Известия Томского политехнического университета: Инжиниринг георесурсов (Apr 2022)

APPLICATION OF GIS-BASED BIVARIATE STATISTICAL METHODS FOR LANDSLIDE POTENTIAL ASSESSMENT IN SAPA, VIETNAM

  • Duong Van B.,
  • Igor K. Fomenko,
  • Trung K. Nguyen,
  • Thi Hong L. Vi,
  • Oleg V. Zerkal,
  • Hong D. Vu

DOI
https://doi.org/10.18799/24131830/2022/4/3473
Journal volume & issue
Vol. 333, no. 4
pp. 126 – 140

Abstract

Read online

The relevance. Predicting and minimizing the impact of natural disasters are critical tasks for governments worldwide, including Vietnam. Landslides are one of the most frequent types of natural disasters in Vietnam, especially in the northern mountainous provinces, resulting in significant loss of life and property. In this study, the GIS-based bivariate statistical methods were applied for assessing landslide potential in Sapa district, Laocai province, Vietnam. For assessing landslide susceptibility, nine landslide-related factors were selected, including elevation, distance to roads, slope, distance to faults, average monthly precipitation, relative relief, land use, crust weathering, and distance to drainage. The main aim of this study is to prepare landslide potential maps for the study area. In addition, the study also demonstrated the effectiveness of bivariate statistical methods for landslide susceptibility assessment. Object of the study is the landslide susceptibility in Sapa district, Laocai province, Vietnam. Methods: GIS-based bivariate statistical methods including frequency ratio, landslide susceptibility analysis, and statistical index. Results. Landslide potential maps were prepared using GIS-based bivariate statistical methods. The study area is divided into five landslide potential zones: very low, low, moderate, high, and very high. The area under the curve of the receiver operating characteristic (AUCROC) was used to evaluate the performance of these models. The success rates of the models for the training data are 74,60 % frequency ratio, 70,82 % landslide susceptibility analysis and 76,36 % statistical index. The prediction rates of the models for the testing data are 77,01 % frequency ratio, 74,36 % landslide susceptibility analysis and 78,11 % statistical index. The performance evaluation of the models revealed that all three techniques are efficient in assessing landslide potential in the study area. Study results are critical for land use planning and economic development, as well as minimizing landslide-related damage.

Keywords