Abstract: In the current evaluation of geological hazards, there is a certain degree of arbitrariness and subjectivity in selecting the best window unit scale for the extraction of relief amplitude (also named relative slope height). The certain error exists as compared the obtained relief amplitude parameters with the actual situation in the study area. During the disaster assessment, using grid-based information models or various popular machine learning methods can lead to a reduction in the reliability of the evaluation results due to the errors of the evaluation factors. In the current study, we use DEM data (2m resolution) of Mizhi County, adopt 10×10, 1×1 rectangular windows for the relief amplitude extraction based on ArcGIS platform, and use the mean change-point analysis to calculate the relief amplitude of Mizhi County from 0–256.60 m, and the best statistical cell is 59×59 with the grid length of 2m, and the side length of the extraction window is 118m and square is 13924 m². Afterwards, according to the statistics on the slope height easily inducing historical landslides and collapses in the loess region of northern Shaanxi Province, the relief amplitude in Mizhi County is divided into five intervals: ＜20 m, 20–40 m, 40–60 m, 60–80 m and ＞80 m. Due to the comprehensive influence of the original terrain conditions, slope cutting and building of houses, factories, etc., 40–80 m is the main interval for the development of disaster hazards, with a proportion of 88.60% of disaster hazard points.The comparison curves of information value and hazard point density of each interval were made, which demonstrates the reasonableness of the selection of the statistical unit and the division of the interval of the relief amplitude. The calculation and analysis results of high-precision DEM data adopted first avoid the drawbacks of visually searching for inflection points, and secondly meet the needs of digital terrain analysis and refined geological hazard investigation in the mountainous terrains of the Loess Plateau region. The method and the results can provide technical support for the evaluation and prevention of geological hazards in the Loess Plateau area and the management of soil erosion and ecological environment protection in the middle and upper reaches of the Yellow River basin.