How bad are these datasets? How sloppy are the data records?
Western Australia (WA) covers 2.5 million square kilometers (1 million square miles, about a third as big as the USA). The average of all WA stations over one month last year was adjusted up by as much as a gobsmacking 0.5 degrees due to a database “bug” – which contributed to August 2009 being the hottest August on record?! That’s one heck of a bug!
Could it get worse? Unbelievably, GISS seems to have lost data for key WA locations that an unpaid volunteer found easily in the BoM online records. GISS only has to maintain copies of records for sixteen stations in WA* which have temperatures current to 2010, but in seven of them they are missing data, and it affects the results. Are they random errors? No, shock me, six errors are upwards: in one case making the spring 2009 average temperatures for Kalgoorlie-Boulder 1.1 C degrees warmer!
But with no-one auditing our BoM or NASA’s GISS, and no team jointly receiving raw data or regulating standards in either agency, temperatures recorded in the field could potentially be listed in official records as being quite different, and who would know? It’s left up to volunteers like Chris Gillham, a freelance journalist and web designer in Perth, to run a sharp eye over the data. Chris has been tracking WA data for the last two years and his site, Average Temperature Trends Across Western Australia, has methodically, neatly exposed some major flaws.
Just how much can we trust any of the pronouncements coming out, and how significant are any of the “records”, even if the adjustments are fair, unbiased and justified? The whole database is surely not “high quality” when bugs of that magnitude are running rampant and data goes missing that professionals can’t find, but people who are not “paid to find warming” dig up without much trouble.
–JN
New questions about reliability of GISS and BoM data
Guest Post by Chris Gillham
Fresh doubts have emerged about the reliability of temperatures within the Goddard Institute of Space Studies Surface Temperature Analysis database with revelations that missing data errors have appeared for various months in the 2009 records of Australian locations, even though the correct mean temperatures are available from the Australian Bureau of Meteorology (BoM).
In turn, the BoM data itself has seen adjustments that might leave researchers wondering about claims that Australia has suffered record high temperatures over the past 12 months.
Western Australia (for scale, it's about 600 km from Perth to Kalgoorlie). Showing some of the sites.
A BoM database bug: Oops, half a degree?
On September 1 last year, the BoM posted mean min and max temperatures on its website for the month of August 2009 at all its recording stations in Western Australia (2.5 million square kilometres).
However, on November 17 the mean temperatures for all WA recording stations were adjusted upward by as much as .5 C for August 2009.
When questioned about the adjustments, the BoM confirmed it had suffered a database bug and the upward shift was a consequent correction for August 2009, which the bureau says was the hottest August ever recorded in Australia.
GISS is “missing”data
The GISS database shows that in the following month, September 2009, there is missing data (999.9) at three Western Australia recording stations:
Esperance | Kalgoorlie-Boulder | Perth Airport
Despite the missing September data and as is evident in their tables, GISS has calculated the Spring (S-O-N) mean temperatures at those three locations as 17.5 C, 20.5 C and 17.7 C respectively.
Trouble is, the data isn’t “missing”. A quick search of the BoM website reveals the September 2009 mean temperatures were:
This in turn means the Spring mean temperatures were actually 16.6 C at Esperance (not 17.5 C), 19.4 C at Kalgoorlie-Boulder (not 20.5 C) and 17.2 at Perth Airport (not 17.7 C).
The GISS database records for Eucla show missing data for December 2009, but the BoM records once again are available and show the mean temperature was in fact 21.6 during that month. The GISS has calculated the Summer 2009/10 (D-J-F) mean at Eucla as 22.8 C, but with the accurate BoM December data included it turns out to be 22.7 C.
Based on evidence available from the GISS and the BoM websites, it appears several WA locations with records current to 2010 have small to significant upward data adjustments.
Wait, there’s more!
I’ve detailed the BoM bug adjustments and the GISS missing data adjustments.
While researching the GISS adjustments, I noticed yet another odd data shift that left me wondering about the reliability of temperature recordings. I had listed the 2009 monthly mean temperatures on October 4, 2010, for Kalgoorlie-Boulder, but when I returned to the GISS website database the following day, October 5, I found that every month in 2009 for that location had been shifted up by .1 C.
This means the newly adjusted GISS record shows Kalgoorlie-Boulder’s average mean for Spring 2009 was 20.6C, not 20.5 C anymore, so this historic mining town’s seasonal temperature record is now 1.2 degrees higher than the reality of the BoM records.
These inexplicable adjustments to domestic and international datasets raise questions about the reliability of record temperatures reported in Australia over the past year and the reliability of official records used by researchers to try to accurately gauge temperature trends.
**UPDATE (Erratum): Text has been slightly altered to reflect new information. Hours after the post was put up, it was discovered that GISS maintains records for sixteen sites around WA, not just five, but in at least three other sites there is also missing data (Port Hedland, Albany, and Geraldton) making it at least 6 of 16 sites where data is missing even though it is readily available, and in the case of Albany, there are holes in the data, but the raw data for times prior to September 2009 is no longer available on the BOM website. Apologies for the error.
UPDATE 2: Geraldton and Port Hedland mistakes are also “upwards”
The Geraldton Airport GISS data shows missing temperature for April 2010 but the BoM data at for April 2010 shows the mean at 20.95, which means the Autumn 2010 M-A-M is 21.3 C, not 21.5 C as calculated by GISS.
The Port Hedland GISS data shows missing temperature for May 2010 but the BoM data at for May 2010 shows the mean at 25.2, which means the Autumn 2010 M-A-M is 29 C, not 29.1 C as calculated by GISS.
Other Notes
Climate Fools Day is on today!
See Carbon Sense Coalition for information on events at The UK House of Commons The event at 2pm includes Christopher Booker: “The most expensive Bill in History”, Piers Corbyn: “Successfully predicting extreme weather events” and the Revd Philip Foster.
Julie Bishop’s poll: Do you want a carbon tax?
She’s running a poll on her site. If you think a carbon tax is a noxious waste of time, it only takes a second to click…
Image adapted from the Map on Australiaforall — tourism information for disabled travelers.