|"Temp record is unreliable"||The warming is the same in rural and urban areas, measured by thermometers and satellites.|
The Bureau presented the Forum with analyses of the raw and adjusted data that suggested a slight difference in the estimated annual mean temperature anomaly compared to the base period 1961–1990. This difference resulted from descriptive quadratic functions fitted to the raw and adjusted data, respectively. The Forum identified several issues with this approach:
- While it is acknowledged that the quadratic function is a valid fit to the available data, it is inconsistent with the language used by the Bureau to describe the temperature pattern as being essentially flat until a particular time, after which it rises progressively. The language used is consistent with a piecewise linear fit rather than a quadratic, and each of the two forms require only a small number of parameters to be estimated (and hence are similarly parsimonious as descriptions of the data).
- A quadratic fit to these data is purely descriptive and not useful for forecasts of future annual mean temperature anomalies. Further, the extent to which the fit is affected by differential variability between distant past data (e.g. 1910–1930) and present data (e.g. 1990–2014) may be material to the parameter estimates, and hence to the precision of estimates based on the fit (for both raw and adjusted cases).
- The use of a quadratic functional form carries the risk, if (mis)used to project temperatures beyond or before the observation period, of seriously overestimating the change. The Forum strongly recommends that the use of piecewise linear fits or nonparametric smoothers such as LOWESS be revisited by the Bureau. However, the Forum notes that all such fits are best considered only as descriptive curves fit to the available data and should not be interpreted as implying any particular underlying physical model for temperature behaviour beyond the period for which data are available. In particular, the Forum cautions against the explicit use of such fitted curves to generate specific forecasts or predictions of future temperature behaviour. NOTE: My bolds and underlines throughout.
Quadratic and other polynomial equations always trend to infinity. That is what they do. Quadratic equations graph parabolas. They should not be used to graph cyclic phenomena such as weather or climate, because whilst they can be made to fit a short section of data of say 100 years, any extrapolation outside this will quickly lead to absurd results.