Neural Network Based Landscape Pattern Simulation in ChangBai Mountain, Northeast China

Abstract

Simulation on evolution of landscape pattern is a hot problem because the evolution of terrestrial landscape pattern will be related directly to the changes of climate. In the study, it presents a neural network model for the evolution of landscape pattern by using the landscape pattern transformation rules and parameters. Owing to the typical vertical zoning of vegetation, Mount ChangBai is taken as an example to demonstrate the application of simulation model. Landsat TM data in 1985 and 1999 are combined with the geographic data. The landscape pattern evolution parameters are built and the transformation rules are confirmed by the help of the three layers Back Propagation (BP) Neural Network. There are 16 neural cells for the input layer and 11 cells for the output cell. The evolution of the landscape pattern in 2013 and 2027 are predicted by the model. The precision of the model in 1985 was 84% by taking the year of 1999 as starting point, while the precision of the model in 1999 was 82% by taking the year of 1985 as starting point. The simulation result was very close to actual situation by comparison with Moran I index in 1985 and 1999.