Introduction
For this lab, I was interested in exploring a dataset about human-created hazards impacting birds. In past projects, I have used bird sighting data from eBird to map trends in bird watching. Inspired by the recent death of Flaco, the owl who escaped the Central Park Zoo in 2023 and lived in Central Park for one year until colliding with a building in the Upper West Side this February, I started to research bird building collisions and played around with a small dataset. As I continued researching, I eventually found the Federal Aviation Administration, which maintains a Wildlife Strike Database that contains records of reported wildlife aircraft strikes since 1990. According to the FAA, wildlife aircraft collisions have been increasing, due to a combination of increased bird populations and aircraft traffic, “a trend toward faster and quieter aircraft,” and increased efforts to report wildlife strikes. While the FAA reports that somewhere between 9,000 – 13,000 birds are struck by aircraft annually, the actual numbers are most likely much higher, since there is no mandate to report.
Inspiration
Searching for graphs and charts on this topic yielded some interesting results. This chart from Cornell University was especially intriguing, and I liked the addition of the plan and bird flock imagery. This chart shows a correlation between bird migration patterns and collisions.
This chart (which was also made in Tableau and with the same dataset) visualized collisions by time of day and month:
This chart shows the proportion of damage/no damage reported:
There are several issues with this chart. While I would assume that the X axis is years and the Y axis is number of strikes, they are unlabelled. I also do not think that the stacked bar chart shows a compelling relationship between the amount of damage reported.
This chart incorporates a nice use of color as well as icons of each bird species:
I like the design and message of this chart. The direct labelling, colors, iconography, and incorporation of the FAA hazard level rating, are all very effective.
Methodology
The Wildlife Strike Database allows you to filter by state, airport name, date range, aircraft type, and more. I filtered to New York state and left all other filters blank, fetching data for the entire date range of 1990 – present and producing a CSV with 14,775 rows. The dataset is very detailed and allows you to look at what part of the aircraft was struck or damaged, the level of damage, the cost of repairs, altitude, phase of flight, weather conditions, time of day, number of human injuries, and more. Before importing in Tableau, I used OpenRefine to tidy my dataset by eliminating columns I wouldn’t use and cleaning up inconsistencies.
After importing my data into Tableau, I spent a while generating different charts to see what aspects of the dataset would be the most interesting to pull out. For example, I was interested in seeing which airports and airlines experienced the most collisions. However, to make these charts meaningful I would need to add an additional dataset containing number of flights.
I was also interested in exploring the relationship between time of day and collisions. While it is clear from the chart below that more collisions occur during the daytime, I believe this analysis would also be made more meaningful by adding data that accounts for the number of flights as well as quantity of migrating birds.
First, I was curious about the general trend in amount of collisions over the years. Other than a dip in 2020 (which I’m confident is due to the decrease in flights that year), there is a general upwards trend in bird collisions.
When I added in the repairs cost data, it was clear that something interesting was going on in 1995 and 2009, as the spikes in the graph were not proportionate to the amount of collisions. Digging a little deeper into the dataset showed that there were several collisions in 1995 which amounted to a high cost of repairs, but only one incident in 2009 that was accounting for the majority of the repairs cost.
It was clear from the line graph of repairs cost that most repairs made to an aircraft after collisions are minor, but there were several incidents that resulted in very high repair costs. Following this line of inquiry, I charted out the damage rating data. The FAA damage rating includes None, Undetermined, Minor, Substantial, and Destroyed. Most collisions (94%) were rated as having caused minor or no damage, 3.5% caused an undetermined amount of damage, and 1.8% caused substantial damage. In the 34 years of recorded collisions in New York state, only 4 incidents resulted in a damage rating of Destroyed (according to the FAA data dictionary, Destroyed means that the damage sustained made it inadvisable to attempt to restore the aircraft).
One of these 4 incidents was the so-called “Miracle on the Hudson River,” which was when US Airways Flight 1549 struck a flock of geese and was forced to make an emergency landing in the Hudson River. The incident occurred on January 15, 2009 and resulted in an insurance payment of 36 million dollars.
After researching this incident more and learning about how its aftermath resulted in a targeted extermination of 70,000 birds in an effort to reduce aircraft collisions, I became interested in using the dataset and Tableau to tell a story from a point of view that emphasizes the cost to wildlife rather than the monetary cost to airline companies. Aircrafts and buildings are man-made creations that interfere with birds’ ecosystems and natural migration patterns. It is clear from analyzing collisions by month and phase of flight that the highest amount of collisions occur at low altitudes (where most birds fly) and during migration season.
Reflection
This lab was incredibly useful in getting started with using Tableau, and I hope to become more comfortable and make more detailed and effective visualizations with it. I was hoping to be able to incorporate imagery (like in the examples that inspired me), but I am not sure if this would be accomplished in Tableau or added afterwards in another software. While I believe I followed best practices for the use of color, it appears a little too simplistic and one-note to me.
I also would like to add some more narrative content to the end of the Data Story about the aftermath of the January 2009 incident. I know there is the option to add a block of text to the data story, but I would like to explore other ways to thoughtfully incorporate text and narrative into the dashboards/data stories without clogging it up.
In my opinion, this lab accomplishes a basic level of data analysis. More robust and meaningful analysis could be achieved by incorporating different and/or multiple data sources. This dataset and research would be informative and useful for bird and wildlife advocates, airline companies, aircraft manufacturers, environmental advocates, policy makers, and more. Studying the patterns and factors involved in aircraft wildlife collisions are important to protecting humans as well as animals and their environments.