Assessing the performance of GIS- based machine learning models with different accuracy measures for determining susceptibility to gully erosion

Christian Conoscenti, Hamid Reza Pourghasemi, Mohsen Sheklabadi, Younes Garosi, Kristof Van Oost

Risultato della ricerca: Article

5 Citazioni (Scopus)

Abstract

The main purpose was to compare discrimination and reliability of four machine learning models to create gully erosion susceptibility map (GESM) in a part of Ekbatan Dam Basin, Hamedan, western Iran. Extensive field surveys using GPS, and the visual interpretation of satellite images, used to prepare a digital map of the spatial distribution of gullies. 130 locations were sampled to elucidate the spatial distribution of the soil surface properties. Topographic attributes were provided from digital elevation model (DEM). The land use and normalized difference vegetation index (NDVI) maps were created by satellite imagery. The functional relationships between gully erosion and controlling factors were calculated using the random forest (RF), support vector machine (SVM), Naïve Bayes (NB), and generalized additive model (GAM) models. The performance of models was evaluated by 10-fold cross-validation based on efficiency, Kappa coefficient, receiver operating characteristic curve (ROC), mean absolute error (MAE), and root mean square error (RMSE). The results showed that the RF model had the highest amount of efficiency, Kappa coefficient, and AUC and the lowest amounts of MAE and RMSE compared with SVM, NB, and GAM. The RF model showed the highest predictive performance (mean AUC = 92.4%), followed by SVM (mean AUC = 90.9%), GAM (mean AUC = 89.9%), and NB (mean AUC = 87.2%) models. Overall accuracy of the models ranged from excellent (NB, GAM) to outstanding (RF, SVM) classes. The capacity of all models for creating GESM was quite stable when the calibration and validation samples were changed through10-fold cross-validation technique. According to variable importance analysis performed by RF model, the most important variables are distance from rivers, calcium carbonate equivalent (CCE), and topographic position index (TPI). The obtained maps can help identifying areas at risk of gully erosion and facilitate the implementation of plans for soil conservation and sustainable management.
Lingua originaleEnglish
pagine (da-a)1117-1132
Numero di pagine16
RivistaScience of the Total Environment
Volume664
Stato di pubblicazionePublished - 2019

Fingerprint

gully erosion
Geographic information systems
Learning systems
Erosion
GIS
Support vector machines
machine learning
Mean square error
Spatial distribution
spatial distribution
fold
Soil conservation
digital map
Satellite imagery
Calcium Carbonate
soil conservation
gully
Calcium carbonate
calcium carbonate
NDVI

All Science Journal Classification (ASJC) codes

  • Environmental Engineering
  • Environmental Chemistry
  • Waste Management and Disposal
  • Pollution

Cita questo

Assessing the performance of GIS- based machine learning models with different accuracy measures for determining susceptibility to gully erosion. / Conoscenti, Christian; Pourghasemi, Hamid Reza; Sheklabadi, Mohsen; Garosi, Younes; Van Oost, Kristof.

In: Science of the Total Environment, Vol. 664, 2019, pag. 1117-1132.

Risultato della ricerca: Article

@article{b6f3deea91374b95b9b10465cf17d6d2,
title = "Assessing the performance of GIS- based machine learning models with different accuracy measures for determining susceptibility to gully erosion",
abstract = "The main purpose was to compare discrimination and reliability of four machine learning models to create gully erosion susceptibility map (GESM) in a part of Ekbatan Dam Basin, Hamedan, western Iran. Extensive field surveys using GPS, and the visual interpretation of satellite images, used to prepare a digital map of the spatial distribution of gullies. 130 locations were sampled to elucidate the spatial distribution of the soil surface properties. Topographic attributes were provided from digital elevation model (DEM). The land use and normalized difference vegetation index (NDVI) maps were created by satellite imagery. The functional relationships between gully erosion and controlling factors were calculated using the random forest (RF), support vector machine (SVM), Na{\"i}ve Bayes (NB), and generalized additive model (GAM) models. The performance of models was evaluated by 10-fold cross-validation based on efficiency, Kappa coefficient, receiver operating characteristic curve (ROC), mean absolute error (MAE), and root mean square error (RMSE). The results showed that the RF model had the highest amount of efficiency, Kappa coefficient, and AUC and the lowest amounts of MAE and RMSE compared with SVM, NB, and GAM. The RF model showed the highest predictive performance (mean AUC = 92.4{\%}), followed by SVM (mean AUC = 90.9{\%}), GAM (mean AUC = 89.9{\%}), and NB (mean AUC = 87.2{\%}) models. Overall accuracy of the models ranged from excellent (NB, GAM) to outstanding (RF, SVM) classes. The capacity of all models for creating GESM was quite stable when the calibration and validation samples were changed through10-fold cross-validation technique. According to variable importance analysis performed by RF model, the most important variables are distance from rivers, calcium carbonate equivalent (CCE), and topographic position index (TPI). The obtained maps can help identifying areas at risk of gully erosion and facilitate the implementation of plans for soil conservation and sustainable management.",
author = "Christian Conoscenti and Pourghasemi, {Hamid Reza} and Mohsen Sheklabadi and Younes Garosi and {Van Oost}, Kristof",
year = "2019",
language = "English",
volume = "664",
pages = "1117--1132",
journal = "Science of the Total Environment",
issn = "0048-9697",
publisher = "Elsevier",

}

TY - JOUR

T1 - Assessing the performance of GIS- based machine learning models with different accuracy measures for determining susceptibility to gully erosion

AU - Conoscenti, Christian

AU - Pourghasemi, Hamid Reza

AU - Sheklabadi, Mohsen

AU - Garosi, Younes

AU - Van Oost, Kristof

PY - 2019

Y1 - 2019

N2 - The main purpose was to compare discrimination and reliability of four machine learning models to create gully erosion susceptibility map (GESM) in a part of Ekbatan Dam Basin, Hamedan, western Iran. Extensive field surveys using GPS, and the visual interpretation of satellite images, used to prepare a digital map of the spatial distribution of gullies. 130 locations were sampled to elucidate the spatial distribution of the soil surface properties. Topographic attributes were provided from digital elevation model (DEM). The land use and normalized difference vegetation index (NDVI) maps were created by satellite imagery. The functional relationships between gully erosion and controlling factors were calculated using the random forest (RF), support vector machine (SVM), Naïve Bayes (NB), and generalized additive model (GAM) models. The performance of models was evaluated by 10-fold cross-validation based on efficiency, Kappa coefficient, receiver operating characteristic curve (ROC), mean absolute error (MAE), and root mean square error (RMSE). The results showed that the RF model had the highest amount of efficiency, Kappa coefficient, and AUC and the lowest amounts of MAE and RMSE compared with SVM, NB, and GAM. The RF model showed the highest predictive performance (mean AUC = 92.4%), followed by SVM (mean AUC = 90.9%), GAM (mean AUC = 89.9%), and NB (mean AUC = 87.2%) models. Overall accuracy of the models ranged from excellent (NB, GAM) to outstanding (RF, SVM) classes. The capacity of all models for creating GESM was quite stable when the calibration and validation samples were changed through10-fold cross-validation technique. According to variable importance analysis performed by RF model, the most important variables are distance from rivers, calcium carbonate equivalent (CCE), and topographic position index (TPI). The obtained maps can help identifying areas at risk of gully erosion and facilitate the implementation of plans for soil conservation and sustainable management.

AB - The main purpose was to compare discrimination and reliability of four machine learning models to create gully erosion susceptibility map (GESM) in a part of Ekbatan Dam Basin, Hamedan, western Iran. Extensive field surveys using GPS, and the visual interpretation of satellite images, used to prepare a digital map of the spatial distribution of gullies. 130 locations were sampled to elucidate the spatial distribution of the soil surface properties. Topographic attributes were provided from digital elevation model (DEM). The land use and normalized difference vegetation index (NDVI) maps were created by satellite imagery. The functional relationships between gully erosion and controlling factors were calculated using the random forest (RF), support vector machine (SVM), Naïve Bayes (NB), and generalized additive model (GAM) models. The performance of models was evaluated by 10-fold cross-validation based on efficiency, Kappa coefficient, receiver operating characteristic curve (ROC), mean absolute error (MAE), and root mean square error (RMSE). The results showed that the RF model had the highest amount of efficiency, Kappa coefficient, and AUC and the lowest amounts of MAE and RMSE compared with SVM, NB, and GAM. The RF model showed the highest predictive performance (mean AUC = 92.4%), followed by SVM (mean AUC = 90.9%), GAM (mean AUC = 89.9%), and NB (mean AUC = 87.2%) models. Overall accuracy of the models ranged from excellent (NB, GAM) to outstanding (RF, SVM) classes. The capacity of all models for creating GESM was quite stable when the calibration and validation samples were changed through10-fold cross-validation technique. According to variable importance analysis performed by RF model, the most important variables are distance from rivers, calcium carbonate equivalent (CCE), and topographic position index (TPI). The obtained maps can help identifying areas at risk of gully erosion and facilitate the implementation of plans for soil conservation and sustainable management.

UR - http://hdl.handle.net/10447/341643

UR - http://www.elsevier.com/locate/scitotenv

M3 - Article

VL - 664

SP - 1117

EP - 1132

JO - Science of the Total Environment

JF - Science of the Total Environment

SN - 0048-9697

ER -