Multiple Kernel Learning with Maximum Inundation Extent from MODIS Imagery for Spatial Prediction of Flood Susceptibility

Identifying flood prone areas is essential for basin management. In this paper, a spatial prediction technology of flood susceptibility based on multiple kernel learning (MKL) is proposed. We establish the flood susceptibility model by using EasyMKL, nonlinear MKL (NLMKL), Representative MKL(RMKL),...

Full description

Saved in:
Bibliographic Details
Published in:Water resources management 2022, Vol.36 (1), p.55-73
Main Authors: Hu, Qiang, Zhu, Yuelong, Hu, Hexuan, Guan, Zhuang, Qian, Zeyu, Yang, Aiming
Format: Article
Language:English
Subjects:
Atmospheric Sciences
Civil Engineering
Earth and Environmental Science
Earth Sciences
Environment
Flood control
Flood management
Flood mapping
Flood predictions
Flooding
Floods
Geotechnical Engineering & Applied Earth Sciences
Historic floods
Historic sites
Hydrogeology
Hydrology/Water Resources
Imagery
Kernels
Learning
MODIS
Performance evaluation
Predictions
Remote sensing
Support vector machines
Susceptibility
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Identifying flood prone areas is essential for basin management. In this paper, a spatial prediction technology of flood susceptibility based on multiple kernel learning (MKL) is proposed. We establish the flood susceptibility model by using EasyMKL, nonlinear MKL (NLMKL), Representative MKL(RMKL), Generalized MKL(GMKL), support vector machine(SVM) with linear kernel and SVM with Gaussian radial base function(RBF) kernel, The spatial prediction of flood susceptibility in Sanhuajian basin of the Yellow River is carried out. We use MODIS remote sensing images to obtain historical flood inundation sites in the study area. Then, ten flood conditioning factors are used as inputs to the flood susceptibility model. The model performance is evaluated in terms of accuracy (ACC), balanced F Score (F1 score), and areas under the curve (AUC). According to the results, MKL significantly outperforms the SVM adopting single kernel, and NLMKL(ACC=0.833,F1=0.841,AUC=0.889) demonstrates the best comprehensive performance. The flood susceptibility map generated by MODIS remote sensing images and MKL, therefore, can provide effective help for researchers and decision makers in flood management.
ISSN:0920-4741
1573-1650
DOI:10.1007/s11269-021-03010-2