Introduction
Every year, millions of people[1] from across the world participate in some form of birdwatching. This can range from admiring the birds in your own back yard, to embarking on an adventure to spot particular species. Whatever your style, it’s important to know both where and when you might expect to find a particular bird, and this map is here to help. Now wherever you travel in New Zealand, at whatever time of year, you know what to look out for when you have your eyes to the sky.
Data
The data for this project was sourced from the New Zealand Bird Atlas, part of the Cornell Lab of Ornithologies eBird project. Although it is possible to download the entire eBird dataset, at 137GB the choice was made to limit to only New Zealand birds, and only for the calendar year 2021 (to account for seasonal movement). All data was sourced by volunteer submission and vetted by the eBird staff, and is amazingly rich. Beyond location and species, it features data including observed behaviour, observation lengths, and interactions with other birds.
The data was imported directly into Tableau with no alterations made. Since it was already broken down by ID, Latitude, Longitude, and Time, Tableau was able to visualise the data with relative ease. Two calculated fields were then created. The first was taking the natural log of the number of birds observed in each event. The data was heavily skewed to lower values but some outliers (including one instance of 500,000) existed and would prove otherwise difficult to visualise. Secondly the birds were aggregated into their taxonomic order, again to aid with visual fidelity. This aggregation was done by hand (due to the relatively small number of entries), and orders with few members were pulled together as “Other”. Kiwi (Aptyerygiformes) was the notable exception, as despite having low numbers they are of cultural significance to New Zealand and thus likely to be looked for.
When exploring the visualisations that were available with the data, most were based on only a single species of bird. This is useful when creating a distribution or migratory pattern, but this project looked to take a step back and try to show a wider variety of data at a glance. The New Zealand Bird Atlas has the points aggregated to 10km2 blocks, which allows for a finer level of detail than aggregating by region, but still lacks fidelity to explore the data presented. Inspiration was drawn from maps of shipping routes, wherein land boundaries can be made out without having them drawn, and the possibility of birds revealing similar geographic information was intriguing.
Results
Within tableau, each event was mapped by Latitude/Longitude, with the points coloured by taxonomic order and sized to the number of birds in the event (see previous section), and filtered to show one month at a time that the user could click through. Tooltips were added to show the basic details of each observation, and a dropdown filter was added to allow the user to view only a specific species of bird. Some consideration was given whether to use scientific or common name for this filter, and it was decided that birdwatchers would prefer the added clarity of scientific nomenclature. Ideally the filter would present both to be consistent with other formatting.
Each event was kept as an individual point for several reasons. Firstly, birds aren’t confined to human geographic boundaries and trying to constrain them as such seems disingenuous. Second, New Zealand has a large number of endangered bird species that can only be found in very specific locations (the dataset has already screened out entries that might pose a hazard). Third, a significant portion of the entires are noted as being over the water, and so would not easily be mapped onto geopolitical boundaries.
The map chosen was Tableau’s built in one, provided by map box. As the data was not aggregated to to any boundaries, the map was stripped back to simply the base layer. Satellite imagery was a consideration but in testing it became difficult to distinguish the datapoints from the map details. The difference between light and dark mode is still contentious.
Discussion
Overall I’m pretty happy with how this turned out. There were some ideas that became abandoned early on that could be worth exploring, such as animating the timeline or switching to a heat map/density based layout, but I wasn’t able to get either working in a state I was happy with. There is also a tension between how the map looks when fully zoomed out vs. zoomed in that I haven’t fully resolved, especially when it comes to the size of the points. Perhaps a more dynamic option could be explored.
Going forward, I would be interested in exploring this concept on a global scale. In theory, nothing presented here would break down with additional data, except the need to expand the list of orders (eBirds does provide this data, albeit separately). Utilising the full breadth of the data available would also allow for more exploratory analysis, as it could show how bird populations may wax and wane over time, and also show migratory patterns. My initial assumption about the birds indicating geographic features seems to have some merit, especially with the shorebirds. All this would likely need to utilise different software however, as even with the single year of New Zealand data it was already beginning to feel sluggish.