Study reveals complexity behind global extreme flooding

Agency
Webp 8m1xlqqp0sh4jl3z2ypfp7shzh86
Dr. Naresh Kumar Executive Director For Atmospheric Sciences | Desert Research Institute Website

A recent study published in Science Advances has revealed that the risk of extreme floods in watersheds globally is complicated by the compounding effects of multiple flood drivers. The research, co-authored by Guo Yu, Ph.D., assistant research professor of hydrometeorology at DRI, emphasizes the complexity involved in understanding flood risks.

"This study demonstrates how machine learning can help disentangle the complex interactions of different flood drivers," said Yu. "We found that extreme floods can be caused by a combination of drivers with moderate magnitudes, rather than a single driver with extreme magnitude. This emphasizes the importance of understanding each individual flood driver in future research."

The study highlights the devastating consequences when rivers overflow their banks, as seen in North Rhine-Westphalia and Rhineland-Palatinate in 2021. Researchers from the Helmholtz Centre for Environmental Research (UFZ) have shown through explainable machine learning methods that floods become more severe when multiple factors contribute to their development.

Several factors such as air temperature, soil moisture, snow depth, and daily precipitation play crucial roles in flooding. UFZ researchers analyzed over 3,500 river basins worldwide from 1981 to 2020 and discovered that precipitation alone was responsible for only about 25% of nearly 125,000 flood events studied. Soil moisture was decisive in just over 10% of cases, while snow melt and air temperature were sole factors in around 3% of cases. Notably, 51.6% of cases involved at least two contributing factors.

Dr. Jakob Zscheischler from UFZ explained that floods become more extreme when more factors are involved: "We also showed that flood events become more extreme when more factors are involved." He noted that regions like eastern Brazil and parts of Europe have high flood complexity due to various flooding mechanisms.

The researchers used explainable machine learning to analyze potential flood drivers such as air temperature and soil moisture to predict run-off magnitude during floods. Dr. Shijie Jiang highlighted the importance of this methodology: "With this new methodology, we can quantify how many driving factors and combinations thereof are relevant for the occurrence and intensity of floods."

The findings aim to improve predictions for future extreme floods by providing a better estimation framework than current extrapolation methods based on less severe events.