I’ve mentioned before how chosen color schemes greatly influence how people see surface temperature data. Frank points out that sea surface temperature presentations suffer from the same problem. – Anthony
Guest post by Frank Lansner
This is no news – but still needs to be told. NOAA can in many contexts come up with the hottest temperatures available. Here we take a look at the European Sea Surface Temperatures as of 3 may 2010.
NOAA vs. UNISYS, SST, Europe. When I look at this compare, again and again I have to check if these SST are from the very same date, 3 may 2010. But they are. Differences are immense to an extend where it hardly makes sense to look after the European SST?
NOAA is hotter than UNISYS in for example these waters:
The Baltic Sea, The North Sea, The Caspian Sea,
And in addition,
The Black Sea has NOAA Approx. 3,5 K warmer than UNISYS, and
“The NOAA hotspot area” North of Scandinavia: NOAA Approx. 4 – 6 K warmer than UNISYS .
Is there a valid sound simple explanation for these great differences?
In addition NOAA uses a colour scheme that makes Europe look as if surrounded by burning lava. It’s quite a difference to the impression you get when looking at the UNISYS graphic.
So which graphic is correct? For the Baltic, here’s what the “jury” says, SMHI (From Sweden) has an updated SST for the Baltic Sea from exactly 3 may 2010:
The 3 graphics agree reasonably for the Northern Baltic Sea, but for the rest of the Baltic Sea, SMHI shows in average around – 1,5 degrees Celsius anomaly. Both UNISYS and NOAA show too warm temperatures, but NOAA far worse than UNISYS. So, NOAA is around 2 K warmer in this area than SMHI – the best estimate.
Europe is not the only area where NOAA has warmer temperatures than UNISYS. NOAA appears markedly warmer than UNISYS on the Northern Hemisphere – but a little colder than UNISYS in areas of the Southern Hemisphere:
http://weather.unisys.com/surface/sst_anom.html
Link to NOAA SST – use “FULL GLOBAL” to see all:
http://www.osdpd.noaa.gov/PSB/EPS/SST/climo.html
Link to SMHI detailed SST for Baltic + Danish waters:
http://www.smhi.se/polarview/
============================
Visit Frank’s blog: Hide the decline
“You neglected to tell us exactly what you thought it communicated”
It communicates exactly what it’s meant to communicate – the distribution of temperature. In CFD we use the same color scheme for the inside of a furnace or a refrigerator. Using warm and cool colors means you don’t have to keep looking at the legend to work out what’s what. And you use a big spread of colors just for clarity.
I don’t know what kind of offence you think NOAA is committing by trying to tell you about SST’s. But if they are going to do it, there’s no point in making it harder for the reader with wishy-washy colors.
Frank:
If the red spot north of Scandinavia is that persistent, well.. it has to go away at some point. Show some plots during the baseline period, if you’ve worked out what it is yet. A place can’t permanently be at +4 C or whatever it is, unless they’ve made some math error.
I wonder if it used to have the white mask, for sea ice, and now it doesn’t? That could throw something off.
As for everything else: I have no idea why you resist doing your readers the basic service of making your comparison of apples to apples. Instead of telling us why we can manage to compare apples to oranges, please just give the apples to apples. It should not be hard. And again, to use a data source, you should probably know what it is. Not knowing the baseline is not knowing what it is.
You do realise that each grid box will have its own separate baseline?
Nick Stokes says,
“But if they are going to do it, there’s no point in making it harder for the reader with wishy-washy colors.”
What are you selling Nick?? Electric washing machines, Furnaces, TimeShares?? These anomaly maps are supposed to transmit information not emotion. Y’all AGW apologists are a HOOT!
Frank:
I’ve poked around a bit. Curious. Could be worth emailing somebody at NOAA, for help.
First, that hot spot is indeed fairly persistent in recent years. Though here, it got frost-bitten:
http://www.osdpd.noaa.gov/data/sst/anomaly/2005/anomnight.12.30.2005.gif
which somehow came from these absolute temperatures
http://www.osdpd.noaa.gov/data/sst/50km_night/2005/image.global50.nightime.12.30.2005.gif
The NOAA baseline is interesting. They interpolate between monthly baselines to get a baseline for a specific day, for any grid location. Anyway, 1985-1993, except 1991-1992.
I can’t figure out how to make trend maps, or download the gridded data, which would make this all much easier. One could probably ask for help on that.
Without downloading the gridded data, one is stuck eyeballing things. But with some diligence, you could check the math and see where that hotspot is coming from.
The anomalies during the baseline period are here
http://www.osdpd.noaa.gov/ml/ocean/sst/monthly_mean_sst_anom.html
And the absolute temperatures during the baseline period here
http://www.osdpd.noaa.gov/ml/ocean/sst/monthly_mean_sst.html
Throughout the baseline period, it looks like there’s a sea ice mask around that patch of ocean north of Scandinavia. Or near it, at least. So during the baseline, you get neither absolute SST nor anomaly at that spot, for much if not all of the time. If I’m looking at the right spot.
How can you find an anomaly when there is no data in the baseline period? I’m puzzled. Here’s an example May month from the baseline period:
http://www.osdpd.noaa.gov/data/sst/mean_mon/May.86.monmean.gif
See that white area with no data, north of Finland?
Now, looking at the absolute temps for May 3,
http://www.osdpd.noaa.gov/data/sst/50km_night/2010/sstnight.5.3.2010.gif
and the anomalies for May 3,
http://www.osdpd.noaa.gov/data/sst/anomaly/2010/anomnight.5.3.2010.gif
Because they changed the map view, it’s hard to compare, but it does look like they are computing anomalies for a bit of ocean where there was little or no data, during the baseline period. So if nothing else, it’d be worth asking them what is going on there.
Finally, UNISYS seems pretty difficult to use. I can’t find any description of the data source, or baseline, or anything. Seems like they want you to pay for stuff; I hope you don’t have to pay for such basic information.
I made an experiment about 6 years ago using a graph from MSU Chanel 2 that shows the warming trend 1979-2002 and shifting the color registers in Photoshop and setting the zero degree in blue, leaving purple and violets to show warming. The visual impact is completely amazing. In the original graph you’d think that Earth was set on fire, and in the modified graph the warming is confined to Siberia. Take a look at:
Original graph
Modified graph
The visual impact of reds and yellows –associated with fire and heat –is 90% of the effect sought by the alarmists.
Kuhnkat,
Here is a collection of CFD videos from Fluent. They’re mostly about bits of machinery. They’re actually fun to watch. But they are, like NOAA, trying to present info in an easily understood way. They aren’t trying to make any point about the hotness of things. And almost every one uses a color scheme like the NOAA one.
Carot eater: As a Scandinave, im quite surpriced to hear that there has been ice just North of scandinavia back in the 1980´ies – have never heard of such ice in that period, probably my mistake.
Im sure this is how NOAA every year can claim that these waters are4-5 degrees warmer than “normal” while UNISYS ssays they are slightly colder than normal.
But if true, we can actually expect those waters to freeze over now when it gets colder? Interesting!
A very local difference can be explained by ice in the baseperiod – but im still puzzled how so many very local differences between NOAA and UNISYS all the local areas where there are no ice. Can you really have 2 baseline periods (for NOAA and UNISYS) that for many years has some small local areas so quite different in temperature?
Can the Black sea in many years of NOAA baselin really be 3-3,5 K colder than all the years of the UNISYS basline period??
well if so, then there should allways be this 3-3,5 difference between NOAA and UNISYS in the black sea.
But there is not in the little test i made, here compare 1 january 2000 NOAA has black sea around +0,5K UNISYS has + 0 approx. At least nothing like 3-3,5 in difference.
http://weather.unisys.com/archive/sst/sst_anom-000102.gif
http://www.osdpd.noaa.gov/data/sst/anomaly/2000/anomnight.1.1.2000.gif
For UNISYS archived animaly charts, see files with the text “anomaly”:
http://weather.unisys.com/archive/sst/
So did UNISYS change baseline underway ? ?
Or there was one reader that sais that SST can shift dramtically from day to day. But for 6 may 2010 we still have a difference between NOAA and UNISYS in the black sea of 3-3,5 K, sp that not it.
This is indeed a “long haired” issue, but Carrot Eater, even though i personally has not spent weeks weeks and week alone on this issue without help from you WUWT readers, i just dont agree that i should not have shown you guys these differnces. They are relevant in any case since users of UNISYS and NOAA simply has a right to know that there are these big differences.
By posting at WUWT, its definetely often the input from the readers i read myself even though we have not 100% uncovered whats going on with these huge differences.
If you only want to read material by scientists that has spent way more time on the issues that I have available, then there are places where only such material are published.
Nick Stokes says:(May 7, 2010 at 1:04 pm)
“Once chosen, you accumulate a whole lot of data and published plots expressed relative to that anomaly base. There’s a big cost in changing, and no reason”
So your answer is (from my list of choices):
c. They have access to different data and do not want to share
d. No one can agree on a standard baseline for each type of anomaly chart
“So it stays, for the same sort of reason the US sticks with F.”
I have to disagree with this one. For the US to change to C from F would involve changing the way everyone thinks not just cost. People think in F not C and it would take several generations for it to become ingrained in our culture. Not so with difference anomaly baselines.
“So GISS, which started Gistemp in the ’80′s, use 1951-80. CRU, which formalised in the ’90′s, use 1961-90. Satellite data people use a period since 1979, because that’s where they have data.”
So again, it is (from my list of choices) :
c. They have access to different data and do not want to share
d. No one can agree on a standard baseline for each type of anomaly chart
with a large dose of:
b. They are arrogant and believe only their results show something special
Mr Stokes
You say that NOAA is innocent – I’ll stand with you on that. You say that skeptics paint the whole organization as a conspiracy dump – no support for you there.
We should question the color scheme use at NOAA because the establishment won’t.
Look at the NOAA scheme yourself , they start using a gradation of yellow, tending towards orange and red for all positive anomalies. Then they use a *different* color family from the other end of the spectrum, abruptly, for all negative values.
Why?
Why did they skip the in-between range of green colors? Please remember, we are just talking about colors now, and not of any connotations as yet.
Look at your CFD diesel engine animation – they have a smooth transiting green range in between blue and yellow. The Unisys map has the same transition.
Look at the NOAA color scale from 0 to 1.5 (anomaly). Can you tell the difference? I cannot. Can you tell the difference, visually, between 1.5 and 2.5? I cannot – these are represented by two very close shades of yellow. This is because they took the yellow-red part of the spectrum, stretched it out to represent one whole half of the entire anomaly range.
This only means, the color scale used is not actually good for conveying information about the anomaly fluctuations/changes of 0.5 which the map purports to depict. On the other hand, the color scale is very good in a binary/dichotomous manner – increased anomaly versus decreased anomaly – you can clearly catch that visually by looking at the NOAA map. The is because of the color family jump in the scale. There is a threshold effect at the ‘0’ point.
Representationally,
Yellow-orange-red belongs to the ‘hot’ end of the spectrum. They all convey a visual impression of warmth, do you agree? Why is a +0.5 anomaly represented by a hot color, when just 1 K away, there is a temperature point, -0.5, represented by blue, a ‘cool color’? Is this appropriate?
It is appropriate only if we believe all positive anomalies, irrespective of their amount, have a different meaning, which needs to be highlighted by a different color.
Apparently the NOAA has never heard of Roy G Biv:
Red
Orange
Yellow
Green
Blue
Indigo
Violet
The NOAA SSTa maps are color-coded mainly to highlight positive vs negative anomalies, just as I wrote, in bimodal fashion. This probably has meaning in coral reef research.
Tom in Florida
“with a large dose of:
b. They are arrogant and believe only their results show something specia”
This doesn’t make any sense. GISS could change their baseline to match CRU’s tomorrow, and the results shouldn’t show anything more or less special. It’d just nudge the absolute value of the anomalies down to match CRU’s.
Look, if everybody used the same baseline for a given sort of chart, then everybody would also have to change their own baselines every so often, as new groups or measurements came into existence to make new versions of those charts. You simply can’t win, in terms of simplicity.
So just do what everybody does now – know what it is that you’re looking at, especially if you’re doing something like the above where you are directly comparing anomaly numbers. Happily, when it comes to trends, none of this matters.
Shub
“Representationally,
Yellow-orange-red belongs to the ‘hot’ end of the spectrum. They all convey a visual impression of warmth, do you agree? Why is a +0.5 anomaly represented by a hot color, when just 1 K away, there is a temperature point, -0.5, represented by blue, a ‘cool color’? Is this appropriate?”
? Of course it’s appropriate. So long as it’s pretty much symmetric. The color at +0.5 should be as ‘warm’ as the color at -0.5 is ‘cold’.
What would be a bad colorbar is one that isn’t even centered in a reasonable way (the unisys one where blue appears on the positive side), or one where the color gradients were different on either side of zero: for example, if +4 were red, but -4 were just light blue, instead of purple.
As for your complaint about distinguishing things: yes, a colorbar without green is one that gives up some scope for visual resolution – if you used greens, then you wouldn’t have to use fine shades of other colors to portray differences. You could use green, you could not. I guess you could use light green for the first bin above 0, and dark green for the first bin below zero, if you really wanted green.
In any case, this is all rather boring. When you see a chart, look at the scale, so you know what you’re looking at. And then look for the definition of the baseline, so you know what you’re looking at.
Frank:
Mind you, I’m not saying the hotspot is there because it’s usually 5 C there, and they are subtracting a null to get an anomaly of +5. If they were, that’s quite an error. But it’s something to ask after. At the very least, I’d ask how to download the gridded data (hopefully it’s possible!); once you do that, you can resolve all these things for yourself, as you can reproduce the anomaly calculation yourself. Just keep in mind that every location has a different baseline, and not only that, but each of the 365 days of the year has its own baseline (somewhat crudely interpolated).
You can only get so far staring at maps and colorbars; at this point you need tabulated numbers.
As for UNISYS: after a reasonable look around the webpage, I could not find information about what they were plotting – the data source, or the baseline. Is the baseline also interpolated from monthly averages to the day? Who knows.
I assume UNISYS does not have its own private commercial weather satellite, so presumably they are getting it from somewhere else. So simply I would not use their map in this way before asking them what the map was. You have to know what it is.
carot eater
One can ‘complain’ only if there is scope to ‘complain’. Thanks for acknowledging (grudgingly) that green could have been used.
The anomaly color scheme is designed to display and convey bimodal information – NOAA says this on its website itself. Their color scheme is *not* suitable to display a range of values
The Unisys one is better in that respect. Yes, we can point out that the Unisys color for -7.5 is the same as its color for +4.5. But atleast it has a differential display for the middle range of anomalies.
Why would a colorbar with blue for positive values be a bad ‘centring’?
Carot eater, I think most of the problem is that NOAA for some reason has picked just 7 years (1985-91 + 1993) for their baseline.
On this SST NOAA world map,
http://www.esrl.noaa.gov/psd/map/images/sst/sst.anom.gif
they use 1982-96 as baseline and, their differences compared to UNISYS is mostly gone, compare with UNISYS:
http://www.esrl.noaa.gov/psd/map/images/sst/sst.anom.gif
So when NOAA uses more years, they start to resemble UNISYS. This does speak for the quality of UNISYS.
The Black sea anomaly for example now are practically the same. A difference of 3 K is gone.
carrot eater says:(May 8, 2010 at 5:56 am)
quoting me: “Tom in Florida:“with a large dose of:
b. They are arrogant and believe only their results show something special”
You say: “This doesn’t make any sense. GISS could change their baseline to match CRU’s tomorrow, and the results shouldn’t show anything more or less special. It’d just nudge the absolute value of the anomalies down to match CRU’s.”
I don’t mean to say the baseline change will show anything special but rather it would make it easier for the average person to understand. Just the fact that GISS won’t change their baseline to CRU’s and vice versa kind of makes my point.
“So just do what everybody does now – know what it is that you’re looking at, especially if you’re doing something like the above where you are directly comparing anomaly numbers. Happily, when it comes to trends, none of this matters.”
No problem with that but everybody does not know what it is they are looking at, reading the posts here is evidence of that, and that is my point. It seems to me that the baselines need to be clearly defined, and made plainly clear and obvious to a even a casual reader on all anomaly charts. It can’t be that hard to do!
I have learned, through reading this blog, to always look for the baseline. But it also seems to me that certain baselines are chosen to fit the argument being presented. I sometimes get the feeling that the creator of the chart is hoping that most people won’t notice a different baseline. And that keeps the old saying “figures lie and liars figure” popping in my brain.
I have attached below an email from John Sapper at NOAA regarding reliability of the SST data in polar regions:
—–Original Message—–
From: John Sapper [mailto:John.Sapper@noaa.gov]
Sent: Thursday, March 05, 2009 1:17 PM
To: Nieuwenhuis, Fred
Subject: Re: SST Anomaly charts
Hi Fred,
I should probably put a disclaimer on these images about the data not
being reliable outside the 70 North to 70 South latitude range.
I don’t think it is a sensor issue. Only that the data are very sparse
in polar regions.
Cheers,
John
Carot eater, you wrote about NOAA, basline (1985-91 + 1993). I was about to update my original article on http://www.hidethedecline.eu – but where did you get this information? I would like to use this link.
Fred: Very interesting info!
K.R. Frank Lansner