Correlation of Quantitative Terrain Factors in Gully Hill Areas of China Loess Plateau
Zhang Ting, Guoan Tang, Yi Long · 2005
Terrain factors,although different in the definition and calculation method,relate each other at different extent.Such relationship can be represented by a correlation index,which reveals the process and stage of terrain development as well.This paper focuses mainly on the correlation between different terrain factors and the mean-slope by means of the Back Propagation model of Neural Network with a latent layer.Furthermore,the regression model and the NN model without a latent layer are compared with the NN model with a latent layer.Fifteen loess gully-hill areas are selected as the experimental area,and the relevant 1∶10 000 scale DEMs(5 m×5 m grid) are applied as the basic data.From the results of the NN model with a latent layer,it is found that roughness and undulation are the most closely correlated with mean-slope.Compared with others,channel density and mean elevation are the least correlated with mean-slope.Experiment results show the NN model with a latent layer is better than the others and it can effectively evaluate the correlation between the terrain factors extracted from DEMs.This method provides a new methodology in the selection of suitable and available terrain factors and the estimation of the relevancy between these factors.