Integration of RS-GIS with Frequency Ratio, Fuzzy Logic, Logistic Regression and Decision Tree Models for Flood Susceptibility Prediction in Lower Gangetic Plain: A Study on Malda District of West Bengal, India

Flood is one of the most commonly occurring natural calamities in the lower Gangetic flood plain region of India. Hence, the present study aims to compare the predictive performance of frequency ratio (FR), fuzzy logic (FL), logistic regression (LR) and decision tree (DT) models for assessing the flood susceptibility in the Malda district situated in lower Gangetic plain of the state of West Bengal of India. Furthermore, the study also depicts the role of different causative factors in manoeuvring flood havoc in this region. A total of 128 flood locations were identified of which 70% were randomly used for data training and remaining 30% were used for data validation. Ten flood causing variables including geology, DEM (Digital Elevation Model), distance from river confluence, distance from river, distance from fault line, rainfall, NDWI (Normalized Difference Water Index), TWI (Topographic Wetness Index), flow accumulation and curvature were determined to construct the spatial database. Collinearity among causative variables was measured using VIF (variance inflation factors) and TOL (tolerance). Subsequently, geographical information systems (GIS) and different predictive models were ensembled together to assess flood susceptibility. Lastly, validation results obtained from ROC (receiver operating characteristics) curve shows that flood susceptibility maps prepared by LR and DT models are much accurate in their prediction as compared to other models. Hence, we believed that flood susceptibility assessment using comparative predictive models is probably the first of its kind in this region and hence, can turn out as a potential resource for future planning and development in the study area.