Does Random Forest outperform the Generalized Additive Model?: An evaluation based on rainfall-triggered landslides in the Styrian Basin, Austria

Raphael Knevels; Philip Leopold; Herwig  Proske; Zhihao Wang; Alexander Brenning

Does Random Forest outperform the Generalized Additive Model? An evaluation based on rainfall-triggered landslides in the Styrian Basin, Austria

Raphael Knevels (Speaker)
, Philip Leopold (Speaker)
, Herwig Proske
, Zhihao Wang
, Alexander Brenning

Transportation Infrastructure Technologies

Friedrich Schiller Universität Jena
Joanneum Research Forschungsgesellschaft mbH

Research output: Poster presentation without proceedings › peer-review

Abstract

Our objective was a comparison between Random Forest (RF, machine learning) and Generalized Additive Mo-del (GAM, conventional method) for landslide susceptibility modeling of rainfall-triggered landslides.
In the Styrian Basin more than 3,000 landslides occurred after heavy rainfall events in 2009 and 2014 causing significant da-mage to human infrastructure (Knevels et al., 2020).
The area affected during a 2009-type event could grow by up to 45% in a 4 K global warming scenario (Maraun et al., 2022; Knevels et al., 2023), making appropriate and robust landslide susceptibility predictions a necessary prerequisite for decision-makers.

Original language	English
Publication status	Published - 14 Nov 2023
Event	World Landslide Forum 6: Landslide Science for Sustainable Development - Palazzo degli Affari & Palazzo dei Congressi, Florence, Italy Duration: 14 Nov 2023 → 17 Nov 2023 Conference number: 6 https://wlf6.org/

Conference

Conference	World Landslide Forum 6
Abbreviated title	WLF6
Country/Territory	Italy
City	Florence
Period	14/11/23 → 17/11/23
Internet address	https://wlf6.org/

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 15 Life on Land

Research Field

Road Infrastructure Assessment, Modelling and Safety Evaluation

Cite this

@conference{4b2176bfdedb45aeae7ef15faf3425ac,

title = "Does Random Forest outperform the Generalized Additive Model?: An evaluation based on rainfall-triggered landslides in the Styrian Basin, Austria",

abstract = "Our objective was a comparison between Random Forest (RF, machine learning) and Generalized Additive Mo-del (GAM, conventional method) for landslide susceptibility modeling of rainfall-triggered landslides. In the Styrian Basin more than 3,000 landslides occurred after heavy rainfall events in 2009 and 2014 causing significant da-mage to human infrastructure (Knevels et al., 2020). The area affected during a 2009-type event could grow by up to 45\% in a 4 K global warming scenario (Maraun et al., 2022; Knevels et al., 2023), making appropriate and robust landslide susceptibility predictions a necessary prerequisite for decision-makers.",

keywords = "landslide; climate change; precipitation; morphology; vulnerability; trigger function; hazard nomogram; observation bias",

author = "Raphael Knevels and Philip Leopold and Herwig Proske and Zhihao Wang and Alexander Brenning",

year = "2023",

month = nov,

day = "14",

language = "English",

note = "World Landslide Forum 6 ; Conference date: 14-11-2023 Through 17-11-2023",

url = "https://wlf6.org/",

}

TY - CONF

T1 - Does Random Forest outperform the Generalized Additive Model?

T2 - World Landslide Forum 6

AU - Proske, Herwig

AU - Wang, Zhihao

AU - Brenning, Alexander

A2 - Knevels, Raphael

A2 - Leopold, Philip

N1 - Conference code: 6

PY - 2023/11/14

Y1 - 2023/11/14

N2 - Our objective was a comparison between Random Forest (RF, machine learning) and Generalized Additive Mo-del (GAM, conventional method) for landslide susceptibility modeling of rainfall-triggered landslides. In the Styrian Basin more than 3,000 landslides occurred after heavy rainfall events in 2009 and 2014 causing significant da-mage to human infrastructure (Knevels et al., 2020). The area affected during a 2009-type event could grow by up to 45% in a 4 K global warming scenario (Maraun et al., 2022; Knevels et al., 2023), making appropriate and robust landslide susceptibility predictions a necessary prerequisite for decision-makers.

AB - Our objective was a comparison between Random Forest (RF, machine learning) and Generalized Additive Mo-del (GAM, conventional method) for landslide susceptibility modeling of rainfall-triggered landslides. In the Styrian Basin more than 3,000 landslides occurred after heavy rainfall events in 2009 and 2014 causing significant da-mage to human infrastructure (Knevels et al., 2020). The area affected during a 2009-type event could grow by up to 45% in a 4 K global warming scenario (Maraun et al., 2022; Knevels et al., 2023), making appropriate and robust landslide susceptibility predictions a necessary prerequisite for decision-makers.

KW - landslide; climate change; precipitation; morphology; vulnerability; trigger function; hazard nomogram; observation bias

M3 - Poster presentation without proceedings

Y2 - 14 November 2023 through 17 November 2023

ER -