The research was aimed to study the efect of choosing different contour interval to produce Digital Elevation Model on a fully raster-based erosion modeling of The Universal Soil Loss Equation using remote sensing data and a geographical information system technique. Methods were applied by analyzing all factors that affecting erosion in GIS environment such data were in the form of raster. Those data were R , K, LS, C and P factors. LS factor was derived from Digital Elevation Model by taking flow direction from each pixel into consideration. Research used 3 contour intervals to produce Digital Elevation Model, i.e. 12.5, 25 and 50 meter. C factor was derived from the formula after applying linearly regression analysis between Normalized Difference Vegetation index of remote sensing data and C factor measured directly on the field. Another analysis was the creation of map of Bulk Density used to convert erosion unit as from Mg ha-1mo-1 to mm mo-1. To know the model accuracy, validation of the model was done by applying statistical analysis and by comparing the result of erosion model (Emodel) with actual erosion (Eactual) which was measured regularly in Merawu watershed. A threshold value of > 0.80 or > 80% was chosen tojustify whether the model was accurate or not. The results showed that all Emodel using 3 countour intervals havecorrelation value of > 0.8. These results were strenghtened with the result of analysis of variance which showing there were no difference between Emodel and Eactual. Among the 3 models, only Emodel using 50 meter countour interval reached the accuracy of 81.13% while the other only had 50.87% (using countour interval 25 meter) and 32.92% (using countour interval 12.5 meter).