HOMOGENIZED SUNSHINE DURATION (1936-2013) AND GLOBAL RADIATION (1959-2013) INSTRUMENTAL TIME SERIES OVER ITALY: VARIABILITY AND TRENDS

Daily sunshine duration (SD) and global radiation (Eg↓) series were recovered for the whole Italian territory for the periods 1936-2013 and 1959-2013, respectively. To obtain reliable series, useful to study the variability and trends, it has been necessary to solve a number of problems concerning the quality of the data, their completeness and their spatial distribution. Specifically, all the corresponding monthly records were subjected to a detailed quality control and homogenization procedure in order to eliminate non climatic signals and then the missing values were estimated. Moreover, all the series were converted into seasonal and annual anomaly series and a 1◦x1◦ gridded version of the dataset has been generated. Finally, two regional records (northern and southern Italy) were obtained by averaging the corresponding grid-point series. Besides to all-sky records, clear-sky anomaly records were also obtained selecting only the clear days in the original series by comparison with corresponding total cloud cover (TCC) series. The all-sky SD and Eg↓ anomaly regional records show a decreasing tendency (”Global dimming”) between the 1950s and 1980s while in the subsequent period they show an increasing tendency (”Brightening period”). The SD series, covering a longer period, show also an increasing tendency between the mid-1930s and the mid-1950s (”Early brightening”). The intensity and the length of the signals depend on the considered variable, region and season. The comparison between SD records with corresponding TCC records shows that the expected negative correlation between these variables is often not evident, especially between the 1960s and 1980s. This suggests that during the dimming period there is an important fraction of SD evolution that cannot be explained by TCC. It must therefore depend on other factors, as for example, changes in aerosol optical thickness. The clear-sky SD and Eg↓ series show longer and more significant trends than the all-sky series for almost all the seasons. The most relevant changes are observed in winter and autumn for both variables, highlighting the important role of clouds under all-sky conditions. The resulting trends under clear-sky conditions are in agreement with changes in anthropogenic aerosols suggesting that they have a relevant role on Eg↓ variability. Nevertheless, an high correlation coefficient between Eg↓ series and the Sahel Precipitation index, especially in the southern region, suggests also a significant contribution of natural aerosols. This can justify the stronger dimming during spring, summer and autumn in the south than in the north. Moreover, the fact that the intensity of the dimming/brightening trends change in all seasons removing the cloud contribution supports the hypothesis that clouds contribute in a significant way to the Eg↓ variability under all-sky conditions. The results highlight that the agreement between SD and Eg↓ decadal variability and long-term trends, over the common period (1959-2013), depends on the considered region, season and period. Specifically, Eg↓ clear-sky series show stronger tendencies than SD, both during the dimming and brightening periods. In order to investigate whether the differences in the clear-sky trends are due to a different sensitivity to atmospheric turbidity changes, a model has been applied with the aim of estimating how large are SD and Eg↓ relative variations when atmospheric turbidity (expressed by means of the Turbidity Linke Factor - TL) changes. For low TL, Eg↓ is expected to be much more sensitive than SD while for high TL, SD is expected to be slightly more sensitive than Eg↓. These results give evidence that the use of SD as a proxy variable for clear-sky Eg↓ may be problematic, especially if TL is low or if it shows significant changes in time. The comparison between the modelled and the observed relative trends highlights a very good agreement with the only exception of the dimming period in winter and autumn in northern region. These disagreement could be both connected to instrumental problems or to the influence of other meteorological variables. Finally, a methodology to estimate Eg↓ normal values for any interval of a period in which a SD anomaly series is available has been set up (it is very useful for example to evaluate the ability of a RCM-GCM (Regional Climate Model - Global Climate Model) in capturing the spatial distribution of Eg↓), and then, for each period during which the Eg↓ normal values are available, to estimate the corresponding climatologies.