Assessing the role of precipitation in logistic regression landslide susceptibility models in the Blue Ridge Mountains of North Carolina
(2026) NGEK11 20261Department of Earth and Environmental Sciences (MGeo)
- Abstract (Swedish)
- Landslide susceptibility models (LSMs) are essential to mitigate the danger of landslides, and with a changing climate posing potential risks to model accuracy, it is vital to understand the role that precipitation plays within LSMs and the effect shifting climatic conditions may have on model accuracy. This study produced a logistic regression LSM in the Blue Ridge Mountain region of North Carolina for the 2001-2020 period in order to understand the significance of precipitation as a predictor, observe the change in landslide occurrence an increase in precipitation intensity could cause, and analyze the change in accuracy between models trained on one precipitation regime when applied to differing climate conditions. The model exhibited... (More)
- Landslide susceptibility models (LSMs) are essential to mitigate the danger of landslides, and with a changing climate posing potential risks to model accuracy, it is vital to understand the role that precipitation plays within LSMs and the effect shifting climatic conditions may have on model accuracy. This study produced a logistic regression LSM in the Blue Ridge Mountain region of North Carolina for the 2001-2020 period in order to understand the significance of precipitation as a predictor, observe the change in landslide occurrence an increase in precipitation intensity could cause, and analyze the change in accuracy between models trained on one precipitation regime when applied to differing climate conditions. The model exhibited high accuracy, with an average AUC score of 0.85, with no signs of overfitting. Precipitation was the second most significant predictor after slope, with extreme statistical significance (p<0.0001) and a 58% susceptibility rise per standard deviation increase. Additionally, a precipitation sensitivity analysis found that a change in precipitation in the study area led to a significant change in the model’s landslide predictions. The results of a decadal cross-validation showed an 8% decrease in overall accuracy and a 9% rise in Type I errors when applied to a decade with differing precipitation conditions. The decadal results did, however, also highlight a potential landslide inventory bias, which could have held undue influence over the model. Overall, the results of the study suggested that precipitation is a vital predictor within the LSM, and that changes in climate could potentially lead to decreased accuracy within landslide models. Further research is encouraged into this vital issue. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9238567
- author
- Samzelius Valentin, Tom LU
- supervisor
-
- Thomas Pugh LU
- organization
- course
- NGEK11 20261
- year
- 2026
- type
- M2 - Bachelor Degree
- subject
- keywords
- Landslide susceptibility models, landslides, landslide susceptbility, precipitation, logistic regression
- language
- English
- id
- 9238567
- date added to LUP
- 2026-06-16 09:33:46
- date last changed
- 2026-06-16 09:33:46
@misc{9238567,
abstract = {{Landslide susceptibility models (LSMs) are essential to mitigate the danger of landslides, and with a changing climate posing potential risks to model accuracy, it is vital to understand the role that precipitation plays within LSMs and the effect shifting climatic conditions may have on model accuracy. This study produced a logistic regression LSM in the Blue Ridge Mountain region of North Carolina for the 2001-2020 period in order to understand the significance of precipitation as a predictor, observe the change in landslide occurrence an increase in precipitation intensity could cause, and analyze the change in accuracy between models trained on one precipitation regime when applied to differing climate conditions. The model exhibited high accuracy, with an average AUC score of 0.85, with no signs of overfitting. Precipitation was the second most significant predictor after slope, with extreme statistical significance (p<0.0001) and a 58% susceptibility rise per standard deviation increase. Additionally, a precipitation sensitivity analysis found that a change in precipitation in the study area led to a significant change in the model’s landslide predictions. The results of a decadal cross-validation showed an 8% decrease in overall accuracy and a 9% rise in Type I errors when applied to a decade with differing precipitation conditions. The decadal results did, however, also highlight a potential landslide inventory bias, which could have held undue influence over the model. Overall, the results of the study suggested that precipitation is a vital predictor within the LSM, and that changes in climate could potentially lead to decreased accuracy within landslide models. Further research is encouraged into this vital issue.}},
author = {{Samzelius Valentin, Tom}},
language = {{eng}},
note = {{Student Paper}},
title = {{Assessing the role of precipitation in logistic regression landslide susceptibility models in the Blue Ridge Mountains of North Carolina}},
year = {{2026}},
}