Torok and Nicholls (1996) produced a high-quality dataset of annual mean maximum and minimum temperature series for Australia. The primary purpose of this dataset was to enable the reliable monitoring of long-term climate trends. Consequently, each station record was corrected for discontinuities caused by changes such as site location and instrumentation. In all, temperature records for 224 observation locations were homogenised, 170 of which were classified as non-urban and believed to be free from the influence of urban warming. Previous to this investigation, the dataset has been updated using data that has undergone basic quality assurance checks. If in the post-1993 period a station in the dataset had closed, its entire record was removed from the dataset to avoid the known biases that occur in compositing another station record. Since 1993 the number of non-urban stations in the dataset has dwindled from 170 to 127 stations due to site moves or closure. This update has re-established approximately 160 non-urban records to the dataset.

This poster presents the techniques used to update the homogeneity of the dataset from 1993 to 2001. The project has replicated the majority of techniques employed by Torok and Nicholls (1996), however a number of new methodologies have been used to help establish the presence of urban warming signals and to determine the magnitude of adjustments using overlap data between sites. An important component in the process of identifying potential discontinuities is the use of station history metadata. Since the development of the initial dataset there has been significant changes in the way station history information is recorded by the Australian Bureau of Meteorology. Traditionally station metadata has been recorded on paper history files. With the development of SitesDB, an electronic metadata database, the process of identifying potential discontinuities has been streamlined, however at present the information in the database is only reliable since the mid to late 1990s. A concerted effort is now underway within the Bureau of Meteorology to populate SitesDB with historical metadata.

In addition to the dataset update details, a method of determining the total measurement uncertainty in the dataset will be given. Generally, the uncertainty in an annual average station record is the combination of measurement error and the error in making adjustments to the annual value in order to homogenise the record. A comparison of uncertainty sources shows inhomogeneity adjustments contribute around 80% of the total uncertainty, which is approximately ±0.5°C in the annual average temperature of individual station records. The treatment of uncertainty is extended to include estimates of uncertainty in temperature areal averages. Preliminary figures show that the uncertainty in Australian annual temperature averages is of the order of ±0.1°C. This figure is largely independent of the uncertainty calculated for individual stations. Hence a method of combining the station uncertainties and spatial average uncertainties is sought. The answer may be found in the application of optimal average techniques.