Monthly mean anomalies of homogenized maximum and minimum temperature from the Australian Climate Observations Reference Network – Surface Air Temperature (ACORN-SAT) dataset for the period 1910-2014 are used. Two sources of uncertainty are considered: “observation error” and “fitting error”. The “observation error” is an estimate of the confidence attached to each station time series. It encompasses measurement errors and the uncertainty due to homogenisation adjustments and spatial representativeness of the station. Estimates of the “observation error” are obtained by modelling the ACORN-SAT inter-station time series correlations as a function of distance and extrapolation to zero separation. The seasonal cycle and spatial patterns of the observation error will be shown.
The “fitting error” is estimated by the Mean Square Error of the mean covariance between ACORN-SAT station locations and the rest of Australia. This is obtained by Empirical Orthogonal Function decomposition of the covariance matrix estimated from the Australian Water Availability Project (AWAP) gridded dataset. The “fitting error” is then estimated by the fractional weight of the ACORN-SAT locations as they enter the network. With increasing stations, this “fitting error” decreases as the stations are increasingly able to characterize the true temperature covariance field. For maximum temperature, there are several step changes in this error, most notably with the introduction of Giles, in central Australia, in 1957. For minimum temperature, the changes are more gradual, although there appears to be a step change in 1965, coincident with introduction of Victoria River Downs in the Northern Territory.