Researchers at the Helmholtz Centre for Environmental Research (UFZ) have conducted a study to better understand the factors that contribute to extreme forms of flooding in order to improve flood risk assessment and limit flood damage. Using explainable machine learning methods, the researchers analyzed over 3,500 river basins worldwide and examined flood events between 1981 and 2020. They found that precipitation, soil moisture, snow depth, and air temperature all play significant roles in the development of floods.
The study revealed that a combination of factors often leads to more extreme flooding events, with 51.6% of cases being caused by at least two factors. The researchers also observed that floods become more extreme when more factors are involved, with flood complexity increasing as the flood magnitude rises. In some cases, three or even all four factors were jointly responsible for a flood event. The findings indicate that the interaction between multiple factors can greatly impact the severity of a flood.
Different regions exhibit varying degrees of flood complexity based on the dominant factors contributing to floods in those areas. River basins in northern regions of Europe and America, as well as in the Alpine region, have low flood complexity due to snow melt being the primary cause of most floods. Conversely, regions such as eastern Brazil, the Andes, and eastern Australia have high flood complexity, with multiple flooding mechanisms at play. Factors such as climate and land surface conditions also influence flood complexity within a particular river basin.
The researchers used explainable machine learning to predict and analyze flood events, quantifying the contribution of different variables to the size and intensity of floods. This approach allows for a better understanding of the relationships between flood drivers and run-off during a flood event. By identifying the relevant driving factors and combinations thereof, the researchers hope to improve predictions of future flood events, particularly extreme floods that are challenging to estimate accurately.
The study findings are expected to enhance flood prediction capabilities by providing a more detailed understanding of the factors influencing flood events. The researchers believe that their methodology can help estimate extreme floods more accurately by considering the specific contributions of each driving factor. By taking into account the interactions between different factors and their impact on flood magnitude, the researchers aim to improve flood risk assessment and better prepare for and mitigate the devastating consequences of flooding events.
