EGU23-12245, updated on 26 Feb 2023
https://doi.org/10.5194/egusphere-egu23-12245
EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Projecting the occurrence of extreme heat-related mortality using long short-term memory networks in cities of Switzerland

Saeid Ashraf Vaghefi1, Veruska Muccione1, Ana Vicedo-Cabrera2,3, Raphael Neukom4,5, Christian Huggel1, and Nadine Salzmann4,5
Saeid Ashraf Vaghefi et al.
  • 1University of Zurich, Department of Geography, Switzerland (saeid.vaghefi@geo.uzh.ch)
  • 2Institute of Social and Preventive Medicine, University of Bern, Bern, Switzerland
  • 3Oeschger Center for Climate Change Research, University of Bern, Switzerland
  • 4WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, Switzerland
  • 5Climate Change, Extremes and Natural Hazards in Alpine Regions Research Center CERC, Davos Dorf, Switzerland

Climate change increases the frequency and severity of heat waves, which can negatively impact human health. Extreme heat can lead to heat stroke, dehydration, and other heat-related illnesses. Heatwaves are more severe for vulnerable populations such as older adults, young children, and people with pre-existing medical conditions. In this study, we analyze the occurrence of compound extreme heat-related mortality in five Swiss cities using neural networks.

To define the excess mortality due to compound heat extremes (Hot day, Tmax>30oC, followed by a tropical night, Tmin>20oC) we compared mortality during the four hot summers of 2003, 2015, 2018, and 2019 with long-term average mortality rates (1981-2020). We trained long short-term memory (LSTM) neural networks on 40-year time series of maximum and minimum temperatures, hot day / tropical night compound events, and mortality in Basel, Bern, Geneva, Lugano, and Zürich.  LSTM neural networks learn the important parts of the sequence seen so far and forget the less important ones. This makes these models predict with greater accuracy than traditional time series analysis methods.

In general, we found that over the past 40 years, more than six percent of deaths were caused by compound extreme heat waves in the five Swiss cities. Geneva and Lugano are the most affected cities by compound heat, but the risk of heat-related mortality has decreased in these two regions over time, which could be a result of the action plans that exist in the Latin regions of Switzerland.

We further used Switzerland's future climate model scenarios (CH2018), to predict mortality rates in Swiss cities in the near-future (2020–2050) and far-future (2070–2100). We projected that the number of people affected by mortality risks associated with heat could increase by three folds by the end of the century in most cities if no further adaptation is taken place.

Our results show how important it is for governments, public health agencies, and individuals to be aware of the potential impacts of climate change on heat-related mortality and to take steps to mitigate and adapt to these impacts.

How to cite: Ashraf Vaghefi, S., Muccione, V., Vicedo-Cabrera, A., Neukom, R., Huggel, C., and Salzmann, N.: Projecting the occurrence of extreme heat-related mortality using long short-term memory networks in cities of Switzerland, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-12245, https://doi.org/10.5194/egusphere-egu23-12245, 2023.