Evaluation of the Flooding Susceptibility in the Telvar River Basin Using the Evidence Weight Models and the Bayesian Logistic Regression

Abstract

Flooding is one of the most dangerous and common natural disasters, which yearly causes many damages and causalities for communities and the environment. A comprehensive flooding assessment and management is essential to reduce the effects of flooding on people's lives and livelihoods. Due to an increase in flooding in recent years, locating and identifying flood-prone areas is very important to predict this environmental disaster as mis-identification of flood-prone areas in a watershed may have devastating effects. The main purpose of this study was to evaluate the performance of the Bayesian Logistic Regression (BLR) and the weight-of-evidence (WOE) models for preparing flooding susceptibility maps in the Talvar Watershed of Province of Kurdistan. The geographical location of 93 flood prone areas identified in the watershed was randomly divided into one group (70%) for calibration and another (30%) for validation. Both models consider ten effective factors in causing flooding, namely: slope, slope direction, curvature, digital elevation model, distance from stream, topographic wetness index (TWI), stream power index (SPI), rainfall amount, geology, and land use. According to the WOE model, about 35.8% of the area is placed in the medium to very high hazard class, and based on the Bayesian logistic regression model, about 45.08% of the area is placed in the medium to very high hazard class. The relative performance detection curve was used to validate the flooding potential maps. Even though both models offered sufficient accuracy, the higher accuracy was assigned to the BLR model (93.4%). Therefore, the BLR model has better performance than the WOE model in terms of the flooding risk potential.

Keywords

• bayesian logistic regression mode
• talvar watershed
• flood