Annual Temperature Plots
UCAR & Keeling CO2 Tab

The UCAR & Keeling CO2 Tab of the GHCN Temperature Plotter tool controls the display of data I obtained from UCAR's The Very, Very Simple Climate Model and CO2 data from Mauna Loa (Keeling data). There are 4 checkboxes, one data entry field, and a year-select field. The purpose of this page is to explain what each of those is for and to provide a simple tutorial on how to use them in combination with some of the other plotter controls. On the default plot (1900 to present, sensitivity equals 3°C) the correlation between temperature increase and CO2 increase is almost perfect - CO2 could be controlling the climate.

However, when the additional data back to 1800 is displayed - not so much (depending on the selected base year and climate sensitivity).

Data | Anomalies | Climate Sensitivity | Temperatures vs Anomalies | Why Filtered? | Keeling Curve | Trends


The UCAR data comes from their The Very, Very Simple Climate Model calculator. It is a flash application that I disassembled to extract the data and the algorithm they used to predict temperature based on CO2 concentration.

Concentration is a bit of a misnomer - it normally refers to the molarity, the number of moles per unit volume.

A better term is abundance which refers to parts per something - CO2 in normally expressed in parts-per-million (ppm).

The Keeling CO2 data comes from the Scripps CO2 Program and is provided as Atmospheric CO2 concentrations (ppm).


A temperature anomaly is found by

In the GHCN Temperature Plotter tool, anomalies are computed for each site and then the anomalies are added to find an average anomaly for each year.

The UCAR data is already averaged with values spaced every 10 years before 1960 and every 5 years after. Therefore, the program provides a combobox to select the baseline (zero anomaly) year. By experimentation, I found that a base year of 1975 and a sensitivity of 3°C/doubling produces an almost perfect overlap between the UCAR historical temperature data and the expected temperature based on the CO2 concentration from 1975 to 2005. This fit (correlation) is so close that everyone could (should) believe that CO2 controls the climate. Expanding the analysis to the period 1900 to 2010 adds a little noise, but the correlation is still pretty good.

Since the UCAR data actually goes from 1800 to 2010, I decided to see what the data would show. As shown in the plots below - Not even close!

To compute the CO2 controlled expected temperature difference, the following equations were used. The base year is 1975, the target is 1800 (which is why ΔT is negative), and the climate sensitivity is 3.0°C/doubling (their default). The discrepancy is the difference in the temperature UCAR provided for 1800 and the one calculated using their formula. To summarize, where the Temperature and CO2 columns contain values provided in the UCAR data As mentioned elsewhere, changing the base year just moves the historical temperature plot up and down. However, it does change the separation between the actual and expected. This is because the ratio between the provided temperature and the log of the ratio of the CO2 abundance (concentration) is not constant.

Since the entire 20th century warming is about 0.7°C, having a delta in 1800 between historical and theoretical temperatures of about 0.603°C pretty much disproves any theory correlating CO2 and temperature. The 2 values are simply too close together.

On the other hand, using 1965 as the base year and setting the sensitivity to 1.8°C/doubling produces a pretty good fit over a much longer period.

The UCAR flash model uses 2000 as the fixed baseline year. (The user can not modify this.)

At any rate, the point of my calculator is that you can vary the parameters yourself and make up your own mind.

By the way, the baseline for the GHCN data is controlled on the Basic Filters tab. The default baseline is the average over the range 1961 to 1990. When changing the baseline year on the CO2 tab, be sure to adjust the GHCN baseline so the 2 temperature curves stay aligned. I chose 1975 as the UCAR baseline year because it was near the middle of the default range and causes the temperature curves have a good overlap.

Climate Sensitivity

Climate Sensitivity is the number of degrees centigrade that the temperature is expected to increase for a doubling of CO2 in the atmosphere. The data entry field on the CO2 tab allows values from -2.0 to 7.0 (it just seemed like a reasonable range). Some extremists suggest the likely value is around 7, the skeptics like positive values under 2. The UCAR The Very, Very Simple Climate Model uses a default of 3.0, which is also the default for my application.

Temperatures vs Anomalies

It is very seldom that AGW articles contain graphs that plot temperatures - instead, it is almost always anomalies. (How they are computed was described above.)

Since the GHCN Temperature Plotter has the ability to plot either anomalies or the actual temperatures, for comparison, I plotted the GHCN temperatures and the UCAR temperatures on the same graph.

I never expected that - there is about a 2°C difference between the 2 data sets!

To be perfectly clear

The procedure to recreate the above plots is a bit more complicated than usual. (The order of the steps is not important except that the Hi/Lo button needs to be clicked after the steps above it are completed.) (This page was written, and the images made, for version 0.10 of the application. Since then, 2 additional filters were added to the table on the Basic Filters tab. In version 0.15, clicking all 5 buttons with their default values limits the selection to 388 sites, but the anomaly plot is almost identical to the one made with 520 sites, and to the one made with only the baseline selected - 4,561 sites. However, the actual temperature plots are quite different. This is because the number of included stations varies a lot each year without the stricter filters. See the next section.)

As you can see, the UCAR data is between 13.5°C and 14.0°C while the "official" GHCN data is between 11°C and 13°C. I have no way to explain this. The official temperature is supposed to be 15°C.

For comparison,

I can't remember ever seeing a plot of real temperatures in any of the reports. All they ever show is "anomalies". Maybe this is why!

Why Filtered?

When working with anomaly data, filtering the stations does not make a very big difference. However, when plotting the actual temperatures - it does.
The difference should be fairly obvious. I have left the UCAR data in the plot for reference only. In the 4,565 station plot, the temperature has 2 step increases (discontinuities)

As a result, it is necessary to only use data filtered to include the 520 stations with over 100 years of data.

(This page was written, and the images made, for version 0.10 of the application. Since then, beside the additional filters mentioned above, 7 sites identified as ship have been removed by default - of these, only 4 meet the default baseline requirement. They can be restored via the Filters tab as well as 2 obviously bad sites that were already removed. As a result, as of version 0.15, the basic baseline filter contains only 4,561 sites, not the 4,565 sites shown in the images.)

Keeling Curve

The only reliable CO2 abundance (concentration) record is the Keeling data, available from 1958 to present and derived from in situ air measurements at Mauna Loa, Observatory, Hawaii. Since I just wanted to see if that data was similar to the UCAR data, I selected one month of the CO2, seasonally adjusted filled values as representative of the year, and used August since those values were close to the UCAR values.

Using the associated checkbox on the CO2 tab, you can see that the temperatures predicted using those values are almost identical to those predicted using UCAR's CO2 values.

The tool does not plot the CO2 values, only the associated predicted changes in temperature which are proportional to ln([CO2]) - logarithm of the CO2 concentration. (The base of the logarithm is irrelevant - it only affects a constant multiplier.)

Note: The UCAR CO2 data before 1958 is highly questionable since there is no long term record from a single site. The same for temperatures before about 1870 - reliable long term records simply do not exist.


The Show trend on period... checkbox on the Hi/Lo graph turns off and on trends on most of the data series - including both UCAR datasets (Temperature and CO2) as well as the Keeling CO2 data. The following charts show trends for the following temperature series

When comparing the "adjusted and "unadjusted" (raw) GHCN data, the current temperatures are about the same while the 1900 raw data is about 0.5°C warmer. The slopes are also significantly different - 0.102°C/decade vs 0.054°C/decade or about half of the measured global warming of the twentieth century. That is a well known problem, but it is never explained. The UCAR data is somewhere between these.

In addition, the following charts compare the use of temperatures verses anomalies.

As you can see, using anomalies hides the fact that the "adjustments" tend to make the past colder. This is not just a result of averaging the data. If you look at data from individual sites, you will see that many have the same characteristic - the adjustments tend to make the past colder.

At some point, those making the adjustments will have to explain why this is happening. But so far, those explanations have not been provided!

Note: All the images on this page can be zoomed by simply using the mouse wheel.
Double click to toggle between full size and default size.

Author: Robert Clemenzi
URL: http:// / Science_Facts / Annual_Temperature_Plots / UCAR_CO2_Tab.html