Biodiversity Seminar: Towards real-time conservation management in all parts of the world

Accurate and up-to-date knowledge about biodiversity is essential for effective biodiversity conservation. Current methods to monitor biodiversity, however, assume strict and well standardized sampling protocols and are sensitive to errors in data, while high quality biodiversity data remains difficult to obtain for many parts of the world. As a result, our knowledge about biodiversity is heavily skewed towards regions with highly developed infrastructure. A major focus of our research is to change that narrative and work towards enabling real-time conservation management in all parts of the world, for which we develop dedicated analysis tools that are robust to variation in the sampling process. I will illustrate three such efforts: 1) The first is a novel method to infer population trends from noisy count data that does not require modeling of variation in detection probabilities across sites. 2) The second is a novel method to infer species distributions from camera trapping data. Such data is often analyzed with occupancy models, which make the strong assumption about closed patches and summarize the species distribution by a simple presence-absence matrix that ignores differences in population densities. We address these issues by modeling capture rates instead of occupancy, which allows us to characterize niche partitioning also for sympatric species. 3) The third effort is regarding the use of Artificial Intelligence (AI) to speed up the labeling process of biodiversity records such as camera trapping images. Many efforts are underway to train AI models for species identification, but they fail to reach the accuracy required for downstream analysis. To enable their use, we therefore develop downstream models that properly learn and account for AI classification errors. Together, these efforts allowed us to enable biodiversity monitoring with relatively fast turnaround times in the Aire de Conservation de Chinko, a huge and understudied but difficult to access region in the Central African Republic.

Daniel Wegmann

I’m Daniel Wegmann, a computational biologist at the University of Fribourg and specialized in the development of probabilistic machine learning methods for noisy data. A major focus of my research is on genetic data and in particular ancient DNA data. Since ten years, however, I’m also heavily involved in the conservation of biodiversity in the Central Africa Republic. What immediately struck me when starting to work in this challenging region was the disconnect between established best practices for biodiversity monitoring and the research realities on the ground, as well as how lousy leading global models on biodiversity predicted what we found. It’s time to change both.

Our seminars are hybrid. Please send an email to events-biodiversity(at)wsl.ch if you would like to get access to the live stream.

The Biodiversity Seminar Series are organized by the WSL Biodiversity Center. Every two weeks, we aim to host a seminar speaker that presents research or outreach on topics relevant to the biodiversity community at WSL. The seminars are public and are usually broadcasted online.

To find out more about the WSL Biodiversity Center and a complete list of events, please visit our website. 

Please send an email to events-biodiversity(at)wsl.ch if you would like to be updated on the activities of the WSL Biodiversity Center.