"Flood risk modeling in southern Bagmati corridor, Nepal" (a study from Sarlahi and Rautahat, Nepal). (December 2022)

• Record Type: Journal Article
• Authors: Shreevastav, Bitu Babu; Tiwari, Krishna Raj; Mandal, Ram Asheshwar; Singh, Bikram
• Is Part Of: Progress in disaster science. Volume 16(2022)
• Publication Date: 2022-12
• Subjects: Natural hazard -- Machine learning -- MaxEnt -- Risk mapping; Disasters -- Periodicals; Disaster relief -- Planning -- Periodicals; Emergency management -- Periodicals; 363.3405
• DOI: 10.1016/j.pdisas.2022.100260
• Languages: English
• ISSNs: 2590-0617

Abstract

Abstract: Flooding is underlying major natural hazard of Nepal and hence socio-economic loss has been paid annually by this hazard. Planning and management of flood based on study is still lack in Nepal. So, the particular study is carried out to assess the flood risk modeling in lower Bagmati river region in Eastern Terai. In this study, total 10 geospatial environment layers and past flood inventory from field were used to run the machine learning model i.e., MaxEnt for risk modeling of flood. The past flood data were separated into 75% for model building and 25% for model validation. The land use land cover change showed the highest contribution (40.8%) to the flood while the lowest contribution was of slope only 0.2%. 9% of total population were in high risk of flood while 39% population were in very low risk. Figure no. 13 shows that 5% of total household were in high risk, 55% were in moderate risk and 20% were in very low risk of flood. Out of total study area about 2.66% of the total area is in very high-risk zone to flood. High risk zone is found to be 4.89%, where as 9.48%, 20.61% and 62.36% are moderate, low, and very low risk zone area. In terms of AUC values, acceptable results were obtained for the test data with 0.931 and the standard deviation 0.019. The values of Area Under Curve (AUC) range from 0.7 to 0.8 and interpreted as fair or good. Finally, this research could directly help in policy, planning, framework, and programming of development intervention to tackle with flood hazard. Highlights: Research's objective was to identify flood risk-prone areas & modeling for predicting flood hazards to minimize flood effects. Flood risk prediction mapping, flood events were studied; geographical locations of previous flood occurrences determined. Total of 124 flood occurrence data were taken randomly. Collection of flood sites was separated into training & validation groups. A training group of 93 flood locations (75%) and a validation group of 31 flood locations (25%) were classified randomly. Flood incidents were given as a raster network (10m*10m) that used in a maximum entropy model. Total 10 geospatial environment layers and past flood inventory from field were used to run the machine learning model i.e., MaxEnt for risk modeling of flood. The past flood data were separated into 75% for model building and 25% for model validation. As per the result, the land use & land cover change showed the highest contribution (40.8%) to the flood while the lowest contribution was of slope only 0.2%. 9% of total population were in high risk of flood while 39% population were in very low risk. The 5% of total household were in high risk, 55% were in moderate risk and 20% were in very low risk of flood. Out of total study area about 2.66% of the total area is in very high-risk zone to flood. High risk zone is found to be 4.89%, where as 9.48%, 20.61% and 62.36% are moderate, low, and very low risk zone area. In terms of AUC (Area Under Curve) values, acceptable results were obtained for the test data with 0.931 and the standard deviation 0.019. The AUC values range from 0.7 to 0.8 and interpreted as fair or good. Research findings could help in policy, planning, framework and programming of development intervention to tackle with flood hazard.