Four radiometric models are compared to study the Angström turbidity coefficient β over Ghardaïa (Algeria). Five years of global irradiance measurements and space data recorded with MODIS are used to estimate β. The models are referenced as Dog β for Dogniaux's method, Louch β for Louche's method, Pinz β for Pinazo's method, Gyem β for Gueymard's method and by modis β for MODIS data. The results showed that Gyem β and Pinz β are very close as the couple Dog β and modis β. Louch β values are between them. Results showed also that all Angström coefficient curves have the same annual trend with maximum and minimum values respectively in summer and winter months. Annual mean values of β increased from 2005 to 2008 with a slight jump in 2007 except for Louch β. The city environment explains it since the urban aerosols predominate over all other types during this period. The jump in 2007 is attributed to the ozone layer thickness that undergoes the same behavior. Some models are then more sensitive to this atmospheric component than others. The occurrence frequency distribution showed that Dog β , Louch β , Pinz β , Gyem β and modis β had their maximum recurrent values near 0.03, 0.07, 0.10, 0.09 and 0.02 respectively. The cumulative frequency distribution revealed also that Dog β and modis β yielded maximum "clean to clear" conditions with respect to others while Pinz β and Gyem β had the minimum. The opposite was observed on the same β pairs with regard to "clear to turbid" and "turbid to very turbid" conditions. Louche's model gave middle values of sky conditions comparing to the other models.