Due to an increasing occurrence of natural hazards, such as floods, a significant number of lives are lost each year worldwide. The risk of experiencing catastrophic losses from flooding is exacerbated due to the changing climate, and the increasing anthropogenic activities. Consequently, predicting the conditions leading to fatalities is crucial in the assessment of flood risk. However, the existing modeling capabilities in this field, are limited, emphasizing the critical need for the development of such tools. Here, we show that the occurrence of flood fatalities can be estimated using a random forest (RF) algorithm applied to nine explanatory variables characterizing each fatality. Furthermore, by converting the RF model outcomes into a user-friendly tool, it is possible to predict the probability of the occurrence of flood-related fatalities, based on variables describing hazard intensity and the environmental and sociodemographic conditions that contribute to such events. Our results represent an initial attempt towards a predictive model of flood fatalities in the Italian context. They reveal the key factors that together influence flood fatalities, enabling the prediction of such occurrences. These findings can serve as a foundational framework for quantitatively assessing the risk to the population from such events and as a valuable resource for identifying strategies to mitigate flood risk.
An empirical flood fatality model for Italy using random forest algorithm
Mina Yazdani;Daniela Molinari
2023-01-01
Abstract
Due to an increasing occurrence of natural hazards, such as floods, a significant number of lives are lost each year worldwide. The risk of experiencing catastrophic losses from flooding is exacerbated due to the changing climate, and the increasing anthropogenic activities. Consequently, predicting the conditions leading to fatalities is crucial in the assessment of flood risk. However, the existing modeling capabilities in this field, are limited, emphasizing the critical need for the development of such tools. Here, we show that the occurrence of flood fatalities can be estimated using a random forest (RF) algorithm applied to nine explanatory variables characterizing each fatality. Furthermore, by converting the RF model outcomes into a user-friendly tool, it is possible to predict the probability of the occurrence of flood-related fatalities, based on variables describing hazard intensity and the environmental and sociodemographic conditions that contribute to such events. Our results represent an initial attempt towards a predictive model of flood fatalities in the Italian context. They reveal the key factors that together influence flood fatalities, enabling the prediction of such occurrences. These findings can serve as a foundational framework for quantitatively assessing the risk to the population from such events and as a valuable resource for identifying strategies to mitigate flood risk.File | Dimensione | Formato | |
---|---|---|---|
1-s2.0-S2212420923005903-main.pdf
accesso aperto
Descrizione: main paper
:
Publisher’s version
Dimensione
2.87 MB
Formato
Adobe PDF
|
2.87 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.