Australia's B.O.M & "Homogenisation" - they must think we're stupid

Australian scientist and blogger Jennifer Marohasy has been leading a small but vocal group campaigning for the Bureau of Meteorology to explain and defend "homogenisation" of temperature data for Australian stations. Their new, improved, value-added and of course, scientifically and mathematically sound database has, in most cases, turned a long-term static or cooling trend into a warming trend. They call it ACORN-SAT. You and I might call it GHCN/GISS outback style - if the present isn't warming to suit their meme, then by god, they'll cool the past, and "prove" local and global warming.

In most cases, the process leads to lowering past temperature data in a series of steps, supposedly corresponding to breakpoints in the series due to station moves, claimed station moves, statistical "discontinuities", or undocumented and therefore unjustified adjustments. Now this was something I could really get my teeth, and my Excel 2000 skills into. Yes, it's the 2000 version - if it ain't broke, don't spend money on a newer version.

I started several draft posts, downloaded shedloads of B.O.M. data, unadjusted GHCN/GISS data, PDF documents, bookmarked scientific literature, created new spreadsheets, chart graphics, googled for images of towns, cities, weather stations. The B.O.M. got very defensive when challenged a couple of years ago. Jennifer and her small group were tilting at a very big and huffily self-righteous windmill. The B.O.M. actually produced some documentation. What right had these upstarts to question their methodology and intent?

Then, a couple of days ago, I had an epiphany; I'd clearly missed something, something very important. The fact was, the B.O.M. has never, ever made the adjustments they'd claimed were necessary! If there's a documented station move, for example from a town out to a rural site, a small temperature change would be expected. In this case, likely to be downward because of the removal of the "urban heat-island" effect. They've identified and quantified such steps in the record at many sites. They believe an appropriate adjustment should be documented and applied, in this case an upward adjustment. The ACORN-SAT database appears not to contain any such adjustments, despite what the B.O.M. has documented, and everyone, until now, has believed.

I'd been looking at the "Station Adjustment Summary" for Amberley, Queensland. There was just one significant adjustment, forward from 1980:

2.  1 January 1980—breakpoint detected by statistical methods.
•  Night-time temperatures started to appear much cooler relative to surrounding                   stations.
•  No accessible documentation for Amberley in 1980, but a breakpoint of this size              would normally be associated with a site move.
•  Min T adjusted by -1.28 °C; no detectable impact on Max T so no adjustments made.

Adjusted by -1.28°C - wouldn't they apply an increase if temperatures "started to appear much cooler relative to surrounding stations"? You'd expect so, but that -1.28°C adjustment was correct - applied to data prior to 1980, thus leaving the "much cooler" data from 1980 onwards unchanged! The shape and trend of the resulting "homogenised" plot becomes identical to what would have resulted from a positive adjustment, but shifted down by the amount of the reversed adjustment. This is the plot of the change to average minimum temperatures for Amberley; for all my graphics, click to see a larger version, or right-click and "save as" to save them:

From the Station Adjustment Summary for Amberley.

The first thing to note is that the "difference" is positive because the author has subtracted adjusted from raw data. If you're comparing A with B, the correct way is to subtract B from A, not vice-versa. Compare 10 with 8, and the difference is 2, not -2. There's no change to raw data after 1980; if post-1980 data was "much cooler relative to surrounding stations", then it remains so. I've plotted ACORN-SAT adjusted against raw for Amberley:

Raw data series (blue) versus ACORN-SAT (red); difference (black)

The raw and adjusted series track together from 1998 rather than 1980 because a second, undocumented adjustment was made after 1996; a 2-step change. From 1980 to 2016, raw minimum trends flat. That's right - no change since 1980. The B.O.M. has what might be termed a "positive attitude" to temperature series; they like to see upward trends - flat or negative has to be "corrected".

Average annual minimum or maximum series are no good metric for measuring or tracking anything, yet they're what the Bureau uses exclusively, and to my knowledge, no other large private nor state-funded meteorological organisation does this. I'll back my up assertions in a future post. In the meantime, here's a plot of monthly average minima for 1978-1983, which spans the Bureau's "breakpoint":

Monthly average minimum temperatures

There's no "step change" evident in 1980. The dip in winter 1982 was caused by a record cold June & July. Minima generally "drifted" down after 1981 to around 1998 - which point coincides with the undocumented adjustment.

Jennifer's campaign focussed on the record for Rutherglen; the story's the same there, though more complicated:

Raw data series (blue) versus ACORN-SAT (red); difference (black)

Raw average minimum and ACORN-SAT adjusted track together (minor differences) from 1974. There's no net upward change due to the claimed adjustments in the Adjustment Summary. I've checked the other Adjustment Summaries, and a few other sites which should show adjustments. In every case, the changes resulted in raising or lowering previous temperatures, not those from the adjustment point forward. The B.O.M. isn't in  the business of correcting anomalies to create ACORN-SAT; they're in the business of changing the past to suit an agenda.

Australia's CSIRO - manufacturing sea level hockeysticks?

The CSIRO's "State of the Climate" report was published today (14th March 2010). Missing a link to the report in the newspaper article summarising the dire consequences of climate change in store for Diggers Down Under, I did a Google News search, and came up with a rather interesting ABC page on the release of the report, titled "Sea Levels are Rising", which is not exactly news, but never mind that. Below the obligatory (and of course irrelevant) statement

It notes that the long-term warming trend has not changed, with each decade having been warmer than the previous decade since the 1950s.

... is an "interactive data visualisation" which "explores sea level rise patterns across Australia using data sets provided to the ABC by CSIRO". What struck me immediately was the uptick in sea level at three of the 62 tide gauge stations on the superimposed charts below the interactive map. I've selected Wyndham (WA) on the map - which shows the largest 2010 uptick. I've almost completed an interactive map for Australia myself (similar to the "South Pacific Seal Level 2011" in the reference pages, top-right sidebar), so I'm familiar with the chart for Wyndham. That uptick is definitely not kosher, not kosher at all. here it is:

Source: ABC

... and the RHS embiggened:

Source: ABC

By reading the data points from the ABC chart, I've established that all the preceding annual data points are correct, but 2010 stands out like a sore thumb Here's the corresponding chart for Wyndham, over the same time period, 1993-2010, with the last two points from the ABC chart added in red:

Data Source: NTC

The lower pair of points are Darwin (346) and Carnarvon (322). If the offsets from the real data for all three stations are added, they total 526, which averaged over all 62 stations comes to 526/62 = 8 mm rounded down. If that's deducted from the all-station average for 2010 (dark green line), it brings it down from 120 mm to 112, which means that the three outliers have increased that 112 figure by almost 11%.

I considered whether this might not be deliberate fiddling (generous and understanding to a fault!), but possibly a result of using incomplete data for 2010, for example January-June, but that doesn't explain it - any average with January as a start point is much lower than the ABC/CSIRO figures. Not only that, but 2009/10 were El Niño years, and there's usually a drop in sea level during one of these events; it shows up (though much less pronounced than 1997/8) at most stations except some in the far south. One such rogue plot might be a genuine error, but three? Here are my plots for Darwin and Carnarvon:

Data Source: NTC

Data Source: NTC

I've checked a random sample of the other stations on that rather clever (I'm green with envy - don't take that as weakness) interactive page, and can detect no further excursions from NTC data whatsoever, even for Hobart, home of the CSIRO. Currently, this is a mystery within an enigma, to paraphrase Churchill's description of Russia at the beginning of WW2.

UPDATE 15th March 2012

I've tracked down the source of two of the "adjustments". The CSIRO (funded by the Australian taxpayer) doesn't obtain its sea level data from the National Tidal Centre (funded by the Australian taxpayer, and where you'd expect them to get it as I do), but from an independent archiver, the Permanent Service for Mean Sea Level, based in Liverpool, UK. It's one which I use regularly for data unobtainable elsewhere, but I'm surprised that the CSIRO uses it. The PSMSL relies on data submission from national or other bodies which collect it, and it's rarely up-to-date, often months or even years of the latest data are missing. This explains why CSIRO web pages and publications quote sea level statistics up to Dec. 2010 for the NTC SEAFRAME stations, even though the NTC publish the data monthly.

Anyway, PSMSL has, apparently arbitrarily, adjusted 2010 monthly data for Carnarvon and Wyndham upwards by 140 mm and 260 mm respectively, with no acknowledgement (as is customary) on the station pages. here's a chart for Wyndham for 2009-10 illustrating what's been done:

Data Sources: PSMSL & NTC

That 260 mm is larger by 63 mm than the change in sea level at Wyndham since the gauge was installed in 1966, and larger than estimates of global sea level rise since 1900, so it's no small matter. What are they playing at? The Darwin data (to 2010) at PSMSL appears to be identical to the NTC data (except measured from a different baseline, which is irrelevant), so where did that 128 mm adjustment come from? Sherlock is on the case, with his (t)rusty magnifying glass at the ready. No violin though. On second thoughts, perhaps I'll just ask them. Watch this space for exciting developments.

A Tale of Two Cities

I recently posted a chart of Los Angeles annual temperatures 1881-2011. Here it is again:

Los Angeles annual temperatures 1881-2011 GHCN via GISS

Looks like a fairly typical temperature chart - a warming trend into the 1930s, a dip in the 1940s, then a climb to a peak in the 1990s, followed by the un-typical dip to lower temperatures this century, the subject of that post. The data is GHCN (Global Historical Climatology Network) provided by the NOAA GISS website, and denoted as "unadjusted" which means no further adjustments by GISS. The GHCN data is most certainly adjusted however, as the following chart shows.

Los Angeles annual temperatures 1881-2011 GHCN vs Unadjusted (raw)

GHCN have lowered the raw data from 1881 to 1930 to produce a convincing upward trend. Note that the adjustment is progressively less, from a massive 1.9°C in 1881, smoothly to zero in 1930. There can be no justification for this "adjustment to fit an agenda", which is exactly what it is. Note the difference in the trend line slopes: from 0.08°C per century (raw), which is effectively zero trend, to 1.62°C per century. Instant global warming sans added heat. I've plotted data from the same sources for San Francisco:

San Francisco annual temperatures 1881-2011 GHCN vs Unadjusted (raw)

This is even more blatant jiggery-pokery, with a progressively increasing negative adjustment from zero for the present all the way back to 2.1°C in 1881. GHCN has turned a slightly negative (effectively zero) trend into a surprisingly (is it a surprise?) similar trend to that for Los Angeles of 1.62°C per century. The precise values are 0.016157 for LA and 0.016144 for San Francisco. Coincidence, perhaps, but I think not. I've seen some large GHCN adjustments, but the blatant fiddling here takes the prize.

As I said earlier, "instant global warming" without added heat. There are those who call sceptics "deniers". To those I say "explain or deny this if you can".

UPDATE 13th Feb 2011

I've just checked the GHCN data for both stations on the GISS site, and both have changed slightly since I downloaded the data on 7th Feb. I also had a dataset for SF from December last, and that's different too. Not just rewriting the past, but doing it regularly. I though only totalitarian regimes rewrote the past?

UPDATE 13th Feb 2011

Just a few minutes ago, I had displayed the GHCN temperature chart for Rochester, Minnesota, among others. With a dozen browser tabs open, I didn't realise that the chart was still displayed, so I reselected it from the GISS station selector, and displayed it again. Unbelievably, the data had changed by a few tenths of a degree (both for early and late years) within ten minutes! Some kind of official investigation is needed of this unjustifiable shenanigans.