This paper analyses the interannual variability of solar radiation and sunshine hours for a large tropical region (Brazil), located between latitude 0degreesS and 30degreesS, in order to improve knowledge on solar resources, generate statistical parameters for model checking or to be used as input data of synthetic time series generation. The statistics for the daily, monthly average solar radiation deviations, and daily, monthly average sunshine hours, for the various localities in Brazil, tested with the Kolmogorov-Smirnov method, show that they are random variables, normally distributed. On the other hand, the sequential properties analysis shows that the auto-correlation coefficients with lag 1 are statistically significant only for a few locals: Fortaleza, Sao Luis, Manaus and Belem. But it is necessary to emphasize that the auto-correlation coefficients with lag 1, though not usually statistically significant, are positive for almost all the locals. The AR-1 is the suggested procedure for monthly solar radiation synthetic time series generation, with auto-correlation coefficients varying from 0.30 to 0.47 for the localities in the north of Brazil and zero for other regions. (C) 2003 Elsevier Ltd. All rights reserved.