The removal of fine materials from surface water represents a challenge. This research tries to improve the efficiency of flocculation process of low turbidity water by dividing the process into stage of gradual descending velocity gradient. A three stages flocculator model of continuous flow was designed and constructed. Combinations of three levels of velocity gradient of 60, 45 and 30 sec-1 were applied in the experiments in descending order. Tigris river water was used as raw water with 8-12 NTU turbidity. Alum was used as a coagulant at the optimum dosage. Turbidity removal percent was considered as an indicator of flocculation efficiency. The results showed a significant increase in turbidity removal percent with the decrease of velocity gradient at stage III. On the other hand, velocity gradient at stage I shows a direct relationship with turbidity removal, while the relationship is not clear at stage II. The best turbidity removal percent of more than 80% was obtained by the combinations 60*30*30 and 45*45* 30 sec-1. A regression model shows that velocity gradient at stage III was the most contributor to turbidity removal variation. The research recommended descending gradual velocity gradient flocculation at two stages