The soil productive forces are mainly affected by soil erosion as soil nutrients and organic carbon are taken away. On the premise of analysis of the last 17 years’ data of rainfalls and vegetation coverage in Xiejiawan small watershed, the grey model was established to forecast and evaluate the losses of organic carbon caused by water erosion in the hilly area of purple soils. Moreover, according to comparison of four-dimensional, five-dimensional and six-dimensional model groups, we found that, six-dimensional GM (1,1) model was the best one, posterior-variance-test accuracy of which amounted to the first class, and maximum simulated relative error ,average simulated relative error and predicted relative error of which were respectively 43.26 % ,16.85% and -17.09%, besides, its simulation and prediction accuracy was also high. Furthermore, the accuracy of simulated forecast value was all satisfactory. So the SOC loss by water erosion in purple areas of Xiejiawan small watershed was simulated and predicted well by grey model, also the unique evolution patterns of changing organic carbon in purple soil by water erosion could be well described and explained by the grey model.