Phosphate sorption is one of the main factors that greatly affects phosphorus (P) availability in soils. The increased concern about precision in fertilizer application emphasizes the importance of quantifying P sorption for better and accurate P recommendations. This study was conducted to examine P sorption by 18 contrasted soil samples of arid and semiarid regions of Morocco. Three isotherm models were used in this study: Langmuir, Cooke, and Freundlich models. Langmuir and Cooke equations were found to accurately describe P sorption isotherms in calcareous soils of Moroccan arid and semiarid zones. Maximum P adsorption (Xm) varied from 146 to 808 mg P kg(-1) soil for the soils used in our study. Averaged across regions, soils from Chaouia adsorbed more P at maximum adsorption compared to Abda and Ben Sliman soils, suggesting that each region has to have specific P recommendation norms. Maximum buffering capacities (MBC) also showed large variation, the values ranged from 35 to 404 mg P kg(-1). This study showed that these buffer indices can be predicted using soil characteristics determined by routine soil analysis especially using clay (CL, %), lime (L, %), and exchangeable calcium (Ca, mg kg(-1)) contents: MBC=-38.1+/-3.3* CL+3.7*L+0.024*Ca (R2=0.98**). The determination of buffer indices requires the establishment of P sorption isotherms which are difficult to adapt to routine analyses. Therefore, this kind of relationships between buffer indices and soil properties possibly could be used to rapidly predict these indices for inclusion in P recommendation models.