Biod.AI.versity Observation & Integration (Bio.AI)

Biod.AI.versity Observation & Integration (Bio.AI)

Dr. Jana Waeldchen

Our mission

Our goal is to simplify, accelerate and increasingly automate global biodiversity monitoring through automated species identification in citizen science projects and remote sensing. Our specifically developed AI technology for species identification provides large amounts of data for this, and targeted science communication additionally contributes to raising public awareness for the protection of biodiversity.

Focus areas of the Bio.AI research group:

The Bio.AI research group combines three focus areas that each complement each other:

  • Automated species identification
  • Biodiversity monitoring
  • Citizen Science/Science communication

Automated species identification

Within this research focus, we develop and use innovative methods of artificial intelligence to automate species identification. As part of our research activities, the Flora Incognita app was developed. To date, the app can identify 4800 wild plants of Central Europe simply by taking photos in the field, resulting in many millions of plant observations including geolocation. By 2022, the app had been installed more than 5 million times.

Building on this successful framework, we are working on the automatic identification of other life forms such as phytoplankton, fungi and butterflies.

Key Publications:

Michael Rzanny et al. (2022)
Image-based automated recognition of 31 Poaceae species: The most relevant perspectives
Frontiers in Plant Science,12: 804140. doi: 10.3389/fpls.2021.804140

Roelvan Klink et al. (2022)
Emerging technologies revolutionise insect ecology and monitoring
Trends in Ecology & Evolution. doi: 
10.1016/j.tree.2022.06.001

Michael Rzanny et al. (2019)
Flowers, leaves or both? How to obtain suitable images for automated plant identification
Plant Methods 15, 77. doi: 10.1186/s13007-019-0462-4

Susanne Dunker et al. (2018)
Combining high‑throughput imaging flow cytometry and deep learning for efficient species and life‑cycle stage identification of phytoplankton
BMC Ecology 18, 51. doi: 10.1186/s12898-018-0209-5

 

Biodiversity Monitoring

The Bio.AI research group aims to simplify, accelerate and increasingly automate global biodiversity monitoring. Therefore, in this research area, we are currently focusing on the question of the extent to which automated species identification in combination with Citizen Science and remote sensing can contribute to biodiversity monitoring. The steadily growing number of plant occurrence data via the Flora Incognita app already provides us with important information:

  • When and where do which species flower? 
  • How much do the characteristics of a plant species vary? 
  • How do the composition of plant communities and the locations of plants change in connection with climate change and the type of underlying land use? 

Key publication

Miguel D. Mahecha et al. (2021)
Crowd-sourced plant occurrence data provide a reliable description of macroecological gradients
Ecography (Editors‘ Choice). doi: 10.1111/ecog.05492

 

Citizen Science & Science communication

Biodiversity monitoring data are often collected by trained persons, and usually only reach a small temporal and spatial coverage. Another limiting factor is the growing public loss in species knowledge. With the developement of automated species idenfication tools it is now possible that people without any species knowledge can collect biodiversity data. Those rarely follow a strict protocol regarding where, when, and how to record a species - resulting in a set of unstructured data.

Within the Bio.AI research group, we run two Citizen Science projects to collect plant species occurences: The Flora Incognita project, which collects opportunistic plant observations, and the Flora Capture project, where Citizen Scientists record plant findings in a set of predefined perspectives, resulting in a set of structured images. 

In order to reach as many people as possible to use our apps and collect plant occurrence data, we invest significant resources into science communication:

  • Social Media channels (Twitter, Instagram, Facebook, Youtube)
  • Flora Incognita website and blog
  • Exhibitions and events (MS Wissenschaft, Long Night of Science)
  • Traditional PR like interviews, podcasts, articles

Key publications

Patrick Mäder et al. (2021)
The Flora Incognita app – Interactive plant species identification
Methods in Ecology and Evolution. 12: 1335– 1342; doi: 10.1111/2041-210X.13611

David Boho et al. (2020)
Flora Capture: a citizen science application for collecting structured plant observations,
BMC Bioinformatics 21, 576; doi:10.1186/s12859-020-03920-9

Projects

    
Flora Incognita - more than plant identification

Flora Incognita - more than plant identification

In order to make change/loss of biodiversity visible, species knowledge is key - not only among experts, but for anyone. We developed the mobile app “Flora Incognita” that leverages modern computer vision techniques such as deep neural networks with a "connected data" approach, using site information (e.g. phenology, location, date and time) and plant morphological traits for semi-automatic species identification. The collected data are used to answer questions of plant species monitoring and phenology. 
The app was released in 2018 and has reached 5M downloads by 2022. It is freely available for iOS and in the Google Play Store. More information can be found on the Flora Incognita website.
NaturaIncognita-creating a workflow platform for AI-based species identification
To address the loss of biodiversity, data on the state and change of biodiversity is needed across many life forms. In this “AI lighthouse” project, the existing frameworks for automated species identification via the "Flora Incognita" project are to be expanded to realise automatic identification for other species groups.
The project's workflow platform will provide a species identification service based on the latest machine learning methods. Pilot projects already include phytoplankton, butterflies and funghi.  more
planktAI
Automatic identification of microscopic algae for biodiversity assessment and trophic status estimation more
AI4Biodiv - Artificial intelligence in biodiversity research

AI4Biodiv - Artificial intelligence in biodiversity research

The project is about creating different benchmark data sets, further developing training algorithms for automatic species identification, interpreting the results and using contextual information in the recognition process. In the area of species and population monitoring, training algorithms and network architectures are applied and further developed specifically for remote sensing data. In modelling, AI-based models are developed that address both the spatial and temporal context at different scales.
BetterWeeds
The aim of the project is to develop a framework for sustainable weed management using autonomous weed detection, AI-based identification of weed species, and geo-referenced weed distribution maps taking site-specific characteristics of the individual field into account. Based on these maps, management strategies for weed control on the respective fields will be developed and tested under field conditions. more
  

Key Papers

Journal Article (19)

1.
Journal Article
Jana Wäldchen, Hans Christian Wittich, Michael Rzanny, Alice Fritz, and Patrick Mäder, "Towards more effective identification keys: A study of people identifying plant species characters," People and Nature (2022).
2.
Journal Article
Roel van Klink, Tom August, Yves Bas, Paul Bodesheim, Aletta Bonn, Frode Fossøy, Toke T. Høye, Eelke Jongejans, Myles H.M. Menz, Andreia Miraldo, Tomas Roslin, Helen E. Roy, Ireneusz Ruczyński, Dmitry Schigel, Livia Schäffler, Julie K. Sheard, Cecilie Svenningsen, Georg F. Tschan, Jana Wäldchen, Vera M.A. Zizka, Jens Åström, and Diana E. Bowler, "Emerging technologies revolutionise insect ecology and monitoring," Trends in Ecology and Evolution 37 (10), 872-885 (2022).
3.
Journal Article
Kevin Karbstein, Salvatore Tomasello, Ladislav Hodac, Natascha Wagner, Pia Marinček, Birthe Hilkka Barke, Claudia Paetzold, and Elvira Hörandl, "Untying Gordian knots: unraveling reticulate polyploid plant evolution by genomic data using the large Ranunculus auricomus species complex," New Phytologist 235 (5), 2081-2098 (2022).
4.
Journal Article
Negin Katal, Michael Rzanny, Patrick Mäder, and Jana Wäldchen, "Deep learning in plant phenological research: A systematic literature review," Frontiers in Plant Science 13, 805738 (2022).
5.
Journal Article
Bernhard Schmid, Martin Schmitz, Michael Rzanny, Michael Scherer-Lorenzen, Peter N. Mwangi, Wolfgang W. Weisser, Andrew Hector, Roland Schmid, and Dan F. B. Flynn, "Removing subordinate species in a biodiversity experiment to mimic observational field studies," Grassland Research 1 (1), 53-62 (2022).
Show more
Follow link for a complete list of publications from the Department of Biogeochemical Integration more
Go to Editor View