Geospatial Analysis of Landslide Susceptibility Through Machine Learning In Relation to Environmental Indicators Across Global Regions

This is a Preprint and has not been peer reviewed. This is version 2 of this Preprint.

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.

Downloads

Download Preprint

Authors

Arjun Nair 

Abstract

Landslides cause over a billion dollars per year in damages and the situation is only exacerbated by climate change. Landslide forecasting, as a result, is key for detecting these disasters. Traditional landslide prediction models often rely on localized data and extensive computational resources, limiting their applicability on a global scale and therefore have an accuracy rate of around 30%. Addressing this gap, our research leverages a dataset which uses an enhanced version of NASA’s Global Landslide Catalog. The dataset spans from 2010-2020, containing 10,100 landslide events with meteorological and geological factors preceding the event. After identifying key drivers of landslides and the minimum slope requirements for different lithologies, a Random Forest model was created using python to predict landslides based on their severity index. An 81% accuracy rate was found when trained and tested, proving to be much more effective when compared to traditional models. The model can now be used to provide a novel, data-driven perspective on landslide prediction allowing for a quicker evacuation of areas prone to landslides and providing enough time for countermeasures to be employed.

DOI

https://doi.org/10.31223/X5471F

Subjects

Physical Sciences and Mathematics

Keywords

Global Landslide Catalog, python, landslide severity index, Random Forest

Dates

Published: 2024-10-30 07:17

Last Updated: 2024-10-30 14:17

Older Versions
License

No Creative Commons license

Additional Metadata

Conflict of interest statement:
None