AcuuWeather meteorologist Joe Bastardi has a question about two datasets and asks: If it is darn warm, how come there is so much sea ice?
Bastardi asks a simple question: how can we have above normal temperatures in the Arctic and the Antarctic and continue to add to the global sea ice trend? After all we’ve been told by media stories that both the Arctic and the Antarctic continue to melt at a frenetic pace. But it looks like this year we’ll see another Arctic recovery as we’ve seen in 2008 and 2009.
Bastardi also wonders about something we routinely ask about here at WUWT: data adjustments. GISS seems to be stuck with Arctic positive anomaly, yet the sea ice isn’t cooperating. Of course just having a positive temperature anomaly doesn’t guarantee melt, but members of the public who are less discerning, who look at red hot color presentations like GISS puts out, usually can’t tell the difference.
For reference here are the images Joe uses in his presentation. I’m going to help out a bit too with some simple comparisons.
First The GISS Dec-Feb 2010 Global Surface Anomaly as Joe presents it in his video:
Note that in the warmest places in the Arctic according to GISS, there are few if any land thermometers:
Above: map of GHCN2 land stations (thanks to commenter Carrick at Lucia’s)
Note the cross section of the GISS data, most of the warmth is at the Arctic where there are no thermometers. The Antarctic comes in a close second, though it has a few thermometers at bases on the perimeter of the continent plus a couple at and near the center. Note the flat plateaus are each pole.
The effects of interpolation become clearer when you do a 250 km map instead of 1200 km:
All of the sudden, the hot Arctic disappears. It disappears because there are no thermometers there as demonstrated by the cross section image which stops at about 80N.
Interestingly, the global surface anomaly also drops, from 0.80°C at 1200km of interpolation to 0.77°C with an interpolation of 250km.
One of the things that I and many other people criticize GISS for is the use of the 1951-1980 base period which they adopted as their “standard” base period. That period encompasses a lot of cool years, so anomalies plotted against that base period will tend to look warmer.
This famous GISS graph of surface temperatures from weather stations, shown worldwide in media outlets, is based on the 1951-1980 period:
Uncertainty bars (95% confidence limits) are shown for both the annual and five-year means, account only for incomplete spatial sampling of data.”]
GISS doesn’t provide a utility to replot the graph above with a different base period on their webpage here http://data.giss.nasa.gov/gistemp/graphs/ but I can demonstrate what would happen to the GISS global maps using a different base period by using their plot selector here http://data.giss.nasa.gov/gistemp/maps/
Watch what happens when we use the same base period as the UAH satellite data, which is 1979-2009. The 1200km interpolated global temperature anomaly for Dec-Jan-Feb 2010 drops more than half to 0.38°C from 0.80°C. That number is not so alarming now is it? As for the graphic, the flaming red is still there in the same places because the anomaly map colors always stay the same, no matter what the absolute temperature scale is. In the first map with the 1951-1980 base period, the max positive anomaly was 6.4°C for 1200km and 8.8°C for 250km, while in the one below with the 1979-2009 base period the max positive anomaly of 7.1C If colors were assigned to absolute temperatures, this map would look cooler than it’s counterpart with the 1951-1980 base period.
And here’s the 250km presentation, note that the global surface temp drops to 0.34°C
So it is clear, with the GISS anomaly presentation, you can look at it many different ways, and get many different answers. Who decides then which maps and graphs with what base periods and interpolations get sent out in press releases? Jim? Gavin?, Reto? Consensus over coffee at Monks?
The answer as to what base period GISS chooses in temperature anomaly maps to present to the public is easily answered by looking at their main page here: http://data.giss.nasa.gov/gistemp/
Here’s a thumbnail of the page, and the full size version of the second graph from the top, note the caption on the top of the graph:
Clearly, they prefer the base period of 1951-1980 as the default base period for the public presentation [as well as 1200 km smoothing] and by choosing that, the GISS results look a lot more alarming than they might be if a different base period was used, such as the 1979-2009 period used by UAH and RSS.
Anomalies can show anything you want based of choosing the base period. For example, a simple thought experiment. I could choose a base period from 11,000 years ago, during the last ice age, and plot maps and graphs of today’s temperatures against that base period. Would we see red? You betcha.
Here’s a graph that shows reconstructed northern hemisphere temps at the end of the last ice age 11k years ago, they were about 4.5°C cooler than today. Granted it’s not a global temp, but close enough for government work.
So if I used a 30 year slice of temperature 11,000 years before the present as a baseline period, our GISTEMP map would look something like this:
Obviously the map above is not an accurate representation, just a visual guesstimate. The more excitable who frequent here will likely cry foul at my abuse of the image. But it does illustrate how choices of colors and baseline periods can have a distinct effect on the final visual. Using a cold baseline period in the past (in this case 4.5°C globally cooler than the present) makes the present look broiling hot.
Anomalies are all about the starting choices made by people. Nature doesn’t give a hoot about anomalies. Generally, people don’t either. Imagine if your local TV weather forecaster gave tomorrow’s forecast in anomalies rather than absolute temperatures. He might say something like:
It’s going to be a hot one folks! Tomorrow we’ll have a high temperature that is 0.8C warmer than the 1951-1980 historical baseline for this city. Dress accordingly.
Useful isn’t it? Even more useful if he’s the weatherman in Svalbaard and people anticipating a heat wave go out in shorts and tank tops in mid February.
While anomalies are fine for illustrating many things, including temperature, bear in mind it’s all about the starting conditions chosen by the individuals doing the analysis. It’s all about choosing a baseline “normal”, which is subjective.
So when Joe Bastardi looks at the GISS map of the world, sees red, and wonders why we have a growing ice presence, the answer is in the choice of baseline and the choice of colors used to calculate and represent the anomaly.
Here is the Goodes.
http://en.wikipedia.org/wiki/Goode_homolosine_projection
Based on Chiefio’s recent work a base period from 1825 to 2005 would seem appropriate.
Why not calculate the baseline over the entire domain instead of just a part of it? That would make the anomoly more meaningful as a deviation from the average of the entire period.
Maybe you should have some fun with GISS charting. I tried, determined to find some way, at some time, to turn the arctic area blue. Well I finally found one:
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2010&month_last=2&sat=4&sst=0&type=anoms&mean_gen=1203&year1=1922&year2=1924&base1=1933&base2=1935&radius=1200&pol=reg
Just change the anomaly base to some arbitrary year span, or even one year as 1980-1980 and year range to something like 1972-1972. Walla (sometimes it seems not to have enough thermometers though).
Some might say that is the anomaly base, it could be, but just south of Leningrad must have been scorching if the arctic was actually hot.
But you know that chart above is quirky, I read an article someone put up here month’s ago of 1922 (I think) was when the arctic was too hot for the seals, ice was gone, etc. And see Perth and Sydney, I’ve done 30 or so charts and the always are opposite. Kind of like they are hardwired, either into GISS data and software or Mother Nature’s view of Earth. And pay attention to that Zonal Map, not sure if it reports or if it controls.
Have fun! Save copies. Never know when you might need them some day to pick you up with a smile! 😉
Oh, the link to draw your own chart is:
http://data.giss.nasa.gov/gistemp/maps/
Please! Temperatures tell nothing about warming on average. Local temperatures cannot be averaged to prove global warming.
Higher temperatures in the Arctic has nothing to do with whether the Earth is heating up or not.
If heat is transferred from a region with a higher temperature to a region with a lower temperature, the average temperature will increase but there is no energy added.
It is easy. Take a VERY simplified thought experiment. Picture two equal regions, one at +25C, another at -25C. Transfer heat from the hotter to the colder so that the hotter remains at +20. The colder region will experience a rise in temperature that is higher than 5 degrees. Because there is so much less humidity there, the relative humidity is much lower and requires less heat in order to rise in temperature.
No wonder when the latitudes around middle Europe and the US cool, then some polar regions will grow much hotter.
Or am I perfectly in the dark? Please tell.
The very NH cold weather seems to be correlated to the Arctic Oscillation and Solar Activity: Ahlbeck sez – provided by the BBC no less.
http://downloads.bbc.co.uk/looknorthyorkslincs/ahlbeck_solar_activity.pdf
Having just returned from a frigid spring in Mongolia I think we can expect more of the same. I thought it was related to the El Nino – maybe not.
Warning: statistics involved.
“So if I used a 30 year slice of temperature 11,000 years before the present as a baseline period, our GISTEMP map would look something like this: (See Chart Above and the one above this sentence.)”
_____________________
Why not use the Holocene Northern Hemisphere temperature average, looks like 15.5C using my calibrated eyeball, and go from there. Looks like the “modern” era (aka – 20th&21st Century) is fairly “average”. Why not have an international agreement, signed in Copenhagen of course, which stipulates that 15.5C is average and Clobal Warming occures when we hit 17C; at which time, the UN will find a Goreacle to guide it and the World on what we should do then.
Geir in Norway (15:10:08) :
No you’re not wrong.
If a humid area cools & then an arid area warms, assuming the energy balance has not changed, the arid area will warm more than the humid area cooled.
Several of us have been pointing this out for some time.
Temperature is a totally useless metric for energy in the atmosphere.
DaveE.
CRS, Dr.P.H. (10:14:06) :
REPLY: Dear Anu, please take a deep, relaxing breath and try to relax.
Reviewing the link you provided (which I also post regularly), we see that the trend of Arctic sea ice extent is approaching levels where it is statistically indistinguishable from the years prior to 2007, when Arctic winds moved much of the ice out of the basin (per Anthony’s post a few days ago).
—————–
“approaching levels where it is statistically indistinguishable from the years prior to 2007” = slightly above 2 standard deviations below this 1979-2000 average line
Let’s look at how important 2 standard deviations are, shall we ?
http://www.iqcomparisonsite.com/IQBasics.aspx
Two standard deviations below normal is an IQ of 68.
Moron is 50 to 69.
In standard deviation terms, the ice is right around “moron” and “borderline retarded”. 2006-2007 was “imbecile” territory.
We’ll be seeing “idiot” in a few years.
http://nsidc.org/images/arcticseaicenews/20100303_Figure3.png
Some people don’t think that’s worrying, in the Arctic ice, or their neighbors, co-workers and fellow citizens.
But I do.
Geir in Norway (15:10:08)
Your comment makes sense at the basic level, and I hadn’t thought of it that way before. However, you’re simply working with air temperature and thermal conductivity. Remember, it’s not just the air temperature that’s changing, it’s also the local surface. Imagine the steady-state (yes, a very simplified example) change in air temperature right about a lake. The total change in energy from -1C to +1C would be enormous and totally dwarf a 5C change elsewhere. Maybe polar cap extents are better measures than temperatures for those regions?
That said, maybe we should just be taking ocean surface temperature measurements worldwide. Because the heat capacity of water is so much higher than most other things, it may not suffer from these effects so badly (notable exception at 0C, of course).
-Scott
Anu (08:50:57)
That is an excellent paper. Yes, the annual changes in temperature are correlated out to a very great distance. That is true. But there is something which neither you nor GISS seem to have noticed about the Hansen paper you cite.
To illustrate the problem, here are five pseudo temperature records:
Figure W2. Five pseudo-temperature anomalies. Note the difference in the trends.
Now, what is curious about this pseudo-temperature data is that they are very similar. However, they are different both in shape and in trends. So how are they similar?
Well, here’s the answer:
Figure W3. Correlations between the five pseudo-temperature anomalies shown in Figure W2.
The similarity is that each one is highly correlated with every other one. Note that the smallest correlation between any pair of these is 0.91 … yet despite that, the trends are all over the board.
This is the problem with the GISS method, and is just the tip of the iceberg. Remember that all of these have a correlation greater than 0.9. But in the paper you cite it shows that the average correlation at 1200 km is only 0.5, and is as low as zero.
So I’m afraid I’m not impressed by Hansen’s logic. He shows annual correlation, and makes an unsubstantiated jump to trends. As Fig. W2 and the correlation table shows, this jump is totally unjustified. In other words, we cannot use the fact that the individual years are correlated to say one single word about the trends, as GISS does.
Thanks for posting that citation, Anu. I have always been skeptical about using one station to represent the trend for a giant area. But until today, I never realized what was wrong with it, so I couldn’t prove mathematically why it was bogus.
Not a scientist just your average garden variety, educated business bloke, that’s me. Some times I can’t make head nor tail out of these conflicting media reports so I fall back on my own observations as a start point.
It seems to me that the warmists are impatient for results to show how correct they are and this in itself is bringing them undone. It must be so frustrating when arctic and antarctic ice won’t follow the plot.
Likewise, in Australia pretty well every month we are told the preceding month has been the hottest on record. But they don’t feel any hotter, and I distinctly remember longer drier spells going back to my childhood. I’m surprised to discover those past temps are now recorded as cooler than reported at the time. As time goes by more and more people are going to see that the reality is widely different from the predictions.
It is 24degrees C in Melbourne today. I wonder in 5 years time what temperature they have recorded for Melbourne 25 March 2010?
Randy
Randy Del Horno (17:50:17) :
BINGO
You’ve just discovered BoM, GISS & CRU.
Congratulations.
DaveE.
Does anyone know of any research on the relationship, if any, between cosmic ray activity and solar activity. Do cosmic rays trigger more nuclear reactions, or less, or what?
Joe Bastardi was our weatherman on local radio (even though he was in State College, Penn) for a good long while, back when I was young. By that experience, I take notes when Joe talks.
He points out the inconsistency that I myself have been mentioning lately:
This link (below) shows you that, 1. Last year (orange line) the acrtic ice was at record highs in May. 2. The summer arctic ice minimum in 2009 was 25% MORE than in 2007. 3. The arctic ice is ABOVE average right now!
http://www.ijis.iarc.uaf.edu/en/home/seaice_extent.htm
Doesn’t sea ice depend on wind patterns too ?
Richard M (09:56:39) :
Phil. (08:52:06) :,
Joe is standing at the top of the hill looking down. You’re still standing at the bottom of the hill looking up. Come on up, the air is fine. 😉
Well I don’t know what this gibberish is supposed to mean. I however am looking at the data, Joe apparently is not.
Benjamin (19:41:42) :
“Doesn’t sea ice depend on wind patterns too ?”
Absolutely. Look a few posts back, mate! 🙂
http://wattsupwiththat.com/2010/03/22/the-guardian-sees-the-light-on-wind-driven-arctic-ice-loss/
Chris
Norfolk, VA, USA
Phil. (20:34:57) : edit
Phil, thanks for the link you gave above. I suspect that what he means is that Joe (and others) are looking at the recent data, while you are looking at a 30-year trend. Here’s the anomaly up to the present.
Figure W4. Arctic sea ice anomaly, from Cryosphere Today.
As you can see, while the decline was steady until 2007, since then the ice has been recovering.
There’s a bigger problem, though, which is that the Arctic ice has only been measured by satellite since 1979. Here’s the longest-term Arctic temperature record I know of, from Polyakov et al.:
Figure W5. Arctic temperatures
I’m sure you can see the problem, which is that when we look at 1979 on we are only measuring a small part of the variable Arctic changes. Looking at the larger picture reduces the need to over-react to what appears to be natural variation in temperature …
@Willis Eschenbach (17:18:05) :
Is Ps2 a constant value of 0.5 ?
Wouldn’t that make the correlation coefficient with Ps1, Ps3, Ps4 and Ps5 undefined ?
But it’s a good idea to look at what the correlations might actually mean. Ps4 and Ps5 look pretty good – if you had no other data for a few decades, it’s a better guess than plugging in the global average temperature anomaly for that grid location, which CRU does. Back in the 80’s, they were trying to adapt existing longterm weather data for new climate concerns. Today, global measurements are getting much better, such as Argo ocean armadas of floats, and multiple satellites measuring various climate data.
Plus, the paper I gave you is from 1987, when satellite data was quite young. GISS now uses such data for ocean coverage, which includes the Arctic Ocean.
So I’m afraid I’m not impressed by Hansen’s logic. He shows annual correlation, and makes an unsubstantiated jump to trends.
Those correlations were based on station pairings that had at least 50 years of data, so correlation over time is exactly what they measured. I’m sure they did similar analyses for monthly correlations too – try Google Scholar if you’re interested, it’s too late for me.
It would be nice to see actual measurements of Arctic temperatures at a high resolution for a few decades, and nail down how the temperature anomalies are related over the months and years. I’m sure someone wouldn’t mind paying tens of $millions to try and show Hansen’s approach is no better than guessing ‘average global warming’ for that month.
http://www.lanl.gov/source/orgs/ees/ees14/pdfs/09Chlylek.pdf
Let’s say the 12 high Arctic ground stations mentioned in this paper had temperature anomaly trends similar to your above Ps1, 3, 4 and 5, with respect to regions out to 1200 km from the station. Let’s say, after the methodology for combining overlapping station coverage given in that paper I cited before, that they guess some monthly Arctic surface temperature anomalies wrong for the old, sparse data years (pre-satellite coverage).
Perhaps this raises or lowers the true baseline temperature anomaly in 1951 to 1980 for these sparse gridpoints by what, 0.1 deg C, 0.5 deg C ?
In the satellite era of measurement, the warming wrt these estimated baselines is real, and the warming in the Arctic shown by this is larger than in other parts of the planet.
Thanks for posting that citation, Anu. I have always been skeptical about using one station to represent the trend for a giant area. But until today, I never realized what was wrong with it, so I couldn’t prove mathematically why it was bogus.
I’m glad you enjoyed the 1987 paper.
Perhaps the Arctic temperature anomalies were not going up and down as much as they thought back in the pre-satellite days. Maybe nobody looked at the agreement of ship-based measurements during these decades with estimates based on far-away land stations. I doubt it, but it’s possible.
Even today, satellite coverage is only up to 82.5 deg N. Perfectly polar orbits are expensive to achieve. Maybe that tiny, sparsely measured bit of Arctic is doing some wild, unexpected temperature changes…
Measurements get better and better, but then you have to relate it to the old data.
@Willis Eschenbach (22:03:20) :
The Arctic ice average 1979-2000 is higher than 1979-2009, since the ice has been melting a lot recently:
http://nsidc.org/data/seaice_index/images/daily_images/N_year_timeseries.png
Show a plot for Northern Hemispher Sea Ice Anomaly with the 1979-2000 mean, and the “recovery” is less:
http://nsidc.org/data/seaice_index/images/n_plot_hires.png
And where did Polyakov et al.: find Arctic temperature anomaly data from 1870 to 1979 that meets your high standards for reliability ? Do they have some secret sources that GISS is not aware of ? Are they merely averaging over a handful of station data from Alaska, Russia and Norway ? I’m surprised you accept this ancient data so readily… Perhaps the true Arctic temperature anomalies lie on the green line all the way back to 1875. Hard to say, given the sparse coverage up there, and the unknown correlation of temperature anomalies 🙂
Anu (22:49:55)
@Willis Eschenbach (17:18:05) :
No, it rises steadily, but very slowly.
I’m not sure what your point is here. Global data is better now, that’s true. I don’t think CRU uses global average temperature anomalies for empty grids, I’d have to see a citation for that. Global measurements are getting better, but there is still little data in the Arctic. There are no Argo floats there, for example.
Nope. They use the Reynolds and Smith data, which doesn’t include the Arctic Ocean.
Measuring the correlation of two datasets always includes time. My pseudo datasets above show the correlation for the period 1990-2010. But that does not mean that they are measuring the correlation of the trends. That’s what I show above, that measuring the correlation of two datasets (whether monthly or annual datasets) does not mean that the trends are similar.
Again, I don’t think anyone uses “average global warming” to infill the Arctic. Here’s Kevin Trenberth from the CRU emails:
So only GISS does this nonsense. Next, you say:
Well, we don’t know, do we, that’s my point … but the issue is not the guess for the old, sparse years. It is the assumption that you can estimate trends in a radius of 1200 km.
Since (as you point out below) the satellites only go to 82.5N, I’m not sure what you mean by this.
True … but what does that have to do with extrapolating trends out 1200 km from the nearest station?
Anu (23:18:40)
The recovery is identical no matter which baseline you use. If it goes up by a million square kilometres, it goes up by a million square kilometres.
Regarding your first plot, you might want to look at something more up to date. Climate data by tradition is evaluated against a 30-year average, and we finally have more than thirty years of satellite data. As a result, the baseline for that chart is 1979-2008. If you want to use a shorter time period, that’s up to you, but thirty years is what is generally used in climate science.
Well, you know, that’s why I provide citations to my sources, so you can examine them and answer these kinds of questions rather than read my (perhaps mistaken) interpretation of them, and make your own judgements on them.
Me, I don’t accept anything “readily”, and if you have a better long-term arctic temperature record, bring it on, I welcome it.
Thanks for your thoughts,
w.
Willis:
You might be interested then in looking at the case of Alert Canada. In the GHCN dataset the data only runs up to 1991, however the Environment Canada website has data through 2005. http://www.climate.weatheroffice.gc.ca/climateData/monthlydata_e.html?Prov=XX&timeframe=3&StationID=1731&Month=1&Day=1&Year=2005&cmdB1=Go
you can get the the gridded trends (actual numbers not the guess by color) by Grid cell from the Gridded map page on the GISS map maker maps at the bottom left corner of the page (You can do the same thing for Anomalies as well, and you can get it in both 250Km and 1200Km infill):
Compare the Trend for the grid Alert is in to the Station data for the period 1951 to 1990 (graph out the GISS adjusted and Environment Canada yearly anomalies for the 51-80 baseline). You will find that the 250Km and 1200Km Trends are the same and they match the Alert Station data trend at -.42°C. When you let GISS infill that Grid cell post 1990 and look at the 1951-2005 trends for 250 Km and 1200Km infill you find a divergence. 250Km gives you -.32°C and 1200km 1.16°C. Then compare those two trends to the Environment Canada data which has a 1951 to 2005 trend of .4°C.
When you do that you find that when GISS had a station in that grid it’s Gridded Cell trend matched the yearly anomaly trend for that station for both 250Km and 1200Km infill. When you don’t have that station after 1990/1 in the GISS data the 250 Km and 1200Km infill diverge, however having that data from Environment Canada lets us see that the GISS interpolation for both 250Km and 1200km do not match the trend of the actual station data.
Because of that unique case of Alert being the only station in that grid cell (According to GISS there is 6 overlapping records for that same location) and having data past when GISS uses data for that Grid you can compare GISS infilling to actual data.