Hi Roman – I’d appreciate a bit of background to your figure on UK temps and sunshine. Please can you identify the data sets you used and also the basis and verification stats of the (I assume?) trendlines with each month? Thanks. C

The data that I used came from the UK Met Office web site and was complete up to and including November, 2007 when I made that
graph. I was able to find the most recent versions of the data sets here. Be sure to click on “Switch to year order” first.

In particular, this page leads to these (updated to February, 2011) text versions of UK hours of sun and temperature.

My calculations were done using R. The graph requires the lattice library to be downloaded and installed on your computer. I redid this for the updated data and the graphs look pretty much the same. This time I used 1 to 12 for the months instead of Jan, Feb, etc.

This is the script.

library(lattice)

#Data is a data frame called ukx with the three variables Temp, Sun and Month (all starting in January, 1929)
attach(ukx)
xyplot(Temp ~ Sun | Month, ylab = “Temp(C)”, xlab = “Monthly hours of sun”,
main = “UK Temperatures vs. Hours of Sun (Jan., 1929 – Dec. 2007)”,
panel = function(x,y) {panel.xyplot(x,y); panel.lmline(x,y, col=”red”)})

#This calculates all of the regressions by month:
by(ukx, Month, function(x) summary(lm(Temp ~ Sun , data = x)))

It does look interesting as an example of non-linear relationship at the micro level. I would, however, point out that the cloud cover at different hours of the same day is going to be very strongly related (clouds don’t just appear and disappear at short notice). There is a further complicating factor in that the relationship between cloud cover and temperature is also affected by seasonal variation which depends on geographical location..

I was somewhat disappointed in that the graphs did not seem to tell me the units for cloud cover (although I did not read all of the different posts in the series.

Hi Roman – I’d appreciate a bit of background to your figure on UK temps and sunshine. Please can you identify the data sets you used and also the basis and verification stats of the (I assume?) trendlines with each month? Thanks. C

The data that I used came from the UK Met Office web site and was complete up to and including November, 2007 when I made that

graph. I was able to find the most recent versions of the data sets here. Be sure to click on “Switch to year order” first.

In particular, this page leads to these (updated to February, 2011) text versions of UK hours of sun and temperature.

My calculations were done using R. The graph requires the lattice library to be downloaded and installed on your computer. I redid this for the updated data and the graphs look pretty much the same. This time I used 1 to 12 for the months instead of Jan, Feb, etc.

This is the script.

library(lattice)

#Data is a data frame called ukx with the three variables Temp, Sun and Month (all starting in January, 1929)

attach(ukx)

xyplot(Temp ~ Sun | Month, ylab = “Temp(C)”, xlab = “Monthly hours of sun”,

main = “UK Temperatures vs. Hours of Sun (Jan., 1929 – Dec. 2007)”,

panel = function(x,y) {panel.xyplot(x,y); panel.lmline(x,y, col=”red”)})

#This calculates all of the regressions by month:

by(ukx, Month, function(x) summary(lm(Temp ~ Sun , data = x)))

Thanks – follow up appreciated! 🙂

Hi Roman

Following on from the above and tAV correlation discussions – have you seen these?:

http://gustofhotair.blogspot.com/2009/12/analysis-of-australian-temperature-part.html

Done a while ago but it looks as if Jonathan will be putting some more posts up soon. Might be interesting to keep an eye out.

C

It does look interesting as an example of non-linear relationship at the micro level. I would, however, point out that the cloud cover at different hours of the same day is going to be very strongly related (clouds don’t just appear and disappear at short notice). There is a further complicating factor in that the relationship between cloud cover and temperature is also affected by seasonal variation which depends on geographical location..

I was somewhat disappointed in that the graphs did not seem to tell me the units for cloud cover (although I did not read all of the different posts in the series.