Hourly global solar irradiation data useful for the design of solar energy conversion systems is generated using a new satellite based model called SICIC (solar irradiation from cloud image classification). it is a model built by processing high resolution visible Meteosat images and ground measurements of solar radiation flux collected in various locations of France during the 1994/95 period. Taking into account the hourly variability of solar radiation flux, SICIC begins by sorting the sun elevation angles into representative classes. For each location and each class, the clearness index is then computed, and the grey levels of the pixels of the visible Meteosat images are converted into cloud cover indexes. Next, two adjustable thresholds are used to split the set of cloud cover indices into three subsets representing clear, partly covered and overcast skies, respectively. Finally, three regression equations linking the clearness and cloud cover indices are obtained for the three sky categories. In these equations, the SICIC parameters (the regression coefficients and both thresholds) vary against the solar elevation angle. SICIC then yields a good estimate of the hourly global solar irradiation flux for every site displayed in the Meteosat images. This model is compared to GISTEL, which is another model governed by the same hypotheses as SICIC but differing by the invariability of its parameters in time. SICIC is found to be more accurate than GISTEL. In particular, the hourly solar data are estimated with an error (RMSE) varying from 32% to 12% for SICIC and 44% to 14% for GISTEL when the sun altitude increases from 15 degrees to 75 degrees. At the daily scale, SICIC is also more efficient than GISTEL. It has been satisfactorily applied to other sites of France and Algeria. The tests made on Wefax and B2 Meteosat images show that this model can be easily extended to other satellite images. (c) 2007 Elsevier Ltd. All rights reserved.