Guest post by Steven Goddard
Several people keep asking why am I focused on winter snow extent. This seems fairly obvious, but I will review here:
- Snow falls in the winter, in places where it is cold. Snow does not generally fall in the summer, because it is too warm.
- Winter snow extent is a good proxy for winter snowfall. Snow has to fall before it can cover the ground.
So what about summer snow cover? Summer snow cover declined significantly (from the 1970s ice age scare) during the 1980s, but minimums have not changed much since then. As you can see in the graph below, the overall annual trend since 1989 has been slightly upwards.
Data from Rutgers University Global Snow Lab
Note in the image above that there has been almost no change in the summer minimum snow extent since 1989, and that the winter maximums have increased significantly as seen below.
Summer snow cover is affected by many factors, but probably the most important one is soot, as Dr. Hansen has stated.
The effects of soot in changing the climate are more than most scientists acknowledge, two US researchers say. In the Proceedings of the National Academy of Sciences, they say reducing atmospheric soot levels could help to slow global warming relatively simply. They believe soot is twice as potent as carbon dioxide, a main greenhouse gas, in raising surface air temperatures. … The researchers are Dr James Hansen and Larissa Nazarenko, both of the Goddard Institute for Space Studies, part of the US space agency Nasa, and Columbia University Earth Institute.
http://news.bbc.co.uk/2/hi/science/nature/3333493.stm
The global warming debate has until now focused almost entirely on carbon dioxide and other greenhouse gas emissions, but scientists at the University of California – Irvine, suggest that a lesser-known problem – dirty snow – could explain the Arctic warming attributed to greenhouse gases….The effect is more conspicuous in Arctic areas, where Zender believes that more than 90 percent of the warming could be attributed to dirty snow.
http://www.scienceagogo.com/news/20070506202633data_trunc_sys.shtml
In summary, winter snowfall is increasing and currently at record levels, and summer snow extent is not changing much. Earlier changes in summer snow extent were likely due primarily to soot – not CO2.
Why Is Winter Snow Extent Interesting?
Several people keep asking why am I focused on winter snow extent. This seems fairly obvious, but I will review here:
1. Snow falls in the winter, in places where it is cold. Snow does not generally fall in the summer, because it is too warm.
2. Winter snow extent is a good proxy for winter snowfall. Snow has to fall before it can cover the ground.
So what about summer snow cover? Summer snow cover declined significantly (from the 1970s ice age scare) during the 1980s, but minimums have not changed much since then. As you can see in the graph below, the overall annual trend since 1989 has been slightly upwards.
Data from Rutgers University Global Snow Lab
Note in the image above that there has been almost no change in the summer minimum snow extent since 1989, and that the winter maximums have increased significantly as seen below.
Summer snow cover is affected by many factors, but probably the most important one is soot, as Dr. Hansen has stated.
The effects of soot in changing the climate are more than most scientists acknowledge, two US researchers say. In the Proceedings of the National Academy of Sciences, they say reducing atmospheric soot levels could help to slow global warming relatively simply. They believe soot is twice as potent as carbon dioxide, a main greenhouse gas, in raising surface air temperatures. … The researchers are Dr James Hansen and Larissa Nazarenko, both of the Goddard Institute for Space Studies, part of the US space agency Nasa, and Columbia University Earth Institute.
http://news.bbc.co.uk/2/hi/science/nature/3333493.stm
The global warming debate has until now focused almost entirely on carbon dioxide and other greenhouse gas emissions, but scientists at the University of California – Irvine, suggest that a lesser-known problem – dirty snow – could explain the Arctic warming attributed to greenhouse gases….The effect is more conspicuous in Arctic areas, where Zender believes that more than 90 percent of the warming could be attributed to dirty snow.
http://www.scienceagogo.com/news/20070506202633data_trunc_sys.shtml
In summary, winter snowfall is increasing and currently at record levels, and summer snow extent is not changing much. Earlier changes in summer snow extent were likely due primarily to soot – not CO2.
Steve,
You got your data from Rutgers University Global Snow Lab.
Have you taken the time to examine their snow cover anomaly chart? Its well done, and tells a bit of a different story:
http://climate.rutgers.edu/snowcover/chart_anom.php?ui_set=0&ui_region=nhland&ui_month=1
Similarly, just winter extent:
http://climate.rutgers.edu/snowcover/chart_seasonal.php?ui_set=nhland&ui_season=1
As Anthony is fond of saying, another entry for the “weather is not climate” department.
Michael Beenstock’s paper (http://wattsupwiththat.com/2010/02/14/new-paper-on/) and my brilliant analysis (http://www.2bc3.com/warming.html) seem to show that there is a feedback mechanism that makes the temperature increase due to CO2 temporary. Why this should be so is in question. Perhaps show cover has a role in this somewhere.
@DirkH
Did we read the same post?
Snow is atmospheric humidity and temperatures below 32°F – or am I wrong?
We have more snow in winters since 1989 now – on the Northern Hemisphere, that’s what the graph suggests and what it feels like – but only this winter in Scotland, not the last 10 or so years.
Are you suggesting that we have more snow because we have the same amount of humidity in the atmosphere but the temperatures dropped considerably so what was rain is now snow? Is there any data on that?
If the temperatures have not dropped and the humidity has not increased then where does the snow come from?
From supercharging?
Leif Svalgaard
I grabbed the data rather crudely from the graph with an on-screen ruler, and looked at the curve in Excel.
If we accept an approximation as a two factor scatter, then we get an R2 of 0.307.
Taking the square root of this as r the correlation coefficient, we get 0.554.
Applying an r of 0.554 to the correlation table, taking (also a crude approximation) the number of years minus 2 as the degrees of freedom, then the P value is between 0.01 (r=0.537) and 0.001 (r=0.652).
If I’ve cocked this up, let the vultures descend!
“John Innes (08:12:49) :
For coal-fired power stations, and perhaps to a lesser extent for other soot-generators, it should be cheaper to reduce the soot emissions than to extract and sequester the CO2 generated and emitted. …
Surely we can work from this as a starting point, and get more bang for our warming reduction buck by reducing soot instead of CO2? …”
Well put. Let’s save people who are now inhaling enough soot to kill a half million people per year instead of worrying about what people 4 generations down the road will do to respond to gradual sea level rise of less than a foot per century.
R^2 on the winter graph since 1989 is 0.298514013
R^2 on the winter graph since 1999 is 0.312822303 Slope is 317,000 km2/year
R^2 on the weekly graph since 1966 is 0.004697147
R^2 on the weekly graph since 1989 is 5.25176E-05
Zeke Hausfather (09:55:07) :
Notice they always use Anomalies, but anomalies from what, there is no “normal” in earth’s climate. They always choose the “Base”. Try using the MWP as the base and everything changes. Try using the period when the vikings were in Greenland instead and see what you get.
There are good men in The Netherlands too;
http://www.probeinternational.org/files/UKVersieHenkTennekes.pdf
Leif,
Couple points – 1. its very rare to use an R^2 value in a time series, unless you expect the relationship to be linear or monotonic. 2. Applying an R^2 to a linear fit of an oscillating system also makes no sense whatsoever! Its exactly things like this that cause skeptics to not be taken seriously – you fail in basic statistics, but you expect to be able to understand a complex climate system? Give me a break!
Christian A. Wittke (10:01:43) :
The world is pretty big, and the climate is wonderfully complex. Dirk H. knows his physics, and is gentle with folks that don’t know it as well.
I tend to be a bit more blunt. Be careful using single point variables like atmospheric humidity and temperature to try to describe a giga-variabled system.
A C Osborn,
Setting a different period as the ‘base’ wouldn’t change the shape of the graph, only the zero-intercept of the data. The result is still the same.
Steve,
If you plot the average Dec-Feb snow extent against the PDO index, you get a very interesting correlation…
Winter Snow Extent vs PDO
As far as the r-squared goes… It’s a cyclical stochastic process. Unless you have a long enough time series, no reasonable trend line will yield a decent r-squared.
phlogiston (10:08:42) :
if we accept an approximation as a two factor scatter, then we get an R2 of 0.307.
A good way to think of R2 is that it is simply the fraction of the variation that is due to the trend [if the data points are independent – which they are not]. So, an R2 of 0.1 means that at least 90% of the variation is not due to the trend, and R2 of 0.3 means that at least 70% of the variation is not due to the trend. The ‘at least’ comes from the fact hat the autocorrelation at log 1 is not zero.
The y-axis for the PDO is reversed (negative is up) in the chart attached to my last post.
Leif Svalgaard (10:43:31) :
The ‘at least’ comes from the fact that the autocorrelation at lag 1 is not zero.
Leif S.
“As you can see in the graph below, the overall annual trend since 1989 has been slightly upwards.
And what is R^2 for that graph?”
Interesting. I don’t recall your EVER asking this for testing out the significance of temperature data. As I have been doing, using Student’s T, Mann Whitney, etc. to decide if certain observed changes (comparing year to year, decade to decade and century to century) are truely “statistically significant” considering the observed standard deviation. (Of course, these is the concept of normalacy and the validty of Student’s versus the non-parametric Mann Whitney test, my judgement is out on that. Because if you treat temperatures by “seasons” they are virtually “normal” in distribution. If you go over a chronological year, they appear as a bifurcated normal distribution.)
Steve Goddard (10:20:59) :
R^2 on the winter graph since 1989 is 0.298514013
R^2 on the winter graph since 1999 is 0.312822303
An interesting graph would be R^2 as a function of start year. And another one the same but without 2009-2010.
Just so I cam see if I am following this, I’d appreciate if anyone would tell
me if I have misunderstood Steve’s post.
Steve Goddard (10:20:59) :
R^2 on the winter graph since 1989 is 0.298514013
————Means snow cover has increased somewhat since 89.
R^2 on the winter graph since 1999 is 0.312822303 Slope is 317,000 km2/year
———Means considerable increase sine 1999
R^2 on the weekly graph since 1966 is 0.004697147
——–Means when averaged out since 1966, no big deal
indicating (opinion) that maybe the climatologists have mistaken natural variability ( caused by xxxxx?) for one way climate change.
R^2 on the weekly graph since 1989 is 5.25176E-05
—–Showing that measured from 1989 there has been a significant increase in snow cover.
Do I at least read this right ?
I don’t see the growth represented on second graph on the first one, it looks pretty much constant.
Also, 2010 isn’t quite finished, it just started, so how was calculated 2010 point on second figure? it seems it will look pretty much the same as 2008.
Any way is hard to look for differences on sinusoidal signals, I like better the anomalies graph as in http://moe.met.fsu.edu/snow/ though their record isn’t as long, it only goes down to year 2000
Hmm. Living in a “snow” country as I do, I’m less than convinced that snow extent is anything more than a weak proxy for flow of cP and Arctic air masses. Its certainly not a prioria linear proxy for temperature nor for global cooling. While we’ve observed snowfall along a lower latitudinal limit this year than in the previous few years, behind the snow line there are large areas with much lower than usual snowfall, and more moderate temps. At my place in S. Ont., for example, we’ve way below normal snowfall and temps have stayed moderate all winter (but not warm, i.e. +/- average without the usual large variation swings).
I’m a little less sanguine about “records” as well. I seem to recall the Washington record this winter was ~ 1/2 in. more snow than the previous record. 1/2 in? A 1/2in of snow measure is meaningless in a pile several inches thick, as anyone who lives with snow will tell you. This is an anecdotal record, not a scientific one.
vukcevic (07:36:13) :
Keeps deep frost away from the ground surface. Any farmer will know benefits of the winter snow cover.
Henry chance (08:05:49) :
Snow cover is God’s gift to winter wheat fields. It is so many times better than mere rain fall.
The only advantage to snow versus rain is in cases where slow snowmelt occurs leading to increased percolation resulting in reduced runoff in spring (March, April, May).
Frost actually mellows the ground due to frost heave and aids in spring sowing. It also kills insect eggs and larvae in the soil, and some weeds and weed seeds.
A persistent, deep snow cover winterkills winter annuals (wheat, rapeseed, winter peas, etc.).
So no, we don’t want increased long lived snow cover in the US at least.
Robert (09:24:32) : “At least 80% of the soot is caused by BURNING FOSSIL AND BIOFUELS. So his theory is very interesting, and I hope it holds up, since it suggests the Arctic ice could rebound much more quickly if we took corrective action. But that corrctive action continues to be drastically reducing burning carbon compounds for energy.”
…by inefficient soot-producing methods and instead burning them in modern efficient power plants, and switching from wood/grass/dung cooking fires to natural gas, and producing abundant energy from nuclear power, etc.
Fixed that for you, Robert. No thanks necessary.
Only thing that makes any sense here is:
delta area/delta time
or to make more sense trend expressed as % pa, which is roughly 0.22% pa
(no calculator or excel needed).
Perhaps a cherry-pick-proof graph would help clarify this morass.
Here is the trend to present for every week since 1966 for each season.
http://i81.photobucket.com/albums/j237/hausfath/Picture41.png
Matt (10:36:58) :
Applying an R^2 to a linear fit of an oscillating system also makes no sense whatsoever!
If you consider the system to be an oscillation riding on a linear trend, then it does make sense to compute R^2 as a measure of the significance of the trend. In interpreting R^2 you must then take into account the non-zero autocorrelation due to the oscillation. There are standard techniques to do all this. One way of looking at this is to average over each cycle and plot the cycle averages as a function of time. This makes perfect sense. R^2 will still show how much of the variation is due to the trend. Here is a little experiment: http://www.leif.org/research/Cycles-with-Trends.png
I generated 17 perfect sine waves [black curves] and ran a linear regression. The result was y = 0.00 x + 0 with R^2 = 0 as expected. No trend. Then I added a linear trend of 0.01 per step in x [red curve], result of linear regression: y = 0.01 x + 0 with R^2 = 0.1525. Perfectly reasonable as the trend explains 15.25% of the variation. Then I added a trend of 0.02 [green curve], result y = 0.02 + 0 with R^2 = 0.4185, i.e. the trend [which is recovered perfectly from the analysis] explains 41.85% of the variation, etc.
No problem here whatsoever. No failing of statistics. For the green curve, t is 27.3 and p = 4*10^(-124). The significance is enormous for this case [I had 1037 points]. With fewer data points the significance drops dramatically: with 23 points t=7.92 and p= 0.00063, but then there is no noise, etc.
R^2 is a good measure of how much of the variation is due to a linear trend, no matter what the data looks like.