The value of the cloudiness data was limited for modelling
because of small variation (ten of the fourteen nights were
completely or mostly overcast(Fig.3b)) and close co-variation of
cloudiness with temperature (bright nights were cold (Fig.3a,b)).
The effects of temperature and cloudiness were found to be largely
congruent,and since it was not possible to separate their effects we
limited the analysis to temperature,which improved its efficiency.