## Abstract

A quantitative model for landslide hazard assessment on natural slopes is presented for a transportation corridor of the Nilgiri Hills in southern India. The data required for the hazard assessment were mostly obtained from historical records. For the hazard modeling, a landslide inventory map was prepared from technical reports and maintenance records for a road and a railroad, available for a 21 year period from 1987 to 2007. Most landslides are shallow translational debris slides and debris flow slides triggered by rainfall. On natural slopes landslides occurred as first-time failures.

A logistic regression model was used to determine the spatial probability of landslides for each pixel by taking the source area of the existing landslides as dependent, and slope angle, aspect, regolith thickness and land use as independent variables. The temporal probability of landslides was estimated indirectly using the exceedance probability of the rainfall threshold required to trigger landslides for the first time on natural slopes. The probability of landslide size was estimated as frequency percentage of landslide volume, a proxy for landslide magnitude, and the percentage values were then expressed as probability. By assuming independence among the three probabilities, a quantitative estimate of a landslide hazard was obtained as the joint probability of landslide volume, of landslide occurrence in an established time period and of spatial probability of a landslide initiation. The models were validated using the rainfall and landslide events that occurred during 2008 and 2009.

Total 12 specific landslide hazard maps were generated considering six time periods (1, 3, 5, 15, 25 and 50 years), and two landslide volumes (volume exceeding 1000 m3 and 10,000 m3).

A logistic regression model was used to determine the spatial probability of landslides for each pixel by taking the source area of the existing landslides as dependent, and slope angle, aspect, regolith thickness and land use as independent variables. The temporal probability of landslides was estimated indirectly using the exceedance probability of the rainfall threshold required to trigger landslides for the first time on natural slopes. The probability of landslide size was estimated as frequency percentage of landslide volume, a proxy for landslide magnitude, and the percentage values were then expressed as probability. By assuming independence among the three probabilities, a quantitative estimate of a landslide hazard was obtained as the joint probability of landslide volume, of landslide occurrence in an established time period and of spatial probability of a landslide initiation. The models were validated using the rainfall and landslide events that occurred during 2008 and 2009.

Total 12 specific landslide hazard maps were generated considering six time periods (1, 3, 5, 15, 25 and 50 years), and two landslide volumes (volume exceeding 1000 m3 and 10,000 m3).

Original language | English |
---|---|

Pages (from-to) | 236-250 |

Journal | Engineering geology |

Volume | 116 |

Issue number | 3-4 |

DOIs | |

Publication status | Published - 2010 |

## Keywords

- ITC-ISI-JOURNAL-ARTICLE
- 2024 OA procedure