Binary logistic regression versus stochastic gradient boosted decision trees in assessing landslide susceptibility for multiple-occurring landslide events: application to the 2009 storm event in Messina (Sicily, southern Italy)

L. Lombardo, M. Cama, C. Conoscenti, M. Märker, E. Rotigliano

Research output: Contribution to journalArticleAcademicpeer-review

56 Citations (Scopus)

Abstract

This study aims to compare binary logistic regression (BLR) and stochastic gradient treeboost (SGT) methods in assessing landslide susceptibility within the Mediterranean region for multiple-occurrence regional landslide events. A test area was selected in the north-eastern sector of Sicily (southern Italy) where thousands of debris flows and debris avalanches triggered on the first October 2009 due to an extreme storm. Exploiting the same set of predictors and the 2009 event landslide archive, BLR- and SGT-based susceptibility models have been obtained for the two catchments separately, adopting a random partition (RP) technique for validation. In addition, the models trained in one catchment have been tested in predicting the landslide distribution in the second, adopting a spatial partition (SP)-based validation. The models produced high predictive performances with a general consistency between BLR and SGT in the susceptibility maps, predictor importance and role. In particular, SGT models reached a higher prediction performance with respect to BLR models for RP-modelling, while for the SP-based models, the difference in predictive skills dropped, converging to equally excellent performances. However, analysing the precision of the probability estimates, BLR produced more robust models around the mean value for each pixel, indicating possible overfitting effects, which affect decision trees to a greater extent. The assessment of the predictor roles allowed identifying the activation mechanisms which are primarily controlled by steep south-facing open slopes located near the coastal area. These slopes are characterised by low/middle altitude downhill from mountain tops, having a medium-grade metamorphic bedrock, under grassland and cultivated (terraced) uses.

Original languageEnglish
Pages (from-to)1621-1648
Number of pages28
JournalNatural hazards
Volume79
Issue number3
DOIs
Publication statusPublished - 1 Aug 2015
Externally publishedYes

Fingerprint

landslide
logistics
catchment
debris avalanche
decision
debris flow
bedrock
pixel
grassland
mountain
prediction
modeling

Keywords

  • Forward logistic regression
  • Landslide susceptibility
  • Messina 2009 disaster
  • Prediction spatial transferability
  • Sicily
  • Stochastic gradient treeboost
  • ITC-ISI-JOURNAL-ARTICLE

Cite this

@article{4da09e72c38d43a1846c9686a7614853,
title = "Binary logistic regression versus stochastic gradient boosted decision trees in assessing landslide susceptibility for multiple-occurring landslide events: application to the 2009 storm event in Messina (Sicily, southern Italy)",
abstract = "This study aims to compare binary logistic regression (BLR) and stochastic gradient treeboost (SGT) methods in assessing landslide susceptibility within the Mediterranean region for multiple-occurrence regional landslide events. A test area was selected in the north-eastern sector of Sicily (southern Italy) where thousands of debris flows and debris avalanches triggered on the first October 2009 due to an extreme storm. Exploiting the same set of predictors and the 2009 event landslide archive, BLR- and SGT-based susceptibility models have been obtained for the two catchments separately, adopting a random partition (RP) technique for validation. In addition, the models trained in one catchment have been tested in predicting the landslide distribution in the second, adopting a spatial partition (SP)-based validation. The models produced high predictive performances with a general consistency between BLR and SGT in the susceptibility maps, predictor importance and role. In particular, SGT models reached a higher prediction performance with respect to BLR models for RP-modelling, while for the SP-based models, the difference in predictive skills dropped, converging to equally excellent performances. However, analysing the precision of the probability estimates, BLR produced more robust models around the mean value for each pixel, indicating possible overfitting effects, which affect decision trees to a greater extent. The assessment of the predictor roles allowed identifying the activation mechanisms which are primarily controlled by steep south-facing open slopes located near the coastal area. These slopes are characterised by low/middle altitude downhill from mountain tops, having a medium-grade metamorphic bedrock, under grassland and cultivated (terraced) uses.",
keywords = "Forward logistic regression, Landslide susceptibility, Messina 2009 disaster, Prediction spatial transferability, Sicily, Stochastic gradient treeboost, ITC-ISI-JOURNAL-ARTICLE",
author = "L. Lombardo and M. Cama and C. Conoscenti and M. M{\"a}rker and E. Rotigliano",
year = "2015",
month = "8",
day = "1",
doi = "10.1007/s11069-015-1915-3",
language = "English",
volume = "79",
pages = "1621--1648",
journal = "Natural hazards",
issn = "0921-030X",
publisher = "Springer",
number = "3",

}

Binary logistic regression versus stochastic gradient boosted decision trees in assessing landslide susceptibility for multiple-occurring landslide events : application to the 2009 storm event in Messina (Sicily, southern Italy). / Lombardo, L.; Cama, M.; Conoscenti, C.; Märker, M.; Rotigliano, E.

In: Natural hazards, Vol. 79, No. 3, 01.08.2015, p. 1621-1648.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Binary logistic regression versus stochastic gradient boosted decision trees in assessing landslide susceptibility for multiple-occurring landslide events

T2 - application to the 2009 storm event in Messina (Sicily, southern Italy)

AU - Lombardo, L.

AU - Cama, M.

AU - Conoscenti, C.

AU - Märker, M.

AU - Rotigliano, E.

PY - 2015/8/1

Y1 - 2015/8/1

N2 - This study aims to compare binary logistic regression (BLR) and stochastic gradient treeboost (SGT) methods in assessing landslide susceptibility within the Mediterranean region for multiple-occurrence regional landslide events. A test area was selected in the north-eastern sector of Sicily (southern Italy) where thousands of debris flows and debris avalanches triggered on the first October 2009 due to an extreme storm. Exploiting the same set of predictors and the 2009 event landslide archive, BLR- and SGT-based susceptibility models have been obtained for the two catchments separately, adopting a random partition (RP) technique for validation. In addition, the models trained in one catchment have been tested in predicting the landslide distribution in the second, adopting a spatial partition (SP)-based validation. The models produced high predictive performances with a general consistency between BLR and SGT in the susceptibility maps, predictor importance and role. In particular, SGT models reached a higher prediction performance with respect to BLR models for RP-modelling, while for the SP-based models, the difference in predictive skills dropped, converging to equally excellent performances. However, analysing the precision of the probability estimates, BLR produced more robust models around the mean value for each pixel, indicating possible overfitting effects, which affect decision trees to a greater extent. The assessment of the predictor roles allowed identifying the activation mechanisms which are primarily controlled by steep south-facing open slopes located near the coastal area. These slopes are characterised by low/middle altitude downhill from mountain tops, having a medium-grade metamorphic bedrock, under grassland and cultivated (terraced) uses.

AB - This study aims to compare binary logistic regression (BLR) and stochastic gradient treeboost (SGT) methods in assessing landslide susceptibility within the Mediterranean region for multiple-occurrence regional landslide events. A test area was selected in the north-eastern sector of Sicily (southern Italy) where thousands of debris flows and debris avalanches triggered on the first October 2009 due to an extreme storm. Exploiting the same set of predictors and the 2009 event landslide archive, BLR- and SGT-based susceptibility models have been obtained for the two catchments separately, adopting a random partition (RP) technique for validation. In addition, the models trained in one catchment have been tested in predicting the landslide distribution in the second, adopting a spatial partition (SP)-based validation. The models produced high predictive performances with a general consistency between BLR and SGT in the susceptibility maps, predictor importance and role. In particular, SGT models reached a higher prediction performance with respect to BLR models for RP-modelling, while for the SP-based models, the difference in predictive skills dropped, converging to equally excellent performances. However, analysing the precision of the probability estimates, BLR produced more robust models around the mean value for each pixel, indicating possible overfitting effects, which affect decision trees to a greater extent. The assessment of the predictor roles allowed identifying the activation mechanisms which are primarily controlled by steep south-facing open slopes located near the coastal area. These slopes are characterised by low/middle altitude downhill from mountain tops, having a medium-grade metamorphic bedrock, under grassland and cultivated (terraced) uses.

KW - Forward logistic regression

KW - Landslide susceptibility

KW - Messina 2009 disaster

KW - Prediction spatial transferability

KW - Sicily

KW - Stochastic gradient treeboost

KW - ITC-ISI-JOURNAL-ARTICLE

UR - https://ezproxy2.utwente.nl/login?url=https://doi.org/10.1007/s11069-015-1915-3

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2015/isi/lombardo_bin.pdf

U2 - 10.1007/s11069-015-1915-3

DO - 10.1007/s11069-015-1915-3

M3 - Article

VL - 79

SP - 1621

EP - 1648

JO - Natural hazards

JF - Natural hazards

SN - 0921-030X

IS - 3

ER -