I’ve moved to WordPress. This post can now be found at Notes On Polar Amplification#############################
This post shows, based on GISS LOTI data, there is nothing unusual about the Polar Amplification taking place during the current warming period and also shows Polar Amplification exaggerates the cooling during periods when global temperatures decline.
This post is a follow-up to my post Can Most Of The Rise In The Satellite-Era Surface Temperatures Be Explained Without Anthropogenic Greenhouse Gases?
I used GISS Land-Ocean Temperature Index (LOTI) data for the latitudes of 60S-60N in my post Can Most Of The Rise In The Satellite-Era Surface Temperatures Be Explained Without Anthropogenic Greenhouse Gases? Even though I explained why I had excluded the polar data, the fact that I had deleted it was not well received by some around the blogosphere. Why did I exclude the polar data?
First, GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data. GISS then extends land surface data out over the Arctic and Southern Oceans. Since land surface temperatures warm more than sea surface temperatures during warming periods (and cool more during cooling periods), replacing sea surface temperature data with land-based data as GISS does creates a bias during seasons of sea ice melt.
Second: As I wrote in that post, Keep in mind that the Arctic is amplifying the effects of the rise in temperature at lower latitudes. This is the basis of the concept of polar amplification. If the vast majority of the change in temperature at the lower latitudes is natural, the same would hold true for the Arctic.
My simple explanation of Polar Amplification received complaints as well, but it does not differ significantly from other definitions.
For example, the Wikipedia definition of Polar amplification : “Polar amplification is defined by International Arctic Science Committee on page 23 of the Arctic Climate Impact Assessment ‘Polar amplification (greater temperature increases in the Arctic compared to the earth as a whole) is a result of the collective effect of these feedbacks and other processes.’ It does not apply to the Antarctic, because the Southern Ocean acts as a heat sink. It is common to see it stated that ‘Climate models generally predict amplified warming in polar regions’, e.g. Doran et al. However, climate models predict amplified warming for the Arctic but only modest warming for Antarctica.”
So let’s take a look at the GISS data and see if my description of Polar Amplification is confirmed or contradicted by it.
THE GISS LOTI DATA INDICATES POLAR AMPLIFICATION EXAGGERATES TEMPERATURE INCREASES AND DECREASES
Figure 1 shows Annual (January to December) GISS LOTI data (1880 to 2010) that has been detrended. (Source: Global-mean monthly, seasonal, and annual means webpage.) Figure 1 serves as a reference for the multidecadal periods of warming and cooling used in this post. Based on the detrended annual GISS LOTI data, the recent warming period began when global temperature anomalies rose from their minimum in 1976. The cooling era runs from its maximum in 1944 to the 1976 minimum. And the early 20th Century warming period starts at the minimum in 1917 and ends in 1944. (Later, I explain why I used detrended data in Figure 1.)
The GISS website allows users to create two different types of maps at their Global Maps webpage:(1) global temperature anomalies, and (2) “trends,” which are the changes in surface temperature over user-defined periods (based on local linear trends).
GISS also creates “Zone Mean” data in their Surface Temperature Analysis output page. If you were to scroll down to the plot below the GISS map, you’d find a graph with “Latitude” as the x-axis and “Zonal Mean” as the y-axis. When “trend” maps are created using the GISS map-making feature, GISS calculates the average surface temperature change for every 2-degree latitude band from pole to pole, where the temperature change data is based on the local linear trends for the period selected. The data for those graphs is also available toward the bottom of the page.
Figure 2 shows the average Change in Annual Surface Temperature Anomalies per latitude band for the periods of 1944 to 1976 (the 20th Century cooling period) and 1976 to 2010 (the current warming period), according to the GISS data. Arctic temperatures dropped more than the rest of the hemisphere or globe during the cooling period. The reverse holds true during the current warming period. Figure 2 illustrates that, since the early 1940s, Polar Amplification exaggerates the multidecadal change in global temperature when temperatures rise and when they cool.
THE DATA FOR THE EARLIER WARMING PERIOD REVEALS THE CURRENT ARCTIC AMPIFICATION IS NOT UNUSUAL
Referring to Figure 1 again, Global Temperatures warmed from 1917 to 2010. In Figure 3, I’ve added the Zonal Mean data for the period of 1917 to 1944 to the graph. The change in Arctic surface temperature anomalies during the early warming period is comparable to the current warming period. One could conclude that the Polar Amplification during the current period is not unusual. It’s a natural response to warming.
ON THE USE OF DETRENDED GISS LOTI DATA IN FIGURE 1
Figure 4 shows the GISS LOTI data without the detrending. I used the detrended GISS LOTI data because it was difficult to determine when the early warming period started in the “raw” data, Figure 4. Was it 1907 or 1917? But based on the detrended data, the minimum occurred in 1917.
If we were to use 1907 as the start year for the early 20th Century warming period, it would make a significant difference in the Arctic warming. The GISS Zonal Mean data with that start year indicates the Arctic warmed much more during the earlier warming period than the current warming period. Refer to Figure 5. (Wouldn’t want someone to accuse me of cherry-picking the start year, would I? So I used the 1917 start date).
This post illustrated that Polar Amplification not only causes the Arctic to exaggerate a rise in global surface temperature, it also causes the Arctic to amplify a decrease in global surface temperature. It additionally showed the current warming of the Arctic is not unusual.
Someone might try to argue the Arctic data is very sparse during the early warming period, and the lack of data prevents a true comparison of the two warming periods. Realistically, one needs to consider the fact that most of the Arctic data presented by GISS in all periods is make-believe data. During seasons with sea ice, GISS uses 1200km radius smoothing to infill the vast majority of the Arctic data. That part of the GISS 1200km radius smoothing process for the Arctic is logical.
BUT (big but) sea ice melts every year, some years more than others. During seasons of sea ice melt, the fact that GISS deletes the SST data makes a significant difference, because SST varies less and has a lower trend than land surface data. This biases the Arctic (and Southern Ocean) data. GISS also deletes the SST data so they can extend the land data over open waters; that is, if they were to include the sea surface temperature data when it’s available, GISS could not extend the land data over the open ocean to reach the pole.
The Arctic amplifies the lower latitude trends when the globe is warming and when it is cooling. The Arctic should amplify any multidecadal warming or cooling signal regardless of source. If a major portion of the current warming period is due to natural variability, as suggested in the post Can Most Of The Rise In The Satellite-Era Surface Temperatures Be Explained Without Anthropogenic Greenhouse Gases?, then I believe my statement in that post was correct. And that statement was, If the vast majority of the change in temperature at the lower latitudes is natural, the same would hold true for the Arctic.
You say: “The change in Arctic surface temperature anomalies during the early warming period
(1917 to 1944) is comparable to the current warming period (1976 to 2010). One could conclude that the Polar Amplification during the current period is not unusual. It’s a natural response to warming.” Some time ago I plotted surface air temperature changes at about 900 GISTEMP stations over the periods 1910-1940, 1940-1970 and 1970-2000 against latitude and got the same general relationships as you show in your Figures. However, there was a difference between the 1910-1940 and 1970-2000 warming periods. The 1910-1940 warming was confined almost entirely to higher northern latitudes, but the 1970-2000 warming went a lot farther south. (I could send you the plots if you are interested.)
You say: “climate models predict amplified warming for the Arctic but only modest warming for Antarctica”. The models actually do a very poor job of hindcasting temperature changes by latitude - in fact they hindcast continued Arctic warming during the strong 1940-70 cooling period (I could send you these plots too if you want).
You say: “Keep in mind that the Arctic is amplifying the effects of the rise in temperature at lower latitudes.” But is it? The "Polar Amplification" theory seems to be based largely if not entirely on climate models that simulate the impacts of GHGs (but nothing else) and which get the sign of the amplification wrong for about a third of the time. Clearly this theory doesn't fit observations.
Here’s a alternative theory that does fit observations. There is no “Polar Amplification”. We aren’t
looking at a low-latitude effect that gets amplified in the Arctic. We are looking at an effect, most likely related to sea-ice extent changes, that begins in the Arctic and diminishes to the south. Or if you like, "Extra- Polar Decay.”
The quote you’re taking exception to, “climate models predict amplified warming for the Arctic but only modest warming for Antarctica,” is from Wikipedia, not me. Just a clarification.
You wrote, “The models actually do a very poor job of hindcasting temperature changes by latitude - in fact they hindcast continued Arctic warming during the strong 1940-70 cooling period (I could send you these plots too if you want).”
Interesting. That could imply a lot of things. The first that comes to mind is that the models are missing the impact of the AMO and the multidecadal North Pacific SST variations, which run in and out of synch with the AMO.
You questioned my statement, “Keep in mind that the Arctic is amplifying the effects of the rise in temperature at lower latitudes.” But as I showed in the post, they also amplify the effects of a cooling period.
With respect to your suggestion that polar amplification does not exist, and that it starts in the Arctic, there are a couple of things you could look at. Detrend Arctic temperatures and lower latitude temperatures, and see which one leads. The other point to consider is that there was a recent paper that showed a dipole relationship between the North and South poles. One warms, the other cools.
Another thought: The detrended Arctic temperatures (based on GISTEMP LOTI north of 65N) runs in a cycle that’s similar to the AMO, but it leads the AMO at times. Is that a function of how poorly the Arctic is sampled? Dunno.
The GISS 1200 km extrapolation from land to polar regions overestimates temperatures at the poles. When the ice extent is low with open seas on north coasts, adjacent land temperatures are higher than when iced in as the water is warmer than ice. Extrapolating these elevated temperatures over sea ice at the poles overestimates temperatures over these sea ice areas that are not effected by open water.
I left another comment yesterday but it hasn't appeared yet. Should I re-send it?
Roger: I don't recall seeing your second comment yesterday, and as you can see, there aren't a lot of comments here. I also checked the spam filter. Sorry. I hope you saved what you wrote in the second comment.
No, I didn't save it, but here's more or less what it said.
I tried your suggestion of detrending temperatures and comparing them at different latitudes, and I did indeed find evidence to suggest that the warming and cooling effects begin first in the Arctic and take a few years to make their way down to the tropics. I wouldn't call this evidence conclusive, but so far the theory holds up.
Second, the 5-10 year lag in the AMO relative to Arctic surface air temperatures is real. It's not a result of bad data - the series in fact track each other very closely when the lag is removed. However, this lag means that the AMO couldn't have caused the air temperature changes (much as I would like to believe otherwise) which brings up the question of what did? As you would say, Dunno.
However, the AMO is estimated from Atlantic SSTs, and it would be interesting to see whether these also show evidence for a "time delay" as we move south. Do you have any data on this?
Some confirming info:
“The warmest year in the extended Greenland temperature record is 1941, while the 1930s and 1940s are the warmest decades.” (from: Extending Greenland temperature records into the late eighteenth century, B. M. Vinther,1 K. K. Andersen,1 P. D. Jones,2 K. R. Briffa,2 and J. Cappelen3; JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D11105, doi:10.1029/2005JD006810, 2006)
And I recommend section 162: “warming of te Arctic” from:
Roger: You said you detrended temperatures at different latitudes. I'll try to comfirm what you're seeing, breaking the data down into latitude bands per continent and ocean. Give me a couple of days.
A just-published paper concludes that the radiative forcing from changes in Arctic sea ice extent is much larger than previously supposed. Could be what drives the global temperature changes?
The paper is at:
Roger: I was curious also so I looked at a few more things. In the following graphs, all datasets have been detrended and smoothed with 121-month filter.
This is a comparison of GISS Land Surface Temperature anomalies (250km smoothing) for Eurasia and North America/Greenland (both 60N-75N) with scaled global GISS Land-Ocean Temperature Index (scaling = 5). From the 1940s to present, global temperatures have led basically both Arctic datasets.
Same comparison, but I’ve used Northern Hemisphere LOTI data for comparison.
I haven’t a clue why the Arctic LST data leads before 1940.
And per your request, here’s a breakdown of detrended North Atlantic SST anomalies for different latitude bands. I didn’t scale two of the latitude bands, but scaled the other three. The scaling factors are in the parentheses.
Your third figure shows that the decrease in the AMO during the 1970s occurs about 5 years sooner at high latitudes that an low latitudes. This lag seems to be confirmed by the results I got using different data sets, but I don't know how to post the figures here. Can I send them to you directly?
Roger: Have yo tried a picture-posting site? All you have to do is save your graph as a picture graph of some type and upload. The picture-posting site (like tinypic) creates the link.
Regardless, if they confirm or agree, there's really no reason to send them.
Polar Amplification (as you've defined it) is expected both in AGW and natural variation scenarios. It's not that the Arctic is warming faster, but that the Arctic is generally where the heat ends up. So heat which starts in the ENSO regions, eventually ends up there. A little different with cooling. The Arctic,is receiving less heat and it's losing heat via the atmosphere. So if the ENSO regions (for example) are not generating heat, there is simply no heat to send to the Arctic and it naturally loses the heat it's holding at the same time.
I was surprised to see that the earlier period had more heat loss than the later period.
The following are interesting graphics of the warming trends in latitudinal band.
I had my own go at looking at the rate of polar amplification. This is defined as the temperature rise in the arctic region relative to global temperature change. I graphed the NCDC temp data for the whole globe (red) and for 65oN-90oN (blue). In order to remove the long term trends in both data sets I simply subtracted global from arctic temp (green). This left a rather nice residual which confirms what you found that the 1910-1940 period looks pretty much similar to the recent warming.
I had a quick and dirty go at calculating the polar amplification rate. I simply measured the distance from the bottom to the top of each phase of warming and cooling, crude I know. This is done for both global and arctic data and the ratio calculated. This gives very similar results for each of the warming phases, about 2.5oC arctic warming/1oC global warming. This is in the range of published estimates. As somebody posted earlier it doesn't really matter what's driving the 1oC warming the theory goes the amplification rate will increase the same because the process of arctic amplification is driven by the sea ice's response to that 1oC.
The interesting result is the 1940-1970 cooling phase. This has a arctic amplification rate of around 7-9oC. The cooling of the arctic is about the same magnitude as in the warming phases but the global cooling in this phase is much reduced. I was curious whether arctic amplification rate should be constant in both warming and cooling phases?
This result struck me as have a couple of possible implications.
1) if polar amplification rate should be similar during warming and cooling phases then this might represent independant evidence that adjusted temperature in this mid 20th century period are problematic. I noticed a bit of chatter about further adjustments around the 1940's which might have the effect of further reducing the magnitude of the cooling phase.
2) the second possibility is if the temp data sets are OK then it would suggest polar amplification during a cooling phase is much greater than during warming phase. This would suggest just a small drop in global temperature could have an enormous impact on arctic temperature.
I don't know that the second one works well with real physical processes and there are probably many other possible explanations of this but polar amplification as an independant confirmation of temperature records might be an interesting approach to take.
HR: Thanks for the discussion and the links to the graphs.
Regarding your discussion on polar amplification during cooling and warming periods, there are measurement problems: sparseness of readings, location, moves in locations, etc. Those would have to be accounted for before I would venture into any other possible reasons for the differences.
Post a Comment