Have automatic weather stations corrupted Australia's temperature records?Extreme temperatures spike with platinum resistance probesThe frequency and averages of extreme temperatures before and after automatic weather stations were introduced can be used to test Australian Bureau of Meteorology claims that AWS platinum resistance sensors are calibrated and their observations averaged to match the response time of liquid-in-glass thermometers. It is argued that these AWS platinum resistance probes introduced since the late 1980s measure air temperature every second and thus create an artificial warming bias. The argument is that the one second response time is much faster than that of mercury or alcohol liquid-in-glass thermometers used previously, and that brief extremes such as a hot or cold wind gust is recorded by probes but would not have registered in a thermometer. The bureau (BoM) asserts that AWS electronic sensors incorporate time constants designed to have similar response characteristics and are essentially identical to thermometers, with each one second measurement of the air temperature not being instantaneous but an average of the previous 40 to 80 seconds. The BoM explains (source) that with its common Almos AWS, all valid one second temperatures within a minute interval are assembled into an array and a range of generated one minute statistics includes:
This analysis will examine whether there is or is not a significant shift in extreme temperatures by using a methodology of correlating, or synchronising, the years before and after both original and replacement probe installation at all ACORN weather stations. If AWS probes influence how hot and cold temperature extremes are recorded, there should be a rapid and significant increase or decrease in their frequency. Maximum results | Minimum results Diurnal temperature range | Average temperatures Replacement probes | Conclusion 50th percentile - the warmest half Background In 2020, Burt and Podesta published Response times of meteorological air temperature sensors in which they reference a laboratory study by Benbow et al. (2018) showing response times for the three most commonly used BoM standard liquid-in-glass thermometers compared to 4 mm diameter platinum resistance sensors :
Within its 2020 publication An updated long-term homogenized daily temperature data set for Australia, the BoM tests one second variations within the minute at 6am and 3pm for replacement temperature probes at 98 ACORN stations.
Alice Springs is one of Australia's most influential weather stations in the ACORN calculation of national average temperatures because of the BoM's area averaging methodology that influences temperature homogenisation at neighbouring stations hundreds of kilometres away. Having originally been installed in 1991, the Alice Spring Airport AWS platinum resistance sensor was replaced with a Temp Control probe on 11 November 2011. Although there were 27 fewer days of rainfall in 2012-2013 than in 2009-2010, there were 66 more days at or above 30C, 65 more days at or above 35C, and 56 more days at or above 40C. Cloud cover influences solar exposure, which was overall the same from 2009-2010 to 2012-2013. There were more days above average in 2012-2013 than 2009-2010 but those days had a lower MJ/m2 intensity in 2012-2013. Comparing Alice Spring Airport from 2009-2010 to 2012-2013, average maxima at or above 30C warmed 1.0C, average maxima at or above 35C warmed 1.1C, and average maxima at or above 40C warmed 0.6C. The average maxima for all days throughout the years increased 1.9C. The average minimum for all days cooled 0.6C from 2009-2010 to 2012-2013. Decreased rainfall is an influence. However, Alice Springs Airport had 76.8mm of rainfall in 2009 with an average maximum of 30.0C, but 194.2mm in 2013 with an average maximum of 30.8C. There were 201 days above 30C in 2009 and 223 in 2013. The probe replacement influence on maxima can be illustrated with charts:
Results such as these at Alice Springs give cause to more thoroughly analyse the frequency of extreme temperature percentiles across all ACORN weather stations. Methodology Original and unadjusted daily temperature observations are downloaded from the Bureau of Meteorology's Climate Data Online web platform at 102 AWS locations since 1910 within the Australian Climate Observation Reference Network (ACORN-SAT). BoM data downloads suggest that 102 of the 112 weather stations in ACORN are AWS. Their averaged start year was 1998 but 50 were operational in 1996 when the BoM made AWS the primary instrument for use within the official temperatures at each station. Of the 102 AWS installations, 13 were from 1986 to 1991, 66 from 1992 to 1999, and 23 from 2000 to 2018. At each station, spreadsheet formulas calculate the 99th percentile (hottest 1%), 95th percentile (hottest 5%) and 90th percentile (hottest 10%) of days since their individual start years beginning 1910, as well as the 1st percentile (coldest 1%), 5th percentile (coldest 5%) and 10th percentile (coldest 10%). Percentiles are calculated using all daily observations throughout all years since the start of each station and inclusive of 2022, not per year, month or season. The great majority of hottest and coldest percentiles were observed during summer or winter, but this analysis is based on all daily observations since 1910. Calculating percentiles based upon all daily observations on record, regardless of month or season, is commensurate with their application to all days in each year. Alternative methods of percentile calculation will produce different results but similar trends, as would using the 3rd, 4th, 9th, 93rd or 97th percentiles instead of the 1st, 5th, 10th, 90th, 95th and 99th percentiles. All relevant percentiles for each ACORN station are itemised in the minimum and maximum temperature spreadsheets linked in the analysis below. This percentile temperature is then used by spreadsheet formulas to calculate the frequency count and average temperature within each of these three extreme percentile days for every year since 1910. These are the average temperature within each percentile, not the average temperature of such days, throughout each year. The collectively correlated 1st, 5th, 10th, 90th, 95th and 99th percentile frequencies and temperature averages are calculated by aligning/synchronising all 20 years before and after an AWS was first installed at each station. The correlation of 20 years before and after AWS installation in this analysis means their actual year of installation is irrelevant. For example, Bridgetown's AWS was installed in 1998 and Birdsville's AWS was installed in 2001. To calculate averages starting 20 years before installation, Bridgetown RAW temperature data in 1978 is aligned with Birdsville data in 1981, and so forth across all 102 AWS within the ACORN dataset. Despite various AWS being installed beforehand at different locations, the BoM only made them the primary instrument from November 1996 (an exception Cape Otway where the thermometer was removed when an AWS was installed in 1994). As a result, temperatures publicly available via Climate Date Online are manual observations from liquid-in-glass thermometers before November 1996 (Cape Otway 1994) and AWS temperatures thereafter at stations where such instruments have been installed. In this analysis, stations with AWS installed earlier are deemed to have usable information only since 1996, and temperatures from all such stations have a 20 year preceding start of 1976. The process is then repeated to calculate averages in the years before and after replacement platinum resistance probes were installed at relevant stations. These daily temperatures are analysed by spreadsheet formulas at each station and all extreme percentile temperature counts and temperatures are calculated to produce a correlated table of collective averages. Stations with increasing and decreasing extreme temperature frequency are determined through measurement of the 90th percentile for maxima and 10th percentile for minima in the five years before and after AWS original platinum resistance sensor installation, as this is the broadest percentile range. Results The analysis shows the 102 ACORN AWS stations had a rapid increase in the average annual frequency and temperature of 90th percentile (hot) and 95th percentile (very hot) days when their first platinum resistance probes were installed, with 99th percentile (extremely hot) days showing a frequency increase but little influence on average temperatures.
Among the 102 ACORN stations, 65 had an increase in 90th extreme percentile frequencies.
Among the 102 ACORN stations, 37 had a decrease in 90th extreme percentile frequencies.
Averaged among 90th, 95th and 99th percentiles comparing the five years before and after original AWS installation, there was a 43.0% frequency increase and 0.31C temperature increase among the 65 more frequent stations, and a 25.8% decrease and 0.31C temperature decrease among the 37 less frequent stations. The shift in average temperature among the 37 decreased frequency stations is biased by the fact there was a significant 0.62C temperature decline within the 99th percentile, although these annually averaged just 4.09 days in the five years prior to and 2.70 days in the five years after AWS installation. The station examples below show that increased extreme percentile frequency usually results in increased average temperature within that percentile, but not always.
Note : Annual maximum temperature percentile calculations for original probe ACORN weather stations can be downloaded as an Excel (xlsx 7.9mb) or an Apple Numbers (13.1mb). The results can also be viewed in a PDF (10.2mb). Minimum temperature percentiles The analysis shows the 102 ACORN AWS stations had a rapid but relatively minor increase in the average annual frequency and a decrease in temperature of 10th percentile (cold) and 5th percentile (very cold) days when their first platinum resistance probes were installed. Among 1st percentile (extremely cold) days, six stations had no occurrences or temperatures either before or after AWS installation so pre-post averages could not be compared at these stations. The remaining 96 stations showed a frequency increase but a decrease in average temperatures.
Among the 102 ACORN stations, 55 had an increase in 10th and 5th percentile frequencies. 53 stations could be compared within the 1st percentile as two had no occurrences or temperatures either before or after AWS installation, but these also had an increase in average frequency.
Among the 102 ACORN stations, 47 had a decrease in 10th and 5th percentile frequencies. 43 stations could be compared within the 1st percentile as four had no occurrences or temperatures either before or after AWS installation, but these also had a decrease in average frequency.
Averaged among 90th, 95th and 99th percentiles comparing the five years before and after original AWS installation, there was a 40.5% frequency increase and 0.13C temperature decrease among the more frequent stations, and a 19.0% decrease and 0.20C temperature increase among the less frequent stations. The shift in average minimum percentile temperatures among decreased frequency stations is biased by the fact there was a significant 0.60C increase within the 1st percentile, although these annually averaged just 3.80 in the five years prior to and 2.85 days in the five years after AWS installation. Note : Annual minimum temperature percentile calculations for original probe ACORN weather stations can be downloaded as an Excel (xlsx 7.8mb) or an Apple Numbers (13.3mb). The results can also be viewed in a PDF (10.1mb). Diurnal temperature range Diurnal temperature range (DTR) is the difference between average maxima and minima in any given month or year. High rainfall levels should reduce DTR due to warmer minima caused by cloud cover trapping overnight heat, and cooler maxima as a result of more daytime cloud cover. As illustrated in the chart below of the BoM’s official ACORN 2.3 DTR anomalies at all 112 ACORN automatic weather stations, this accepted meteorological principle can be seen in 1910-1995 rainfall anomalies being 15.883mm below and DTR being 0.104C above their 1961-1990 averages. However, rainfall increased significantly (+62.83mm) from 1986-1995 to 1997-2006 yet ACORN anomaly DTR also increased significantly (+0.191C) from 1986-1995 to 1997-2006, with chart data comparing before and after 1996 when many but not all of the 102 stations converted to AWS observations :
The same DTR shift can be seen in the chart below of homogenised absolute ACORN 2.3 annual temperatures at the 102 automatic weather stations :
A similar shift is apparent when comparing unadjusted RAW temperatures with rainfall anomalies at the 102 automatic weather stations :
This change in the DTR reaction to increased rainfall is due to minima decreasing and maxima increasing, a probable reflection of the significant increase in the frequency of extreme minima and maxima following 1996 when automatic weather stations became the primary instrument, as well as the gradual introduction of AWS instruments and observations in the following 20 years. The charts below show DTR when comparing 10th v 90th, 5th v 95th and 1st v 99th extreme temperature percentiles at the 102 ACORN automatic weather stations :
These results suggest the abrupt increased frequency of hot and cold extreme temperature observations following AWS introduction affected a broad range of extremes rather than just the coldest and hottest 1% of all recordings, and that the change also influenced overall average temperatures. Among all 102 automatic weather stations, 37 had an increase in the frequency of both hot 90th percentile and cold 10th percentile temperatures following AWS installation, and these stations experienced a rapid 0.51C increase in overall DTR when comparing the decades before and after 1996 :
Average temperatures The charts below illustrate and detail the average minima and maxima in the 10 years before and after 101 of the 102 locations were originally converted from manual to automatic weather stations with the installation of platinum resistance temperature probes.
Screen size It should be noted there are 91 ACORN AWS where the station has had its 230 litre Stevenson Screen replaced with a small 60 litre screen, which may influence how the instruments within them react to brief periods of extreme temperature. Among these 91 stations, 43 had an original AWS and small screen installation in the same year. These 43 had an average annual increase of 2.8 days per year in 90th percentile observation frequency comparing the five years before and after installation, with a 90th percentile temperature increase of 0.1C. The 48 other stations that had a small screen installed in a different year to AWS installation had an average annual increase of 4.2 days per year in 90th percentile observation frequency comparing the five years before and after installation, with a 90th percentile temperature increase of 0.2C. The 43 stations had an average annual increase of 1.9 days per year in 10th percentile observation frequency comparing the five years before and after installation, with a 10th percentile temperature decrease of 0.1C. The 48 other stations that had a small screen installed in a different year to AWS installation had an average annual increase of 3.5 days per year in 10th percentile observation frequency comparing the five years before and after installation, with a 10th percentile temperature decrease of 0.1C. The simultaneous installation of small screens when AWS were first introduced creates an unknown influence on overall averages of extreme percentile temperature changes, with a possibility that the screens were either wholly or partially responsible for increased frequency. Whether or not the abrupt frequency changes were caused by screen size or automatic weather stations themselves, their influence is artificial rather than natural or due to climate change. Note : Tables detailing small screen installation and influence at all 91 stations can be downloaded as an Excel (xlsx 25kb) or an Apple Numbers (324kb). The tables can also be viewed in a PDF (440kb). Discussion When the frequency of extreme maximum percentiles increases, average temperatures within those percentiles usually also increase, and when the frequency of extreme minimum percentiles increases there is usually a decreased average temperature within those percentiles. There are inconsistent results among many stations. For example, at many stations the frequency increases sharply in the 1st, 5th, 10th, 90th, 95th or 99th percentile range when an AWS is installed, as does the average temperature within those percentiles. However, at some stations the frequency increases but there is no change or a cooling of temperatures within the percentile. Alternatively, the frequency might drop while the temperature also drops, remains the same or increases. Similarly, a station might show a significant increase in 90th percentile frequency or temperature averages, but a decrease in 99th percentile frequency or temperature averages. This may be indicative of different platinum resistor brands being installed and/or each probe of whichever brand having different reaction characteristics in how frequency or temperature is logged. Also, the AWS installation might involve another environmental variable such as a slight change in location (e.g. to an airport) or the commensurate installation of a small Stevenson screen which alters exposure to hot or extremely hot air in different ways. Only through the collective averaging of all synchronised observations can the influence of AWS installation be established. At most stations there is a rapid shift, whether higher or lower, in frequency, average temperature or both within the different extreme percentiles when an AWS is installed, suggesting platinum resistors do not log these extremely hot and cold temperatures in the same way as preceding manual thermometers. A majority of ACORN stations (65 v 37) experienced an increase in the frequency of 90th, 95th and 99th percentile days. Conversion from manual to AWS observations caused a frequency and temperature increase at some stations and a decrease at others, but the AWS extreme temperature influence is biased toward heating rather than cooling. Maxima 90th percentile days five years before and after AWS 65 AWS stations : 30.6% frequency increase / 0.30C temperature increase 37 AWS stations : 17.9% frequency decrease / 0.25C temperature decrease 95th percentile days five years before and after AWS 65 AWS stations : 39.7% frequency increase / 0.22C temperature increase 37 AWS stations : 24.1% frequency decrease / 0.07C temperature decrease 99th percentile days five years before and after AWS 65 AWS stations : 58.6% frequency increase / 0.40C temperature increase 37 AWS stations : 35.4% frequency decrease / 0.62C temperature decrease 90th percentile days 10 years before and after AWS 65 AWS stations : 29.2% frequency increase / 0.20C temperature increase 37 AWS stations : 4.0% frequency decrease / 0.15C temperature decrease 95th percentile days 10 years before and after AWS 65 AWS stations : 41.3% frequency increase / 0.14C temperature increase 37 AWS stations : 6.9% frequency decrease / 0.02C temperature increase 99th percentile days 10 years before and after AWS 65 AWS stations : 58.4% frequency increase / 0.16C temperature increase 37 AWS stations : 10.1% frequency decrease / 0.33C temperature decrease Minima 10th percentile days five years before and after AWS 55 AWS stations : 28.4% frequency increase / 0.16C temperature decrease 47 AWS stations : 15.1% frequency decrease / 0.01C temperature increase 5th percentile days five years before and after AWS 55 AWS stations : 40.3% frequency increase / 0.11C temperature decrease 47 AWS stations : 16.8% frequency decrease / 0.02C temperature increase 1st percentile days five years before and after AWS 55 AWS stations : 52.9% frequency increase / 0.11C temperature decrease 43 AWS stations : 25.0% frequency decrease / 0.60C temperature increase 10th percentile days 10 years before and after AWS 55 AWS stations : 31.0% frequency increase / 0.20C temperature decrease 47 AWS stations : 6.1% frequency decrease / 0.06C temperature decrease 5th percentile days 10 years before and after AWS 55 AWS stations : 44.9% frequency increase / 0.14C temperature decrease 47 AWS stations : 6.3% frequency decrease / 0.05C temperature decrease 1st percentile days 10 years before and after AWS 55 AWS stations : 70.8% frequency increase / no temperature change 43 AWS stations : 7.6% frequency decrease / 0.31C temperature increase Following the initial spike in maximum extreme percentile frequencies and in many cases average temperatures of these percentiles at a majority of stations, there was a gradual or sporadic increase rather than plateau in the frequency and temperature of hot, very hot and extremely hot days in the 20 years following AWS installation. The spike and ensuing increases were far less pronounced within minimum temperatures. Increases following the spike among maxima may be due to locations, particularly in southern Australia, experiencing a rainfall decline since the year 2000 with a resultant increase in the likelihood of hot days. It is worth noting that among all 102 stations and with 1996 the year when automatic weather stations became the primary instrument for observations in the unadjusted Climate Data Online platform (Cape Otway 1994), the average year of AWS installation was 1998. After 1996, 52 stations were converted from manual to AWS with their averaged year of installation being 2000. After the year 2000, 17 stations were converted to AWS with their averaged year of installation being 2004. Four stations were converted from manual to AWS after 2005 with their averaged year of installation being 2010. AWS platinum resistance probe replacements Another possible reason for the gradual or sporadic increase in maxima is that among the 102 ACORN automatic weather stations, 57 have had their platinum resistance probes replaced at different times since the original installation of an AWS. Below are the annual average frequency and temperature trends of maxima 90th, 95th and 99th percentile days and minima 10th, 5th and 1st percentile days in the years preceding and following probe replacements at the 57 stations. Only six years are compared before and after probe replacement because this ensures all 57 stations are compared, with the most recent replacement being six years before and inclusive of 2022.
Maxima 90th percentile days six years before and after replacement probe 57 AWS : 20.8% frequency increase / 0.20C temperature increase 95th percentile days six years before and after replacement probe 57 AWS : 29.7% frequency increase / 0.15C temperature increase 99th percentile days six years before and after replacement probe 57 AWS : 50.6% frequency increase / 0.37C temperature increase Note : Annual maximum temperature percentile calculations for replacement probe ACORN weather stations can be downloaded as an Excel (xlsx 7.3mb) or an Apple Numbers (12.4mb). The results can also be viewed in a PDF (9.8mb). Minima 10th percentile days six years before and after replacement probe 57 AWS : 1.6% frequency increase / 0.07C temperature decrease 5th percentile days six years before and after replacement probe 57 AWS : 4.5% frequency increase / 0.06C temperature decrease 1st percentile days six years before and after replacement probe 57 AWS : no frequency change / 0.09C temperature decrease Note : Annual minimum temperature percentile calculations for replacement probe ACORN weather stations can be downloaded as an Excel (xlsx 7.4mb) or an Apple Numbers (12.3mb). The results can also be viewed in a PDF (9.9mb). Replacement probes at the 57 ACORN AWS stations had a significant influence on extreme maximum percentile frequency and average temperatures, particularly extremely hot 99th percentile days. However, the replacement platinum resistance probes had comparatively little influence on the frequency of extreme minima percentiles and mostly resulted in a slight temperature decrease among these cold, very cold and extremely cold percentiles. The results suggest that both original and replacement platinum resistance probes tended to increase the frequency and average temperature of maximum extreme percentiles more than their minimum extreme percentiles. In this context it is noteworthy and possibly relevant to consider ACORN 2.3 maximum and minimum temperature anomalies since November 1996 when automatic weather stations became the BoM primary instrument, and AWS rather than manual thermometer observations were used for official average national temperature calculations. Comparing the first and second halves of this time period, the ACORN 2.3 maximum temperature anomaly has increased 0.345C (November 1996 to November 2009 - 0.616C v December 2009 to December 2022 - 0.961C) while the minimum temperature anomaly has increased 0.189C (November 1996 to November 2009 - 0.488C v December 2009 to December 2022 - 0.677C). With maxima, the original AWS installation at all 102 ACORN stations in the first five years resulted in an annual average 3.65 additional 90th percentile days, 2.10 more 95th percentile days and 0.50 more 99th percentile days at each station. At the 57 replacement probe stations, this instrument upgrade resulted in an annual average 8.39 additional 90th percentile days, an additional 5.96 95th percentile days and an additional 2.01 99th percentile days during the ensuing six years. With minima, the original AWS installation at all 102 ACORN stations in the first five years resulted in an annual average 2.16 additional 10th percentile days, 1.76 more 5th percentile days and 0.35 more 1st percentile days at each station. At the 57 replacement probe stations, the instrument upgrade resulted in an annual average 0.54 additional 10th percentile days, an additional 0.76 5th percentile days and no change to 1st percentile days during the ensuing six years. These averages do not reflect the extent of the extreme percentile frequency increase at many stations. For example, at the weather station in Kalumburu during the five years following AWS installation, there were an average 25.4 more 90th percentile maximum days annually than in the five years before. In its 62 years of operation from 1942 to 2004, Kalumburu Mission 1021 recorded 221 days at or above 40.0C, or an average 3.56 such days per year. An AWS was installed in 1998 and in its 23 years of operation from 1999 to 2022, Kalumburu 1019 recorded 168 days at or above 40.0C, or an average 7.30 such days per year. The hottest day ever recorded at Kalumburu since 1942 was 43.4C on 15 October 2018, but the validity of this record might be questioned in light of the extreme percentile frequency increase after AWS installation. At the nearest weather station of Truscott, approximately 35 kilometres distance, the maximum temperature on 15 October 2018 was 36.6C. At the manual Doongan weather stations about 125 kilometres away the maximum on 15 October 2018 was 39.4C. In relation to the upward shift in extreme maximum percentiles in the years following original AWS installation visible in charts above, the average time between original and replacement probe installation was 11.6 years. The charts suggest a 90th, 95th and 99th percentile frequency and temperature spike about 15 years after original AWS installation, but these are averages and the duration until replacement probes varied among the 57 affected stations between two and 19 years (see spreadsheet downloads in addendum). Among the 57, 30 had a duration between 13 and 19 years before their original probe was replaced. This analysis suggests the maximum extreme percentile frequency/temperature spike some 15 years after original AWS installation may be due to replacement probes that exaggerated extremes more than their predecessors (with about a dozen changing from the Rosemount to Temp Control brand of probes). What about the 50th percentile, the warmest half of all days? It is worth averaging the warmest 50th percentile of all days at ACORN stations to test if automatic weather stations and their replacement probes had a broader influence on frequency and temperatures and not just on the hottest 10%, 5% and 1% of observations.
The 0.17C increase in 50th percentile maximum temperatures in the five years following AWS installation is greater than the 0.13C increase within the 90th percentile and suggests the instruments influenced more than just extreme temperatures. The chart below shows the 57 stations that had an increased frequency of 50th percentile observations in the five years following AWS installation.
The chart below shows the 44 stations that had a decreased frequency of 50th percentile observations in the five years following AWS installation.
As within the hottest 10% of extreme temperature observations, there is an upward plateau shift in 50th percentile frequency and temperatures in the years following original AWS installation. The average time between original and replacement probe installation was 11.6 years, and the chart below shows trends combining stations that have never had a probe replacement with the 57 stations that have had a probe replacement.
The chart below shows trends among only the 57 stations that have had a probe replacement.
Note : Annual 50th percentile maximum temperature percentile calculations for original probe ACORN weather stations can be downloaded as an Excel (xlsx 1.2mb) or an Apple Numbers (6.5mb). The results can also be viewed in a PDF (3.6mb). Annual 50th percentile maximum temperature percentile calculations for replacement probe ACORN weather stations can be downloaded as an Excel (xlsx 1.1mb) or an Apple Numbers (6.2mb). The results can also be viewed in a PDF (3.5mb). Tables summarising all 50th percentile frequency and temperature changes at all 102 ACORN stations, as well as years of original/replacement probe installation and probe type, can be downloaded as an Excel (xlsx 18kb) or an Apple Numbers (321kb). The tables can also be viewed in a PDF (428kb). Conclusion The sharp frequency increase in extreme percentile daily temperatures following the installation of AWS original and replacement platinum resistance probes at 102 ACORN weather stations raises questions about the accuracy of Australian temperature averages since 1996. It is unknown how this affects the BoM calculation of overall temperature averages but artificial warming is likely as an increased frequency of the hottest 10% of all daily observations should influence monthly and/or annual averages. For example, at Port Perpendicular the 90th percentile frequency had an annual increase of 5.2 days in the five years after AWS installation in 2001, and the average temperature of these days increased by 0.3C. The average 2021/2022 summer temperature at the station was 23.8C. If the five hottest days of Port Perpendicular's 2021/2022 summer are cooled by 0.3C, the average was 23.7C. In averaging terms based on the five years pre and post AWS, the 30.6% frequency and 0.30 temperature increases within the 90th percentile maximum observations at 65 AWS are partly offset by the 17.9% frequency and 0.25C temperature decreases at 37 AWS, as well as the 28.4% frequency increase and 0.16C temperature decrease within the 10th percentile minimum observations at 55 AWS. However, averaged across all 102 stations the 10.8% frequency increase and 0.13C temperature increase in maximum 90th percentile observations is greater than the 6.5% frequency increase and 0.08C temperature decrease in minimum 10th percentile observations. The frequency and temperature increases/decreases vary significantly within the 1st, 5th, 10th, 90th, 95th and 99th percentile at different stations, possibly related to different AWS brands and local climate conditions affecting observations during changing timespans. This analysis also raises questions about the validity of record extreme maximum and minimum daily observations since 1996 wherein the artificial influence of automatic weather stations may have caused the record rather than actual heat or cold. The research methodology of synchronising the years before and after AWS installations suggests there was a significant and rapid shift in daily extreme temperature observations caused by their introduction. Among 90th percentile maximum temperatures, or the hottest 10% of days since stations opened, the original AWS installation at all 102 ACORN stations saw an annual average 3.65 additional such days within five years. At the 57 stations that had their probes replaced, there were an annual average 8.39 additional such days within six years. This suggests that on average a majority of ACORN AWS stations had more than 11 extra 90th percentile days annually after their probes were either first installed or replaced. Among 10th percentile minimum temperatures, or the coldest 10% of days since stations opened, the original AWS installation at all 102 ACORN stations saw an annual average 2.16 additional such days within five years. At the 57 stations that had their probes replaced, there were an annual average 0.54 additional such days within six years. This suggests that on average a majority of ACORN AWS stations had less than one extra 10th percentile days annually after their probes were either first installed or replaced. It is difficult to determine whether extreme percentile average temperatures are affected by either the increased/decreased frequency of such days or by how maximum/minimum temperatures are actually logged and the one second readings of original and/or replacement platinum resistance probes are averaged. However, this analysis suggests a confirmation of the claim that AWS one second observations affect the frequency and temperature of extremes when compared to liquid-in-glass thermometers. Further, the analysis of 50th percentile observations suggests that automatic weather stations had a broader influence on overall averages, with replacement probes potentially causing an increase of about 0.3C within the warmest half of all daily observations. Addendum It should be noted that both manual thermometer and automatic weather station observations are subject to errors caused by issues such as parallax liquid on glass and human tendencies to round their observations up or down, as well as one second spot readings or inaccurate duration averaging of one second readings taken by platinum resistors to correlate with manual thermometers. It is also worth noting that there are differing numbers of missing observation days at a majority of weather stations, creating an unknown margin of calculation error for all percentile frequency and average temperature estimates within this analysis. In 1991-1995, the five years before the most common year of original AWS installation in 1996, collectively among all 102 AWS stations there was an average annual total of 883 missing maximum observation days. In the five following years (1997-2001) the annual collective average was 1,438 missing days, or 62.9% more than the preceding five years. In 1991-1995 there was an average annual total of 1,136 missing minimum observation days. In the five following years (1997-2001) the annual collective average was 1,520 missing days, or 33.8% more than the preceding five years. These missing days include an unknown proportion of extreme percentile hot and cold days that cannot be counted, and the significantly greater number of post 1996 missing days suggests the calculations in this analysis are most likely a slight underestimate of their rapidly increased frequency following AWS installation. Note : Tables summarising all hot and cold percentile frequency and temperature changes at all 102 ACORN stations, as well as years of original/replacement probe installation and probe type, can be downloaded as an Excel (xlsx 53kb) or an Apple Numbers (513kb). The tables can also be viewed in a PDF (1.8mb).
This website produced by Scribeworks 2008-2023 |
Back To Top
