AUTHOR=Pandey Manish , Arora Aman , Arabameri Alireza , Costache Romulus , Kumar Naveen , Mishra Varun Narayan , Nguyen Hoang , Mishra Jagriti , Siddiqui Masood Ahsan , Ray Yogesh , Soni Sangeeta , Shukla UK 

TITLE=Flood Susceptibility Modeling in a Subtropical Humid Low-Relief Alluvial Plain Environment: Application of Novel Ensemble Machine Learning Approach

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 9 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2021.659296

DOI=10.3389/feart.2021.659296

ISSN=2296-6463

ABSTRACT=Study region: The study is based on the sub-tropical Middle Ganga Plain (MGP) of central India having a high risk of flood susceptibility. 
Study focus: Using highly accurate flood inventory and 12 flood predictors (FPs), two machine learning (ML) ensemble models constructed by applying frequency ratio (FR) and evidential belief function (EBF) with classification and regression tree (CART): CART-FR and CART-EBF, this study aims to quantitatively compare the performance of two machine learning ensemble models, one first time built and the other one used in other natural hazards but not for floods, in mapping the flood susceptible zones in the subtropical fluvial basin of the Middle Ganga Plain.
Hydrological insights: The MGP is an ideal natural laboratory that allows testing all the genres of susceptibility prediction models. Randomly generated flood and non-flood points, apportioned in a 70:30 ratio for training and validation of the ensembles and different evaluation matrices suggest that CART-EBF (AUCSR=0.843; AUCPR=0.819; AUCSR & AUCPR refer to the area under the curve – success rate and prediction rate respectively) was better performing than CART-FR (AUCSR=0.828; AUCPR=0.802) which shows more accuracy in mountainous terrain. The result of this study indicates that both the ensembles used delineate flood susceptible zones in low-latitude, subtropical monsoonal regions like MGP with fairly good accuracy and precision, the application of CART-EBF used in the study for the first time needs to be tested in different topoclimatic settings for viability.