Roger Andrews: Sunshine and Temperature study following up Euan Mearns article

Guest post from Roger Andrews, who says: ” This is a review that extends Euan Mearns’ article on sunshine hours, cloud cover and SAT in the UK over mainland Europe and the North Atlantic. It reveals some interesting features that I make no attempt to explain – basically because I can’t – but someone else may have some ideas.” Apologies to Roger A for the delay in getting this article posted.

SUNSHINE, CLOUD COVER AND SURFACE AIR TEMPERATURES IN EUROPE
by Roger Andrews

The recent “UK temperatures since 1933” post discussed the relationships between sunshine hours, which were assumed to be an inverse cloud cover proxy, on surface air temperatures (hereafter SAT) at 23 UK stations. Here I summarize the relationships between sunshine hours, cloud cover and SAT over  Europe using observations from ~30 stations selected from the European Climate Assessment (ECA) data set (acknowledgement as requested to Klein Tank, A.M.G. and Coauthors, 2002. Daily dataset of 20th-century surface air temperature and precipitation series for the European Climate Assessment. Int. J. of Climatol., 22, 1441-1453.) Station locations are shown in Figure 1:

The results are preliminary but indicate that the relationships between sunshine hours, cloud cover and  SAT become more complex when areas larger than the UK are considered, as illustrated in the following figures:

Figure 2. Mean annual SAT is well correlated with mean annual sunshine hours in Europe after 1985 but not before (the data were normalized by dividing by two standard deviations):

Figure 3. Mean annual SAT is poorly correlated overall with mean annual cloud cover, with mean SAT increasing since 1950 but mean cloud cover showing no significant change:

Figure 4. Mean annual sunshine hours and mean annual cloud cover show short term peak-trough matches with ~2 year periodicity but are also poorly correlated overall:

Figure 5. Monthly sunshine hours means are generally well correlated with monthly SAT means at individual stations, but expressed to the nearest whole month sunshine hours lead SAT by two months at stations adjacent to the North Atlantic Ocean and by zero months in Southeastern Europe:

Figure 6. Monthly cloud cover means are at best moderately correlated with monthly SAT means at stations in mainland Europe with no leads/lags except in the Central Baltic. At stations adjacent to the North Atlantic cloud cover and SAT are effectively uncorrelated (“U”):

Figure 7. Monthly sunshine hour means are generally well correlated with monthly cloud cover means at stations in Central Europe but not well correlated at stations adjacent to the North Atlantic (there are no significant leads or lags):

Figure 8. The correlations between SAT and sunshine hours /cloud cover decrease with increasing elevation in the Alps, with the SAT/cloud cover correlation turning positive (i.e. more clouds, higher temperatures) above ~2,500m. However, the match with increasing duration of snow cover is equally good. (The ten stations shown are 1=Wien, 2=Basel, 3=Geneva, 4=Hohenpeissenberg, 5=Feldberg, 6=Davos, 7=Wendelstein, 8=Saentis, 9=Zugspitze and 10=Sonnblick):

Comments as to what these results may be telling us are solicited.

Some brief notes on data and data treatment:

The ECA data set is available at http://eca.knmi.nl/indicesextremes/customquerytimeseriesplots.php

It contains multiple variables for over 8,000 stations between Greenland and Kyrgyzstan and Svalbard and the Canary Islands, although the majority are in Germany. It stores monthly means for different variables in different files, so obtaining all the data for one station requires a number of separate downloads.

I downloaded monthly means of maximum temperature, minimum temperature, sunshine hours, cloud cover and snow depth (for the Alpine stations only). I calculated mean temperature by averaging Tmax and Tmin and also calculated Tmax-Tmin. This yielded six variables (sun, clouds, Tmax, Tmin, Tavg and Tmax-Tmin) that could be compared 15 different ways, and I limited the comparisons to sunshine hours and clouds versus Tavg partly for simplicity, partly because the (alleged) impacts of increasing CO2 are quantified in relation to average rather than maximum or minimum temperatures and partly because while Tmax tends to be most closely correlated with sunshine hours and clouds the correlation isn’t much stronger than the correlation with Tavg. The table below summarizes the means and standard deviations of correlation coefficients measured at 21 stations, with no allowance for leads or lags:

CORRELATION COEFFICIENTS, 21 EUROPEAN STATIONS
Variable	Versus Variable	Mean R Value	Standard Deviation
Sunshine Hours	Average Temperature	0.73	0.15
	Maximum Temperature	0.76	0.15
	Minimum Temperature	0.68	0.14
	Max-Min Temperature	0.66	0.39
Cloud Cover	Average Temperature	-0.29	0.41
	Maximum Temperature	-0.34	0.41
	Minimum Temperature	-0.24	0.40
	Max-Min Temperature	-0.59	0.23
Sunshine Hours	Cloud Cover	-0.62	0.26

The analyses are based on unadjusted monthly means except for the the annual means shown in Figures 2, 3 and 4, which are arithmetic averages of the monthly means. I discarded obviously suspect data, such as Liepaja  temperatures after 1999 and Bjornoya cloud cover. I downloaded data back to 1901 but there were too few stations to estimate reliable annual means before 1950.