This week I’ll be starting work on my first independent data science project. After quite a few rabbit hole sessions throughout the internet, I’ve finally settled down on a topic: I’ll be exploring the correlations between incidents of armed conflict and measures of climate change.
For some background, here’s a short video describing a Stanford study on the topic, published last year. You can find links to an article about the study and the study itself at the end of this post.
Disclaimer: just want to put out there that I’m neither an expert on climate change nor on international security studies. My main goal with this project is to apply my geographical expertise to a topic I’m curious about and for which I could find relevant, public data. Any and all feedback from people who know more about this than me is very welcome!
I’ll be combining two datasets for this project. To measure armed conflict I‘ll be using the Uppsala Conflict Data Program [UCDP] Georeferenced Event Database covering 1946 to the present. This dataset includes over 225,000 incidents of armed conflict, including measures like starting and ending dates, parties involved, lat- and longitude, type of violence, and number of fatalities.
I‘ll be combining this with the Global Historical Climatology Network [GHCN] dataset which includes daily global climate measurements such as rainfall, temperatures, snowfall, average wind speed, etc., taken from more than 100.000 stations around the world.
Fine-grained Spatial Data
Both datasets include fine-grained spatial references of the measured variables. For the GHCN climate data, the exact latitude and longitude of the measurement stations are provided. The UCDP conflict data also includes geo-references for each event, although the accuracy and precision of these spatial indicators varies. A categorical variable is included that measures the precision with which the coordinates and location assigned to the event reflect the location of the actual event, ranging from exact location is known and coded, to event happened within 25km, to only the country where the event took place is known.
With this data, I hope to be able to construct a model which precisely maps the two phenomena by their respective locations, thereby going beyond the country-level descriptions of conflict we are used to hearing or reading about in the news. The contribution of my project, I hope, will be in the fine-grained detail of the analysis, allowing us to notice new patterns in the data. To do all this, I’ll be diving into the GeoPandas package.
Questions to Explore
But for the moment, I’ve got more questions than answers. For example:
- How will I deal with the differences in precision regarding the geo-spatial references of the two datasets, i.e. at what spatial scale will I aggregate the data?
- How will I account for the fact that climate change is a planetary phenomenon while the data is measured locally?
- And of course the biggest question of all: can I even say anything about causality?
As the experts of the Stanford study point out, while there seem to be indications that incidents of armed conflict are increasing as Earth’s climate becomes more extreme, it is very difficult to draw a direct causal link between the two phenomena. What we should strive for, then, is “[u]nderstanding the multifaceted ways that climate may interact with known drivers of conflict”, such as socioeconomic development, the strength of government, inequalities in societies, and a recent history of violent conflict.
So to say something about causality, I will have to include data on these ‘known drivers of conflict’ as well. Let’s see how I far this rabbit hole goes!
This story is the first in a linked series documenting my progress through my first independent data science project. Find the next story here.
Does climate change cause armed conflict? | Stanford News
Intensifying climate change will increase the future risk of violent armed conflict within countries, according to a…