Weather v BGR completions
Following on from my blog about BGR completions for 2019 (and previous years) I have had look at weather data for the last 10 years to see if one can infer any relationship between weather in the Lakes and BGR completion levels. [NB: I am quite prepared to accept that this analysis shares some pretty dodgy statistical analysis]
Firstly, I made an assumption (extrapolating from the monthly data of completions) that the most significant months of any year would be May to September. I then calculated the average rainfall (in mm) for those months for each year from 2010 to 2019, and plotted this against the numbers of completions for those years. The hypothesis being that wetter years would have lower numbers of completions.
In the resulting graph [above] the orange line is numbers of completions (not % but absolute numbers) and the blue line is rainfall (in mm, as per calculation above). For the hypothesis to be true the high rainfall years would have less completions. It can be seen from the right of the graph that for 2017, 2018 and 2019 this does not follow. 2017 and 2019 are high rainfall – high completions, and 2018 is conversely low rainfall – low completions. For 2010, 2011 and 2012 it varies from fitting the hypothesis for 2012 (high rainfall lower completions – but see 2013 onwards), and sort of fitting for the 2010 (low rainfall, moderately high completions), to no pattern for 2011. For 2013 to 2016 it seems that it is counter-intuitive, in that the completions goes down with low rainfall and up with higher rainfall.
Being not satisfied with this aspect I moved on to look at one year, month by month, to see if the was any inferable weather effect there.
Taking the monthly completions data and monthly actual rainfall data, I plotted the two against each other for the year of 2019.
In the resulting graph [above] the left axis and blue line is rainfall and the right axis and orange line is the number of completions for that month. [The months are numbered from1 to 12 along the bottom axis, representing Jan to Dec] The completions line is a normal curve, showing the normal distribution you would expect of completions data (ie more in summer months and less in winter, peaking this year in June). The rainfall shows three peaks in March, Aug/Sept and Dec. At a stretch you could argue that the March peak drags the completions data lower than it might have been – there is a small downward bulge in the curve for that month. Similarly the August peak also produces a dip in (expected) completions.
At the end of the day it has proved to be not a particularly meaningful analysis. I have made some pretty big assumptions, for instance that rainfall would have an immediate effect on completion numbers, and used a fairly small data set – so will accept any criticisms of my methodology. Looking back, it may have better to plot, say, June/July/August rainfall against the completions data as that might well have a better correlation, as that is when more people plan their rounds. Thoughts or counter suggestions welcomed, through the ‘Comments’ link.
If anyone wants to look further, the data is available. See: here for BGR completions data [this year is the first time by month data has been published), and here for the weather data [which is for the weather station at Newton Rigg, and includes max temp, min temp, and hours of sunshine for each month, going back to 1959].