Historic climate comparison with 2000-2021
Based primarily on <1931 temperatures collated by the Council for Scientific and Industrial Research and Australia's 1953 and 1954 Year Books
These estimates are based on a comparison of Meteorological Data for Certain Australian Localities, published in 1933 by the Council for Scientific and Industrial Research, and unadjusted temperatures from the BoM's Climate Data Online platform.
The CSIR document collates unadjusted minima and maxima from the earliest records at all Australian stations up to and including 1931.
Most weather station thermometer shelters in the 1800s were different to the Stevenson screens introduced since the late 1800s/early 1900s, creating heating or cooling biases among minima and maxima observations. These influences are considered below - see Temperature accuracy.
The BoM RAW temperature data are monthly averages from 2000 to June 2021, with some relocated stations having time periods starting after 2000 to ensure the most recent data from 2021 locations, and those monthlies are then averaged from January 2000 to June 2021.
The largest mean temperature increase from <1931 to 2000-2021 was in Tasmania (0.9C) and the smallest increase was in South Australia (0.2C).
New South Wales had the most weather stations before 1931 (72) and that state saw mean temperatures increase 0.4C by 2000-2021.
Among the 225 stations, Windorah in Queensland had the largest mean temperature increase of 2.4C from <1931 to 2000-2021, and Bridgetown in WA had the largest decrease of -1.2C.
The table below presents summary findings from the comparison, including rainfall averages at all 225 stations from their start years to 1931, in periods matching with temperature observations, and in 2000-2021:
Download CSIR Excel spreadsheet (272kb)
Monthly temperatures and rainfall
The table below breaks down total temperature trends in the table above into monthly trends, showing month-by-month maxima, minima, their warming from <1931 to 2000-2021 including means, and national monthly rainfall comparing all years before 1931, years before 1931 matching when temperatures were observed, and 2000-2021:
There is a correlation between monthly temperatures and changing rainfall patterns since the 1800s.
For example, the warm months of October to March saw an averaged 3.3mm increase in rainfall, with averaged maxima increasing by 0.5C and averaged minima increasing by 0.8C - a difference to be expected from more rainfall cloud cover warming nighttime temperatures due to trapped heat.
In contrast, the cool months of April to September saw an averaged -9.5mm decrease in rainfall, with averaged maxima increasing by 0.7C and averaged minima increasing by 0.3C - again in line with reduced rainfall warming daytime temperatures and reduced nighttime cloud cover trapping less heat.
It is noteworthy that if comparing 2000-2021 with periods when pre-1931 temperatures were recorded at each station, the largest rainfall increases at the 225 stations were in February (up 11.3mm) when their averaged maxima increased 0.1C and averaged minima increased 0.7C, and in November (up 7.6mm) when their averaged maxima increased 0.4C and averaged minima increased 0.7C.
The biggest rainfall decreases at the 225 stations were in May (down 13.8mm) when their averaged maxima increased 0.7C and averaged minima increased 0.3C, and in June (down 12.7mm) when their averaged maxima increased 0.6C and averaged minima increased 0.2C.
The table below shows the jurisdictions with the largest rainfall decreases inconsistently but generally had the largest mean temperatures increases.
Rainfall increased significantly in the Northern Territory but this was based on only four weather stations.
Nationally, there was an average 1mm rainfall decrease at the 225 stations from their start years (earliest Adelaide 1839) to the periods when pre-1931 temperatures were recorded at each station, suggesting Australian rainfall has been declining since the 1800s.
The averaged annual rainfall decrease of -3.1mm at all 225 weather stations in 2000-2021 compared to years corresponding with CSIR temperature recordings is at odds with BoM calculations that national rainfall has increased since 1900.
BoM data suggest an annual rainfall increase of 3.98mm from 1900-1931 to 2000-2021, mostly in northern Australia where average annual rainfall increased from 497.01mm in 1900-1931 to 544.09mm in 2000-2021 (47.08mm increase). BoM figures suggest that in southern Australia, average annual rainfall increased from 370.20mm in 1900-1931 to 371.20mm in 2000-2021 (1mm increase).
The discrepancy between a 47.08mm increase in northern Australia and the national increase of 3.98mm is because of the distribution of BoM rain gauges being heavily biased to southern Australia, as seen in the map below showing the number of years between 1870 and 2019 for which sites reported rainfall within a 0.25o grid cell (source).
This CSIR analysis suggests 48 weather stations in northern Australia (see BoM definition of northern/southern) saw averaged maxima increase 0.73C and averaged minima increase 0.83C from <1931 to 2000-2021 - a mean temperature increase of 0.78C.
The 177 weather stations in southern Australia saw averaged maxima increase 0.56C and averaged minima increase 0.45C from <1931 to 2000-2021 - a mean temperature increase of 0.51C.
Average rainfall in the 48 northern Australia stations reduced by 1.25mm from <1931 to 2000-2021, and average rainfall in the 177 southern Australia stations reduced by 3.58mm.
Although there is a far more comprehensive density of weather stations in southern Australia where there is more urban heat, this finding suggests there was a greater reduction in nighttime rainfall cloud cover trapping daytime heat in southern rather than northern Australia - resulting in a minima increase almost half that experienced in the 48 northern stations.
ACORN homogenisation doesn't take account of rainfall trends or urban heat (apart from eight major cities and regional locations that are excluded from national averages), and decreased rainfall since the 1800s at the 225 weather stations in this CSIR analysis suggests a precipitation influence on temperature trends.
Year Book <1950
The 1953 and 1954 Year Books of the Commonwealth of Australia present maxima and minima prior to 1950 at 79 Australian weather stations that can be compared with those stations' temperature observations during 2000-2012.
This data shows their average mean temperature has increased 0.41C from <1950 to 2000-2021 (18.15C > 18.56C).
As in the CSIR <1931 dataset, there was a variety of screen shelters used at the 79 Year Book stations dating back to the mid 1800s. Almost all used Stevenson screens by the first decade of the 1900s.
Although 79 stations is far fewer than the 225 in the <1931 CSIR analysis, the unadjusted Year Book temperatures reinforce the validity of the CSIR estimate that Australian mean temperatures increased 0.6C by 2000-2021.
If accurate, the Year Book temperatures suggest maxima cooled 0.07C from <1931 to <1950 and warmed 0.64C from <1950 to 2000-2021. The data suggest minima warmed 0.12C from <1931 to <1950 and warmed 0.16C from <1950 to 2000-2021..
The table below presents summary findings from the Year Book comparison and includes CSIR <1931 temperatures for further comparison:
Tasmania had the largest mean temperature increase of 0.67C from <1950 to 2020-2021 and the Northern Territory had the smallest increase of 0.11C.
Western Australia had the largest maximum temperature increase of 0.94C from <1950 to 2020-2021 and NSW had the smallest increase of 0.42C.
Tasmania had the largest minimum temperature increase of 0.66C from <1950 to 2020-2021 and the Northern Territory had the largest decrease of -0.33C.
In the warm months of October-March, average maxima warmed 0.61C (27.94C > 28.55C) from <1950 to 2000-2021 and, in the cool months of April-September, average maxima warmed 0.68C (20.23C > 20.91C).
In the warm months of October-March, average minima warmed 0.40C (15.56C > 15.96C) from <1950 to 2000-2021 and, in the cool months of April-September, average minima cooled -0.08C (8.88C > 8.80C).
Year Book temperature extremes
The 1953 and 1954 Year Books also present hot and cold daily temperature extremes recorded prior to 1950 at 79 weather stations dotted around Australia.
The table below compares hottest and coldest days in each month prior to 1950 with the hottest and coldest days recorded at the weather stations operating currently at or near these locations (from their first day of observation to 31 July 2021).
The table also shows extreme hot and cold day averages and most extreme days at 34 ACORN stations within the broader Year Book dataset of 79 stations, as well as trends at locations with large and small surrounding populations that might influence urban heat:
The table above shows the average temperature of the hottest day each month among all 79 stations before 1950 was 0.05C warmer than the average of the hottest days observed in each month at the current stations during their years of operation up to 2021.
Among the hottest days recorded at any of the 79 stations each month before 1950, their average was 0.76C hotter than at the current stations during their years of operation up to 2021.
The table above shows the average temperature of the coldest days each month among all 79 stations before 1950 was 0.30C colder than the average of the coldest days observed in each month at the current stations during their years of operation up to 2021.
Among the coldest days recorded at any of the 79 stations each month before 1950, their average was 0.04C colder than at the current stations during their years of operation up to 2021.
Among all 79 stations, the average temperature of extreme hot days in the warm months of October to March was 0.09C warmer before 1950 than in the stations open in 2021 (41.30C > 41.21C), and in the cool months of April-September they were exactly the same before 1950 as the stations open in 2021 (30.36C > 30.36C).
Among all 79 stations, the average temperature of extreme cold days in the warm months of October to March was -0.63C colder before 1950 than in the stations open in 2021 (5.85C > 6.48C), and in the cool months of April-September they were 0.03C colder before 1950 than in the stations open in 2021 (-0.46C > -0.49C).
A notable point is that although the average of all cold days in each month at all 79 stations was 0.26C colder before 1950 than in the stations open in 2021, the coldest days in Melbourne averaged 1.1C before 1950 and 5.1C in 2013-2021 - a 4.0C warming indicative of urban heat.
Among the 79 stations, 34 are within the Australian Climate Observation Reference Network (ACORN) maintained by the BoM to estimate national and regional temperature trends, with numerous daily temperature adjustments back to 1910. The adjusted ACORN 2.1 averages and most extreme hot/cold days can be compared in the table above with the unadjusted Year Book and daily extremes recorded at those stations open in 2021.
The average adjusted temperature since 1910 of hottest days at the 34 ACORN stations was 0.50C warmer than original RAW temperatures in current stations during their years of observation, and the adjusted coldest days averaged 1.33C cooler than in the original RAW observation periods of the current stations.
Among the hottest days recorded anywhere in Australia at a single station since 1910 (different to averages of the monthly hottest/coldest day average at all stations), the average adjusted temperature of hottest days at the 34 ACORN stations was 0.36C warmer than original RAW temperatures in current stations during their years of observation, and the adjusted coldest days averaged 2.54C cooler than in the original RAW observation periods of the current stations.
The average adjusted temperature of hottest days since the start year of the 34 currently operating ACORN station was 0.11C warmer than original RAW temperatures in those stations during their years of observation, and the adjusted coldest days averaged 0.38C cooler than in the original RAW observation periods of the current stations.
Among the hottest days recorded anywhere in Australia at a single station since the start year of the 34 currently operating ACORN station (different to averages of the monthly hottest/coldest day average at all stations), the average adjusted temperature of hottest days was 0.22C warmer than original RAW temperatures in current stations during their years of observation, and the adjusted coldest days averaged 1.41C cooler than in the original RAW observation periods of the current stations.
Year Book and BoM 24 hour rainfall extremes
The 1954 Year Book collates the most extreme 24 hour rainfall recorded in each Australian state and territory since records began up to 1951.
The table below extracts 74 weather stations with equivalent stations at the same locations which were operating in 2021, as well as their preceding stations in the BoM records, and compares their most extreme 24 hour rainfall recordings.
The comparison shows the most extreme 24 hour rainfall recordings at the 74 stations nationally averaged 413.0mm before 1951, compared to an average 300.4mm at the current stations up to 2021, with total rainfall on those extreme days at 30,558.6mm before 1951 and 22,232.7mm at the current stations.
The wettest ever day was 907.0mm in 1893 and the wettest day since recordings started at the 74 stations operating in 2021 was 800.9mm at Port Douglas in 1911.
The vast majority of Australia's most extreme rainfall days are not recorded because there isn't a rain gauge at the precise location of the downpour. The frequency and volume of known extreme rainfall days is dependent upon the number and density of rain gauges in Australia, and the maps below (extracted from Australian Daily Temperature and Rainfall Extremes, 2009) show how the number and distribution of gauges significantly increased from 1930 to 2000:
The table below shows decadal blocks of average and total volumes for the 10 most extreme 24 hour rainfall recordings each year in Australia since 1910 (1,135 wettest days analysed - see BoM tables).
This data shows that the average and total volumes of rainfall in the wettest days since 1910 have been in decline since peaking in the 1970s.
This is in agreement with the 1970s having the highest rainfall levels ever recorded in Australia. There might also be a connection to Australia's reduced frequency of severe and non-severe tropical cyclones since 1970/71 with their associated extreme and flooding downpours, as charted below (source BoM - Tropical cyclone climatology):
In 2000-2019, the average rainfall volume of the wettest days was -11.3% less than in 1980-1999, and the total rainfall volume was -7.8% less than in 1980-1999.
To determine the frequency of extreme 24 hour rainfall events, the average 413.0mm among wettest days before 1951 at the 74 Year Book stations can be used as a benchmark (554 days analysed), as calculated in the table below:
Again, the frequency of extreme rainfall days above 413.0mm peaked in the 1970s and has declined since.
BoM data (source) show that both the frequency and volume of extreme precipitation days in Australia has been decreasing:
The BoM's State of the Climate 2020 states:
"Heavy rainfall events are typically caused by weather systems such as thunderstorms, cyclones and east coast lows. Daily rainfall totals associated with thunderstorms have increased since 1979, particularly in northern Australia. This is due to an increase in the intensity of rainfall per storm, rather than an increase in the number of storms in general.
At the same time, the number of low-pressure systems that can bring heavy rainfall to heavily populated parts of southern Australia have declined in recent decades. This could have implications for recharging water storages and water resource management.
As the climate warms, heavy rainfall events are expected to continue to become more intense. A warmer atmosphere can hold more water vapour than a cooler atmosphere, and this relationship alone can increase moisture in the atmosphere by 7 per cent per degree of global warming. This can cause an increased likelihood of heavy rainfall events. Increased atmospheric moisture can also provide more energy for some processes that generate extreme rainfall events, which further increases the likelihood of heavy rainfall due to global warming."
The history books and BoM records to 2021 indicate the frequency, average and total volume of extreme 24 hour rainfall events are not indicative of climate change increasing the risk of extreme rainfall days prone to cause flooding.
Commonwealth Meteorologist Henry Hunt <1912
The Climate and Weather of Australia was published in 1913 by Commonwealth Meteorologist Henry Hunt, showing temperatures recorded at various Australian capital city and regional locations prior to 1912.
These suggest mean temperatures increased 0.6C from <1912 to 2000-2021 at 27 locations (19.2C > 19.8C).
This is the same as the 0.6C mean temperate increase in the CSIR <1913 dataset.
The 20 regional locations in the Hunt dataset saw a 0.55C mean temperature increase (19.50C > 20.05C) while the seven capital cities saw a 0.95C increase (18.25C > 19.20C).
This 0.4C greater warming in the major cities again suggests an urban heat influence.
The Hunt dataset also has rainfall data showing pre-1912 average annual rainfall at all 27 locations was 627.4mm, which compares to 617.6mm in 2000-2021 - a 1.6% reduction.
The 20 regional centres had an average 504.6mm before 1912 and 512.0mm in 2021 (a 1.5% increase) while the seven capital cities had an average 943.1mm before 1912 and 889.0mm in 2000-2021 (a 5.7% reduction), possibly contributing to the different increase in mean temperatures.
The table below presents summary findings from the Henry Hunt dataset comparison and includes annual monthly rainfall averages:
Yass-Canberra district <1908
On the Climate of the Yass-Canberra District was published in 1910 by Commonwealth Meteorologist Henry Hunt, showing temperatures recorded at 10 weather stations within that district prior to 1909.
These suggest mean temperatures increased 0.2C from <1909 to 2000-2021 at the 10 locations (14.0C > 14.2C).
Maximum temperatures in the Yass-Canberra district increased 0.7C from <1909 to 2000-2021 (20.7C > 21.4C) while minimum temperatures decreased 0.2C (7.2C > 7.0).
Ten weather stations is a small sample and 2000-2021 averages are sourced from the nearest stations available within the BoM's CDO database. Yass temperatures were only recorded from 2011 to 2018.
Nine of the 10 stations were using non-Stevenson screens before 1909, most likely Glaisher.
The one station in the Yass-Canberra district with a Stevenson Screen was Laverstock, which had six years of observations <1909 with mean temperatures increasing 0.4C by 2000-2021 (14.1C > 14.5C).
The table below presents summary findings from the Henry Hunt dataset for the Yass-Canberra district:
Australian climate data <1863
Some Data Concerning the Climate of Australia and New Zealand was published in 1863 by William Stanley Jevons and included a compilation of mean temperatures calculated by himself and government authorities prior to 1863.
These are among the earliest known temperature records in Australia and provide 12 locations that can be compared with modern equivalents.
The Jevon dataset suggests mean temperatures at those 12 locations increased 0.8C from <1863 to 2000-2021 (16.7C > 17.5C).
However, five of the 12 locations are capital cities and these had a 1.3C mean temperature increase from <1863 to 2000-2021 (15.8C > 17.1C).
The seven regional locations had a 0.5C mean temperature increase from <1863 to 2000-2021 (17.3C > 17.8C).
None of the 12 stations were using Stevenson screens before 1863 so there is likely some instrument influence, and the Jevon averages should be considered unreliable but the best available at that time of Australian colonisation.
The fact that capital cities have warmed at well over twice the rate of regional centres supports an assumption that urban heat is a significant contributor to measurements of climate warming.
The table below presents summary findings from the William Jevons dataset for 12 nationwide weather stations:
Western Australia climate data <1899
The Climate of Western Australia 1876-1899 was published in 1901 by Government Astronomer W. Ernest Cooke.
The Cooke dataset provides 20 WA weather stations that can be compared with the the 2000-2021 mean temperatures of their modern equivalent stations.
The 20 stations had a 0.8C mean temperature increase from <1899 to 2000-2021 (19.7C > 20.5C).
This is exactly the same as the 0.8 mean temperature increase among the 36 WA stations in the CSIR <1931 dataset.
In the Cooke dataset of 20 WA stations, the mean maximum increased 1.0C from <1899 to 2000-2021 (25.2C > 26.2C), while the mean minimum increased 0.5C (14.2C > 14.7C).
The 20 stations were using a combination of Stevenson, Glaisher and other screen types before 1899 so there is likely some instrument influence.
The table below presents summary findings from the Cooke dataset for 20 WA weather stations:
Cooke temperature extremes
The Climate of Western Australia 1876-1899 also presents hot and cold daily temperature extremes recorded in various time periods prior to 1899 at 16 weather stations that can be compared with their contemporary locations.
The table below compares hottest and coldest days in each month prior to 1899 with the hottest and coldest days recorded at the weather stations operating currently at or near these locations (from their first day of observation to 31 July 2021).
The table above shows the average temperature of the hottest days each month among all 16 stations prior to 1899 was 1.55C cooler than the average of the hottest days observed in each month at the current stations during their years of operation up to 2021.
Among the hottest days recorded at any of the 16 stations each month prior to 1899, their average was 0.7C hotter than at the current stations during their years of operation up to 2021.
The table above shows the average temperature of the coldest days each month among all 16 stations prior to 1899 was 0.45C warmer than the average of the coldest days observed in each month at the current stations during their years of operation up to 2021.
Among the coldest days recorded at any of the 16 stations each month prior to 1899, their average was 0.7C warmer than at the current stations during their years of operation up to 2021.
Among all 16 stations, the average temperature of extreme hot days in the warm months of October to March was 1.64C cooler prior to 1899 than in the stations open in 2021, and in the cool months of April-September they were 1.45 cooler prior to 1899 than in the stations open in 2021.
Among all 16 stations, the average temperature of extreme cold days in the warm months of October to March was 0.21C warmer prior to 1899 than in the stations open in 2021, and in the cool months of April-September they were 0.71C warmer prior to 1899 than in the stations open in 2021.
The pre-1899 durations of observation by the 16 stations were in some cases as short as just three years and instrument shelter differences are a factor, but nevertheless the data suggests WA residents in the late 1800s endured extremely hot days that were on average cooler than in recent decades, but the extreme days in any given month at different locations across the state were more likely to be hotter.
The data also suggests WA residents in recent decades have endured colder average and specific extreme days than in the late 1800s.
The average start year for the 225 CSIR locations was 1901 and approximately two thirds of recordings to 1931 were from thermometers in Stevenson screens. Click here for pop-up CSIR weather station start years.
The approximate third of recordings from Greenwich/Glaisher screens or open shade thermometers tended to record warmer maxima and cooler minima in different seasons.
Adelaide's West Terrace weather station used a combination of Greenwich/Glaisher and Stevenson screens from 1887. Nichols et al in Historical Thermometer Exposures in Australia summarised the Adelaide experiment results:
"The minima measured in the two exposures are similar throughout the year, with Stevenson screen minima being about 0.2C warmer. The maxima measured in the Stevenson screen are cooler than those in the Glaisher stand and the difference varies seasonally. The difference is only about 0.2C in winter, but reaches nearly 1C in summer."
In Exposure, instrumentation, and observing practice effects on land temperature measurements, ACORN architect Blair Trewin from the BoM states:
"In particular, a number of studies found that the Glaisher stand, in widespread use in English speaking countries prior to Stevenson screen introduction, had a warm bias in maximum temperatures, ranging from 0.2C to 0.6C in annual means and reaching up to 1.0C in mean summer maximum temperatures and 2–3C on some individual hot days. Minimum temperatures tended to have a cool bias of 0.2–0.3C all year. These results were based on, among others, a 60-year set of parallel observations at Adelaide, Australia. A warm bias in maximum temperatures, particularly on sunny days and/or in summer, was also common to numerous other pre-Stevenson exposures."
In Effects of Changing Exposure of Thermometers at Land Stations, D.E Parker from the Hadley Centre Meteorological Office in the UK states:
"The average difference for the four investigations quoted here was, however, only 0.01C, with the Glaisher value colder. The indications are thus for an insignificant annual mean difference between the exposures, but for a significant raising of daytime and summer-mean temperatures, and a significant lowering of night-time and possibly winter-mean temperatures, in the Glaisher stand relative to the Stevenson screen."
Comparisons show the monthly Glaisher screen influences on Stevenson maximum recordings in Adelaide were:
Dec -1.1C, Jan -1.0C, Feb -1.0C
Trewin reported that the Glaisher upward maxima bias averaged 0.4C and downward minima bias averaged -0.25C over a year, resulting in an average upward mean temperature bias of 0.075C which if applied to approximately one third of all temperatures in the CSIR dataset has no influence on total averages to 1931. This is supported by the Parker finding of a 0.01C difference between the Glaisher and Stevenson screens, with Glaisher cooler.
In Manual of Meteorology published in 1932, a comparison is made of Stevenson and Glaisher screens at Kew University from 1881 to 1915, with the Stevenson mean temperature recording an average 0.2C warmer:
In the <1931 CSIR dataset comparisons above, the average start years and number of stations open before or since the year 1900 should be considered when estimating any likely Glaisher or other screen influence:
Australia - 1901 (69 open before 1900, 157 since)
It is also worth noting that automatic weather stations have been installed since the mid-1990s and by 2021 were the majority instrument across Australia. AWS stations use electronic probes that take one second recordings of air temperature, a much more rapid response than occurred in manual liquid in glass thermometers before the 1990s, and there is evidence this may have caused an artificial warming since then. See Climate change or instrument change?
Urban Heat Island
Urban development replaces natural surfaces and vegetation with heat-absorbing concrete and bitumen materials such as roads, footpaths, roofs and buildings, resulting in what is known as an Urban Heat Island influence on natural temperatures.
UHI is likely to have influenced some Australian temperature recording locations before 1940, particularly in cities and large regional towns, but the sparsity of motorised transport, asphalt and concrete construction during this era suggests the influence would be minor compared to modern urban environments.
It is possible that the UHI influence on temperatures was already fully developed at various but not all locations by the first year of recordings in the CSIR and BoM RAW datasets.
In 2012 and 2013 it was anticipated that UHI warming in south-eastern Australia will continue to intensify by approximately 1C per decade over and above that caused by global warming (Voogt 2002), with tests in 1992 showing a UHI influence up to 7.2C between the Melbourne CBD and rural areas (source).
Smaller but significant UHI influences were found in regional towns, with a 1994 test observing a UHI intensity up to 5.4C between the centre of a Victorian town and its rural outskirts (source).
Compared to 225 stations in the CSIR dataset, the average long-term RAW mean temperature at the corresponding locations in 2000-2021 was 0.6C warmer, or 0.573C to be precise.
The jurisdictions with hottest to coolest mean temperature increase were Tasmania (0.9C), Western Australia (0.8C) Queensland (0.7C), Victoria (0.5C), NSW (0.4C), Northern Territory (0.4C), South Australia (0.2C).
Nationally at all 225 stations, average maxima and minima both increased by 0.6C from <1931 to 2000-2021.
NSW maxima increased by 0.3C and minima by 0.5C; Victoria maxima increased by 0.6C and minima by 0.3C; Tasmania maxima increased by 0.8C and minima by 0.8C; Northern Territory maxima increased by 0.2C and minima by 0.2C; Queensland maxima increased by 0.7C and minima by 0.9C; South Australia maxima increased by 0.9C and minima by 0.2C; Western Australia maxima increased by 1.1C and minima by 0.4C.
Based on known years of installation, about two thirds of temperatures in the CSIR dataset were recorded in Stevenson screens. Published estimates of Glaisher/Greenwich/open shade temperature influence suggest an insignificant bias below 0.1C on the overall averages, so the original records should not need adjusting on the basis of thermometer stands.
The CSIR <1931 mean temperature averages are validated by similar results within the <1912 dataset of Commonwealth Meteorologist Henry Hunt and the <1950 Year Book dataset.
The Henry Hunt dataset of <1909 for the Yass-Canberra district is small and possibly unreliable due to the majority use of Glaisher rather than Stevenson screens. However, the tabulated comparison with 2000-2021 is included above to provide a more comprehensive insight to reliable and unique recordings prior to the ACORN start year of 1910, and the mean temperature increase of just 0.2C over more than a hundred years reinforces the point that Australian temperatures are unlikely to have increased 1.4C over that time.
Year Book extreme temperatures
Within the analysis above of Year Book extreme hot and cold days prior to 1950, it should be noted that more than half the temperatures recorded in those decades were rounded to .0F (see Metrication influence on Australian temperature trends). This is also true of all daily Fahrenheit recordings prior to 1972 metrication.
These recordings might be observers rounding up or down to the closest whole degree or a possible tendency to truncate temperatures so that their observations seem correct (e.g. 78.7F rounded to 78F as it seems more accurate than 79F).
This rounding could mean that the averaged maximum and minimum observations before 1950 are cooler than if decimals had been included.
Although the BoM presents charts claiming an increase in the frequency of very hot (40C+) days in Australia (see here), this is based on percentile threshold levels calculated relative to 1961-90 using ACORN daily data which is adjusted and reduces the number of such days in the early 20th century.
Rather than using percentiles relative to 1961-90, the frequency of very hot 40C+ days can alternatively be calculated using only the temperatures recorded on each day.
See No more extreme hot days in Australia than 100 years ago which compares original RAW observations with adjusted ACORN data for the annual number of 40C+ days from 1910 to 2017 (final year to accommodate the ACORN 1 dataset) at all 112 ACORN stations used to calculate Australia's temperature trends, as charted below in RAW, ACORN 1, ACORN 2 and ACORN 2.1.
The chart below shows the average temperature of all 40C+ days at the 112 ACORN stations, with subtle differences between how each ACORN version has adjusted the temperature of very hot 40C+ days observed since 1910.
The charts above show an increase in the frequency and temperature of very hot 40C+ days within original RAW observations over the final 20 years (1998-2007 : 9.66 frequency, 41.43C temperature; 2008-2017 : 11.14 frequency; 41.54C temperature).
Most ACORN stations have been replaced since the mid 1990s with automatic weather stations which take one second air temperature recordings at a much faster rate than was physically possible in the earlier liquid-in-glass thermometers, increasing the likelihood that brief extreme temperatures of 40C or more can be be observed.
The charts above show the average number and temperature of very hot days each year that are dependent on how many ACORN stations have opened over time since 1910 (total 40C+ count divided by number of stations each year), but the frequency of 40C+ days is influenced by many of the stations that opened from 1910 to 1976 being in northern or inland locations where very hot days are more frequent than in southern or coastal areas.
For example, hot northern stations include Camooweal which opened in 1939, Rockhampton in 1939, Oodnadatta in 1940, Townsville in 1940, Kalumburu in 1941, Horn Island in 1950, Wittenoom in 1951, Birdsville in 1954, Giles in 1956, Thargomindah in 1957, Weipa in 1959, Barcaldine in 1962, Victoria River Downs in 1965, Rabbit Flat in 1969 and Learmonth in 1975.
From 1957 to 1976 the number of northern Australia stations increased by five from 30 to 35, while the number of southern stations increased by four from 73 to 77.
The average number of 40C+ days (unadjusted RAW) at northern stations increased from 14.96 in 1937-1956 to 15.66 in 1957-1976, while the average number of 40C+ days in southern stations reduced from 4.58 to 4.39. With adjusted ACORN 2.1 daily temperatures, the average number of 40C+ days at northern stations increased from 12.16 in 1937-1956 to 14.12 in 1957-1976, while the average number of 40C+ days in southern stations increased from 3.78 to 3.85.
The sum total of 40C+ days (unadjusted RAW) at northern stations increased by 31.8% (2,458) from 7,724 in 1937-1956 to 10,182 in 1957-1976, while the sum total of 40C+ days in southern stations increased by 19.3% (1,074) from 5,577 to 6,651. With adjusted ACORN 2.1 daily temperatures, the sum total of 40C+ days at northern stations increased by 46.3% (2,909) from 6,287 in 1937-1956 to 9,196 in 1957-1976, while the sum total of 40C+ days in southern stations increased by 4.7% (264) from 5,577 to 5,841.
A more accurate gauge of extreme hot day frequency in Australia can be calculated from the 60 long-term ACORN stations that were in operation in 1910 (with minor relocations such as post office to airport).
The chart below shows the average temperature of all 40C+ days at the 60 long-term ACORN stations.
Within the 60 long-term ACORN stations, the average annual RAW frequency of 40C+ days was 7.89 in 1910-1963 and 7.65 in 1964-2017, and the average annual ACORN 2.1 frequency of 40C+ days was 6.04 in 1910-1963 and 7.19 in 1964-2017.
The average annual RAW temperature of 40C+ days was 41.21C in 1910-1963 and 41.30C in 1964-2017, and the average annual ACORN 2.1 temperature of 40C+ days was 41.33C in 1910-1963 and 41.35C in 1964-2017.
ACORN adjustments have reduced the temperature increase of extreme hot days from 0.09C to 0.02C (and warmed 1964-2017 by 0.05C), but have reversed the original RAW frequency from an average decrease of 0.25 to an average increase of 1.15 from 1910-1963 to 1964-2017.
It is apparent that Australia has not suffered an increase in extreme 40C+ hot days since 1910 unless within the context of ACORN adjustments to observations made on every day.
ACORN v non-ACORN
Among the 225 weather stations sourced from the <1931 CSIR dataset, 73 are designated by the BoM as non-urban ACORN stations.
These 73 ACORN stations saw average unadjusted maxima increase 0.63C and average unadjusted minima increase 0.58C from <1931 to 2000-2021, representing a mean unadjusted temperature increase of 0.61C.
The remaining 146 non-ACORN stations saw average unadjusted maxima increase 0.56C and average unadjusted minima increase 0.50C from <1931 to 2000-2021, representing a mean unadjusted temperature increase of 0.53C.
This suggests the weather stations selected by the bureau for inclusion in ACORN have warmed almost 0.1C more than the average among other Australian stations in terms of mean unadjusted temperatures.
The 104 ACORN stations used by the BoM to calculate warming trends across Australia and national/state averages suggest 1.44C warming since 1910 following temperature adjustments.
Urban heat and smog
Apart from thermometer screen shelters, another common artificial element likely to influence temperatures over time is UHI from urban development. This is more likely to influence recent rather than historic temperatures and published tests from the early 1990s show an upward bias of several degrees between some central and outer city or town locations.
Other artificial temperatures not associated with CO2 include the absence or sparsity of motor vehicles prior to 1931, along with associated bitumen road infrastructure, as well as city/township smog based on coal/wood burning for energy and heating requirements.
Smog levels in Australian cities and major regional towns increased significantly after 1931 due to vehicle use, potentially cooling temperatures through albedo reflection of solar heat, but strict pollution controls have reduced vehicle emissions since the 1980s and this has substantially reduced smog levels with a consequent reduction in albedo cooling.
The 225 historic weather station temperature averages include all capital cities so their 2000-2021 estimates have a warming bias that doesn't as greatly influence ACORN national averages which exclude urban hotspots such as Sydney and Melbourne.
It should be noted that although the data calculated by William Jevons prior to 1863 is unreliable, comparison with the CSIR <1931 data suggests that in the two major cities of Sydney and Melbourne the averaged mean temperature increased 0.7C from <1863 to <1931 (15.3C > 16.0C).
It is also noteworthy that Alice Springs in the Northern Territory is the most influential station in Australia for the ACORN process of area averaging to determine neighbouring station temperature adjustments, and the CSIR dataset shows the Alice Springs mean temperature increased 0.5C from <1931 to 2000-2021 (20.9C > 21.4C), with maxima warming 0.8C and minima warming 0.2C. It's believed Alice Springs had a Stevenson screen from 1894, with a Glaisher used from 1879 to 1893 (see here).
The eight urban or urban-influenced ACORN stations excluded from BoM assessments of the warming trend across Australia and its calculation of national and State averages are Sydney, Melbourne, Adelaide, Hobart, Laverton, Richmond, Townsville and Rockhampton (see 14. How does the urban heat island effect impact the climate data).
The CSIR and Year Book analyses on this page include most of these eight urban hot spots, meaning their estimates of warming by 2000-2021 are more biased by this artificial influence.
UHI population comparison
The CSIR table above shows that from <1931 to 2000-2021, locations with a 2021 population above 100,000 had an average maximum temperature increase of 0.85C and an average minimum temperature increase of 0.83C, while locations with a 2021 population below 100,000 had an average maximum temperature increase of 0.58C and an average minimum temperature increase of 0.51C. This suggests the larger locations had maximum warming 0.27C greater and minimum warming 0.33C greater than smaller locations.
The Year Book table above shows that from <1950 to 2000-2021, locations with a 2021 population above 100,000 had an average maximum temperature increase of 0.81C and an average minimum temperature increase of 0.52C, while locations with a 2021 population below 100,000 had an average maximum temperature increase of 0.62C and an average minimum temperature increase of 0.09C. This suggests the larger locations had maximum warming 0.19C greater and minimum warming 0.43C greater than smaller locations.
The CSIR table above shows that from <1931 to 2000-2021, locations with a 2021 population above 50,000 had an average maximum temperature increase of 0.71C and an average minimum temperature increase of 0.74C, while locations with a 2021 population below 50,000 had an average maximum temperature increase of 0.58C and an average minimum temperature increase of 0.20C. This suggests the larger locations had maximum warming 0.13C greater and minimum warming 0.54C greater than smaller locations.
The Year Book table above shows that from <1950 to 2000-2021, locations with a 2021 population above 50,000 had an average maximum temperature increase of 0.63C and an average minimum temperature increase of 0.46C, while locations with a 2021 population below 50,000 had an average maximum temperature increase of 0.65C and an average minimum temperature increase of 0.07C. This suggests the larger locations had maximum warming 0.02C less and minimum warming 0.39C greater than smaller locations.
These population comparisons add weight to the evidence that the urban heat island effect has a significant influence on temperature trends.
The Year Book table above shows that from <1950 to 2000-2021, at the 18 locations locations with a 2021 population above 50,000 the annual average temperature of the hottest days recorded cooled -0.26C and the annual average temperature of the coldest days warmed 0.19C. At the 61 locations locations with a 2021 population below 50,000 the annual average temperature of the hottest days recorded warmed 0.02C and the annual average temperature of the coldest days warmed 0.33C.
The Year Book table above shows that from <1950 to 2000-2021, at the 18 locations locations with a 2021 population above 50,000 the hottest days recorded cooled -0.10C and the coldest days recorded cooled 0.45C. At the 61 locations locations with a 2021 population below 50,000 the hottest days recorded cooled -1.33C and the coldest days recorded cooled 0.15C.
Rainfall, airports and AWS
Rainfall patterns might also be influencing temperature trends. The <1912 rainfall dataset from Commonwealth Meteorologist Henry Hunt shows that at 25 locations across Australia between <1912 and 2000-2021, rainfall has increased from November to March and decreased from April to October.
Mean temperatures at the 25 locations between <1912 and 2020-2021 increased 0.57C from November to March and 0.67C from April to October, suggesting a possible 0.1C cloud cover influence that has mostly involved a rainfall increase in northern Australia and a decrease in southern Australia.
42.7% of the 225 weather stations have over the years been relocated to airport sites since 1931. This may have lessened the influence of UHI, dependent upon saturation levels at the original urban sites and acknowledged heat islands that also occurs in airport environments where many of the thermometer stands were relocated.
However and as contained in the table above, this analysis shows no evidence that airport weather stations have had greater warming than non-airport stations since <1931.
Another influence may be the advent of automatic weather stations in the 1990s which are said to exaggerate extreme temperatures in particular due to their second-by-second recording of air temperature that didn't occur in traditional mercury or liquid-in-glass thermometers.
However and as contained in the table above, this analysis shows no evidence that automatic weather stations have had greater warming than manual stations since <1931.
If thermometer stands, airport locations and automatic weather stations are discounted, the remaining influence likely to result in an 0.6C increase in mean temperature is either UHI or climate change, or a combination of the two.
Various international studies of temperature trends since the 1800s have found similar mean temperature increases as the 0.6C warming since <1931 in the CSIR dataset and 0.4C warming since <1950 in the Year Book dataset.
For example, the extract below from the 2021 paper How much has the Sun influenced Northern Hemisphere temperature trends? An ongoing debate. by Connolly et al suggests 0.55C per century warming since 1850 at northern hemisphere rural weather stations:
Most weather stations in the CSIR and Year Book datasets are rural but include cities and major regional centres with urban heat - unlike the northern hemisphere chart above and the national temperature estimates at the 104 non-urban stations within ACORN.
The CSIR, Year Book and other temperature datasets are unadjusted records compiled by Australia's leading scientists and weather experts since the 1800s, and are accurate but differ from BoM records that are adjusted within the 112 weather stations in ACORN.
Their dataset timescales include the first 85 years of temperature recording at most weather stations across Australia in a network more than twice as large as ACORN, and their averages are a legitimate historic record indicating climate warming has been significantly less than calculated with ACORN adjusted data since 1910.
Changes to instrument shelters and weather station relocations are likely to influence temperature observations over 180 years but averages from more than 225 locations should be sufficient to smooth artificial changes which might warm or cool the recorded maxima and minima at each individual station.
With 225 weather stations suggesting 0.6C warming from <1931 to 2000-2021 and 108 non-urban ACORN weather stations suggesting 1.4C warming from 1910 to 2021, it can be argued that the BoM's observation reference network is either an unrepresentative distribution of weather stations and/or ACORN adjustments have added about 0.8C warming to Australia's temperature trends.
Note: the January-July ACORN average mean temperature anomaly in 2021 was +0.44C and the January-July ACORN average mean temperature anomaly in 2000-2020 was +0.63C.
This website produced by Scribeworks 2008-2022