Abstract: A dynamic model for forecasting forest gap size was established using a modified Elman neural network for overcoming the disadvantage of timevariability, uncertainty, and complex nonlinear relationship with its impact factor of forest gap size. Firstly, the structural features, mathematical model and learning algorithm of the modified Elman neural network were examined, and a dynamic model for forecasting forest gap size was then set up based on the modified Elman neural network by selecting Larix principisrupprechtii, Spruce, Pinus tabulaeformis forest in Pangquangou Nature Reserve as the test objects. The fitting and simulation results showed that the established prediction model is feasible to forest gap size, and three kinds of forest gaps were lastly investigated using this model