Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland

Smartphone apps and citizen science platforms have enabled large-scale public contributions to biodiversity monitoring, with image recognition readily allowing automated species identification. Yet, practical evaluations of the best use of such identification tools for nationwide plant community mon...

Full description

Saved in:
Bibliographic Details
Main Authors: Manuel R. Popp, Niklaus E. Zimmermann, Philipp Brun
Format: Article
Language:English
Published: Elsevier 2025-12-01
Series:Ecological Informatics
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1574954125003255
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849233582214610944
author Manuel R. Popp
Niklaus E. Zimmermann
Philipp Brun
author_facet Manuel R. Popp
Niklaus E. Zimmermann
Philipp Brun
author_sort Manuel R. Popp
collection DOAJ
description Smartphone apps and citizen science platforms have enabled large-scale public contributions to biodiversity monitoring, with image recognition readily allowing automated species identification. Yet, practical evaluations of the best use of such identification tools for nationwide plant community monitoring remain limited. We conducted a stratified field survey covering diverse habitats and different growth forms at 151 sites in all biogeographic regions of Switzerland, collecting > 5000 labelled photos of different plant parts for about 20% of the Swiss Flora. We assessed the performance of leading, publicly available identification services, differing in geographic scope and taxonomic resolution. Up to 85% of species observations were correctly identified when multiple images were supplied. While all identification providers correctly identified > 90% of the species at least once, we found models with a narrower geographic scope and coarser taxonomic resolution to reach higher identification accuracy. Identifications of photographs depicting multiple plant parts were often more successful than those of images focusing on single plant parts. While manual control remains indispensable for high-quality data, involving automated identification services can boost efficiency in compiling standardised floristic data, especially when (1) several images are taken, (2) an identification provider with matching taxonomic concepts and scope is selected, and (3) auxiliary information such as environmental context is considered.
format Article
id doaj-art-e31d957525cb4ed0b2a03a95e4d1ba96
institution Kabale University
issn 1574-9541
language English
publishDate 2025-12-01
publisher Elsevier
record_format Article
series Ecological Informatics
spelling doaj-art-e31d957525cb4ed0b2a03a95e4d1ba962025-08-20T05:05:43ZengElsevierEcological Informatics1574-95412025-12-019010331610.1016/j.ecoinf.2025.103316Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in SwitzerlandManuel R. Popp0Niklaus E. Zimmermann1Philipp Brun2Corresponding author at: Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, Birmensdorf, 8903, Zurich, Switzerland.; Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, Birmensdorf, 8903, Zurich, Switzerland; Federal Institute of Technology Zurich ETHZ, Universitätstrasse 16, Zürich, 8092, Zurich, SwitzerlandSwiss Federal Institute for Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, Birmensdorf, 8903, Zurich, Switzerland; Federal Institute of Technology Zurich ETHZ, Universitätstrasse 16, Zürich, 8092, Zurich, SwitzerlandSwiss Federal Institute for Forest, Snow and Landscape Research WSL, Zürcherstrasse 111, Birmensdorf, 8903, Zurich, Switzerland; Federal Institute of Technology Zurich ETHZ, Universitätstrasse 16, Zürich, 8092, Zurich, SwitzerlandSmartphone apps and citizen science platforms have enabled large-scale public contributions to biodiversity monitoring, with image recognition readily allowing automated species identification. Yet, practical evaluations of the best use of such identification tools for nationwide plant community monitoring remain limited. We conducted a stratified field survey covering diverse habitats and different growth forms at 151 sites in all biogeographic regions of Switzerland, collecting > 5000 labelled photos of different plant parts for about 20% of the Swiss Flora. We assessed the performance of leading, publicly available identification services, differing in geographic scope and taxonomic resolution. Up to 85% of species observations were correctly identified when multiple images were supplied. While all identification providers correctly identified > 90% of the species at least once, we found models with a narrower geographic scope and coarser taxonomic resolution to reach higher identification accuracy. Identifications of photographs depicting multiple plant parts were often more successful than those of images focusing on single plant parts. While manual control remains indispensable for high-quality data, involving automated identification services can boost efficiency in compiling standardised floristic data, especially when (1) several images are taken, (2) an identification provider with matching taxonomic concepts and scope is selected, and (3) auxiliary information such as environmental context is considered.http://www.sciencedirect.com/science/article/pii/S1574954125003255BotanyCitizen scienceDeep learningiNaturalistPlantNetFlora incognita
spellingShingle Manuel R. Popp
Niklaus E. Zimmermann
Philipp Brun
Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
Ecological Informatics
Botany
Citizen science
Deep learning
iNaturalist
PlantNet
Flora incognita
title Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
title_full Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
title_fullStr Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
title_full_unstemmed Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
title_short Evaluating the use of automated plant identification tools in biodiversity monitoring—a case study in Switzerland
title_sort evaluating the use of automated plant identification tools in biodiversity monitoring a case study in switzerland
topic Botany
Citizen science
Deep learning
iNaturalist
PlantNet
Flora incognita
url http://www.sciencedirect.com/science/article/pii/S1574954125003255
work_keys_str_mv AT manuelrpopp evaluatingtheuseofautomatedplantidentificationtoolsinbiodiversitymonitoringacasestudyinswitzerland
AT niklausezimmermann evaluatingtheuseofautomatedplantidentificationtoolsinbiodiversitymonitoringacasestudyinswitzerland
AT philippbrun evaluatingtheuseofautomatedplantidentificationtoolsinbiodiversitymonitoringacasestudyinswitzerland