Method

Data were obtained from the GHCND for the longest continuous time series in the UK at Oxford. A GAM was fitted taking year day as input in order to represent the seasonal trend. The data for 2020 were removed as some anomalous records with very low temperatures were present, perhaps due to quality control issues.

The dygraphs are very dense. To use them select small time windows of interest using the mouse. The anomalies are the values minus the trend. The data were then split into months and a fitted regression added per month to show trends in the anomalies for each month over the time series.

Results

Maximum temperatures and seasonal trend

Anomalies

Regression coefficients for each month

A linear regression of the anomalies for each month is fit using year as input. The coefficients are multiplied by 100 to represent the change over a century. Upper and lower 95% confidence intervals are supplied. This provides a formalisation of the trends shown on the previous tab.