Credal decision tree based novel ensemble models for spatial assessment of gully erosion and sustainable management

Abstract We introduce novel hybrid ensemble models in gully erosion susceptibility mapping (GESM) through a case study in the Bastam sedimentary plain of Northern Iran. Four new ensemble models including credal decision tree-bagging (CDT-BA), credal decision tree-dagging (CDT-DA), credal decision tr...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Alireza Arabameri, Nitheshnirmal Sadhasivam, Hamza Turabieh, Majdi Mafarja, Fatemeh Rezaie, Subodh Chandra Pal, M. Santosh
Formato: article
Lenguaje:EN
Publicado: Nature Portfolio 2021
Materias:
R
Q
Acceso en línea:https://doaj.org/article/2b232a1b707744089733828da19af3ad
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Descripción
Sumario:Abstract We introduce novel hybrid ensemble models in gully erosion susceptibility mapping (GESM) through a case study in the Bastam sedimentary plain of Northern Iran. Four new ensemble models including credal decision tree-bagging (CDT-BA), credal decision tree-dagging (CDT-DA), credal decision tree-rotation forest (CDT-RF), and credal decision tree-alternative decision tree (CDT-ADTree) are employed for mapping the gully erosion susceptibility (GES) with the help of 14 predictor factors and 293 gully locations. The relative significance of GECFs in modelling GES is assessed by random forest algorithm. Two cut-off-independent (area under success rate curve and area under predictor rate curve) and six cut-off-dependent metrics (accuracy, sensitivity, specificity, F-score, odd ratio and Cohen Kappa) were utilized based on both calibration as well as testing dataset. Drainage density, distance to road, rainfall and NDVI were found to be the most influencing predictor variables for GESM. The CDT-RF (AUSRC = 0.942, AUPRC = 0.945, accuracy = 0.869, specificity = 0.875, sensitivity = 0.864, RMSE = 0.488, F-score = 0.869 and Cohen’s Kappa = 0.305) was found to be the most robust model which showcased outstanding predictive accuracy in mapping GES. Our study shows that the GESM can be utilized for conserving soil resources and for controlling future gully erosion.