Machine-supported monitoring of plant communities and habitats

In Switzerland, 48% of habitats and more than one third of all species are threatened. Nevertheless, only a small proportion of them are adequately monitored, as the reliable identification of habitat types and their associated species is time-consuming and requires expert knowledge. Our goal is to expand habitat monitoring in Switzerland by developing algorithms that can semi-automatically identify plant communities from videos and assign them to habitat types.

While the rapidly growing popularity of citizen-science reporting platforms has led to an enormous increase in species observations, the majority of these observations are concentrated on a small number of charismatic species and easily accessible locations. In the SpeciesID project, we have additionally demonstrated the potential of modern machine learning for species identification by developing image- and location-based identification tools for plants, butterflies, fungi, and moths. These tools are playing an increasingly important role in supporting biodiversity monitoring and reducing biases in the resulting data.

Here, we lay the foundations for rapid monitoring of plant communities and habitats. Specifically, we aim to optimize vegetation surveys by developing the algorithms for a smartphone app that can (1) derive preliminary vegetation surveys from videos, (2) suggest species that are expected but have not yet been observed, and (3) propose suitable habitat types. All developments will be made freely available through the WSL-hosted SpeciesID API.