One of the tactics of climate deniers to cast doubt on climate science is to attack the accuracy of the atmospheric temperature record. See this post for discussion of these types of attack.
There are five main groups that produce global temperature time series. Three of the groups use data from surface weather stations - GISS, Hadley and NOAA - and two of them use microwave measurements from sattelites - UAH and RSS.
The attacks on the temperature record are invariably aimed at the surface temperature groups and rarely or never at the satelite groups. This provides us with a simple method of checking the claims, rather than reading the many links in the previous post.
If the surface measurements are significantly different to the satelite ones then maybe the climate deniers have a case.
All of the five groups freely provide their data on the Internet. They can be found at the following locations: GISS, Hadley, NOAA, UAH and RSS
Charting the data
I downloaded all five sets of data and charted them using the statistical programming language R. The resulting chart can be seen below:
 |
Figure 1: All five atmospheric data sets presented on one chart - with data unadjusted
At first sight this chart gives some support for the denier claim, as the surface measurements show a warmer world than the satelite measurements. On closer analysis though it is clear that the temperature trend lines are quite similar, just offset from eachother. Note that the temperature values (on the Y-axis) are in fractions of a degree. They are temperature anomalies not absolute temperatures.
The following explanation for the way the anomalies are calulated comes from this post, by Tamino:
Anomaly is the difference between temperature at a given time, and the average temperature for the same time of year during some reference period. So temperature anomaly doesn’t really tell you, in absolute terms, how hot or cold it is — it tells you how much hotter or colder it is, than it was (on average) during the reference period.
It is important to realise that the five different temperature groups develop their data using four different reference periods:
* GISS : 1951 to 1980
* Hadley : 1961 to 1990
* NOAA : 1901 to 2000
* UAH and RSS : 1979 to 2000
These diffeent periods had differnet average temperatures. GISS and NOAA are the coolest (roughly the same), followed by Hadley and then the satelite groups (UAH and RSS) the warmest periods. As GISS and NOAA are calculating their anomalies by comparing with a the coldest of the reference periods is would be expected that they would show the greatest warming, followed by Hadley and then the satelite groups. This is exactly what can be seen in the chart above.
The appropriate way of placing all of the temperature groups on the same chart is by computing the anomolies for each group using a common reference period. I chose the reference period of the satelite groups as I would only have to re-calculate the anomolies for the surface groups - ie only three sets of calculations.
The method of computing the new anomolies was to calculate the mean anomoly for each of the temperature series during the chosen reference period and subtract it from each of the monthly values. I saved this data as a new file.
Figure 2 below shows the data charted. Note that the spread is much reduced and that the trend lines are quite close together and are relatively similar.
 |
Figure 2: All five atmospheric data sets presented on one chart - with data adjusted to account for different reference periods
My confidence with the accuracy of figure 2 increased when I compared it with this chart that was produced by Tamino.
Both charts are crowded and therefore a little difficult to read, so I decided to produce charts just comparing two of the data sets at a time. Figure three compares the two satelite data sets - RSS and UAH.
 |
Figure 3: RSS and UAH since 1979
These use the same data so it is not surprising that they are relatively similar - the differences can be attributed to different processing of the data. The main diffeence is in the trends - RSS has a steeper trend than UAH.
One of the claims about the CRU during the email controversy was that the CRU scientists were fudging the data to enhance the warming trend. If this were true you would expect that the Hadley data would have a significantly warmer trend than the satelite data. Figure 4 compares the RSS and Hadley data sets after adjusting them to the same reference period. What is most interesting about this chart is the similarity of the data with virtually identical trends! Note that these two groups use fundamentally different methods of calculating their data sets. If Hadley are fudging their data and processes, RSS must be too!
 |
Figure 4 : RSS and Hadley data sets compared when adusted to the same reference period
Figure 5 shows that Hadley has a stronger warming trend than UAH. This is hardly a surprise in the light of the two previous figures.
 |
Figure 5: compares Hadley with UAH adusted to the same reference period.
Lastly, how does Hadley compare to GISS? Figure 6 shows that GISS has a stronger warming trend. This is largely due to the fact that GISS attampts to take account of the Arctic (which is warming very strongly) while Hadley ignores the Arctic.
 |
Figure 6: hadley compared to GISS (same reference period)
A Skeptic Agrees
Roy Spencer, who with John Chrisite produces the UAH satelite data, agrees that there is very little difference between his data and that of HadleyCRU. In the quote below he discusses that attacks on Phil Jones and the Hadley data:
"He says he's not very organised. I'm not very organised myself," said Professor Spencer. "If you asked me to find original data from 20 years ago I'd have great difficulty too.
"We just didn't realise in those days how important and controversial this would all become - now it would just all be stored on computer. Phil Jones has been looking at climate records for a very long time. Frankly our data set agrees with his, so unless we are all making the same mistake we're not likely to find out anything new from the data anyway."
Conclusion
It is clear from this analysis that the temperature data sets are quite similar, showing warming over the last 30 years. The claims that the surface temperature data sets are manipulated to exaggerate the warming trend are clearly not supported by the evidence.
UPDATE: 16/12/2010
Tamino has written an updated post on this topic at this link.
_______________________________________________________________
Post Script
It is not necessary to download data and begin to learn charting with R to analyse the temperature data sets. Charts can be easily generated at the Wood for Trees website.
Some of my confidence in the charts that I generated for this post comes from the charts that I produced at Wood For Trees, which are virtually identical.
To view my Wood for Trees charts click on the links below:
UAH Vs RSS
Hadley Vs RSS
Hadley Vs UAH
Hadley Vs GISS
Post Post Script
Here are some of the scripts that I wrote to produce the charts.
Figure 2
R Scripts
########### All Temp time series since 1979 adjusted ##############
## STEP 1: SETUP - Source File
par(las=1)
link <- "C:\\Learn_R\\All_temp_groups_ADJUSTED.csv" ## STEP 2: READ DATA my_data <- read.table(link, sep = ",", dec=".", skip = 0, row.names = NULL, header = T, colClasses = rep("numeric", 2), na.strings = c("", "*", "-", -99.99,99.9, 999.9), col.names = c("yr","GISS","HAD","NOAA","UAH","RSS")) ## STEP 3: MANIPULATE DATA Title <- "GISS, RSS, HAD, UAH, NOAAA Temp series since 1979 adjusted" ## Ceate Plot plot(GISS ~ yr, data = my_data, type="l", col = "red", xlab = "year", ylab = "temp", main = Title) points(RSS ~ yr, data = my_data, type="l", col = "darkblue") points(HAD ~ yr, data = my_data, type="l", col = "green") points(UAH ~ yr, data = my_data, type="l", col = "black") points(NOAA ~ yr, data = my_data, type="l", col = "slategrey") ## Insert text annotations text(1980,0.65, "GISS", col="red") text(1980,0.6, "RSS", col="darkblue") text(1980,0.55, "HADLEY", col="darkgreen") text(1980,0.5, "UAH", col="black") text(1980,0.45, "NOAA", col="slategrey") ## Plot trends abline(lm(GISS ~ yr, data = my_data), col = "red") abline(lm(RSS ~ yr, data = my_data), col = "darkblue") abline(lm(HAD ~ yr, data = my_data), col = "darkgreen") abline(lm(UAH ~ yr, data = my_data), col = "black") abline(lm(NOAA ~ yr, data = my_data), col = "slategrey") Figure 4
#################### Hadley and RSS temp calc adjusted #################
## STEP 1: SETUP - Source File
par(las=1)
link <- "C:\\Learn_R\\All_temp_groups_ADJUSTED.csv"
## STEP 2: READ DATA
my_data <- read.table(link,
sep = ",", dec=".", skip = 0,
row.names = NULL, header = T,
colClasses = rep("numeric", 2),
na.strings = c("", "*", "-", -99.99,99.9, 999.9),
col.names = c("yr","GISS","HAD","NOAA","UAH","RSS"))
## STEP 3: MANIPULATE DATA
Title <- "RSS and Hadley Temp series since 1979 (Hadley adjusted)"
## Ceate Plot
plot(GISS ~ yr, data = my_data, type="l", xlim = c(1979,2010 ), ylim = c(-0.4,0.8 ), col = "white", xlab = "year", ylab = "temp", main = Title)
points(RSS ~ yr, data = my_data, type="l", col = "blue")
points(HAD ~ yr, data = my_data, type="l", col = "green")
## Insert text annotations
text(1980,0.4, "HAD", col="green")
text(1980,0.5, "RSS", col="blue")
## Plot trends
abline(lm(RSS ~ yr, data = my_data), col = "blue")
abline(lm(HAD ~ yr, data = my_data), col = "green")
No comments:
Post a Comment