Drainage basin object-based method for regional-scale landform classification: a case study of loess area in China

Landform classification is one of the most important procedures in recognizing and dividing earth surface landforms. However, topographical homogeneity and differences in regional-scale landforms are often ignored by traditional pixel- and object-based landform classification methods based on digital elevation models (DEMs). In this work, a drainage basin object-based method for classifying regional-scale landforms is proposed. Drainage basins with least critical areas are first delineated from DEMs. Then, terrain derivatives of mean elevation, mean slope, drainage density, drainage depth, and terrain texture are employed to characterize the morphology of the drainage basins. Finally, a decision tree based on the topographical characteristics of the drainage basins is constructed and employed to determine the landform classification law. The experiment is validated in the landform classification of regional-scale loess areas in China. Results show that clear boundaries exist in different landforms at the regional scale. Landform type in a specific region shows significant topographical homogeneity under its specific regional geomorphological process. Classification accuracies are 87.3 and 86.3% for the field investigation and model validation, respectively. Spatial patterns of classified landforms and their proximity to sediment sources and other factors can be regarded as indicators of the evolutionary process of loess landform formation.