Abstract: The Fuzzy Clustering Method(FCM) is widely used in agricultural land gradation (ALG). However, the FCM may usually bring into an accumulated information bias during the evolution of fuzzy similar matrix. The Fuzzy ISODATA Clustering Method(FISOCM), which can refine the clustering center dynamically, is researched in the present paper. Integrated with GIS, the FISOCM is applied to make the land gradation in Jintan city, Jiangsu Province. The area proportion of gradation from grade I to grade IV is 42.20%, 22.41%, 20.35% and 15.04% respectively. The result shows that the improved method FISOCM is better than the traditional agricultural land gradation method, which valued by weighted sum of indices, and gradation is classified by axis based method or sum frequency curve method. The test result can relate the existent result of agricultural land classification hierarchy more closely.