Introduction

The BBC’s reporting on the recent high temperatures appears to have provoked something of an online backlash.

https://www.bbc.co.uk/news/uk-62323048

This divisiveness is very unfortunate. It seems to be the result of an unhelpful polarisation of public views on the issue of climate change. However, the manner in which the BBC and the met office reported the temperatures did not help. The problem stems in part from an over reliance on single data points. The impression was given that on the 19 July 2022 all parts of the UK experienced record breaking temperatures simultaneously. In reality only a few meteorological stations broke records. The BBC was upset by comparisons with the summer of 1976. However the comparison was justified. In 1976 a very warm and dry spell was prolonged throughout most parts of the country. The peak temperatures tended to be lower, but the heat lasted much longer.

The GHCND data set from NOAA does not yet include the last two years for the long running station at Oxford. Most station records have not been updated to 2022. However there are two records from Heathrow. One of these series contains the new record high. The Heathrow data on GHCND unfortunately has many missing values for 2020 and 2021. See the dygraphs below to visualise the Oxford data and the Heathrow data. You can click on them to open up the whole available time window.

Oxford

Heathrow

This consists of two time series, with some data gaps.

Was the high temperature unprecedented?

The BBC and Met office repeated use of the word “unprecedented” to describe the record high was particularly unfortunate and may have led to the online abuse they complained about. In many parts of the UK 19 July was not the hottest day on record. The Oxford records go back to 1815. Double clicking on the Oxford dygraph will show that single days with peak temperatures very similar to 19 July 2022 have occurred often throughout the long time series. In many ways a single hot day does not make a true heatwave. It would be preferable to reserve the term heat wave for a period when high summer temperatures persist for several days or weeks in a row. Leaving aside the fact that Heathrow airport is clearly an urban heat island, there is another problem with reporting “record highs”. A record can only be set within a recorded time series. Fewer stations were recording data in the past. There is clearly a great deal of intrinsic year on year variability with respect to the hottest day. If the were no change in mean annual temperatures the record hottest day of the year would be distributed at random along the series. If, for example, the period between 1990 and 2022 was used as a time window there would be 32 years in which the record could fall. If a search were made within data from a large number of stations it would be quite likely, just by chance, to find one in which the record was set in 2022, while others would have set their records in previous years.

Nevertheless aggregated global time series do show a trend in rising mean annual temperature from 1990 to 2022. This should make record temperatures rather more likely. The hottest days should tend to occur in more recent years.

Rank correlation for Oxford

The dramatic style of reporting adopted by the BBC produces an impression that new record temperatures will not just become more likely but will become inevitable. It may even be tempting to think that next year is almost certainly going to record higher temperatures than this year. In reality the upward trend is likely to be quite small with considerable noise (scatter) around it. One way of looking at the data is to use a simple rank correlation. The longer the time period the more likely it is that the correlation will be statistically significant. So if the years since 1990 are used for Oxford there is an upward trend in ranks, but this is not significant. If years since 1970 are used the trend is significant.

Oxford since 1990

## 
##  Spearman's rank correlation rho
## 
## data:  yrly$max and yrly$year
## S = 3489.6, p-value = 0.2347
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##       rho 
## 0.2236813

Oxford since 1970

## 
##  Spearman's rank correlation rho
## 
## data:  yrly$max and yrly$year
## S = 12808, p-value = 0.005773
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
##       rho 
## 0.3849461

Interpretation of the rank correlation

The shorter time series starting in 1990 does not produce a statistically significant p-value, under the traditional cut off point of p<0.05. The longer time series does show a significant relationship. In both cases there is a slight upward trend and the correlation coefficient (rho) is positive. However this does not imply that next year’s hottest day is highly likely to be the hottest on record at Oxford, even if the data from 2022 was available and added to the series.

How widespread are positive correlations?

There are 65 stations that have sufficiently complete daily records from 1970 to 2020 to allow a rank correlation analysis to be run. The same procedure was carried out for each of them and the results saved. Unfotunately the Heathrow data had too many gaps to be included. Cutting the correlation coefficient into broad classes allows the results to be visualised as a map.The classes are labelled as negative correlation, 0 to 0.2 = very weak,0.2 to 0.4 = weak, o.4 to 0.5 weak to moderate.

Its noticeable that almost all the stations showing a weak to moderate correlation are clustered in the South East of England. So, even if there is a slight trend in the UK towards “record breaking” high temperatures it is not evenly distributed. The west of the UK does not follow the trend and some stations even show weak negative correlations.

Map

Table

Conclusion

The selective use of single daily records from individual meteorological stations is misleading. Some of the comments directed towards the BBC and the Met office may have been abusive. Others may have little grounding in data. However in many respects the over hyped reporting was to blame for this. Understanding climate and the way the climate may vary and change requires taking a wide temporal and spatial perspective.