No more extreme hot days in Australia than 100 years ago
The BoM defines a very hot day as having a maximum of 40C or greater, and the bureau’s own official data show that the ACORN 2 dataset (see analysis) has significantly increased the frequency of very hot days compared to its predecessor, ACORN 1, mostly by decreasing 40C+ days in the first half of the 1900s (see Jo Nova post).
Analysis of the annual 40C+ average numbers and temperatures at the 112 stations allows a comparison between original RAW daily observations and the homogenised ACORN 1 and ACORN 2 datasets. The following analysis is from 1910 to 2017 as this is the final full year of ACORN 1 daily temperatures.
A formula script extracts all days with a maximum of 40C or more each year within the RAW, ACORN 1 and ACORN 2 datasets of daily temperatures sourced to the BoM website. Annual counts and temperatures are calculated at 104, 112 and 60 stations. To estimate annual averages, these are divided by those stations that contributed to the sum in each consecutive year.
The first analysis compares the three datasets at the 104 non-urban ACORN stations used by the bureau to calculate national and regional average temperatures :
It’s clear that ACORN 1 homogenisation reduced the number of RAW very hot days in the early 1900s, and ACORN 2 adjustments have created a steep slope of very hot days by further cooling the early temperature observations.
However, the animation suggests a RAW increase in the annual number of very hot days in Australia. Significant observable step changes coincide with 1972 metrication (see analysis) and the introduction of automatic weather stations in the mid to late 1990s.
Automatic weather stations are believed to increase maxima because of their instant electronic response to warmth compared to the slower responsiveness of liquid thermometers. The very hot day increase in the new millennium also coincides with the introduction of smaller Stevenson screens that decreased their internal space by almost 74%.
The trend and dataset differences in the 104 non-urban stations are similar among all 112 ACORN stations :
The charts above are calculated from the bureau's daily temperatures in the RAW, ACORN 1 and ACORN 2 datasets.
ACORN 1 v ACORN 2 adjustments
The chart below shows the bureau's own official calculation of average annual very hot days from 1910 to 2015 in just the ACORN 1 (archived before the update to ACORN 2) and ACORN 2 datasets sourced from the BoM website :
The chart above shows how ACORN 2 significantly reduced the number of very hot days in the early 1900s calculated by ACORN 1, which itself homogenised and cooled those historic temperatures in the original RAW observations, and eliminated the record average number of very hot days observed across Australia in 1952.
ACORN 1 very hot day (40C+) annual averages (ACORN 1 archived 1910-2016):
ACORN 2 very hot day (40C+) annual averages :
The average annual number of very hot 40C+ days in the archived ACORN 1 dataset for the decade of 2007-2016 was 13.13. This average was exceeded in 2006-2015, 2005-2014, 2001-2010, 2000-2009, 1987-1996, 1986-1995, 1985-1994, 1984-1993, 1983-1992, 1982-1991, 1981-1990, 1980-1989, 1979-1988, 1978-1987, 1977-1986, 1956-1965, 1952-1961, 1951-1960, 1950-1959, 1949-1958, 1948-1957, 1938-1947, 1937-1946, 1936-1945, 1935-1944, 1934-1943, 1933-1942, 1932-1941, 1931-1940, 1931-1939, 1929-1938 and 1926-1935.
The average annual number of very hot 40C+ days in the current (March 2020) ACORN 2 dataset in the decade of 2007-2016 was 12.87. This average was exceeded in 2006-2015, 2005-2014, 2001-2010 and 2000-2009. The average annual frequency of very hot days in the decade 2009-2018 in ACORN 2 (most recent available in March 2020) was 14.44, the highest number in that dataset but less frequent than 1951-1960 (14.49) in the ACORN 1 dataset.
The original weather station observations since 1910
Apart from the metrication and AWS artificial influences, what if many of the ACORN stations that opened after 1910 were in hotter local climates?
That would presumably cause an increase in the number of very hot days recorded as the network expanded from 60 in 1910 to 112 in 1976, when Learmonth opened.
One way to gauge this influence is to look at the years from 1977 (a few years after metrication) to 1996 (before the AWS influence) and compare the 60 fledgling stations with the 52 newcomers :
37.5% more very hot days at the 52 newcomers than at the original 60 stations in the same timeframe suggests that the 52 are in hotter locations.
To avoid the influence of these 52 newer stations that have more very hot days, the original 60 weather stations that were open in 1910 should be analysed.
How many more very hot days do they have nowadays when compared between the RAW, ACORN 1 and ACORN 2 datasets?
The animation shows that Australia used to swelter through more very hot days in the early 1900s than nowadays. It is noteworthy that the original RAW observations were all taken from Stevenson screens.
However, ACORN 1 reduced this historic number of very hot days, and ACORN 2 has adjusted and cooled the past so much that many people are convinced Australia suffers through more “extreme” very hot days than ever previously recorded.
It should be noted that at some weather stations, ACORN actually increased the number of very hot days in the early 1900s. However, these were dwarfed by the number of stations where ACORN adjusted and mostly cooled daily temperatures, and decreased the frequency of 40C+ days in the early 1900s.
The BoM claims Australia's number of extreme days, measured either as the warmest 1% of days for each month or as an annual comparison of absolute temperatures, shows an increase since 1910.
However, these claims are based on homogenised ACORN temperatures rather than original RAW temperatures.
Two examples are Boulia in Queensland and Marble Bar in Western Australia, which provide many of the hot and very hot days within the total for all 112 ACORN weather stations. In 1910-1963, the 60 ACORN stations with maximum observations available since 1910 had 131,519 days above 35.0C in original RAW, and 115,202 in homogenised ACORN (-12.4%). Among these, 21,500 were at Boulia and Marble Bar in RAW (16.4% of Australian total), and 13,587 in ACORN (11.8% of Australian total):
For many decades Marble Bar has held the world heatwave record for the longest number of consecutive days with a maximum above 37.8C (100F or a “century” in the Fahrenheit days), with 160 such days from 31 October 1923 to 7 April 1924.
ACORN 2 has reduced the 160 consecutive century days to just 128 from 1 November 1923 to 7 March 1924, with the total from 31 October 1923 to 7 April 1924 cut to 153 days at or above 37.8C. It’s not known if the advent of ACORN 2 means Marble Bar no longer holds the world record, but this archived BoM article about the 1920s Pilbara heatwave is now wrong if ACORN 2 is considered a more accurate representation of Australia’s temperature history (click here for more detail).
The data analysis above suggests Australia doesn’t have more, probably about the same and maybe even fewer very hot days than it used to.
An influence on the annual average number of very hot days may be Australia's shifting patterns of annual rainfall. The more rainfall, the more likely that cloudy days will prevent the temperature reaching 40C.
The chart below is of very hot days at the original 60 ACORN weather stations open in 1910 (original RAW observations, bearing in mind that most stations have been relocated since 1910 - usually from town centres to nearby airports). This avoids the influence of the other 52 newer stations in hotter locations. The annual rainfall (in millimetres) is sourced to the Bureau of Meteorology from a far greater number of rainfall recording stations across Australia than the 60 ACORN temperature stations, and are a proxy to indicate total rainfall patterns across Australia and different regions.
The chart shows a clear correlation between annual Australian rainfall and the annual average number of very hot days recorded at the 60 original weather stations.
It is established that total rainfall has increased in the northern half of Australia since 1910, mostly since the early 1970s.
The chart below shows the annual average number of very hot days at the 19 ACORN weather stations in northern Australia that were open in 1910, and annual northern Australia rainfall levels sourced to the much larger number of northern Australia rainfall stations maintained by the BoM :
Above can be compared with the 41 ACORN stations open in 1910 in southern Australia :
The chart above suggests a slight decline in southern Australia annual rainfall since the 1970s, albeit comparable with the first half of the 20th century.
The rainfall figures are also influenced by a negligible decline in south-eastern Australia rainfall but a significant rainfall decline in south-western Australia where rainfall levels have dropped by 11% since 1970 within the BoM's defined "southwestern" area bounded roughly by Geraldton in the north to the southern coast just east of Albany (708.1mm in 1910-1969 v 630.4mm in 1970-2017).
Rainfall levels have declined at similar rates across the entire WA region south of Geraldton (1877-1969 464mm / 1970-2017 403.3mm - down 13.1%) and east to Esperance (1883-1968 676.5mm / 1970-2017 618.1mm - down 8.6%). The average of these two outlying stations is a 10.85% rainfall decline.
There are nine ACORN stations in the south of WA with a start year of 1910 - Albany, Bridgetown, Cape Leeuwin, Esperance, Geraldton, Kalgoorlie, Katanning, Perth and Wandering. Their collective annual rainfall averages are 1910-1969 705.0mm and 1970-2017 627.3mm - an 11% reduction (exactly the same as the BoM’s southwestern rainfall data). Average annual very hot days at these nine locations were 1910-1969 4.01 and 1970-2017 4.85 - a 20.9% increase.
The chart below compares the annual average number of very hot days at the nine southern WA ACORN stations open in 1910 with their average annual rainfall :
It appears that Australia's national increase in very hot days at all 112 ACORN stations (or 104 non-urban ACORN stations) can be partly attributed to southern Western Australia, where the frequency of such days has increased because of declining rainfall - not because of more extreme weather caused by CO2. However, the increase from 4.01 before 1970 to 4.85 since that year at the nine long-term southern WA ACORN stations effectively means there is an average one more day each year that reaches or exceeds 40C.
Climate change advocates will attribute the northern rainfall shift to CO2, but must acknowledge that most of Australia enjoys more rainfall than in the early 20th century.
It should also be noted that there are 19 long-term weather stations in the north of Australia where rainfall has increased since 1910, and 41 long-term weather stations in the south of Australia where rainfall has declined, particularly in the south-west.
There are 17 ACORN stations in southern Western Australia (including those that have opened since 1910), and this creates a rainfall bias within estimates of national very hot day trends since 1910.
The four charts above suggest the year-by-year correlation between annual rainfall and very hot days is abundantly clear in the national and northern Australia comparisons, but less so in southern Australia and particularly the south of WA.
This may be because northern Australia receives mostly monsoonal rainfall in the summer months when it is also the hottest. This isn't the case in the southern third of Australia, where rainfall usually follows a traditional pattern of wet winters and dry summers.
However, many southern Australia locations have only minor differences between summer and winter average rainfall, with ample summer rainfall to influence the occurrence of very hot days warmer than 40C. For example, Sydney Observatory receives an average 101.4mm in January but 95.7mm in July, Kalgoorlie averages 27.5mm in January and 24.2mm in July, and even Melbourne averages 46.8mm in January and 47.5mm in July.
Annual rainfall is a proxy for annual daily cloud cover and is an imperfect comparator. The seasonal rainfall differences in southern Australia and southern WA possibly explain why there is a less obvious correlation between annual rainfall and very hot days in these regions.
Nevertheless, the national and northern Australia correlations are too consistent to believe that annual rainfall isn't an indicator that changing rainfall patterns and associated cloud cover are a more likely cause than a warming climate to explain an increase in Australia's increasing frequency of very hot days within the total ACORN network of 112 stations, along with warmer locations that opened between 1910 and 1976 and the historic cooling caused by ACORN 2 adjustments.
Note: Critics might argue that annual rainfall is mostly in winter and creates an inaccurate correlation with the average annual number of very hot days, none of which occur from May to October - even though both datasets are in the same January-December timeframe. Click here to view the four charts above with annual very hot day averages and rainfall averages either in summer or from November to April.
The frequency of very hot 40C+ days at Australia's 60 long-term ACORN stations can also be compared with the number of days of rainfall recorded at those sites in 1910-1963 and 1964-2017 :
Click here for an Excel download with data at all 60 stations for the table above.
Average temperature of very hot days
But what about how hot those very hot days are nowadays compared to the first half of the 20th century? Has the average maximum of all days above 40C increased since the early 1900s?
First, the average annual temperature of very hot days in RAW, ACORN 1 and ACORN 2 at the 104 non-urban ACORN stations :
There has been an increase since 1910, much of it following 1972 metrication and the introduction of automatic weather stations with smaller screens (0.15C warmer in 1964-2017 than 1910-1963, according to RAW). However, several years of record heat in the 1930s persist, despite ACORN adjustments that otherwise cool the temperature of very hot days in the early 20th century.
Again, similar among all 112 ACORN stations :
What about those 52 stations that have opened since 1910 that the earlier comparison suggests were in hotter locations?
Just because they have more very hot days, does that mean their maxima above 40C are hotter than at the 60 original stations? Below compares the annual average of 40C+ days, and the annual average maxima, at the 60 stations with the 52 stations - both from 1977 to 1996 :
The 52 stations that opened between 1910 and 1976 have 37.5% more very hot days which, on average, are 0.24C warmer than the average at the 60 original stations in the same timeframe. The average annual maximum at the 52 newcomer stations is 0.15C warmer than at the 60 original stations that date back to 1910.
The 0.15C warming from 1910-1963 to 1964-2017 at the 104 non-urban stations is now questionable as almost half the contributing stations - those that opened after 1910 and up to 1976 - are in hotter locations.
And how much hotter are the very hot days at the 60 original stations?
Among the 60 original stations open in 1910, the very hot days were 0.09C warmer in 1964-2017 than 1910-1963, according to RAW.
Surprisingly, the ACORN 1 and ACORN 2 datasets have caused a small increase in the average maximum temperature of 40C+ days in the early 1900s.
These illogical quirks of the ACORN area average algorithms saw Albany having Australia's hottest every day in ACORN 1 (51.2C vs 44.8C originally and adjusted back to 49.5C in ACORN 2). In ACORN 2, Oodnadatta becomes the hottest day ever recorded in Australia (51.1C on 2 January 1960 vs unadjusted raw temp of 50.7C), and Carnarvon has Australia's second hottest ever day in 1953 (51.0C vs 47.7C originally).
Between 1910 and 1976, the gradual addition of 52 hotter weather stations has contributed to the trend and the artificial warming of the average historic temperature of very hot days.
Very hot 40C+ days are not the same as average maxima, and analysis shows ACORN 2 cooled RAW by 0.2C in 1910-1963 (RAW 25.03C to A2 24.83C).
It's also worth noting that at the 60 original stations open in 1910, the average annual number of very hot days was 473.2 in 1910-1963 and 459.2 in 1964-2017.
This analysis suggests that despite ACORN homogenisation, Australia has had no increase in the number of extremely hot days since 1910.
Click here for an Excel download containing the annual tallies and average annual maxima at all 112 stations, and the calculations used in the analysis above.
This website produced by Scribeworks 2008-2022