Exploring the deaths by natural disasters since 1900, and possible relations to climate change.
Our whole process is documented in the jupyter notebook main.ipynb.
Here we can see the occurrences and deaths of disasters over time for each decade.
The number of disasters is rising. However, this is likely subject to a reporting bias, since likely just more disasters are reported. The deaths by disasters are sinking, we see that fewer apocalyptic disasters where millions of people die occur. However this could also be likely due to humans being better prepared for natural disasters, and not disasters being less severe.
Different types of disasters kill different amounts of people. Droughts and floods are by far the deadliest. A possible bias here could be that deaths by droughts and floods can be strongly dependent on political decisions, such as the Russian famine of 1921-1922, so floods & droughts might not inherently be more severe on their own.
Asia has by far the most deaths by disasters, total deaths as well as deaths by population.
Here we look at a small selection of countries that are different to each other e.g. in their region or how developed they are.
Usually, droughts are the deadliest disasters. However, France has nearly no deaths by droughts, but here heatwaves are the most deadly.
The temperature rises in all regions, with Europe being the most extreme.
We use temperature anomalies as a measurement for temperature since they are supposedly more representative. Temperature anomalies are simply the temperature, but corrected by a baseline, in our case by the average temperature from Jan 1951 to Dec 1980. Additionally, we smoothed the temperature with lowess.
Here we can see the correlation of deaths by natural disasters with the temperature, foreach type of natural disaster.
For earthquakes the correlation is very small at 0.028, for extreme temperature events it is relatively high at 0.8.
However, when looking closer at specific disaster types in specific countries, e.g. heatwaves deaths in France, the correlation drops to 0.18, and in Spain even to -0.49.
This kinda contradicts scientific evidence which suggests that heatwaves increase in severity.
There is also the effect that when looking at the country level and disaster subtypes just very few recorded instances remain, and outlines are quite influential (thus the negative correlation in Spain).
So here more research is necessary to reach substantial conclusions.
Since 1900 the number of recorded disasters rises, and the number of deaths by disaster is sinking.
When looking at the deaths from natural disasters there is a strong difference between different disaster types, and between different regions and countries.
There seems to be a correlation between some natural disaster types and rising temperatures, but further research is necessary for this area.
Overall everything should be taken with a grain of salt. Because there are quite some biases, like better reporting of natural disasters, better preparation for natural disasters, and too few instances to derive meaningful results.
More details are available in our jupyter notebook main.ipynb.
One dataset that we use for our analysis, the Emdat dataset, is not allowed to be shared publicly. Thus for reproducing, it has to be manually downloaded.
First, create an account and download the dataset from the Emdat website.
Next put the dataset to the following location, with the exact filename:
data/raw/disaster/emdat.xlsx. We used the version from 2022-12-22, but newer versions should also be fine, since we exclude entries after 2020.
Finally, you can run the jupyter notebook main.ipynb to reproduce our results. Don't forget to check if you have the correct python version as specific in the notebook.
This project was created by students as part of a lecture at the Vienna University of Technology.
Sebastian Fürndraht, Hannes Rokitte, Paul Schmitt, Lukas Wieser · 24.01.2023