Loess landform variability across large spatial extents needs to be analyzed to understand the formation and evolution of loess landscapes. This is becoming increasingly possible via the automated analysis of remotely-sensed data. Here, we quantify loess landforms using an object-based image analysis (OBIA) method and use this classification to describe the spatial variability of loess landforms. Quantitative indicators are used to drive the spatial variability analysis of loess landforms and explain their spatio-temporal evolution. Moreover, the hypsometric integral (HI) and topographic interpolation are employed to investigate soil erosion and development patterns of loess landscape. Results show that the OBIA method classified loess landforms to an accuracy of 88.7 %. The derived metrics in terms of the area, slope and complexity of landform shape allow the determination of the spatial structure of the loess landscapes. The HI value of the entire basin is 0.486, representing the mature stage of landform development, with relatively severe surface erosion. Correlation analysis of HI values and related indicators in the sub-basins shows that HI is poorly correlated with the area proportion of loess landform types and the total erosion volume in the basin but shows a relatively strong correlation with the volume of erosion per unit area. A link to the paper is here.
