Use of rainfall data in checking temperature measurements

In this post I show how monthly rainfall data can be used to give additional confidence in monthly Tmax values that may otherwise be regarded as being incorrect or inconsistent with the regional climate. Analysis of rainfall data will play a major role in reconstruction of 19th century climates, partly because rainfall itself is a major part of climate and partly because the relative sparseness of early temperature data forces one to look further afield for confirmation. Differences in rainfall patterns may cause temperature homogenisation errors if such differences are not taken into account.

The figures below show raw data from Melbourne Regional Office (BoM id 86071). The red lines are time series of monthly Tmax anomalies (12 samples per year, January of year 1900 is placed at 1900.0); purely for display purposes the data are smoothed over 3 consecutive months. The blue curve (always above the blue zero line) shows monthly rainfall values in mm/100; multiply the values shown by 10 to get the monthly rainfall in centimeters.

1890 – 1900

In this period the most extreme Tmax values were associated with extremes of rainfall, low rainfall went with high Tmax, high rainfall with low Tmax.

1900 – 1910

Again there was an inverse relationship between Tmax and rainfall, especially in summer.

Breaking the blog prohibition on theoretical speculation by the author I’m going to suggest that cloudiness was the dominant driver of both Tmax and rainfall in summer in this part of Australia. However, theory will play no part  in climate reconstructions presented here, but rainfall plots will be used to guide quality-control and step-change adjustments of temperatures. If rainfall data is ignored some temperature extremes (such as the ones shown above) might be falsely corrected, or a persistent drought or rainy spell might trigger an invalid homogeneity adjustment.