Effects of changes in algorithms used for the calculation of Australian mean temperature

There was a change in November 1994 in the algorithm used for the calculation of mean monthly temperatures reported internationally by Australian stations, from one based on daily maximum and minimum temperatures to one based on threehourly observations. A comparison of the pre and post-1994 methods, along with other methods based on fixed-hour observations which are in use in other countries, was carried out. Averaged over Australia as a whole, mean temperatures calculated using the post-1994 method were found to be 0.14ºC cooler than those calculated using the pre-1994 method, whilst other methods tested based on fixed-hour observations were found to produce means between 0.17ºC and 0.43ºC cooler than those derived from maximum and minimum temperatures. There were regional variations in the observed differences, with the largest differences occurring in coastal areas. The observed differences occurred in all seasons, being smallest in the (southern) spring and similar in all other seasons. The observed differences are substantial in the context of the trend in Australian temperatures over the last century, and reinforce the importance of documenting and taking into account such inhomogeneities in the record in making assessments of observed climate change.