Optimizing Classification Of Landform Element Using Feature Space: A Case Study In A Loess Area

Landform element can be used to reveal the surface morphology, and it can also be classified by a readily approach, Self Organizing Map (SOM), of which the assumed element amount determines the consequence. However, the amount-selected process is not meticulously clarified in literature. To conduct landform element classification of the Chinese Loess Plateau and also introduce this classification approach, we studied the digital elevation model (DEM) of Qingcheng, trained SOM for 11 combinations when the amount was posited from 5 to 10. The suitable amount and subsequent combination could be confirmed through distribution of mean morphometric parameters of landform element in feature space. The optimum amount is suggested to be 8 and combined in two or three dimension due to the best clustering of the parameters. Free from the limitation of element amount, SOM is a flexible way for landform element classification in a distinct region such as the Loess Plateau.