JoNova

With only a million dollars a day it’s hard for the BoM to keep up with their own stations. Luckily Bill Johnston has arrived to help out for free. The BoM announced that it was Canberra’s hottest ever day last week, but forgot to check whether the heat was due to the site moving three times,  changes in thermometers and a wind profiler they installed themselves in 2010.

Normally the BoM would detect and correct for these sorts of things by using Homogenisation Magic (HM). That’s where they spot these effects by comparing a station with surrounding stations. However in this case HM missed all three site moves and the wind profiler.  It looks like those might add up to 2.2°C of artificial warming. Nothing to worry about, but the hottest ever record will have to be shredded, and naturally, the BoM will need to issue a correction with at least as much fuss and coverage as the mistaken headlines. It’s only fair…

After the effect of rainfall is removed there are at least three site moves, a screen change, equipment change, and alterations to the surrounding area that may influence the site. These step changes align with documented moves and account for most of the warming.

Instead of using Homogenisation Magic, Johnston did things like getting site photos from national archives, and from google as well as using the BoM’s own data. He found out that most things have changed about the Canberra site since 1995. In 1995, not only did the thermometer shift to a new spot, the old liquid-in-glass style was converted to an electronic sensor (and we know they can be more sensitive and responsive to temperature changes). On top of that the screen around the thermometer shrank to about a quarter of its original volume (from 0.23m³ to 0.06m³). In the last ten years, large gravel or concrete pads were built near the sensor as well as a thingummygig called a wind profiler (which is a vertical radar array for detecting atmospheric turbulence).

Johnston uses a change in the relationship between temperature and rainfall to spot the site changes (then confirms them with documentation). If a site moves to a warmer or cooler spot the temperatures will shift up or down but the rainfall won’t.  The relationship between rain and temperature will be broken — an effect that shows up during the following years of data. These changes in pattern often occur at the same time as a site change, and Johnston uses the shifting ratio’s to estimate how much effect the site change has.

Johnston has looked at a lot of sites around Australia, and estimates that many modern records wouldn’t be records at all if the BOM bothered to dig out historic photos, maps, and data and took more care to estimate the effect of site changes. The techno magical homogenization tool sounds fancy but allows site problems to be turned into misleading headlines.

— Jo

_____________________________________________________________

Welcome to Canberra airport where it’s always sometimes hotter

A sad day for meteorology but another marketing success story for Australia’s BoM.

Guest post by Dr. Bill Johnston[1]

Main points.

• Canberra’s record “hottest ever September day” is a great headline but not remotely real. The BoM forgot to mention thermometers were moved at least three times, each time to a warmer site.
• Homogenisation is a complete failure. Canberra temperature appears to follow model projections because homogenisation ignores site changes in 1973, 1997, 2004 and 2010.

Introduction

It’s not fair to expect meteorologists who don’t observe the weather to know much about their data.

However, its reasonable for people in Canberra gouged by ACTEW-AGL’s energy prices to expect Ashleigh Lange from the Bureau of Meteorology to have researched the BoM site at Canberra Airport so she knows what she is talking about when claiming “Canberra records its hottest ever September day” (Canberra Times, 23 September 2017).

It was cold as a toad three days before the record was allegedly “smashed”; people and businesses were forced to turn on the heat, which sent ACTEW-AGL laughing all the way to the bank. Which bank? The ones that won’t support low-cost coal-fired electricity generation of course.

Why is site history important?

Moving a Stevenson screen often results in a permanent change in background heat sources and sinks. For example buildings may shield prevailing winds; a site near a runway is likely warmed by tarmac not the weather. So for places like Canberra airport, where the site moved at least three times (See those details below) its important to know if temperature changes are due to site effects or the climate.

Canberra Airport Figure 1. The original 1939 Canberra airport Aeradio site was near the hanger (S1) where the office was located. The site moved to S2, probably in 1973; then to an AWS south of the meteorological office at S3 (December 1995); then S4 in December 2008.

Rainfall can be used to help analyze the effect of changes in the site and equipment

Evaporation cools the environment and there is a simple relationship between average Canberra maximum temperature (Tmax) and annual rainfall. Dry years are warm years, and the drier it is, the hotter it gets [Figure 2 (a)]). [It’s not the same for minimums where things are more complicated. A wet year can have slightly warmer nights.  Clouds and humidity reduce heat loss by radiation so temperature is warmer in the early morning. (Exposure of the Stevenson screen to down-slope drainage of cool air is also important.) As foggy and cloudy winter days often don’t bring rain in Canberra, relationships between Tmin and rainfall are less clear-cut.]

The variation in Tmax can be split into the portion attributable to rainfall (the component described by the linear regression line [Figure 2 (a)]) and the residual variation, which is the portion not explained by rainfall. The relationship is robust if it is statistically significant and more than 50% of Tmax variation is accounted-for (R²_adj, the coefficient of determination, is greater than 0.50). Although random in the rainfall-domain, Tmax-rainfall residuals may embed a hidden time-signal. As rainfall effects are removed the trajectory of residuals (rescaled by adding grand-mean Tmax) is unaffected by persistent weather effects and is analysed for step-changes using an independent statistical test [Figure 2 (b)].

Residual step-changes show that either the S1-site changed in 1973 (due to works nearby, for example) and it moved to S2 later (about 1975); or the move in the ACORN-SAT catalogue is miss-specified. Many things changed in 1995 [the site moved to S3, an AWS replaced thermometers, and it is likely that a small Stevenson screen (0.06m³) replaced a previous large one (0.23m³)]. In 2005, a vertical radar array for detecting atmospheric turbulence (wind profiler) was installed on a 350 m² gravel or concrete pad 30 m north of the screen and a bitumen car-park replaced a former playing field 25 m east (Figure 3). However, the site didn’t move to S4 until December 2008, afterwards, in 2010, the wind-profiler relocated 40 m north of the S4 screen.

Figure 2. Tmax depends on rainfall (a); however, due to embedded site changes, variation explained (R²_adj) is only 0.36 or 36%.  As the naïve regression (Tmax = 21.91 – 0.34^oC/100 mm) accounts for rainfall; rescaled residual variation (b) is due to non-rainfall factors including site and observer inconsistencies data in-filling etc.; and site changes and relocations. (Out-of-range outliers (o/r; red squares) are excluded from analysis.) Segments defined by step-changes are analysed separately in (c). Except for 2005 to 2016, relationships are significant. Differences in R²_adj reflect both the number of cases and closeness of data to respective regression lines and are not comparable. Dotted lines indicate median rainfall (vertical) and overall average Tmax (horizontal), which provides a visual reference. (A P value of 0.10 is not statistically significant.)

Factored on step-changes, simultaneous analysis[3] shows regressions [individually free-fit in Figure 2 (c)] are parallel; rainfall reduces Tmax 0.33°C/100 mm and 78.8% of Tmax variation is explained (vs. 36% for rainfall alone). Moving the site beside the runways in 1972; then to the AWS and small screen in the vicinity of the new met-office in 1995; then installation of the 2005 wind-profiler array; sealing the car-park; then in December 2008 moving 400 m south along the eastern airport boundary, where a wind-profiler array was installed in 2010 caused data to warm 1.75°C[4] (0.43°C + 0.76°C + 0.56°C [Figure 2 (b)]). With those changes and rainfall accounted-for no Tmax trend remains that is indicative of changes in the climate.

The history of Canberra airport’s moving weather station

Temperature measurements started at Canberra airport in 1939 when Aeradio was set-up to monitor aircraft and advise pilots of inclement weather. In those days planes flew through turbulence not above it like jets do and Aerado ensued aviation safety. Plans at the National Archives of Australia show the Aerado and meteorological offices were in the northwest corner of the original hanger, still standing at RAAF (Royal Australian Air Force) Fairbairn. In June 1940 weather observers and radio operators were conscripted to the RAAF and after WWII, in July 1946, air traffic control merged with the Department of Civil Aviation and Weather Bureau staff transferred to Department of the Interior. As technology changed and demand for services increased facilities across Australia were up-graded in the 1950s. Radar was introduced to monitor aerosonde balloons; the busy met-office in the hanger at Canberra airport was provided with forced ventilation in 1954.

The National Library of Australia holds aerial photographs from 1956 and 1960 that show a meteorological enclosure between the north-south runway and the hanger (at about Latitude ‑35.3049, Longitude 149.2014) (S1 in Figure 1). The 2012 ACORN-SAT[2] catalogue states vaguely that the “original site (070014) was on the eastern side of the airport”; and “there are indications … of a possible move in about 1975”. The Canberra Aero Comparison (70228) (December 1995 to 1997) site summary locates the second site (S2) in the centre of the airport near intersection of the N-S and E-W runways (Lat. -35.3083, Lon. 149.1936). The site moved 800 m northeast to an automatic weather station (AWS) in December 1995 (S3); then 400 m south in December 2008 (S4). Canberra Airport Comparison (70014) metadata doesn’t mention either of the two earlier sites so perhaps the Bureau doesn’t really know what happened.

Discussion and conclusions

Daily temperature fluctuates around average site temperature. For September, average S1 Tmax from 1939 to 1972 is 15.8°C; from 1973 to 1996 (S2) it is 16.0°C; from 1997 to 2004 (S3), 17.3°C and at the current S4 site it is 18.0°C. Due to site changes and moves which have nothing to do with the climate, average September Tmax has shifted-up by 2.2°C!

World’s worst-practice compounds multiple site and instrument changes. Although the 1973 S2 site operated in parallel with S3 until December 1997; reply to an FOI request confirmed that data for thermometers observed in parallel with the AWS until March 2010 at S3 (which would enable the change to the AWS/small screen to be cross-referenced using the same instrument) are not available. Internal Bureau policy directed that manually observed data were discarded without being databased and that paper records were destroyed. In the absence of comparative data statistical inference and interpretation of site factors (Figure 3) is the only evidence that the heat signature of the MO-site (S3) is different to the runway (S2) site.

…		Canberra
Figure 3. Google Earth satellite images of the Stevenson screen (sc) and meteorological office (mo) at S3 beside a sports oval on 11 March 2004 (left); and with the wind profiler array (wp) and sealed car-park on 31 March 2008 (the site did not relocate until December 2008).

Homogenisation makes no Tmax adjustment despite the weather station moving three times to situations whose background heat signatures are different; and changes such as to the AWS and small Stevenson screen in 1995, which is documented; and installation of wind profiler arrays in the vicinity in 2005. Bureau meteorologist Ashleigh Lange who claimed 30.2°C on Saturday 23 September  “smashed a 52-year record” and was the “hottest ever September day” may not have visited the site, analysed any data and may not know that Tmax is affected by background warming independent of the climate. Even if it’s an above-average anomaly, inflated by site factors it is not a valid 52-year record.

Homogenisation provably does not improve site records. Choosing not to adjust for site and instrument changes in 1973, 1997, 2004 and 2010 causes Canberra data to warm like the models claim it should. Although Canberra airport data are not used directly to calculate Australia’s warming, Bureau meteorologists use the site to stir alarm about the climate.

Furthermore, airport data spread their faults far and wide: to ACORN-SAT sites at Bathurst Agricultural Research Station (63005), Dubbo airport AWS (65070), Moruya Pilot Station (69018) and Nowra RAN (68072). Data for those, adjust others, including Sydney Observatory, whose numerous faults find their way to Alice Springs (via. Tibooburra). Homogenisation of faulty ACORN-SAT site data using other data that are faulty is flawed. Instead of being recycled into yet another version, ACORN-SAT should simply be abandoned.

Bureau marketing of record-heat that doesn’t exist is the reason for the RET; the rivers-of-cash ACTEW-AGL takes from the Canberra community and the ACT Labor-Greens government’s hard left-turn to consumer-subsidised unreliable “renewables”.

Trends don’t exist; the cost is too high; people are hurting and losing their jobs. An open public inquiry into the Bureau of Meteorology and its dodgy homogenisation methods is long-overdue. In the meantime, the Bureau should cease making claims about record temperatures, which like at Canberra are provably false.