Guest essay by Sheldon Walker
It can be difficult to accurately estimate the warming rates of a temperature series, from a graph of temperature versus time. A global warming contour map makes this task easy, by colour coding all of the trends in the temperature series.
Global warming contour maps will be displayed for each of the 6 temperature series used in “Assessing recent warming using instrumentally homogeneous sea surface temperature records” by Zeke Hausfather, Kevin Cowtan, David C. Clarke, Peter Jacobs, Mark Richardson, and Robert Rohde.
The 6 temperature series are:
- Satellite radiometer record from 1997
- Buoy only record from 1997
- COBE-SST (Japanese record) from 1997
- HadSST3 from 1997
- ERSSTv3b from 1997
- ERSSTv4 from 1997
A detailed analysis of the contour maps will not be done in this article. The idea is for people to do their own analysis by looking at the contour maps. Some general comments will be made to assist people in understanding how a global warming contour map works. For example, how to work out if a temperature series has had a recent slowdown or pause.
First, the legend for global warming contour maps is displayed. This is used to convert a colour into a warming rate range, or a warming rate range into a colour. The same legend is used for all global warming contour maps.
Now the global warming contour maps for each of the 6 temperature series:
1) Satellite radiometer record from 1997
2) Buoy only record from 1997
3) COBE-SST (Japanese record) from 1997
4) HadSST3 from 1997
5) ERSSTv3b from 1997
6) ERSSTv4 from 1997
The first thing that should be noticed when looking at the 6 contour maps, is how similar they all are. Given that all of the temperature series are attempting to measure the same thing, it is reasonable to expect some similarity. But the contour maps are so similar, that you need to look closely to tell them apart.
It is worth comparing ERSSTv3b carefully with ERSSTv4. This will show you what effect “Karlization” has had on the ERSST temperature series (also used by GISTEMP).
If I had to separate these 6 contour maps based on whatever differences that there are between them, then I would group:
- the ERSSTv4 series,
- the Buoy only series, and
- the Satellite radiometer series, (this one could be in either group)
in one group.
And
- the COBE-SST (Japanese record)
- the HadSST3 series, and
- the ERSSTv3b series
in the other group.
This grouping is based on the size of the mid-green (pause or cooling) area near the centre of each contour map. The first group has a smaller mid-green area, and the second group has a larger mid-green area.
A quick comment on slowdowns and pauses. Has the recent slowdown or pause been removed?
Let me just say that:
- the light-green colour represents a warming rate of 0.0 to +1.0 degC/century, and
- the mid-green colour represents a warming rate of -1.0 to 0.0 degC/century (this is either a full pause (warming rate = zero) or cooling (warming rate < zero)
Have a look at any of the global warming contour maps above.
If you can’t see any green, then I suggest that you get your colour vision checked.
References:
“Assessing recent warming using instrumentally homogeneous sea surface temperature records” by Zeke Hausfather, Kevin Cowtan, David C. Clarke, Peter Jacobs, Mark Richardson, and Robert Rohde. Science Advances 04 Jan 2017: Vol. 3, no. 1, e1601207 DOI: 10.1126/sciadv.1601207
The datasets: data available here in a ZIP file (17KB)
Am I the only person who fails to see the utility of heat maps?
It depends on what is being conveyed. Heat maps are certainly useful since they can show multiple axes of data and make otherwise hidden patterns recognizable. However, in this case, the axes are temperature trend duration, amplitude, and date of onset and I can’t see any use for that information unless it is includes other conditions, at least current temperature, at onset.
I am colour blind.
It’s not much use to my dog, either.
Yes, nobody not being that is able to really understand what it means in everyday’s life.
By accident, I saw that Kevin Cowtan, author of a much more professional trend computing software, is color-blind too.
http://www.ysbl.york.ac.uk/~cowtan/applets/trend/trend.html
A click on ‘Further Information’, and you see it:
http://www.ysbl.york.ac.uk/~cowtan/colour/colour.html
When I saw the graphic depiction, I at first thought it was a post by Willis. – Sorry Willis…
Thank you. This makes it very easy to interpret what those contour maps are demonstrating. And yes, the effect of unjustifiable Karlization is readily visible.
well, since the tip of each of the pyramids is yellow or light green, what matter if I see darker green or blue?
Mods
i have posted a couple of comments that have disappeared. Presumably they have gone into moderation. But they should be uncontentious and do not contain any dodgy wording.
Please will you look out for these.
many thanks.
Two suggestions for the graphs:
1. Why not cut the bottom five years of it? If we are looking for middle to long term trend, the trends over only a few years are not relevant. Also all the “extreme” trends (both cooling and warming) colors are seen on the bottom and are not relevant. This would allows to groups trends into smaller categories (instead of “from 0 to 1 deg/century”, having “from 0 to 0.5 deg/century, for example) Obviously the categories must stay bigger than the uncertainty (note that you didn’t give us any uncertainty on your data)
2. I would prefer to see one of the category centered on 0deg/100y, for example “from -0.5 to 0.5 deg/100y”, instead of “from 0 to 1 deg/100y”, because this creates a “no warming nor cooling” category, while your current ranges are only “warming” or “cooling”.
I’m 72 years old. Before I die, I would like to see one revision study, just ONE, where the temperature record actually is adjusted downward. Is that even possible, or is it just that every “improvement” in data analysis necessarily leads to an upward revision? I’ve studied this subject for years. What am I missing?
Is it possible that such a study could be funded in today’s research world, a world in which Kevin Trenberth can be proven wrong about the pause (“there really wasn’t one”) but whose credibility is actually enhanced?
As much as Donald Trump worries me, I am hopeful that science can evolve so that the study I want to see would be conducted by a traditional scientific source, who might even admit that a departure from the consensus is something other than denialism?
Nick Stokes, TonyL, come on, you’re reasonable people. Is the above even possible, or is the status quo just the way science works?
I’mmm, shorry but how musch more Absinthe doIhave to drink befour the colors….
2PetitsVerres on January 6, 2017 at 3:23 am
… note that you didn’t give us any uncertainty on your data
This is exactly what I miss here – and everywhere S. Walker presents his childish toy.
If you want a more professional approach, 2PetitsVerres, you might visit really useful trend viewers like
– Nick Stokes’ Trend Viewer
https://moyhu.blogspot.de/p/temperature-trend-viewer.html
– Kevin Cowtan’s Trend Computer
http://www.ysbl.york.ac.uk/~cowtan/applets/trend/trend.html
Nick for example had a beautiful idea: that of weakening colors upon an increasing loss of significance.
Here is for example a trend & significance chart for HadSST3 from 1997 till 2016:
http://fs5.directupload.net/images/170107/bsup6tbd.jpg
Kevin has a thoroughly different approach I appreciate as much as Nick’s, because I’m often enough rather interested in singular observations than in a twodimensional trend view.
Below you see two RSS3.3 TLT trend charts, one for 1979-2016:
http://fs5.directupload.net/images/170107/j67s778h.jpg
and one for 2009-2016:
http://fs5.directupload.net/images/170107/xzaru428.jpg
You immediately see here that especially for time series originating from satellite observations, the 2σ increases exponentially with the trend period getting shorter.
That you see best when you compute, for e.g. a monthly time series, each linear estimate with its 2σ month by month, and plot the resulting monthly trend time series with the increasing 2σ around it.
So in my opinion, one could give Nick’s tool as additional parameter a 2σ level above which trends become totally insignificant, what then would be expressed by the tool as white areas in the trend charts.
Thx mod 🙂
Great graphs, showing any trend over x years for any year. Just one suggestion: as the contrast visually between yellows/reds and greens/blues is quite marked, would it not be better to make the 1st green band in the legend -1.0 to 0? This would allow immediate interpretation between heating and cooling.
Many thanks for this presentation of data.
[-> mod]
I have nothing against any comment being refused whatever the reason: I’m a guest here, not more.
But
– to give a [snipped -mod] hint would be fair;
– refusing to publish even a second one-line comment indicating that I’m missing the bigger preceeding one? Watts that? Strange, unusual behavior.
Sheldon: If you are going to compare these triangles, perhaps you should create triangles representing the DIFFERENCE between the trends in these records. That way we could see that the buoy trend is similar to ERSST4 during some period, but perhaps not during all periods. etc.
The triangle with the greatest amount of area with the color that represents a zero difference would be the best match.
If you included confidence intervals, all differences might be insignificant.