Climate, Vol. 11, Pages 229: Flood Hazard Assessment in Australian Tropical Cyclone-Prone Regions

This study investigated tropical cyclone (TC)-induced flooding in coastal regions of Australia due to the impact of TC Debbie in 2017 utilising a differential evolution-optimised random forest to model flood susceptibility in the region of Bowen, Airlie Beach, and Mackay in North Queensland. Model performance was evaluated using a receiver operating characteristic curve, which showed an area under the curve of 0.925 and an overall accuracy score of 80%. The important flood-influencing factors (FIFs) were investigated using both feature importance scores and the SHapely Additive exPlanations method (SHAP), creating a flood hazard map of the region and a map of SHAP contributions. It was found that the elevation, slope, and normalised difference vegetation index were the most important FIFs overall. However, in some regions, the distance to the river and the stream power index dominated for a similar flood hazard susceptibility outcome. Validation using SHAP to test the physical reasoning of the model confirmed the reliability of the flood hazard map. This study shows that explainable artificial intelligence allows for improved interpretation of model predictions, assisting decision-makers in better understanding machine learning-based flood hazard assessments and ultimately aiding in mitigating adverse impacts of flooding in coastal regions affected by TCs.

Apr 23, 2024 - 23:04
 0  145
Climate, Vol. 11, Pages 229: Flood Hazard Assessment in Australian Tropical Cyclone-Prone Regions

Climate, Vol. 11, Pages 229: Flood Hazard Assessment in Australian Tropical Cyclone-Prone Regions

Climate doi: 10.3390/cli11110229

Authors: Michael Kaspi Yuriy Kuleshov

This study investigated tropical cyclone (TC)-induced flooding in coastal regions of Australia due to the impact of TC Debbie in 2017 utilising a differential evolution-optimised random forest to model flood susceptibility in the region of Bowen, Airlie Beach, and Mackay in North Queensland. Model performance was evaluated using a receiver operating characteristic curve, which showed an area under the curve of 0.925 and an overall accuracy score of 80%. The important flood-influencing factors (FIFs) were investigated using both feature importance scores and the SHapely Additive exPlanations method (SHAP), creating a flood hazard map of the region and a map of SHAP contributions. It was found that the elevation, slope, and normalised difference vegetation index were the most important FIFs overall. However, in some regions, the distance to the river and the stream power index dominated for a similar flood hazard susceptibility outcome. Validation using SHAP to test the physical reasoning of the model confirmed the reliability of the flood hazard map. This study shows that explainable artificial intelligence allows for improved interpretation of model predictions, assisting decision-makers in better understanding machine learning-based flood hazard assessments and ultimately aiding in mitigating adverse impacts of flooding in coastal regions affected by TCs.

What's Your Reaction?

like

dislike

love

funny

angry

sad

wow

Lema Kefyalew Studying a Ph.D. in Civil Engineering (Hydraulic Engineering). Studying Master of Project Management (MPM). Graduated with MSc in Civil Engineering (Hydraulic Engineering). Followed a 12 month Software Engineering program at ALX_Africa (Holberton school). Graduated with BA degree in Management. Graduated with BSc degree in Civil Engineering.