**7. Year to year variation of seasonal and monthly totals and trends**

At the study site the summer half-year of 1955-2010 contributed 80 % of the annual global irradiance, about 87 % of direct irradiance and about 75 % of sunshine duration on average. In the dark winter period the amplitude of variations was larger before year 1977, including the extreme lowest value recorded in 1961 and the extreme highest one in 1963.

Fig. 10. Typical histogram of monthly distribution of relative daily sum of global irradiance G/Gclear in bright half-year

Fig. 11. Time evolution of the ratio Mean/Trimean of G/Gclear in July and in November 1955-2010

In the dark winter period the amplitude of variations was larger before year 1977, including

0.0 0.2 0.4 0.6 0.8 1.0

Fig. 10. Typical histogram of monthly distribution of relative daily sum of global irradiance

*G/Gclear*

July November

1950 1960 1970 1980 1990 2000 2010 Year

Fig. 11. Time evolution of the ratio Mean/Trimean of G/Gclear in July and in November

the extreme lowest value recorded in 1961 and the extreme highest one in 1963.

0.00

G/Gclear in bright half-year

0.6

1955-2010

0.8

1.0

Mean/Trimean of *G/Gclear*

1.2

1.4

1.6

0.05

0.10

0.15

0.20

Frequency

0.25

0.30

0.35

0.40

Often linear trends are used for characterizing the tendencies of variation in time. We have no reason to expect some overall trend of increase or decrease in G/Gclear during the whole observed time interval of 1955-2010 in any month. However, there were shorter intervals manifesting quite linear tendency of brightening or dimming. Atmosphere is essentially a non-linear system exhibiting quasiperiodic and nonsymmertic variability in different timescales. Ascending and descending branches of those cycles could be successfully approximated by linear trends.


Table 4. Slopes of linear trends of G/Gclear by the conventional mean, trimean and median during time intervals allowing linear approximation

Usually the slopes of calculated linear trend in G/Gclear, both positive and negative, happened to be larger for the trimean and median than for the conventional mean. A sample of estimated linear trends in all the three general trends measures for time intervals, manifesting the trends most evidently is presented in Table 4. The time evolutions of the ratio of Mean/Trimean in summer and winter months are different. In Fig. 11 the time evolutions of the ratio for relatively sunny month, July and cloudy month, November are presented. In November to March the dimming trends prevail at the site in all months and negative slopes for trimean and median are larger than those for the conventional mean. In January there appeared an overall positive tendency in the ratio Mean/Trimean in 1967- 2008. Relatively large positive peaks of the ratio have been encountered in separate years with an increasing frequency after 1988. As was noticed above, those large values of ratio appear when G/Gclear values were small, presenting sharp and strong positive skewed distribution. In February they were about two times smaller than those in January, manifesting also overall negative trend. During the last two decades, on the average, the ratios Mean/Trimean were larger, and the values of all three central trend measures of G/Gclear were smaller than before.

Fig. 12. Trend of dimming for January in 1963-2008

Fig. 13. Switching from brightening to dimming trend in October

In March, a dimming trend was obvious in 1955-1995 and in April of 1955 – 1977 the trend also appeared, but then has changed to the brightening in 1977-1995. In May to August small positive trends in G/Gclear were observed. In June the positive trend in 1955-1973 has changed to negative in 1973-2003. July and August exhibit small overall positive trends due to the more frequent fine weather conditions since 1994. In September the trend of dimming in 1955-1992, has changed to the brightening later. In October quite significant tendency of brightening has changed to more strong dimming during the last two decades (see Fig. 13). In November, the dimming tendency was observed during the whole observation period. The negative slope of the trend has increased significantly after 1983. In December, similar increase of the linear dimming trend happened around 1977.
