After beavers were almost wiped out in the early 19th century, efforts to reintroduce these large rodents began in 1922. These attempts were successful, and beavers can now be found regularly in many places.
However, much of what we know about beavers comes from a time when they were able to spread virtually unchecked. Today, this is no longer so easy. In the Swiss Plateau, for example, territories are very close together in many places and migrating beavers do not find it so easy to find their own territory. Given this increased territorial density, the question arises as to whether our knowledge of beavers still corresponds to our observations.
Our main focus is on family size and age structure, as both factors are extremely important for an accurate assessment of population size. If family size has adapted to high population densities, this could have a significant impact on population estimates.
To count the beaver families, we use camera traps that provide us with photos of their tail flukes. These are similar to fingerprints and are unique to each individual beaver. The photos enable us to distinguish between the animals and count them. We can also measure the tail flukes to estimate the age of each individual.
The camera traps are mounted at a height of 80 cm using a custom-made frame. To ensure that they can take sharp images even at short distances, we attach small dioptres in front of the camera lenses.
The camera traps are located in areas where we expect high activity. These include areas between two bodies of water or near a beaver lodge.
Since there is no truly suitable software solution for our data, I began working on automating the methodology step by step and optimising it in as many areas as possible during my bachelor's degree. During my master's degree, I developed a more or less independent end-to-end solution. BeaverFindr allows you to download photos directly from the camera trap and edit them accordingly. BeaverFindr provides you with a range of tools to evaluate all data as efficiently as possible. The programme consists of four modules.
An initial preselection takes place in ‘Procezr’. Here, your photos are grouped into events and you can select all the images that are important for further analysis. In order to be able to identify individual beavers, we usually use images of their tails.
These are now automatically forwarded to the ‘Recognizr’ module, where an algorithm can recognise and measure the tail. Calibration images can be used to keep the measurements as accurate and comparable as possible. These images can also be used to correct any perspective distortions.
The results of the measurements and a picture of the tail are then sent to Matchr. There, the images are finally assigned to individuals. Using an algorithm, the programme automatically suggests which photos belong together, greatly facilitating the identification of individual animals. To do this, the software uses both the pattern and scars on the tail and the measurements.
Once all cameras and territories have been evaluated, the data can be analysed using Analyzer. To do this, all individuals are first assigned to a territory based on their observations. They can then be assigned to an age category based on the measurements of their tail fan. The result is a range of statistics and a cleaned data table for advanced analysis. In this context, we are also in contact with other research groups, for example to model spatial capture-recapture with our data.
