Artificial Neural Network Model of the Hydrological Effects of the Pinus tabulaeformis Plantations in Loess Plateau

Taking the Pinus tabulaeformis plantation in Pianguan County, Shanxi Province, with typical semiarid loess landform and climate characteristics, as the research object, the 5∶q∶1 BP artificial neural network model was established based on the correlations of the variables in the water balance formula by the hypothesis of ‘brought by the stand itself’, integrating with the actual local conditions of the experimental plots to determine the input and output variables. The neural network model was testified with 567 groups of data observed from 2008 to 2011, and the optimum network structure was obtained as 5∶6∶1. The results showed that the mean square error was 0.002 888, with the general fitting accuracy as 93.87%, and the testified fitting accuracy was 93.35%.