, 2007]
And what can one say about quality control as practiced by Phil Jones and Jim Hansen if they let this sort of stuff into their network? Maybe they should spend a little less time going to conferences and a little more time doing quality control on the temperature data that they’re being paid to report on.
Here are plots of the GISS adjustments for Marysville and Orland. I haven’t tried to wade through GISS adjustments yet. At a first impression, the adjustments look to implemented in 5-year steps. This is obviously something different than Time-of-Observation adjustments or step adjustments. Any thoughts on what they’re doing would be welcome.
Here is the difference between the GHCN raw monthly data and the USHCN raw (“Areal”) data.
A GISS Rural Cooling Adjustment?
Here’s another odd plot, which I’ll have to triplecheck. The first diagram below shows 3 calculation stages: GHCN Raw- GISS Raw (which should deal with all the time of observation and station history stuff); GISS Raw – GISS Adjusted (the GISS urban heat adjustment); the combined GHCN Raw – GISS Adjusted. I’ve done this first for Marysville GISS population 12,000, periurban lighting; and Orland GISS rural; unlit. We’ve seen a strong UHI effect at Marysville and GISS makes a substantial adjustment in the early portion of this series – up to 2 deg C, notwithstanding Jones view that UHI does not exceed 0.05 deg per century.
Now the same things with Orland CA. This time the GISS adjustment goes the other way – I presume that the rationale for this adjustment is the well-known “rural cooling” effect, but I’ll ocntinue investigating.