4.1.1 Managing your own data

Collecting good quality data is a key element of any seabird project and is a fundamental aspect of research and conservation work. In this manual we have discussed a wide range of methodologies used to collect data on your target seabird populations. It’s at this point that you may find yourself staring at your reams of acoustic files, tracking points and survey sheets and thinking, well now what?

It may seem like there is no such thing as too much data, and broadly speaking that’s true. However, if you are not managing your data properly you can find yourself quickly swamped, making it difficult to drill down to the data you really need and the analyses you should be focusing on. This becomes even more true if you are using multiple methods – cameras, song meters, drones, auditory surveys, tracking devices etc. – with data in different formats. Below are some considerations for how to manage your data properly, from collection through to the end product. While we cannot provide detailed guidance on how to set up a robust and efficient data management system for every project, there are a few key steps that you should consider.

• Understand your data collection needs before you start. Before embarking on a project, step back and consider the core focus of what it is you are trying to achieve. Then assess exactly what data you require to tackle these objectives. This is a key component of study design and ensures that you are collecting all the data you need (there is nothing more frustrating than being a year into your project and then realising you have not collected a key piece of data) and that you aren’t wasting valuable time collecting data that you won’t actually use.
• Ensure your data collection methods are clear. When collecting data in the field, make sure everyone is on the same page about what needs to be collected, and exactly how that data is collected. For example, if you are measuring morphometrics on captured birds, make sure that everyone knows what you mean by ‘beak depth’ or ‘flattened wing chord’. If you are carrying out nocturnal surveys and counting bird calls, make sure your definition of what you are considering a single call is clear (e.g. a call is a single unbroken note or series of notes). If you are mapping distribution and have several types of polygons (i.e. ‘hotspot heavy’, ‘hotspot light’, ‘hotspot ground calling’), make sure your team has a clear definition of those polygon types. Clear and concise definitions lead to good quality data collection.
• Don’t wait to digitise! If you use field notebooks, make sure the field data is digitised as soon as possible; don’t wait until the end of the season to start digitising. If you can’t, then at least make a PDF copy so that data collected earlier will not be easily lost. By that point, you probably won’t remember what some of your notes meant (especially cryptically written notes taken at 2am), and your notebooks may be covered in mud or be wet, or they may go missing altogether. Once a field trip is over, digitise! Also, consider collecting data digitally in the field on mobile devices, which can by uploaded when you return to your base of operations. There are multiple apps (both open source and proprietary) that are now available for data collection, and these make the whole step of digitising field notes redundant.
• Digitise only what you will use. There is often the impulse to digitise and collect everything when it comes to data. This might not actually be necessary. Consider, for example, a project that has 50 burrow cameras, with cameras set to take photos whenever there is movement in front of the burrow. If these cameras are deployed throughout the breeding season, the number of photos being taken will rapidly build up – even more so during the pre-fledging period when a chick may spend hours outside a burrow exercising and thus ramping up the photos. A project such as this will have hundreds of thousands of photos to digitise and this can be truly overwhelming. In this scenario the impulse may be to try to collect data on every single photo, but do you really need to? If you are assessing the impact of predator control on seabird fledging success, you may only need to digitise predator visitations and confirm chick fledging dates, rather than collecting data on every entry and exit of each burrow by an adult bird. Narrowing this down saves time and allows for a more focused analysis. Another option to consider are cloud-based repositories and processing facilities for your camera trapping images (e.g. https://docs.gbif.org/camera-trap-guide/en/).
• Design and use a good database. Choosing the right database is key. This is where your data will be stored, and where you will be going to ask questions and understand relationships about your data. An unsuitable database will create extra work - making it hard to find the data you are looking for or answer even simple questions. Carefully consider your data needs, and resources available to you. Do you have funds to purchase something proprietary or should you go the open-source route (e.g., Microsoft Access vs. PostgreSQL)? Do you work with a lot of spatial data and want to automate spatial queries and store spatial data (i.e. PostgreSQL with PostGIS or Oracle Spatial)? Does your project have the personnel expertise to implement powerful solutions, or do you need something simple and userfriendly out of the box to just answer questions (e.g. Metabase)? This step is so important that it is worth consulting professionals to assist with the design. Avoid organic solutions of storing data from different years, species, colonies in separate files hoping that they will be interoperable and always remain in the same folder – ideally all of the data should be contained in a single (or very few related) database(s).
• Back up your data! It may seem obvious, but make sure you back your data up frequently, to multiple sources, one of which includes a cloud-based solution. Don’t store everything on a single laptop, as inevitably it will crash, swirling your data away into the ether, never to be seen again. Have an external hard drive (or several) kept somewhere away from your computer or office database or data server. These hard drives can be small and portable, so you can also take them into the field with you if you are on extended field trips with base camps. For larger teams, use an office server with regularly scheduled backups to keep your data safe, while also frequently backing up to a cloud-based storage service. Cloud back-up solutions are not only safe storage places, but they also offer data processing that would be very laborious to do by hand. For example, seabird tracking data can be safely stored on both Movebank (https://www.movebank.org) and the BirdLife Seabird Tracking Database (www. seabirdtracking.org), which facilitates analysing the data via R packages and annotating data with environmental remote-sensing products. Raw sound recordings can be safely stored on ecoSound-web (https://ecosound-web.de/ecosound_web/collection/index/118) where you can also automatically run recognizer algorithms (BirdNet) over your data to extract calls of target species, which can also be done for camera trap data on several online repositories (e.g. https://docs.gbif.org/camera-trap-guide/en/).
• Automatically produce reports and validate. Most projects will need to report summaries of what they have found to partners, funders, government authorities etc. Creating such reports can be a significant burden, but a well-designed database can facilitate the automated creation of reports at regular (e.g. quarterly, annually) intervals (e.g. https://github.com/steffenoppel/GoughReports). These automated reports are not only useful for external reporting, but also immediately flag up potential errors in the database for data validation. For example, a plot of reproductive success over time that lists a value of >100% for one year will immediately indicate an error in the data, whereas finding the one erroneous entry where somebody accidentally entered an incorrect value will be much harder to detect. Producing and inspecting reports at regular intervals therefore allows you to efficiently assure the quality of your data.