The Topographic Wetness Index (TWI) is frequently used to simulate the soil moisture conditions in a watershed quantitatively. The index that extracted from DEM is an important factor with explicit physical significance in the rainfall-runoff process models. Some previous studies discovered that the TWI would change with the resolution of DEM from which it was derived and this change, and terrain heterogeneity of DEM grids were employed to explain the scale dependency. In this paper, we explored the resolution effects of TWI and the influence of terrain heterogeneity of DEM grids. The research area is located in Xiyuanxi watershed, Xiamen City of Fujian Province, which includes different terrain types such as hill, platform, plain, etc. A fine-resolution DEM with 2.5m grid size was used to investigate the scale dependency of TWI values when converting DEM with resolution of 20m. Four terrain and heterogeneity attributes were employed as the quantitative indices of terrain heterogeneity of DEM grids, i.e. standard deviation of elevation, relief, landscape fragmentation index and landscape diversity index. The correlation and regression analysis was performed to identify sensitive and correlative responses between each terrain heterogeneity index and TWI variance as resolution was changed. The results indicated significant negative correlations between the terrain heterogeneity indices and the TWI variance (the Pearson correlation coefficients were -0.707, -0.712, -0.779 and -0.841 respectively, all based on a high confidence level of 0.01). By extrapolating the fitted curve for the terrain heterogeneity indices and TWI variance, the logarithmic curves fit the optimal equations well (the coefficient of determination are all greater than 0.9, F>F_0.01). So, these indices can be used to evaluate the impact of the terrain heterogeneity on the TWI. Meanwhile, the regression models can improve the accuracy of the TWI derived from a coarse resolution DEM. The effect discussed in this study is helpful in providing a more accurate data for the TWI applications.