Wait til you see what Lance Pidgeon has found. He was looking at the BOM website temperature archive maps of Australia for early last century (using AWAP data). He was wondering how the Bureau of Meteorology could possibly create maps this detailed for specific days that long ago. He was especially curious about the remote, vast areas where there were no thermometers, yet there were wiggly jiggly temperature lines on the map, shaded as if they had meaning. I’ve heard that more people have visited the South pole, than have stood at the point in central Australia where the three large western and central states meet.
Then he noticed something positively strange — April 14th in 1915 and one year later in 1916 looked almost identical, as did the same day in 1917. The more he looked, the weirder things got. He plodded, year after year, all the way from 1911 to 1917, then through Jan, Feb, March, and so on. Worse, he tells me he could keep going right through to 1956 without seeing much change (though there are interesting exceptions). After that, temperatures of the area start to vary from year to year, like the “weather” we’d expect if we had multiple thermometers in the site with the black square, which we still don’t have. Even in the modern era there are only two sites.
It is not just “April 14ths” each year that are suspiciously similar, it’s pretty much all days of the same month. In this blockbuster graph below, he looks at one spot in central Australia, about the size of Tasmania (which is 65,000 km2), and tracks the temperature profile of that same area, on the same day, year after year. The BOM tells us they have good temperature records. They tell us the AWAP analyses are based on “in situ” surface observations, and they make much about AWAP trends being “unadjusted“. Yet here in an AWAP Map, presumably derived from the same data as those “unadjusted trends” there’s an area with no thermometers before 1940 (when Warburton opened) and we see detailed temperature lines that are identical year after year.
Do AWAP maps and AWAP data matter?
AWAP maps are used in press releases and in the news. The detailed wiggles send a message to the world that “we have very accurate data”. But when the BOM tells us we set an “area averaged” record across the whole of Australia since 1910, they don’t mention that it’s compared to “calculations” of estimated wiggles over hundreds of thousands of kilometers where there are barely any roads, let alone thermometers. Nor do they mention that suspiciously, magically, in the early part of the AWAP record the temperatures in remote central Australia appear to be the same year after year — or at least they are in the maps*.
Significantly, the BOM use trendlines from the AWAP data as justification that their all-homogenized ACORN data is virtually the same as the “unadjusted” data. It’s their excuse for why their massive adjustments in the ACORN set look neutral (when other analysis of ACORN shows the adjustments that warm the trend are much more common). The AWAP maps are created from this data too (but obviously the maps themselves aren’t “raw” because they must have an area weighting algorithm run over the data, plus elevations, plus who knows what else?). The question then is what is the state of the AWAP “unadjusted” data? The maps generated from it suggest quality control is awful, weekly data disagrees with daily data, and the program used to do area-weighting and to generate the maps is not producing results that look credible. How “unadjusted” are the trendlines that are called “unadjusted”?
Below Lance Pidgeon has graphed the squares that fit in the black box in central Australia (shown in the map below this) from Jan 14th each year, then Feb 14th, then March 14th… you get the picture. Astonishing. Thank citizen science for telling you what the BOM doesn’t mention.
Guest Post by Lance Pidgeon
Building the past — BoM style
To produce an area averaged temperature for Australia a fine matrix of squares on a map can be used. Just add the values in the squares and divide by the total after correcting for latitude. Anomalies and trends can be produced over time. Simple right?
The BoM “Australian Water Availability Project” (AWAP) maps have a fine lat-long grid over many years of daily data. But in parts of Australia thermometer sites are hundreds of kilometers apart, especially in the first half of last century. To make a complete picture gaps need to be filled — but with what exactly? Between thermometers MUST be derived values. Does the fill come from raw data, estimates and future averages or a desired outcome? Is the gridded data that was used to generate this map called “raw” data by the BOM?
In the map below, we’ll take a close look at the area marked by the black square. Horizontal lines drawn from the same island in W.A. shows two mapping methods.
Pasted squares in the graph pictures have been chosen to show a problem that would effect the whole map to some degree. Tasmania is about the same size as Ireland, Switzerland or the state of West Virginia in the USA.
These maps can be found here. Select the day, month and year etc from the drop down menus.
Thermometer location map: Each dot currently has a site. There are more dots if you tick the box for “closed” stations.
Media and the maps
Early 2014, long before all the data had been Q.A. checked, the BoM and ABC were quick to hit us with headlines.
Long stretches of hot weather, with little respite at night, have combined to make 2013 Australia’s warmest since records began over a century ago.
The ABC quoted the BoM, repeated the long time span and used AWAP maps as a display of evidence.
According to Bureau of Meteorology data, the Australian area-averaged mean temperature for 2013 was 1.20 degrees Celsius above the average since 1910.
Is it evidence or art with a desired outcome?
Can you tune raw data? Should you tune raw data? From a 2009 paper about the AWAP methods we read:
The Barnes analysis technique has a number of advantages, including being efficient, robust (coping with strong gradients and data voids) and highly tunable. Jones and Trewin (2000a) have previously shown that the accuracy of this method is similar to that of more sophisticated techniques, and avoids some of the weaknesses such as extrapolation of unrealistic values into data voids and the dampening of variance.
Bold mine. Oh and why is it important to “avoid” “unrealistic values”?
Is it obvious to all that this tuning has been performed? In the Report of the Technical Advisory Forum June 2015 we read:
The chart below (Fig 4.1) shows the difference in mean temperature anomalies between the
homogenised ACORN-SAT and unadjusted AWAP datasets for Australia. Bold Mine.
Then the big question. Does what you see in the paste pictures affect those differences? It appears that the fill contains a terrain elevation adjustment which is then treated as being part of the “gridded” raw data. This could explain some of the fine detail in the wiggles in the data void. From The Centre for Australian Weather and Climate Research technical report no 050 we read:
“These climatological analyses have embedded within them climatological temperature-elevation relationships.”
That does not explain where either the underlying value or repeating step sequence in anomaly came from. See below.
A picture speaks a thousand words
Below are colour patterns from approximately the same place as the black square on the maximum temperature map above for the 14th day of the month and year. The first picture has sections from maps of a time before weather station sites opened at Giles, Warburton Airfield, Blackstone Range and Yowalga. The second picture is from recent years. The difference is stark. In recent years the weather is different year after year. In past years, not. A tad “unrealistic”?
In the last seven years the same spot on the same day varies each year (as we’d expect):
Exceptions are mysterious when the “hottest” day of the week is not the hottest day that week.
During one interesting week in 1943 ending on Feb 16, the highest maximum temperature recorded across this same area of Australia was between 36 and 39C. Check individual days that week, like the temperature of February 14 and thousands of square kilometers recorded a hotter maximum of 39 – 42C in the same area of central Australia.
Yes that’s right. In that location, February 14 that year was hotter that itself. Is anyone silly enough to defend this rather than wanting it to be fixed?
An overlapping half square step to the left (i.e. West) from those above, we see a yearly square step sequence in anomaly through the months on the earlier sections below but not on the more recent sections. This step sequence could be related to the similar mysterious square wave adjustments found in ACORN by Bob Fernley Jones.
Over the last seven years, the weather changes on the same day from year to year:
Far, Far less seriously
Jo asks me “What does raw data mean anyway?” and “why bother with real stations”?
“Oh but the colours in the later section pictures may be somewhat influenced by real data.” I reply. Though I concede the squares in the earlier ones are more like God playing tricks on the BoM.
Mulling it over for a while I begin to wonder if it’s a small group of very young abstract artists at the BoM each with their own colour, favorite detail and month (The Weather Squiggles?). It is not hard to imagine kiddies who have been asked to paint the temperatures of a mythical place on the map in W.A. It’s the locality of “Thin Air”. A mining town along yellow brick road at the end of the rainbow where they dig down to the twilight zone to extract the blue. Out of the blue from Thin Air, processing yields herring red, leprechaun green, and purple haze etc. The kiddies then stick twelve photocopiers shut with the result. [As good as all modern art should be, says Jo. Like a Rorschach test for climate-nerds? Tell me what you see!]
So far we have been able to deduce that the data in the earlier squares is produced using the F.A.K E. method but are unable to determine what F.A.K.E. stands for. Theories include “Fecalithic Absentaneous Kenophobic Esemplasy” and “Fast Autogenerated Klimate Estimates”. The assumed Squiggles BoM colouring in group may also enjoy square dancing on a grid after a little nap.
Jo says: We laugh, because if we don’t laugh, we cry.
Thank Lance for the hours of research it took to find and put these graphs together. The Bureau of Magic’s budget is $300m +. Lance gets no payment.
*Edited — added “remote central Australia” to clarify.