CO2 and Temperature Trends

In a previous post I discussed problems with an Excel based CO2 and temperature trend chart that used 2 Y axes. Double axis charts can be misleading because they may distort the Y axis for one of the series.

In this post I show another example of am ineffective double Y axis CO2 and temperature trend chart and present 2 alternative ways to show the same data more effectively. Links to my  R scripts and Google spreadsheet based data file are included.


In this previous post,  I discussed problems with Australian Senator Steve Fielding’s double Y axis chart of temperature and CO2.  Peter Gallagher commenting on my post, used the on-line climate data charting tool to compare the 2 time series trendlines. In the post comments, I added a link to a normalized and rescaled trend line chart using the on-line tool.

Another Example of Double Axis Chart

I thought that I was done with the double axis issue until I looked at a number of  temperature anomaly trend charts on the website. If you scroll down their page, you will see a series of thumbnails showing a number of temperature anomaly trends. When you click on a thumbnail, like the UAH chart, you see the following:


I suggest  you click the image to see it in full size. A few things to notice:

  1. Chart includes both UAH temperature anomaly on the left axis  and CO2 on the right axis
  2. CO2 scale ranges from 0 t0 550 ppmv
  3. CO2 data ranges from 320 to 380, only a small fraction of the CO2 scale range
  4. JunkScience’s note at bottom: “These script-generated graphs are auto-scaled by default. Users must be aware that scales vary such that the Y1-values virtually fill the plot area and that this might (probably will) distort the impression gained from casual perusal.”

While acknowledges  that their “script-generated charts” .. “might (probably will) distort” ,  they clearly need to improve their temperature and CO2 charts.

The combination of and Senator Fielding’s double Y axis charts tells me that misuse of the dual axis chart technique is broader than I thought. So how do we handle time series with significantly different Y axis value ranges?

Charting Data With Different Ranges

Two techniques can be used in this situation to make sure that the Y axis scales do not distort data relationships:

  • Panel chart –  each data series gets a separate chart with a Y axis scale appropriate to its data range
  • Rescale each series to fit into same range -0.5 – 0.5 in the example.

Let’s demonstrate both approaches using the same data that used for their chart. Here’s a link to a Google spreadsheet that I developed to make it easy  for you to download the CO2 and UAH data to try this yourself.

Here are my  panel and rescaled data charts:

Panel Chart

Panel Chart

Rescaled Data

Rescaled Data

The panel chart shows both series in their natural scale ranges. The rescaled data shows the series on the same scale, in this example over a range of -0.5 to 0.5. The rescaled data chart let’s the reader compare the trend rates directly. The rescaled data chart shows that both CO2 and temperature are rising and the CO2 trend rate is nearly 3 times the temperature trend rate. Quite a different impression than provided by the chart.

While the proper type of chart depends on what you are trying to evaluate or show, the burden is on the chart maker to select the proper display technique to accurately  reflect the data. Senator Fielding and JunkScience.Com dual Y axis charts misrepresent the CO2 and temperature trend relationships.

R Scripts and Data Available On-line

I used 2 R scripts for this post:

  • Download and process the CO2 and UAH data and generate the panel chart – link here
  • Rescale Data Series and generate Rescaled Data Chart – link here

The CO2 and UAH data is available as a Google spreadshee: web page here, csv file here.

About these ads

2 responses to “CO2 and Temperature Trends

  1. Nice investigating and charting! Thank you.

  2. I wish that I could share a typical graph from work. This dual-axis technique is frequently used to combine about fifteen data series. Series 1 through 10 are in the same units and scale. The eleventh is in the same units but in different range, resulting in a split panel effect. The remaining series are all in different units bu similar ranges. The net effect is chaotic, but decipherable to the initiate. This design is partially driven by the choice of tools: Excel and Powerpoint. I’ve often wondered how many bad decisions have been made as a result.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s