TY - JOUR
T1 - Novel hybrid models by coupling support vector regression (SVR) with meta-heuristic algorithms (WOA and GWO) for flood susceptibility mapping
AU - Rezaie, Fatemeh
AU - Panahi, Mahdi
AU - Bateni, Sayed M.
AU - Jun, Changhyun
AU - Neale, Christopher M.U.
AU - Lee, Saro
N1 - Funding Information:
This research was supported by the Basic Research Project of the Korea Institute of Geoscience and Mineral Resources (KIGAM), Project of Environmental Business Big Data Platform and Center Construction funded by the Ministry of Science and ICT.
Publisher Copyright:
© 2022, The Author(s), under exclusive licence to Springer Nature B.V.
PY - 2022/11
Y1 - 2022/11
N2 - Schools as social bases and children’s centers are among the most vulnerable areas to flooding. Flood susceptibility mapping is very important for flood preparedness and adopting preventive plans for reducing the school vulnerability to flooding. To achieve this, there is a need for the models that can be used in vast areas with high predictive accuracy. This study aims to develop the innovative hybrid models by coupling the support vector regression (SVR), statistical approaches, and two meta-heuristic algorithms, whale optimization algorithms (WOA) as well as grey wolf optimizer (GWO). According to the proposed methodology, a hybrid feature of SVR and frequency ratio (FR-SVR) is optimized by applying the GWO and WOA optimization algorithms to generate the maps related to flood susceptibility. The method was utilized for the Ardabil Province located in southwestern Caspian Sea precincts of which faced devastating floods. The GIS database including 147 ground control locations of flooded zones and nine factors which influence flood were utilized to learn and ascertain the validity of the models. Three statistical metrics namely, mean absolute error (MAE), root mean square error (RMSE), and the area under the receiver operating characteristic curve (AUC) were computed for the developed models in order to estimate prophetically. The results indicated that the meta-optimized FR-SVR-GWO as well as FR-SVR-WOA models exceeded the FR-SVR and FR models in training (RMSEFR-SVR-WOA = 0.2016, RMSEFR-SVR-GWO = 0.1885, AUCFR-SVR-WOA = 0.87, AUCFR-SVR-GWO = 0.88) and validation (RMSEFR-SVR-WOA = 0.2025, RMSEFR-SVR-GWO = 0.1986, AUCFR-SVR-WOA = 0.87, AUCFR-SVR-GWO = 0.87) phases. The FR-SVR-WOA and FR-SVR-GWO models were very competitive regarding AUC and RMSE values, but the FR-SVR-WOA model reproduced greater flood susceptibility rates and was considered for identifying vulnerability of schools to flood events. To this end, number of schools, number of students in each school, and the area of the school building were taken into account to generate the vulnerability map. The results demonstrated that schools with the highest and lowest vulnerability to flooding were mostly located in southeastern and central parts of the Ardabil Province, respectively.
AB - Schools as social bases and children’s centers are among the most vulnerable areas to flooding. Flood susceptibility mapping is very important for flood preparedness and adopting preventive plans for reducing the school vulnerability to flooding. To achieve this, there is a need for the models that can be used in vast areas with high predictive accuracy. This study aims to develop the innovative hybrid models by coupling the support vector regression (SVR), statistical approaches, and two meta-heuristic algorithms, whale optimization algorithms (WOA) as well as grey wolf optimizer (GWO). According to the proposed methodology, a hybrid feature of SVR and frequency ratio (FR-SVR) is optimized by applying the GWO and WOA optimization algorithms to generate the maps related to flood susceptibility. The method was utilized for the Ardabil Province located in southwestern Caspian Sea precincts of which faced devastating floods. The GIS database including 147 ground control locations of flooded zones and nine factors which influence flood were utilized to learn and ascertain the validity of the models. Three statistical metrics namely, mean absolute error (MAE), root mean square error (RMSE), and the area under the receiver operating characteristic curve (AUC) were computed for the developed models in order to estimate prophetically. The results indicated that the meta-optimized FR-SVR-GWO as well as FR-SVR-WOA models exceeded the FR-SVR and FR models in training (RMSEFR-SVR-WOA = 0.2016, RMSEFR-SVR-GWO = 0.1885, AUCFR-SVR-WOA = 0.87, AUCFR-SVR-GWO = 0.88) and validation (RMSEFR-SVR-WOA = 0.2025, RMSEFR-SVR-GWO = 0.1986, AUCFR-SVR-WOA = 0.87, AUCFR-SVR-GWO = 0.87) phases. The FR-SVR-WOA and FR-SVR-GWO models were very competitive regarding AUC and RMSE values, but the FR-SVR-WOA model reproduced greater flood susceptibility rates and was considered for identifying vulnerability of schools to flood events. To this end, number of schools, number of students in each school, and the area of the school building were taken into account to generate the vulnerability map. The results demonstrated that schools with the highest and lowest vulnerability to flooding were mostly located in southeastern and central parts of the Ardabil Province, respectively.
KW - Flood susceptibility map
KW - Frequency ratio
KW - Grey wolf optimizer
KW - Iran
KW - SVR
KW - Whale optimization algorithm
UR - http://www.scopus.com/inward/record.url?scp=85132243730&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85132243730&partnerID=8YFLogxK
U2 - 10.1007/s11069-022-05424-6
DO - 10.1007/s11069-022-05424-6
M3 - Article
AN - SCOPUS:85132243730
SN - 0921-030X
VL - 114
SP - 1247
EP - 1283
JO - Natural Hazards
JF - Natural Hazards
IS - 2
ER -