Apr 07 2008
Using Color Contours to Improve Public Interpretation of the Temperature Record
In my last post, I described how I thought that color can be used to visualize changes in sea ice extent. The use of color is often underutilized when presenting figures for consuption by the general public. The GISTEMP plots of the global temperature index is an example where I think that color could be used to help improve Joe Somebody interprete the graph.
There are 3 things that are presented in the graph:
- The actual temperature index for each year (black box with connecting dashed line)
- The 5 year running mean of the temperature index (solid red line)
- Error bars on 3 of the data points (green ‘I’ shaped things)
|
All three are important and need to be included. I like that the 5 years running mean is in red. This draws the eye to this line on the graph, and not the underlying black points. These “raw” points have too much scatter, and it’s way too easy to unintentionally (or intentionally) pick dates that show something that is not evident in the large-picture view.
Error bars are almost a necessity on graphs for journal articles, but they can be extremely confusing. What exactly are they showing in this case? “The green bars show uncertainty estimates.” That wouldn’t help a user that much. The link provided to the actual paper is also unhelpful; it points to the wrong one. Hansen, et al. (2006) is the correct link if anyone wants to send a note to GISS to correct it. Hansen et al. write:
Estimated 2σ error (95% confidence) in comparing nearby years of global temperature (Fig. 1A), such as 1998 and 2005, decreases from 0.1°C at the beginning of the 20th century to 0.05°C in recent decades. Error sources include incomplete station coverage, quantified by sampling a model generated data set with realistic variability at actual station locations, and partly subjective estimates of data quality problems.
My main concern is that users will visualize the data as points or lines. This is not good. I’ve stated in the past that global warming is boring. Changes will not easily be seen in the span of one human lifetime. If a person lives to be 100, the expected change in global temperature would be around 2C. We routinely see much larger temperature changes due to the daily and yearly cycles. The natural weather fluctuations represent much of the scatter in the above GISTEMP graphs. This is unimportant noise when talking about climate.
Instead of using the traditional scatterplot with connected points with error bars, I’ve been thinking at it might be better to present the data in such a way that would make it easier to understand the meaning of the graph. I’ve plotted the same GISTEMP data using a different technique which I think may be easier to understand.
|
Sorry for the lack of labels on the axes. The vertical axis is temperature anomaly in degrees C. The blue line is the raw GISTEMP, and is the same as the black boxes plotted on the GISS pages. But instead of plotting the error as standard error bars, it is represented by the varying shades of red. I assumed that the errors in the measurements did not change from year to year, thus this is only a representation of the noise inherent in the climate system.
Each of the colors provides an envelope around the data. It’s the changes in the envelopes that represent changes in the climate system. The actual changes in the individual data points is immaterial. There can be a very hot year, such as 1998, which has a peak of around 0.6C, but the middle of the envelope (dark red part) is lower at around 0.4C.
The plot is actually a contour plot where the colors (z axis) represent a Gaussian distribution around each data point. I also applied a 7 year running average similar to the 5 year average on the GISS plot.
A plot like the one above offers several advantages to the ones presented on the GISS website. For a casual viewer, it will provide an easier way to the take-home message; it focuses the attention on the long-term trends and not the interannual variability. It also would allow an estimation of the error “bars” at all locations instead of just three. The errors could be altered to show that estimations of measurement errors have decreased since 1900s.
(No Ratings Yet)
Loading ...Related Posts:
6 Responses to “Using Color Contours to Improve Public Interpretation of the Temperature Record”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
global warming is boring! I laughed so hard at that I didnt even need to click on it. Thanks for the nose enema. It was milk .my nostrils will recover.
Before I got to you colored version I thought you might do the spread as a gradient… like one shade of blue from 100% down to 0% at 3sig… eh whatever.
I wanted to use shading, but the program that I generate the graphics with doesn’t have that capability (to my knowledge).
[Added a short time later:]
Oops. Actually it does. I’d have to read up on how to implement it though.
Starting in 1880 is a great start, give yourself credit for that. Nothing screams global warming like tracking temps right after the Little Ice Age.
Why does the darkest part disappear here and there?
It’s an artifact of the process. I contoured then smoothed. If I had smoothed then contoured the darkest part would still be there. The other way would have looked nicer.
[...] posted a couple thoughts on the use of color when presenting figures to the public; for example Using Color Contours to Improve Public Interpretation of the Temperature Record and Using Color to Visualize Decreases in Sea Ice Extent. Color is not just a way to make a boring [...]