Background:

High elevation plant communities are rapidly changing due to climate warming, with species potentially changing their ranges in elevation. Pollinators respond to changes in flowers' density and type, however it is difficult to anticipate future changes in the structure of alpine interaction networks. Currently, information on plant-pollinator interactions is often based on direct observations by researchers in the field, which often provides only limited data. The availability of affordable digital video recording devices and advances in computer vision has recently sparked the development of several automated pollinator monitoring systems. When coupled with powerful machine learning algorithms, these systems can enable the identification of floral visitors from images, but their application remains limited. In this project we will test and further develop methods for observing pollinators in the field, to study community changes, species interactions networks and ecosystem function.

About the Project:

Goals of the Project:

Our goals for this project are twofold:

1. To advance the functionality of automated monitoring systems in a novel environment
   • We used the methodology outlined by Droissart et al. 2021 as a guide to building a standardizable insect camera trap
2. To use our data to answer a biological question:
   • Does the floral neighborhood (i.e. the plants that surround the plant of interest) impact the visitation rates of pollinators differently for generalist versus specialist flower species?

Field Site

The field site where we collected our images was located in the central Swiss Alps at the Alpine Research Station Furka (ALPFOR) at Furka Pass (46.58° N, 8.42° E) in Southeastern Switzerland. Our field site sits at approximately 2400m asl between the Cantons of Uri and Valais. Here, the weather is representative of high alpine conditions, which can provide unique challenges for camera systems used in monitoring including sudden summer storms, high winds, hail, direct sun and even late season snow. We set up six plant interaction camera traps (PICT) in two sites: one at 2200m and one at 2400m in order to monitor the pollinator visitation rates for a generalist (Leontodon helveticus) and specialist (Campanula barbata and C. scheuchzeri) flowers.

Your role!

Now that we have collected our field images with the cameras, we are recruiting help from the Zooniverse community to begin to annotate and build a database of insect identities captured by our cameras. Your job will be to identify the location and type of insect in the frames, which can then be used to train an artificial intelligence (AI) model to automate the rest of the footage captured in the field. We have collected millions of frames during our field work, and by eventually automating the identification process, we hope to increase the amount of usable data significantly, leading to more robust and comprehensive analyses of this system.

Project Stages:

Affiliations

• Swiss Federal Research Institute WSL
• Swiss Data Science Center
  

Funding
This project was financed through the WSL internal innovative project funding scheme.