For this project, our team wanted to explore one of the ways that climate change is affecting the world as we know it: with warming sea temperatures and rising sea levels. As seasonal storms and flooding feel like they're getting worse every year, we decided to look for data that would examine the costs, both monetary and otherwise, of the way our planet is changing.
Our team entered this exploration with the goal of understanding the complex relationships between floods, people, and money. Some of the details we were interested in included specific flooding events, the causes behind those events, the monetary cost of the damage (in terms of flood insurance payouts), and the migration caused by flooding (in terms of people who lose their homes or are displaced by flooding). Our questions were:
Is it more common to deal with flooding events near oceans, due to rising sea levels?
Are floods becoming more common in general, due to climate change?
Are floods that affect large areas more devastating than those that hit smaller areas?
These questions proved almost impossible to answer conclusively, especially with the constraints of time and publicly-available data that we were able to use, but we were still able to make some connections within the bounds of this research project.
In order to answer our questions, we went hunting for official sources on climate change-related data, as well as flood-related data. While these questions could apply to the world at large, we decided to limit ourselves to the United States, a scope that could provide a large amount of data without being too overwhelming.
We found data from FEMA (the Federal Emergency Management Agency), EPA (the Environmental Protection Agency), and the Dartmouth Flood Observatory (described as "a global active archive of large flood events"). These datasets include information like the location of flooding events worldwide, anonymized information regarding specific flood insurance claims in the U.S., data on the average surface temperature of the ocean over time, and the number of natural disasters of several different types over time.
Links to our data can be found at the bottom of this page.
Once we found data that we thought could answer some of our questions, we created some visualizations to explore the topic. In order to process and visualize the data, we used OpenRefine, Tableau, QGIS, and ArcGIS Online.
Our research team decided to divide the visualizations among us, since we were unable to meet in person to work together. In order to make our visualizations look cohesive and part of the same project despite being worked on by different people in different programs and environments, our first step was to choose a color palette. We chose our colors to be primarily shades of blue, since blue is the color most associated with water; then we chose an accent color of yellow, which provides a contrasting tone in order to draw attention to important data points. This color palette complies with recent Web Content Accessibility Guidelines, meaning that each color contrasts with the others enough to be discerned even by people with visual impairments like colorblindness.
We also worked on creating a narrative surrounding our visualizations, by creating consistent tooltips and pop-ups, as well as editing the explanatory text that goes with the visualizations. This helped to emphasize each visualization as part of a larger whole.
Our visualizations and the conclusions we were able to form from them are detailed below.
This graph is a simple overview of the number of different types of natural disasters declared by FEMA from 1953 until now. Since our questions focused on floods, the flood line is highlighted in dark blue below.
Although the number of floods fluctuate over this period, it can be seen that the number of floods annually has increased in the past decade, with large peaks in alternate years. In addition, flooding that is caused by other weather phenomena such as hurricanes, coastal storms, etc are not included in these numbers, but have also been increasing over this period.
The significance of this increase is greater than the immediate damage of property and lives. Floods can cause drinking water to become contaminated, and with the water shortage that is already being predicted to happen as a result of climate change, this makes the problem even worse. They can also cause hazards such as disease-carrying animals and spills of chemicals or other hazardous materials. Therefore, even if the total number of floods that occur annually have not increased much, their frequency means that recovering from the previous flood and anticipating the next is going to occur in shorter and shorter intervals.
Another dataset we looked at shows the change over time in the surface temperature of all oceans worldwide. Here, we can see that from 1880 to 2015, sea temperatures have definitely been rising, especially in the last 20 years or so covered by the chart.
In this visualization, sea surface temperature is represented by a difference (in degrees Fahrenheit) from the global average sea temperature from 1971-2000, which is the conventional way to depict this data. (The global average sea surface temperature is about 60.9°F.) Sea surface temperature increased during the 20th century and continues to rise. From 1901 through 2015, temperature rose at an average rate of 0.13°F per decade. Sea surface temperature has been consistently higher during the past three decades than at any other time since reliable observations began in 1880. Based on the historical record, increases in sea surface temperature have largely occurred over two key periods: between 1910 and 1940, and from about 1970 to the present.
This chart also includes a band that indicates some uncertainty in measurement based on the instruments used to measure this temperature anomaly during the time. Sea temperature measurement technology has greatly improved since 1880, and increased frequency of measurement in the later decades means that the tracking of anomalies have gotten more accurate over time.
One dataset that came in handy with this project catalogued roughly 2.4 million flood insurance claims (anonymized to protect the privacy of claimants) over the course of the past 50 years. This data comes from the National Flood Insurance Program (NFIP). While we couldn't process all 2.4 million records, we used a randomized sample of about 10% of the data in order to generate the visualizations that follow.
Firstly, the map below shows the shape of the affected areas of various floods, colored to represent the number of flood insurance claims that came with those floods. This map spans the years from 1985 to 2020.
Lighter colors mean fewer insurance claims, while darker colors mean more claims. The overall shape of the polygons in the map represent the area covered by the flood, which is why there are shapes that overlap in places that are more prone to flooding. In addition, the yellow dots represent the severity of the floods: larger dots for more severe floods, smaller dots for less severe floods.
Interestingly, some of the floods that cover the largest areas have few insurance claims associated with them, while some smaller areas clearly impacted more properties with flood insurance. This would seem to answer one of our questions: floods that affect larger areas don't always cause more damage to people. Sometimes, the more concentrated flooding events hit the hardest.
In addition, there does appear to be a correlation between coastal areas and severe flooding events. While there are rainfall-caused floods towards the middle of the North American continent, many flood areas overlap in coastal regions. In particular, the Gulf of Mexico and the northeastern corner of the U.S. appear to have several flood events that also correlate with darker colors, meaning more flood insurance claims filed.
Similarly to the map above, the one below takes a look at the expenses of those flood insurance claims. Once again spanning from 1985 to the current year, this time the areas colored in darkest had the largest payouts, while lighter areas had smaller payouts resulting from their flood insurance claims.
This map makes it easy to see just how expensive flooding is, especially in coastal areas. Huge areas of the map are colored in dark shades, meaning that the average insurance payout from those floods was in the tens of thousands of dollars. Note that these are averages, not sums, which could mean that some areas colored in lightly might have had many insurance payouts of smaller sums, rather than a smaller number of insurance payouts in higher amounts.
Another dimension it's important to consider, perhaps even more important than the monetary cost of large flooding events, is the number of people who are affected by floods. The data we explored contains variables for the number of casualties associated with each flood, as well as people who lose their homes. These are known as "displaced people," and their lives are often affected for years after the flooding event that initially causes their displacement.
In the map below, the color of the flood area corresponds with the flood's severity: light colors mean a milder flood, and dark colors mean a more severe one. The size of the yellow dot corresponds to the number of people displaced by the flood.
Many of the largest yellow dots on this map are clustered in coastal regions, showing that flooding from the ocean tends to displace more people than flooding in landlocked areas. While this doesn't prove that sea level rise and climate change are responsible for all of these displacements, it's possible that that is one of the factors that has led to some of our country's most devastating floods in recent decades.
The following charts help to emphasize that the area that a flood affects doesn't always correspond to a higher expense in insurance payouts. The years 2005, 2012, and 2017 below have the highest total payouts in flood insurance; but they don't have anywhere near the highest affected areas for flood events.
The spikes in insurance claims can be explained by larger, well known flooding events in the country. 2005 was the year of Hurricane Katrina and the devastation of levees breaking in New Orleans. Due to the small area and the city being below sea level, it is clear that asset losses were at an all-time high along with the well publicized displacement of people. The 2012 peak could have been from the damage brought by Tropical Storm Lee in late 2011, from which water levels set a record at 42.66 feet in the relatively small area where the water stagnated. Finally, 2017 was the year of Hurricane Harvey, publicized as one of the costliest disasters in recent history. Each of these spikes points to the fact that the most water damage is done when the water stays in a smaller area, leading to higher levels and ultimately more loss of insured property.
We presented a preliminary iteration of our project to fellow researchers working on other projects. Based on the survey we conducted, users were satisfied with the visual elements, colors, and data used, and they felt the narrative provided enough context for the visualizations. We received compliments on our implementation of consistent color palettes, as well as the interesting theme and topic for our research.
On the other hand, some users felt confusion at the overlapping shapes in the maps above, and expressed that some of our visualizations were clearer than others. To that end, we edited the visualizations and accompanying text in an effort to clarify the meanings of the maps and other visualizations. It's our hope that this also brought to light the conclusions we were able to draw from the data.
The original datasets we used for our project can be found at the following links:
FEMA: https://www.fema.gov/openfema-dataset-disaster-declarations-summaries-v2
FEMA: https://www.fema.gov/media-library/assets/documents/180374
Flood Observatory: http://floodobservatory.colorado.edu/Archives/index.html
EPA: https://www.epa.gov/climate-indicators/climate-change-indicators-sea-surface-temperature
The original data files, as well as our modified files and workbooks that we used to create these visualizations, can be found here.
This data is very rich in information, and there is plenty more to explore relating to this subject.
Dartmouth Flood Observatory. Flood Observatory: Global Active Archive of Large Flood Events. Retrieved from http://floodobservatory.colorado.edu/Archives/index.html 27 April 2020.
EPA. Climate Change Indicators: Sea Surface Temperature. Retrieved from https://www.epa.gov/climate-indicators/climate-change-indicators-sea-surface-temperature 27 April 2020.
FEMA. FIMA NFIP Redacted Claims Data Set. Retrieved from https://www.fema.gov/media-library/assets/documents/180374 28 April 2020.
FEMA. OpenFEMA Dataset: Disaster Declarations Summaries. Retrieved from https://www.fema.gov/openfema-dataset-disaster-declarations-summaries-v2 27 April 2020.
Wikipedia, the free encyclopedia. Sea surface temperature. Retrieved from https://en.wikipedia.org/wiki/Sea_surface_temperature 5 May 2020.
Wikipedia, the free encyclopedia. Floods in the United States: 2001–present. Retrieved from https://en.wikipedia.org/wiki/Floods_in_the_United_States:_2001%E2%80%93present. 5 May 2020.