Yes, You Can Do This At Home
OK, I was doing a GHCN V1 vs GHCN V2 comparison for “practice” while waiting for GHCN V3. And I decided I needed a direct comparison between the two. That meant the dT/dt method could not be used since it ‘starts time’ (and the baseline) in the present. And “the present” ends in 1990 for V1 and in 2010 for V2. What to do. What to do.
So I decided to use a ‘large chunk of common data’ for the baseline. The results of that are in “other hands” right now and may become a posting later. But along the way I realized that I had a way to validate dT/dt and/or show the “shape of the data” in a way anyone could do at home.
There is a very strong rise out of the Little Ice Age at the start of the thermometer series in 1701. Partly this is due to there being only one thermometer and it was in a cold spot. So I’ve chosen to “start time” in 1800 for this graph. This way it includes the “start of time” for GIStemp (1880) and CRU (1850) along with some context (1800-1850) so we can see if there was any cherry picking going on. It also includes 1816 “the year without a summer” just because I’m interested in it.
This one chart says so much. You can click on it to get a larger version with much more detail.
OK, I’d like to jump right into analysis, but first a couple of notes on the process.
It’s very simple, and yes, you can do this at home.
For each thermometer, I average all the data in any given month. That is the “baseline”. It then gets subtracted from each monthly data value to make that value an “anomaly” or “delta from the average”. These “monthly anomalies” are then averaged together for all thermometers that exist in any given year (Those data are the thin “hair” lines in the graph for each month.), at the same time an annual average is calculated by adding up the anomalies for all thermometers in a given year. You could do this in a spreadsheet.
(And I’d love it if someone were to do so and validate this. This is presently grotty FORTRAN code hacked together from preexisting bits in a particularly cavalier manner and I’ve not gone back yet to do my usual clean up and QA runs. Yeah, I think GIStemp code is starting to corrupt my style ;-)
FWIW, this graph fairly closely duplicates the dT/dt graphs, giving more confidence in the validity of both.
So What About The Graph?
Well, first off, notice that a trend line fit to the annual data is just about dead flat right up until about 1990. (That is the very thin blue line at near the zero line) The change of “duplicate numbers” or “modification history flag” on thermometers starts to hit in about 1986-1987 but those records have a matching entry from the prior “duplicate number” until about 1990 when “the reveal” is done and the older series of “duplicate numbers” are dropped. You could put the segment break at 1987 and get similar results, but I chose 1990 as that is when the data series are left to stand on their own.
At that point we see a dramatic increase in the slope of the trend line as “AGW suddenly begins”. But, IMHO, it’s not the world that’s warming, it’s the data…
That is the most striking thing I see. Just look at the slope of that thin Hot Pink trend line! We go from nearly flat trend up to 1990 (the thin blue line) to dramatic warming (thin hot pink line) just as the “modification history” or “duplicate number” change indicates a change of processing of the data. Hmmmm.
But wait, there is more…
Look at the period from about 1950 to 1980. It’s cold. But it gets cold not by having a lot of cold excursions (those monthly “hair” lines don’t go down very far) but by having no hot excursions (the “hair” barely gets above zero in several years). This, as they say, is odd… Some of those years were cold, but not the whole lot of them. There ought to be some warm excursions in some months. It’s almost like the data were tailored to have no hot excursions in those decades. Perhaps by way of the thermometers that were added coming into those years, then the other thermometers that were deleted when leaving…
Some Mathematical Expectations
Now, mathematically speaking, you would expect a broader range of excursions in the early years when there are few thermometers. If you look at the 1800 to 1940 part of the chart, you see both high and low excursions get compressed as the thermometer count rises. Not dramatic, but clearly happening. Then in that 1950 – 1980 part, volatility goes WAY down. It would be very interesting to do a statistical analysis of the degree of compression of range and ask if that is in keeping with the number of thermometer count change, and if not, it implies thermometer selection bias. To my eye, that’s what the chart says, but it needs a rigorous statistical treatment by someone other than me. Is that volatility reduction really justified by the effect of averaging a few more representative thermometers?
Though: it is worth noting that the volatility does start to rebound in about 1977 with a cold spike. I remember it “snowing where it doesn’t snow” in a couple of years around then. But I also remember a lot of “110 F in the shade and there ain’t no shade” years around the ’60s and ’70s. This makes that very low volatility, and no hot excursions at all, look very wrong to me. It was 117 F near Marysville one of those years, the hottest in the area ever and not reached again, IIRC. So those clipped hot peaks are very, very suspicious.
To me, this is NOT just a result of a few more thermometers. It looks like both MORE thermometers and LESS VOLATILE thermometers. Less volatility happens at warmer places (lower latitudes and elevations) and near the water. The same pattern of thermometer change over time that we’ve seen in the prior analysis of thermometer location changes over time. Our “thermometers on the Beach” problem is showing up in the volatility compression. Then a Very Odd Thing happens.
Industrial Revolution 1700s to 1990
As the thermometer count drops, we have a sudden rise in the trend to a spectacular warming. One Small Problem. The Industrial Revolution had been under way since “The Late 18th Century” (or sometime in the late 1700s). So we’ve had burgeoning CO2 production for over 200 years by then. And a Big Fat Nothing. Then suddenly, right when “Global Warming” becomes a popular point (and when NOAA / NCDC, Hadley CRU, and NASA / GISS get buckets of money and True Believers at the helm) we suddenly have a ‘hockey blade” form in the data. As someone recently said in another posting: “Hmmmm”…
Either CO2 has a 200 year lag in effect, or it’s not CO2 causing the data to change.
But Wait, There Is Even More
So take just a moment to look closely at the “hair” in that 1990 to date segment.
It doesn’t look at all like prior decades. In fact, it looks decidedly Non-Physical. There is an increased range, but only to the upside. Downward anomalies are substantially gone. Not even reaching zero. Yet in all prior time, the range is more or less symmetrical, though with a slightly larger “cold going” range (that is very physical, as hot air rises, so hot going anomalies are limited while cold air just pools around the thermometer. So cold going excursions can go further, at least in the real world.)
That the post 1990 range is so non-physical is, in my professional opinion, evidence of data tampering. Perhaps deliberate, perhaps out of error. But the data are now very non-physical. It has clear onset in time. That time matches a recognized change in the processing at NOAA / NCDC (when many / most stations increment the “Duplicate Number”) and match a date when new “Quality Control” procedures were introduced for the USHCN data set that will preferentially suppress cold going anomalies in that data set. Could something similar have happened to GHCN?
Further, 2009 had many cold locations, much snow around the world (it snowed on the French Mediterranean coast…) and was NOT a warm year. Yet the “Hair” for 2009 does not even make it down to the zero normal line. To me, that says these data are cooked. Not enough for a conviction (yet…), but enough to justify a detailed Forensic Investigation.
And yes boys and girls, you can try this at home.
Graph With Legend
I realized that the first graph does not have the legend on it. Here is a version with the legend.
UPDATE: Added Graph With Trends By Month
You really do need to click on this graph and get a bigger copy to see the equations and see the detail of the lines.
I’ve added fat trend lines to each of the monthly “hair” plots. The equations are at the bottom of the graph in “month order”. So if I’ve kept all this straight, you ought to be able to pick out, for example, the August lime green line and the 8th equation in the list and see that August has not warmed. At all. The other summer months are not much warmed either. The “warming trend” is all concentrated in the winter months.
Frankly, I wish this AGW were actually happening. I think more of use would be quite happy to live in perpetual Spring and Summer and skip a lot of the worst cold of winter. Especially the folks dying in South American cold right now…
I find it very interesting that the “hair” plunges below their collective trend lines rather dramatically around 1970 and collective jumps way above the trend lines in the last decade. Very unusual when compared to prior years. Just wrong for a natural process, IMHO.
These trends, though, argue that CO2 and tipping points is simply a fantasy. There is no tipping point to hotter or you would have summer months warming (there would be an acceleration of warming in the warmer months with positive feedback). Instead we have August going nowhere (perhaps slightly DOWN per the equation) with May and July almost as flat while June and Sept are nearly so. All the positive slope of merit is in Nov, Dec, Jan, Feb, and March. Funny stuff this CO2, only warms the winter months of the Northern Hemisphere…
And yes, I know that the data series has far more N.H. thermometers than southern hemisphere and that without areal weighting they will dominate the trend lines.
It would be particularly interesting to put up a graph of the monthly trend lines after 1990. Perhaps I’ll do that one for another posting. But I think it’s pretty clear that the “all data” graph has interesting variation from month to month in trend. Wonder how the CO2 theory can be warped such that August doesn’t warm?…