본문 바로가기
  • Home
Basic Search

Development of an Automated Algorithm for Analyzing Rainfall Thresholds Triggering Landslide Based on AWS and AMOS

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2024, 29(9), pp.125-136
  • DOI : 10.9708/jksci.2024.29.09.125
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : July 15, 2024
  • Accepted : September 10, 2024
  • Published : September 30, 2024

Donghyeon Kim 1,  Song Eu 2,  Kwangyoun Lee 1,  Sukhee Yoon 1,  Jongseo Lee 1,  Donggeun Kim ORD ID 3

1한국치산기술협회
2국립산림과학원
3경북대학교

Accredited

ABSTRACT

This study presents an automated Python algorithm for analyzing rainfall characteristics to establish critical rainfall thresholds as part of a landslide early warning system. Rainfall data were sourced from the Korea Meteorological Administration's Automatic Weather System (AWS) and the Korea Forest Service's Automatic Mountain Observation System (AMOS), while landslide data from 2020 to 2023 were gathered via the Life Safety Map. The algorithm involves three main steps: 1) processing rainfall data to correct inconsistencies and fill data gaps, 2) identifying the nearest observation station to each landslide location, and 3) conducting statistical analysis of rainfall characteristics. The analysis utilized power law and nonlinear regression, yielding an average R² of 0.45 for the relationships between rainfall intensity-duration, effective rainfall-duration, antecedent rainfall-duration, and maximum hourly rainfall-duration. The critical thresholds identified were 0.9-1.4 mm/hr for rainfall intensity, 68.5-132.5 mm for effective rainfall, 81.6-151.1 mm for antecedent rainfall, and 17.5-26.5 mm for maximum hourly rainfall. Validation using AUC-ROC analysis showed a low AUC value of 0.5, highlighting the limitations of using rainfall data alone to predict landslides. Additionally, the algorithm's speed performance evaluation revealed a total processing time of 30 minutes, further emphasizing the limitations of relying solely on rainfall data for disaster prediction. However, to mitigate loss of life and property damage due to disasters, it is crucial to establish criteria using quantitative and easily interpretable methods. Thus, the algorithm developed in this study is expected to contribute to reducing damage by providing a quantitative evaluation of critical rainfall thresholds that trigger landslides.

Statistics

  • 107 Viewed
  • 72 Downloaded

Tools

Issue List

Citation status

* References for papers published after 2023 are currently being built.

This paper was written with support from the National Research Foundation of Korea.
Basic Search Advanced Search