(5)

Loess landforms in the Loess Plateau are typical landforms in arid and semiarid areas and have a significant impact on the environment and soil erosion. Quantitative analyses on loess landform have been employed from various perspectives. Peak intervisibility can provide the potential topographic information implied in the visual connectivity of peaks, however, its application in loess landform analysis remains unexplored. In this study, the interwoven sightlines among peaks, representing peak intervisibility, were extracted from the digital elevation model and simulated into a peak intervisibility network (PIN). Nine indices were proposed to quantify the PIN. Through a case study in Northern Shaanxi, China, three tasks were conducted, including, landform interpretation, spatial pattern mining, and landform classification. The main findings are as follows: (1) PIN responds to terrain morphology and is beneficial for loess landform interpretation. (2) The spatial patterns of PIN indices are heterogeneous and strongly coupled with the terrain morphologies, showing anisotropy and autocorrelation in spatial variations. (3) Using the light gradient boost machine classifier, the PIN index-based classification reaches a mean accuracy of 86.09%, an overall accuracy of 86% and a kappa coefficient of 0.84. These findings shed light on the applicability of PIN in loess landform analysis. Peak intervisibility not only enriches the theories and methodologies of relation-based digital terrain analysis, but also enhances our comprehension of loess landform genesis, morphology, distribution, and evolution.

来源平台：JOURNAL OF MOUNTAIN SCIENCE