We have analyzed the influence of global radiation temporal variability on diffuse fraction correlations. The variability of a zenith angle independent clearness index within an hour is defined by means of three parameters, from a 5-minute basis data set, and assigned to the corresponding hour. Although description of sky condition is usually based only on the clearness index k(t) leading to Liu-Jordan type correlations, various authors use kt(t) along with solar altitude, which is considered as the second most important variable in the diffuse fraction predictions The use of a variability parameter also improves the correlations. More specifically, our analysis reveals that the variability parameters play a role as important as the solar altitude. The influence of the variability parameters is more effective under clear sky conditions and high k, values. The linear correlations obtained by using the clearness index, the solar altitude and a variability parameter reduce the RMSE of the diffuse fraction estimation by about 15%, relative to the k(t)-only correlation. The analysis used two series of data recorded in Catalonia (NE of Iberian Peninsula), which are 3 years and 1 year long respectively.