The trend of SSD for the 10 stations in East China from 1908 to 1936 was 0.16 h d^？1 per decade. And it showed different seasonal trends and the most significant increasing trend (0.33 h d^？1 per decade) occurred in autumn. Given that clouds are one of the most important factors modulating SSR and, therefore, SSD is expected to be inversely related to TCC. As shown in Figure, a strong inverse relationship was found between SSD and TCC during 1908–36. The TCC time series showed a decreasing trend of ？0.93% per decade from 1908 to 1936. The decreasing trend of TCC was in good agreement with the brightening, mentioned above, indicating that TCC change accounted for the brightening. The correlation coefficients between SSD and TCC were ？0.79, ？0.64, ？0.78, and ？0.81 in spring, summer, autumn and winter, respectively, and ？0.79 for the annual mean.  

Why the minimum correlation was found for summer can be explained by using the prevailing local convective clouds. These clouds would have contributed to the TCC observations but would not have greatly affected SSD observations. Compared with the coefficient between SSD and TCC (0.23) during 1954–2005 (Xia, 2010), TCC showed an obvious inverse relationship with SSD during 1908–36. TCC could explain the SSD variation well during 1908–36, which was totally different to the period after the 1950s. We suggest that, owing to industrialization having been quite limited before the 1950s, and because of the relatively minor influence of human activities on the atmospheric environment at that time, cloud cover variation would thus have been the main factor driving the decadal trend of surface solar radiation.