Forecasting of flash flood susceptibility mapping using random forest regression model and geographic information...

Area of Science:

• Environmental Science
• Geographic Information Systems (GIS)
• Climate Change Research

Background:

• Flash floods are increasing due to climate change, impacting communities and environmental management.
• Accurate flood susceptibility mapping is crucial for effective disaster mitigation and urban planning.

Purpose of the Study:

• To develop a Flood Susceptibility Map (FSM) for the Ibaraki prefecture, Japan.
• To investigate and predict flood-prone areas using advanced modeling techniques.

Main Methods:

• Utilized a Random Forest (RF) regression model integrated with GIS.
• Incorporated 11 environmental variables (e.g., elevation, slope, land cover, proximity to water bodies and roads).
• Validated the model using Receiver Operating Characteristics (ROC) curve analysis with 224 flood/non-flood instances.

Main Results:

• The FSM revealed that approximately two-thirds of Ibaraki prefecture has low to very low flood susceptibility.
• Approximately one-fifth of the region was identified as having high to very high flood susceptibility.
• The RF model demonstrated high accuracy with an Area Under the ROC Curve (AUC) of 99.56%.

Conclusions:

• The developed FSM is a vital tool for policymakers in Ibaraki prefecture.
• The map supports informed spatial planning and the implementation of targeted flood mitigation strategies.
• This research contributes to climate change adaptation and sustainable land management practices.