EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

From Flash Flood Vulnerability and Risk Assessment to Property Damage Prediction: the Value of Machine Learning

Atieh Alipour, Peyman Abbaszadeh, Ali Ahmadalipour, and Hamid Moradkhani
Atieh Alipour et al.
  • Department of Civil, Construction and Environmental Engineering, The University of Alabama , AL, USA

Flash floods, as a result of frequent torrential rainfalls caused by tropical storms, thunderstorms,
and hurricanes, are a prevalent natural disaster in the southeast U.S. (SEUS), which frequently
threaten human lives and properties in the region. According to the U.S. National Weather
Service (NWS), flash floods generally initiate within less than six hours of an intense rainfall
onset. Therefore, there is a limited chance for effective and timely decision-making. Due to the
rapid onset of flash floods, they are costly events, such that only during 1996 to 2017 flash
floods imposed 7.5 billion dollars property damage to the SEUS. Therefore, estimating the
potential economic damages as a result of flash floods are crucial for flood risk management and
financial appraisals for decision makers. A multitude of studies have focused on flood damage
modeling, few of which investigated the issue on a large domain. Here, we propose a systematic
framework that considers a variety of factors that explain different risk components (i.e., hazard,
vulnerability, and exposure) and leverages Machine Learning (ML) for flood damage prediction.
Over 14,000 flash flood events during 1996 to 2017 were assessed to analyze their characteristics
including frequency, duration, and intensity. Also, different data sources were utilized to derive
information related to each event. The most influential features are then selected using a multi
criteria variable selection approach. Then, the ML model is implemented for not only binary
classification of damage (i.e., whether a flash flood event caused any damage or not), but also for
developing a model to predict the financial consequences associated with flash flood events. The
results indicate a high accuracy for the classifier, significant correlation and relatively low bias
between the predicted and observed property damages showing the effectiveness of proposed
methodology for flash flood damage modeling applicable to variety of flood prone regions.

How to cite: Alipour, A., Abbaszadeh, P., Ahmadalipour, A., and Moradkhani, H.: From Flash Flood Vulnerability and Risk Assessment to Property Damage Prediction: the Value of Machine Learning, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-19057,, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 1 – uploaded on 18 Apr 2020
  • CC1: Comment on EGU2020-19057, Chiara Arrighi, 06 May 2020

    Hi, thanks for your material. What kind of info were available for algorithm training? I understood duration and intensity are responsible of higher losses, aren't they?

    • AC1: Reply to CC1, Atieh Alipour, 06 May 2020

      Hi, thanks for the comment. We used different information of each event including the topographic features of the inundated area, vulnerability indices, duration, intensity, time, and median home value to train the algorithm and predict the potential flash flood property damage. Yes, as the intensity and duration of flooding increase, the amount of damage gets higher (please see Figure 5).