Many of you may recall that I got my start in climate skepticism back in 2006 when I started looking at the paint on Stevenson Screens – because there was a change from the original lime-whitewash paint in the 1890s to modern latex paint. I figured there was a bias, and latex paint made the shelter warmer due its different IR signature. Temperature sensor tests over a month proved I was right. But in looking at temperature shelters in my area, I discovered an even bigger problem – most were sited near heat sources and heat sinks, in contradiction to NOAA’s own published siting standards. This started my journey to uncover just how bad the temperature observing network actually was. Comprehensive reports I made in 2009 and again in 2022 showed that surface measurements were a huge warm biased mess. This paper is over 10 years old, but I somehow missed it. I’m correcting that oversight.
Now, to add to that mess, comes this revelation – the Australian Bureau of Meteorology changed the size of Stevenson Screens to something that had just ~ 25% of the volume of the original, and did not run parallel tests to see if the conversion mattered. – Anthony
Craig Kelly of the AFEE in Australia writes on X.com
The peer-reviewed science confirms that shrinking the size of Stevenson Screens increased average temperatures across a year by 0.54°C and, on hot summer days, it can increase the maximum temperature by 1.7°C. https://waclimate.net/stevenson-sizes.pdf Yet the BOM denies the existence of this peer-reviewed science, pretends that it doesn’t exist, and claims that shrinking the screens by 74% had no effect on the recorded temperatures.
Furthermore, at every weather station where the BOM replaced the traditional “large” Stevenson Screens with smaller ones, they ripped out the large ones and replaced them with the small ones on the very same day. This is contrary to long-established practices, which require that when you change measuring equipment, you keep parallel data from both setups to determine whether the equipment change may have introduced a warming or cooling bias into the record.
If you wanted to artificially inflate temperatures and create new “record hot days” to generate propaganda for the climate cult, you’d do exactly what the BOM did: shrink the size of the Stevenson Screens. And if you wanted to cover up your malfeasance and fraud, you’d rip out the large screens and replace them with the small ones on the very same day so there would be no parallel data — exactly what the BOM did.
Attached is a photograph from the Sydney Observatory from 1947 showing the thermometers that officially recorded Sydney’s temperatures housed inside a traditional ‘large’ Stevenson Screen – with an internal volume of approximately 0.23m3. The BOM has shrunk the size of the Stevenson Screens, reducing the internal volume to just 0.06m3 – a 74% reduction. By shrinking the Stevenson Screens in such a manner, how much hotter will the recorded temperatures be inside the smaller screen on a hot and windless day?
Figure 1. Internal view of the large Stevenson screen at Sydney Observatory in 1947 (Top above, black and white image) and the small screen at Wagga Wagga airport in June 2016 (lower above, colour image). While thermometers in the 230-litre screen are exposed on the same plane, the electronic probe at Wagga Wagga is placed behind the frame about 2 cm closer to the rear of the screen, which faces north to the sun. According to metadata, the 60-litre screen at Wagga Wagga was installed on 10 January 2001 and although thermometers were removed on 28 April 2016 inter-comparative data is unavailable. Source: BoMWatch.
From the paper:
The main findings of this research are summarized as follows:
- An overheating of air temperature inside the medium-sized Stevenson screen was detected in comparison to the large-sized Stevenson screen throughout the year. This bias affected daily maximum air temperature records, especially during the warm season (May to October) and at 1300 UTC.
- The weather conditions enhancing this overheating bias (not statistically significant) are associated with clear skies, high solar radiation rates, weak winds and low relative humidity values.
- Comparison to nearby station have revealed that the different size of the naturally ventilated wooden Stevenson screens have an impact on mean, maximum and daily air temperature range. These kinds of investigations are crucial for removing inhomogeneities and accurately assessing the
spatio-temporal variability and long-term trends of near-surface air temperature measurements.
Anthony, your climate skepticism didn’t begin until 2006? Wow! My skepticism began in high school, and that was back in the 1980’s….
Source: ChatGPT (& my questions based on the Anthony’s article above).
”Background from the Spanish Study
A controlled field experiment in Spain compared two naturally ventilated Stevenson screens of different sizes placed side-by-side for one year. The study demonstrated that the smaller screen systematically recorded higher daytime maximum temperatures (Tmax) than the larger traditional screen. The mean annual bias in Tmax was approximately +0.54 °C, while during warm, dry, low-wind conditions typical of late spring to early autumn the difference frequently reached +1.0 to +1.5 °C, with individual days up to +1.7 °C. Night-time minimum temperatures (Tmin) showed almost no effect (around −0.1 °C). The bias was physically consistent with enhanced solar heating and reduced ventilation in the smaller enclosure.
—
### Assumed Scenario for Australia
For the following analysis we adopt a hypothetical but plausible scenario:
1. Replacement of traditional large Stevenson screens with smaller enclosures began around 1990 and progressed gradually over 20–30 years.
2. The size change produces a systematic bias identical to that observed in the Spanish experiment: +0.54 °C on average for Tmax, rising to +1.0–1.5 °C in typical summer conditions.
3. The change occurred station by station, not simultaneously, so the network-wide bias increased progressively.
4. The existing homogenisation procedures are assumed not to correct this specific physical bias.
5. Climatic conditions themselves are assumed unchanged in order to isolate the measurement effect.
Based on these premises we estimate that approximately 10–30 % of stations had changed by the late 1990s, 40–60 % by the 2000s, and 70–90 % by the 2010s.
—
### 1990s – Emergence of Bias
During the 1990s only a minority of stations would have used the smaller screen. The network-mean artificial warming would therefore be modest, roughly +0.05 to +0.16 °C in annual Tmax. Nevertheless, this would already influence statistics based on extremes: the number of hot days would increase and new “records” would appear disproportionately after the replacement. Any real climate signal during this decade would be mixed with a growing non-climatic component.
—
### 2000s – Rapid Distortion
With an estimated 40–60 % of stations converted, the artificial contribution to network Tmax would rise to about +0.22 to +0.32 °C annually. In summer months the effect would be considerably larger, +0.4 to +0.8 °C. This magnitude is comparable to the warming attributed to Australia for the same period in IPCC assessments. Consequently, a substantial fraction of the apparent early-21st-century warming could originate from instrumentation rather than atmospheric change, particularly in analyses focused on daytime heat and heat-wave frequency.
—
### 2010s – New “Normal”
By the 2010s, when most stations are assumed to have adopted the smaller screen, the embedded bias approaches its full value: +0.38 to +0.49 °C in annual Tmax and +0.7 to +1.3 °C during warm seasons. The diurnal temperature range would expand artificially because Tmin remains largely unaffected. Trend calculations over 1990–2020 would therefore show an apparent rise of roughly +0.15 °C per decade even if the true climatic trend were near zero.
—
### Comparison with Reported IPCC Trends
IPCC-referenced analyses indicate Australian surface warming of roughly 0.1–0.2 °C per decade in recent decades. Under the scenario described above, instrumentation alone could generate +0.4 to +0.6 °C of warming over 30 years, equalling or exceeding the reported signal. Metrics relying on Tmax—heat-wave counts, record highs, fire-danger indices—would be especially sensitive, potentially doubling without any physical change in climate.
—
### Implications
If such a transition occurred and remained uncorrected, long Australian Tmax series would be structurally non-comparable across time. The resulting trends would primarily reflect the gradual introduction of a warmer-biased sensor environment rather than atmospheric evolution. Robust reassessment would require parallel measurements with both screen types, physically based adjustment models, and explicit metadata on the timing of replacements. Until then, conclusions drawn from unadjusted Tmax data should be regarded as highly uncertain.
”
This conclusion (“structurally non-comparable “) is vitally important in deciding if trending is a viable option across time.
It is where statisticians declare that bias occurs and that they can fix it by changing data. All the while not even realizing that it is likely that different microclimates are being measured causing the difference. The older measuring system may well be accurate to ±0.1 degree but measuring different things.
It seems reasonable on a case by case basis to want to give the bad actors the benefit of the doubt. “Oh, they probably thought this was a neutral change.” It was just a mistake.
The problem comes when you look at the entire picture of all the ‘mistakes’ and they are almost universally tilted to one side. Mistakes on average should fall approximately evenly on both sides and generally be approximately equivalent. But when you look at all the mistakes, they almost all lean towards proving global warming. That indicates malice rather than simply mistakes.
Thanks for this Anthony. I’ve only been talking about this for a decade. Long ago I did a post …. https://joannenova.com.au/2019/10/shrinking-stevenson-screens-cause-global-warming-and-peeling-paint-long-grass/
It is not just rolling out smaller screens, it is doing it over time so nobody (except a busybody like me) has the field-based knowledge or has developed the protocols to detect such changes on a site-by-site basis.
The BoM has also rolled-out 60-litre PVC screens. These have more louvers, less airflow, and I looked inside one and it was painted matt-black so it radiates from the outside into the screen.
They have also enclosed their ‘enclosures’ with 1.8 m -high heavy duty black security fencing, which traps even more heat on warm windless days.
Check-out https://www.bomwatch.com.au/data-homogenisation/part-9-rabbit-flat-roadhouse-northern-territory-australia/ and the full report: http://www.bomwatch.com.au/wp-content/uploads/2025/07/Rabbit-Flat.pdf
Yours sincerely,
Dr Bill Johnston
https://www.bomwatch.com.au/
I recall back in the days of Gore’s propaganda, “An Inconvenient Truth” reading about the significant effects of changing from whitewash to latex.
Now I am reading about changing the placement of the sensor withing the enclosure.
I have to wonder if the next update will change from wood to aluminum.
What I find most personally interesting is Anthony’s initial concern about the change in color between the lime whitewash , and the latex .
The mathematics of the crucial calculation between the power spectrum of a radiant source ,eg: the Sun , and the color , absorptivity==emissivity spectrum of an object , eg: our planet , is generally all too poorly understood , altho is high school level physics . See :
https://x.com/i/grok/share/hdvZNCjYz0twizt9c1rH0tDna
which explains why simply citing ` albedo is essentially meaningless ,
and https://cosy.com/Science/warm.html#EqTempEq
Which expresses the energy balance in terms of the ratio of dot products of the of the object color spectrum with that of its source(s) and sink(s) .
Thus the effect of a change in color , like from whitewash to latex , can be calculated .
Consider also that it doesn’t matter if it is whitewash or latex, both suffer from degradation over time from expsure to UV from the sun. This will cause a gradual drift in the reflection coefficient of the surface. Of course, climate science doesn’t care which is obvious from the fact that they never publish a measurement uncertainty budget for *any* of their measurement devices. Even microclimate changes such as grass turning from green to brown and back to green influences the radiation impacting the screen from the actual surface of the earth – meaning calibration drift is different over time. Compensating for seasonality won’t catch this.