Could human losses have been lower during the 2021 flood in the Ahr Valley, Germany? A study based on agent-based modelling of alternative early warning and evacuation scenarios

Código

I-EBHE0133

Autores

André Felipe Rocha da Silva, Julian Cardoso Eleutério, Heidi Kreibich

Tema

WG 1.04: From local to large scale human-water dynamics

Resumo

Between July 12 and 19, 2021, a quasi-stationary atmospheric low-pressure system named "Bernd" caused intense precipitation on already saturated soil, leading to severe flooding in Germany, Belgium and the Netherlands. This catastrophic event resulted in 190 fatalities in Germany, surpassing the total number of victims in the past 40 years. The Ahr River valley in Rhineland-Palatinate was particularly affected, with approximately 42,000 residents impacted, around 8,800 buildings damaged, and 134 fatalities recorded. The flood in the Ahr River valley significantly exceeded the scenarios outlined in official hazard maps, leaving decision-makers and the public unprepared. This lack of preparedness led to substantial issues with the content, issuance, and dissemination of warnings, thereby reducing the effectiveness of emergency response. We evaluate how human losses in the Ahr Valley might have differed under alternative flood early warning and evacuation (FEWE) scenarios, using outputs from the agent-based model LifeSim. This probabilistic life-loss estimation model incorporates uncertainties in several input parameters, such as the stability criteria thresholds for people and buildings, drowning due to submergence, and FEWE times. In order to run the LifeSim model for the Ahr flood in 2021, we utilised a reconstructed modelled time series of water depth and flow velocities and estimated the FEWE timeline based on a post-event survey with the affected population. For the reconstructed FEWE timeline, we identified the first flood warning approximately 13 hours before the peak of the flood upstream of the simulated domain; only 17.5% of those affected received a warning with evacuation instructions; most were warned or became aware of evacuation necessities only after flooding had already reached them, so that only about 34% of the population evacuated their homes or were rescued. The mean life-loss estimation of the reconstructed flood overestimated the actual life-loss by 30%. Simulations of alternative FEWE scenarios indicated a potential life loss reduction of up to 80% with timely warning dissemination and increased population evacuation. However, scenarios in which the FEWE prompted the population to evacuate around the moment of the imminent hazard at their buildings resulted in higher human losses. In these cases, vertical evacuation within buildings would have been more effective. The use of a life-loss agent-based model, such as LifeSim, can support decisions on FEWEs, and can help to improve emergency response planning.