This post shows some examples of small step changes in monthly Tmax data from the Melbourne area. These steps were detected and validated visually, no algorithms were involved. The results will contribute to the construction of a “non-urban” composite temperature record for Melbourne.
The following figure shows 12-month moving averages (with small offsets for display purposes) of raw monthly Tmax data from the BoM for the period 1987 to 1997. Stations shown (BoM ids in brackets) are: Melbourne Regional Office (86071), Moorabbin Airport (86077), Melbourne Airport (86282), Bundoora (86351), Geelong Norlane (87117) and Laverton RAAF (87031).
A small (around 0.4C annual average) step change can be seen in the data from Moorabbin Airport, most visible at the end of the change. At that date the Moorabbin data changed its position within the very local hierarchy of warmth, a change that persisted, indicative of a non-climatic change in the measurement system or environment.
To confirm the visual detection I plotted differences between the data using Melbourne Regional as a reference. If all the differences change in a similar fashion at the same time then that indicates a change in the relative warmth of the reference. If only a single difference changes relative to the others then that indicates a relative change in the station in question rather than in the reference. Forming differences takes out the local background climate variations, in general making it easier to see inhomogeneities, though in a more abstract way.
The following figure plots the temperature differences (manually offset for display purposes) of the data shown above:
The temperature differences have a very good level of consistency apart from the one involving Moorabbin, confirming the sudden shift in the data from that station. The consistent variations in the other lines reveal the changes in relative-to-neighbours warmth at the Regional Office.
I have estimated the monthly profile of the step change in 1992 at Moorabbin, after first of all restoring a few missing months of data, and correcting a few anomalous monthly values using data from the neighbours. Full details of the corrections applied at Moorabbin will be published here later.
The following figure shows the resulting set of temperature differences, confirming that I’ve got roughly the right annual average of the step change, and am not wildly wrong about the monthly profile. The monthly profile of the step is harder to establish/confirm, and will not be covered in this post.
There was a net increase in the relative-to-neighbours warmth at Regional Office over the period shown (1987 to 1997), but it is not clear to the author whether that was in step changes or gradual drift or both. The ACORN-SAT version of the Regional Office data says that there was a step change in 1990, which is certainly consistent with the data, though other hypotheses seem to be possible.
Finally I return to the less abstract world of 12-month moving averages, in which the following figure reveals an anomalous warming at Regional Office relative to a consistent and hopefully less Urban Heated background climate: