Is March In The Upper Midwest Losing It’s Freeze? The actual data doesn’t seem to support Climate Central’s recent claim.
Guest post by Steven Goddard
Yesterday, WUWT discussed an article on future regional temperature modeling from Heidi Cullen et. al at Climate Central claiming that most of the upper Midwest will no longer be freezing in March by the year 2090 – as a result of increases in atmospheric CO2 content. This was based on averaging the output of 16 different climate models. Here’s the image included in their press release:
Caption: In blue: projected areas with average March temperatures below freezing in the 2010s (above) compared to the 2090s (below), under a high carbon emissions scenario extending current trends. Click image for an interactive map
As you can see below, CO2 has been increasing rather steadily for the last few decades, particularly the last 30 years. No dispute there.
Source: Scripps Trends in Carbon Dioxide
If Climate Central’s press release theory were correct, we would expect to have already seen an increase in March temperatures, and an increase in number of years above freezing. Below is a graph of NCDC March temperatures for Wisconsin since 1979.
The orange line is the mean and the red line is the freezing line. Note that not only is there no trend towards a warmer March, but the standard deviation is high (3.67) and the range is also large – about 15 degrees difference between the warmest and coldest March.
Source: NCDC Wisconsin March Temperature data
Even so, the 100 year graph of March temperature in Wisconsin seems rather flat also.
The next graph is the number of years above freezing per decade. As you can see, there were fewer years above freezing in the last decade than there were in the 1980s.
Minnesota shows the same patterns – no warming and high variability. The number of years above freezing has also decreased.
NCDC Minnesota March Temperatures
And here is the 100 year March temperature graph, like Wisconsin, pretty flat:
Like Wisconsin, it seems there have been less days above freezing in recent decades:
Conclusion: Based on the NCDC data, there is no evidence that increases in CO2 over the last 30 years have affected March temperatures in the north central region of the USA or moved the freeze line north. Once again, we see a case of scientists trusting climate models ahead of reality.
More on Climate Central:
http://climatecentral.org/about
http://climatecentral.org/about/people/
UPDATE:
Here is Minnesota and Wisconsin with five different trend lines for different start years.
In order to highlight the lack of correlation between year and March temperature, I also made a scatter diagrams:
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Climte Kate (15:08:22) :
Climte Kate (16:20:21) :
Not only if the earth goes into la Nina will the recent warming not mean anything but also if it goes into another Little Ice Age that lasts for decades. Our arguments over trends, and statistical significance, of the last 10 years, or 15, or 30, or 150, won’t mean much.
But I think most people here already know that.
Heidi Cullen should also know that. If she does know it then she must have some ulterior motive for not bringing it up. If she doesn’t know then how did she rise so high at The Weather Channel? The same for Jim Cantore.
Robert Burns (21:39:02) : “I think you are wrong, the models are build in part on past relationships and correlations and with ‘tuning’, not only on physics theory. If the climate models were built on equations based on physics, then all the models would return the same results.”
I don’t disagree but my point was that you seem to overestimate the power of a relatively small number of tuneable parameters when the observed flow one is trying to reproduce is complex.
This can be illustrated using a simpler example, a computational model of turbulent flow in a pipe which involves a single tuneable parameter, p, where, p, is a number. Suppose computations are performed with different values of the parameter and that the best approximation to the observed three-dimensional mean flow is obtained when p=P. In other words computational mean flow V_c(x,y,z,p) that best approximates the observed mean flow V_(x,y,z) is V_c(x,y,z,P). You seem to think that producing a good approximation is not particularly meaningful. You are wrong. It is a VERY big deal. Varying a single number to reproduce a three-dimensional field is next to impossible unless a LOT of the physics is already built into the model.
If I lower my sights and only seek to reproduce the AVERAGE flow through the pipe (i.e. forget about how mean flow varies within the pipe.) Now I only have to choose p so that the computational output U_c(p), a number, matches U, the observed average, another number. In this case you are right, the procedure is worthless. The ability to match says nothing about the quality of the model since a match is all but guaranteed (under a few rather weak constraints).
Just more evidence that the models are broken and shouldn’t be relied on for policy decisions.
What are you nuts? Models are right the real world data is wrong. When in doubt believe the models.
The reason GCMs run over climatic time-scales come under severe, well-deserved criticism is because of the outlandish claims made about their ability to simulate physical reality and predict the future of a chaotic process. Modelers in other fields involving fluid thermodynamics make no such claims!
Steven: I’m going to end my half of the discussion of trends. To confirm your post, let me switch datasets to two that are available through the KNMI Climate Explorer: CPC Snow Cover and GHCN/CAMS t2m Land Surface Temperature:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
I’ve used the coordinates of 43N-49N, 96W-88W to capture Minnesota and Wisconsin (the Upper Midwest) in one dataset. The following is a graph of the March Snow Cover (CPC dataset) for the Upper Midwest from 1972 to 2009. The dataset starts in 1972. As you will note, there is basically no trend:
http://i40.tinypic.com/1626tlk.png
Now let’s look at the CPC’s GCHN/CAMS t2m Land Surface Temperature data for March, also from 1972 to 2009, the same years as the snow cover data. It has a minor positive trend, about 0.03 deg C per decade. That in itself is remarkable since the GCHN/CAMS t2m data has the highest trend of the land surface temperature datasets:
http://i43.tinypic.com/wb2vbq.png
One last graph just to confirm what everyone knows anyway, comparing the two (with the snow cover inverted and scaled), it’s obvious the two are interrelated (correlation = -0.7). Temperature goes up, snow cover drops.
http://i41.tinypic.com/j94h1y.png
March temps peaked in early 90s here in Norway: http://eklima.met.no/metno/trend/TAMA_G0_3_1000_NO.jpg
Bob,
Looks like the two years with the most snow cover (1973 and 2000) were also the two warmest years.
http://i41.tinypic.com/j94h1y.png
Steve Goddard (21:34:23) : You replied, “Looks like the two years with the most snow cover (1973 and 2000) were also the two warmest years.”
http://i41.tinypic.com/j94h1y.png
Look again at the title block. The Snow Cover data is inverted.
What’s the data for Montana, Washington, Upper New York, New Hampshire?
Might as well show the whole set, eh??
Rhys Jaggar (05:10:18) :
Here are all the Northern States from west to east. No March upwards trend anywhere.
http://docs.google.com/View?id=ddw82wws_536gpg969kw
ERRRrrr, I am in the middle of North Carolina on March 17th (chosen at random) the low was 30F. I was planning to plant grass in February per the USDA extension guidelines but could not because of the snow cover…..
Just to illustrate the uncertainties in science and the use of computer models take a look at this recent event and please don’t be having food or drink in your mouths.
March 2nd 2010
“Chilean Quake May Have Shortened Earth Days”
“Using a complex model…”
Source: Science Daily
April 17th 2010
“Earthquake in Chile Causesd Days to Be Longer…”
“…simulation model.”
Source: Science Daily