November — December 2021
Exploring the Relationship Between CO2 Emissions and Natural Disasters
This was my final project for STAT 302: Data Visualization. The project was coded from scratch in R, and involved data cleaning, data wrangling, various visualization techniques and annotations, and coding interaction through shiny Studio.
My project investigates natural disaster data from NASA’s Earth Observing System Data and Information System (EOSDIS) and carbon dioxide emissions data from the World Bank in order to characterize the relationship between carbon dioxide emissions and the frequency and severity of natural disasters.
First, to get a general picture, let’s examine which kinds of natural disasters are most common in general. The above graph takes the sum of each type of natural disaster recorded in each region from 1970 to 1980. The barplot allows us to see differences in the types of natural disasters that occur most frequently in each region, but also allows us to identify patterns in which natural disasters occur most frequently globally. We can also see which regions are more prone to natural disaster.
The most common type of natural disaster across all regions was flooding, followed by storms. The next most significant category was epidemics, which disproportionately occurred in African countries. The top two categories could both plausibly be relevant to carbon dioxide emissions: flooding is related to rising sea levels and global warming, and storms may also be the cause of atmospheric changes due to greenhouse gas emissions.
Next, let’s get a sense of CO2 emissions for each region overall. Above, we see a line graph of CO2 emissions in metric tons per capita (per person in a country) for each region from 1970-2018. The value for each region per year is calculated by summing the emissions value for each country in that region for that year, but note that some missingness in data is to be expected for countries that did not record their CO2 emissions each year. I chose to use a line graph because this was the clearest way to show trends in the total emissions for each region over time, as we can easily see peaks and lows.
From this graph, we can see that the Americas clearly produced the most CO2 per year, though its emissions are roughly decreasing over time. Europe was also a high-emitter, but is experiencing a similar decline in emissions over time. Asia started out as a low-emitter, remaining at levels similar to Oceania and Africa until around 1996, but has been increasing gradually and surpassed the Americas around 2013.
Note that the effects of CO2 emissions have not been distributed equally amongst geographic regions. Despite producing relatively few emissions until recently, Asia experienced a significantly higher number of natural disasters in this time frame. For its minimal contribution to global CO2 emissions, Africa experienced the third most number of natural disasters during the recorded period. In contrast, Europe was the second highest-emitting region for the majority of this period and remains the third highest-emitting region, but experienced the second fewest number of natural disasters.
Severity of natural disasters over time
I created a shiny app to explore the frequency and severity of natural disasters across different regions and using different measures of severity.
An interactive visualization felt the most useful and widely-applicable as the app allows you to look at the data using different timescales and to select or deselect specific regions, countries, and disaster types of interest. Looking at the full range of years available, my data does indicate that the sheer number of natural disasters has been increasing over time: we see more points appear over time for all regions. However, has the severity of natural disasters also been increasing?
Because the number of disasters has been increasing, the number of people affected has naturally also increased. With the exception of Africa, which has seen more frequent disasters on smaller scales, other regions saw their costliest and deadliest disasters occur in more recent years (around or post-2000). This indicates that natural disasters have been increasing in severity over the years. Generally, earthquakes, floods, and storms seemed to be the deadliest; droughts and floods affected the most people in each incident; storms and floods were typically the costliest to recover from.
Based on my data, floods are one of the most destructive natural disasters. As a follow-up, I investigate the relationship between CO2 emissions and flood frequency by comparing the CO2 emissions graphic above to a histogram of the number of floods globally per year.
As we can see, the number of floods annually corresponds extremely well with global CO2 emissions, peaking around the same years that CO2 emissions peaked.
Click the image to access the shiny app.
The graphic above indicates that there is a link between CO2 emissions and natural disaster frequency.
Natural disasters in specific countries
Now, we can take a closer look at six countries to get a better understanding of the overall trends in the relationship between CO2 emissions and natural disasters. I selected at least one country from each region. My choice of an overlay graph allows for direct comparison between a country’s emissions and the impacts it experiences.
Algeria is the largest country in Africa; its CO2 emissions per capita are below average and its frequency of natural disasters is also below average.
Both Australia’s and the United Kingdom’s CO2 emissions per capita are extremely high – well above average – but their natural disaster incidence rates are extremely low.
China’s CO2 emissions per capita are quite high and steadily increasing (considering the size of China’s population, the country’s overall emissions should be amongst the highest), but it also experiences a large number of natural disasters.
The Philippines represents a different area within the Asian region and has an extremely low CO2 emission rate per capita, but a relatively large number of natural disasters.
Finally, the United States – like Australia and the United Kingdom – has a particularly high CO2 emission rate per capita. At the same time, it experiences far more natural disasters than they do.
Severity and cost of natural disaster recovery
I have established that there is a relationship between CO2 emissions and natural disaster frequency, but that CO2 emissions per capita are not a strong indicator of whether or not a region or country will actually suffer the consequences of emissions.
One of the key roadblocks to climate change action is the supposed cost of preventative measures. Has the cost of inaction been increasing over the years? I look at this cost both in monetary terms and in terms of the number of people affected: the number of individuals who lost their lives, were injured, rendered homeless, or overall impacted by the natural disaster.
The graph below allows us to examine the global costs in USD required to recover from damages related to natural disasters each year, alongside the total number of individuals affected by natural disasters that year.
We see that the monetary costs have increased enormously over time, as has the number of individuals affected; there is also a strong relationship between individuals affected and associated costs. We can also see that Asia is one of the regions that typically pays the most for damages but also has the largest number of individuals affected by natural disasters; in contrast, the Americas also pay a significant amount to recover from damages annually but residents are comparatively less affected. The high costs associated with natural disasters should thus serve as an incentive to invest in more preventative measures and emissions-reducing technologies.
Presence of associated disasters
Some natural disasters can be tied to subsequent disasters – such as how tsunamis are caused by earthquakes. Frequent associated disasters are another indication of natural disasters becoming more severe, and our ecosystems growing more fragile and volatile.
Here, I first explore the occurrence of associated disasters across all of my data (all regions, all years from 1970-2018) for each type of natural disaster. As could be predicted from my prior analyses which suggest that floods and storms are particularly destructive, these two types had the most incidences with associated disasters. Next, I look at the number of associated disasters each year for each natural disaster type. Notably, the number of associated disasters each year has grown dramatically in recent years, peaking in the mid 2000s and once again in the mid 2010s.
Seasonality of natural disasters
United States — click image to access shiny app
India — click image to access shiny app
Finally, I created a simpler shiny app that allows one to see when, throughout the year, each type of natural disaster occurred for individual countries. This tile graph allows us to see any abnormal years (for example, spike in flooding in the US around April 1991), seasonal patterns throughout the years, and which types of natural disasters are prevalent in each country.
To the left is the graph for the United States. Again, we see that the sheer frequency of natural disasters started increasing dramatically around 1988. Wildfires and droughts have become more common over time despite not being prominent natural disasters in the earlier years of the dataset, while storms and floods have continued to increase in frequency over time.
This can be contrasted to India’s natural disaster record, which includes more cases of extreme temperatures and drought as well as flooding at consistent times each Fall (likely a reflection of monsoon season). India’s natural disasters tend to occur mid-year or towards the end of the year.
Conclusion
Overall, when dealing with such a large dataset—including global data across several decades, with information about many different types of natural disasters, coupled with emissions data—I found that my ideas were best presented in simpler formats, such as bar graphs with information being conveyed through fill aesthetics.
I found that:
Floods and storms were the most common natural disaster across all regions.
CO2 emissions have generally increased for all countries in the time period studied (1970-2018). Natural disaster frequency and severity (in terms of people affected and cost to recover from damages) has correspondingly also increased.
Increases in CO2 emissions are correlated with increasing frequency of common types of natural disasters, such as floods.
However, CO2 emissions were not necessarily correlated with frequency of natural disasters in a particular country. Countries such as Australia and the United Kingdom had high CO2 emissions per capita but experienced relatively few natural disasters, while countries like the Philippines had a low carbon footprint per capita but suffered from many natural disasters.
Further, associated disasters—that is, disasters linked to a preceding disaster—have dramatically increased in frequency over time, particularly in relation to floods and storms. This once again demonstrates the increasing severity of natural disasters over time.
Each individual country exhibits unique seasonal patterns in natural disasters, but disaster frequency has increased for all countries over time and as such, many are facing types of disasters that they previously did not have to prepare for.
Citations
Dincer, Baris. All Natural Disasters 1900-2021. Earth Observing System Data and Information System. Kaggle, 2021. Web.
CO2 Emissions 1960 - 2018. World Bank Open Data. Kaggle, 2021. Web.