In the process of spatial analysis, the vector data are always rasterized into the raster data because of the convenience for spatial analysis. How the area accuracy loses during rasterizing needs to be answered when different grid sizes are adopted in order to choose a suitable grid size. The question is answered here by rasterizing the landuse vector data of Chongqing City with grid of different sizes. Firstly, the landuse vector data are rasterized with grid of different resolutions such as 30m,60m,100m,200m,300m,400m,500m,600m,700m,800m and 1000m. The value of each grid takes the landuse type code as the dominant area. Secondly, the area for each landuse type is calculated in grid. Thirdly, the area for each landuse type is also calculated in vector and used as the actual area. Both the number and average size of polygons for each landuse type are calculated. Fourthly, the relationship among the accuracy loss, grid sizes and the average block size of each landuse type is analyzed. Finally, models for describing the relationships are formulated. It is shown that: (1)The relationship between the accuracy loss(Y) of each landuse type except the high canopy grassland and grid sizes(X) which can be described by the linear regression models such as, Y = A + BX . Here, A and B are coefficients of the models. The correlation coefficients for each model of each landuse type are all above 0.82?The correlations of the absolute value of accuracy losses and grid sizes are positive. The accuracy losses denote that the area reduces compared to the true area when the area of the average block size of each landuse type is above 101 hectare, which is opposite when the area of the average block size of each landuse type is below 82.(2)The correlations between the accuracy loss and the area of the average block size (S) of each landuse type for each grid size can be described by the regression models such as, Y = A +B ln (S) . Here, A and B are coefficients of the models. The correlation coefficients for each model of each grid size are all above 0.72 except grid size of 30 M. The correlations are positive. (3) Generally, the relationship among the accuracy loss(Y), grid size(X) and the average block size(S) of each landuse type can be described by such model as, Y = 5.366 -0.179 X -0.978ln (S) + 0.0348 X ln (S) ?The correlation coefficient of the model is 0.93.