Tag
Loading...

Biodiversity monitoring involves the observation and measurement of species, ecosystems, and their changes. Monitoring can involve identifying and counting different species, observing their behaviours, noticing species absences, or observing other forms of biodiversity such as genes, taxa, structures, functions, or wider ecosystems.

The outcomes of monitoring activities can have essential impacts on land-use, ecosystem management, and policy . Data is used to assess ecosystem services, motivate urban development, formulate political arguments, and even to predict future biodiversity statistics. Biodiversity monitoring is often undertaken by ecologists or other professionals who observe biodiversity within a defined landscape or research question. However, this monitoring also increasingly involves wider publics who assist in counting, identifying, and mapping species to collect large datasets. Digital technologies, including automated species recognition and prediction algorithms, are playing an increasing role in these practices.

This logbook narrates different practices of participatory biodiversity monitoring and shares fieldwork data carried out in the Netherlands during spring and summer of 2023.

A Forest Tour Becomes Digital

On May 26th, I (Michelle) was able to join a guided walk through the fields and forests of the Peelrandbreuk near Uden (NL). This tour was organised by a regional society for field biology and open to the public. I initially participated to learn more about the Peelrandbreuk land phenomenon and its distinct biodiversity , but I was also able to learn a lot about participants' usage of mobile applications for species identification and the use of automated image recognition.

Once we gathered together at the starting point of our tour, I clearly stood out from the rest of the group due to my younger age and especially because of my lack of semi-professional binoculars. Other participants were clearly very experienced at local biodiversity walks.

"Are you interested in bird or plants?", another participant asked. "Uhm, both, I guess? Though I am not so knowledgeable about either, probably", I replied.

Friendly small talk followed, and I briefly shared that I am a researcher with an interest in learning more about the distinct biodiversity of the Peelrandbreuk, as well as my community research project with the eco village. We didn't further discuss the topics of digital technology in this project, as the conversation moved into other directions.

This later turned out to be helpful, as I was able to observe participants' interactions with mobile species identification technologies without deliberately steering or encouraging user engagements.

As a technology, binoculars are already closely tied to the human biodiversity observer. Especially for those with an interest in watching and identifying birds , walking through the forests is done with binoculars at the ready, safely tied around one's neck, to 'capture' all bird-like movements in the skies and trees . Collective walks such as this one elicit a shared excitement every time a new bird is spotted. Participants join together to share the exact location of a possible-new-bird on the horizon:

"left of the three bushes in the middle of the field, about halfway the small tree, on the branch sticking out... could that be a Kieviet [Lapwing, in English]?".

A discussion on the different possibilities regarding the species identity of the bird follows. Often an agreement is reached, but sometimes the final conclusion is left in the middle. Such conversations involve wonder, shared excitement, and knowledge sharing.

IMG_8402

Some of the participants of the walk through the Peelrandbreuk at the start of the excursion. Image taken by Michelle Westerlaken.

It was only after a while that the first participant took our their mobile phone to identify a plant species. During this recreational walk, identification was not necessarily sought for biodiversity monitoring or for contributing data, but instead inspired by collective curiosity. Or to settle a dispute.

"Witte gij't?"

Participants joked in a local dialect, where the words 'a white goat?' and 'do you know it?' are pronounced the same. No white goats were roaming the forest during that day, but more and more participants started to take out their phones to engage in digitally identifying local plants.

Each time, before the mobile application revealed its results, certain participants made sure to publicly announce their knowledgeable guesses, hoping for the digital technology to confirm their insights. Sometimes, the results obtained through the digital technology took on a role of authority that directly settled species disagreements. A single decision was reached. Other times, participants directly challenged the image recognition feature and the group used their collective plant knowledge to make contradicting suggestions. The accuracy rating of the automated image recognition played a big role in these discussions. Certain numbers were beyond dispute, other percentages opened collective discussion, and lower ratings were an immediate cause for disagreements. Social dynamics shifted. Some people were quick at using the application, so they became the 'identifiers'. Others were 'often correct' in announcing species before the technology confirmed their suggestions, and they were thus seen as more knowledgeable. Several participants, myself included, quietly observed these discussions, possibly reflecting on the proximate flora and fauna in a more personal manner.

"Effen kijken met de app " ["Take a quick look with the app"]

"Hoornbloem, zegt ie" ["mouse-ear, it/he says"], an 'identifier' called out from behind the bushes.

"Oh right, that's also possible", an announcer agreed after hearing the digital identification.

IMG_8414

A participant using their mobile phone to picture a plant and use automated species identification technology. Image taken by Michelle Westerlaken.

These activities also provoked discussions on the use of ObsIdentify versus Pl@ntnet, where people had different preferences. People challenged the scientific objectivity of these apps and reminded each other not to believe everything "these apps say".

"92% Koninginnekruid [Hemp-agrimony], well I don't recognise that leaf... oh actually, I do"

A few meters later, another participant is seemingly identifying the same plant with a different picture:

"29% Koninginnekruid [Hemp-agrimony], but that is very little", one participant remarks after photographing the plant with the app .

"Well, I think it's quite a lot, actually", another participant replies, followed with a conversation on numerical confusion.

IMG_8416

While certain people obtained close up images of the plants 'in question', others engaged in discussing possible species. Image taken by Michelle Westerlaken.

Later, the identification of plants with mobile apps shifted to the identification of birds via acoustics through different mobile apps such as Merlin. This shift also changed the social dynamics of the walk, because quietness is required to lean in and listen to the sounds. It also differentiated people with interests in birds and plants. At first, the bird enthusiasts were still more preoccupied with binoculars rather than phones, but once these acoustic apps became involved, this changed. Quite organically, the group had split up in two smaller groups. At some point one of the groups got rather delayed because they were engaged in an extensive and digitally informed discussion about the identification of a yellow flower that can go by many different names. "It looks like a Paardenbloem [Dandelion], but it isn't".

Near the end of the walk, one participant revealed yet another use of these applications that I had not encountered. He shared that before we gathered today, he had looked up the rare types of plants that have previously be observed in this area through the online database , in order to "know what to look for", during the walk. It is interesting to note as well that people were not specifically engaged in uploading their pictures through the applications to contribute them to the online database, but instead they were mainly preoccupied with automated species identifications and disregarded the images afterwards.



*Due to the public nature of this guided walk, I anonymised all pictures and personal details of participants.

Naturalis Biodiversity Center, Leiden, Nederlands

Testing the Limits of Automated Species Identification

The ARISE-day and conference in March 2023 was hosted at Naturalis Biodiversity Centre in Leiden, Netherlands. Besides a research centre, Naturalis is also a biodiversity museum. After discussing participatory biodiversity monitoring and automated species identification with ObsIdentify, Michelle tested the limits of this software and experience the internal validation process.

IMG_8599

A picture taken of a Tyrannosaurus skeleton hosted at the Naturalis Biodiversity Centre (a 67 million years old fossil, 80% complete, female, >30 years old, 12 meter long, and originally found in Montana, US). ObsIdentify's image recognition software determined that this may be a (common) porpoise.

IMG_8601

The ObsIdentify application indicates an accuracy rating of 75% for this observation.

_

Screenshot 2023-06-06 at 15.01.01

As a reference, these are the (validated) observations for this type of dolphin species that have been made in the Netherlands. Screenshot of open-dataset, taken by Michelle Westerlaken, retrieved on June 6 2023 via https://waarneming.nl/species/380/

_

Screenshot 2023-06-06 at 15.09.40

Within 24 hours, a validator successfully recognised the error in the identification and rejected the observation. Not only did the image include the wrong type of species, the observation also did not follow the platform established rules. Perhaps the "photo shows an animal that has not been seen in the wild or has clearly escaped". Screenshot of open-dataset, taken by Michelle Westerlaken, retrieved on June 6 2023 via https://waarneming.nl/observation/266219698/

BioBlitz Ecodorp Boekel

Between May 18-21, we organised a BioBlitz at Ecodorp Boekel in order to map local species and experience the use of automated species identification applications. The BioBlitz was hosted by Michelle and approximately 11 participants joined in mapping plant, animals, and fungi species via the ObsIdentify mobile application. (More people joined whose data were not collected, including those <18 years old as well as people who joined but didn't register their data online via the application.). During this extended weekend with pleasant weather conditions, we collectively uploaded 310 observations that included 157 different species.

Screenshot 2023-06-06 at 13.51.50

Data visualisation of the BioBlitz at Ecodorp Boekel, screenshot of open-dataset taken by Michelle Westerlaken, retrieved on June 6th 2023 via https://waarneming.nl/bioblitz/bioblitz-ecodorp-boekel/

What immediately stands out when looking at this data is the Ecodorp inhabitants' keen interest in observing and identifying plant species. Contrary to the Waarnemingen.nl user statistics that mostly observe birds, this BioBlitz was much more plant-oriented. This does not only reflect the inhabitants' interest and time spent in growing their own food, but we also found that plants were much easier to document compared to moving insects or nearly invisible fungi .

IMG_8291

A participant observing species in the grass at Ecodorp Boekel. Image taken by Michelle Westerlaken

Interviewing participants while mapping species and using the mobile application revealed many interesting stories and observations. One participant was engaged in training their eyesight and found the activity of finding insects very suitable for training eye movement and shifting perspectives. Another participant was forcefully displaced from Yemen and while observing species locally, we talked about the kinds of plants that occur in Yemen, their names, and the relations of family members with native plants.

IMG_8290

Two participants of the BioBlitz searching the grass for insects. Image taken by Michelle Westerlaken

One participants was concerned about the scientific value of mapping species. She is an experienced volunteer in biodiversity monitoring and shared the principles of systematic monitoring in order to help divide people across the area. A few parents joined the BioBlitz together with their kids as an educational activity. They engaged in conversations together and expressed their excitement about all the different species they encountered. Other participants were more interested in mapping the small gardens in front of their homes to gain knowledge of biodiversity that can be observed close by.

Several participants experienced difficulties in connecting their mobile phones to the application or experienced other limitations such as reaching mobile data limits on their phones. Another much discussed topic involved the precision of the automated species identification. Participants discovered that the application would only automatically record the observation once it reveals >90% accuracy for each observation. This led participants to discard all the 'less successful' observations and delete these images. Participants discussed strategies for 'correctly' photographing plants and observed how certain groups of plants (such as grasses) were much more difficult to identify.

IMG_8284

We found, among others, that the most successful way of picturing a plant for automated species recognition could include clear, isolated images of plants with leaf-like structures in flowering states. Image taken by Michelle Westerlaken

IMG_8608

With the ObsIdentify app, the species was identified as 'kromhals' (bugloss, in English) with a '100% accuracy' rating. Screenshot of the ObsIdentify open-dataset, taken by Michelle Westerlaken

An often recurring comment during the BioBlitz involved people's excitement about using the mobile application. "It is really addictive", on participant mentioned. She explained how she worked on maintaining the food garden during the day, but she was continuously drawn to picturing and recording new observations as different species drew her attention. This way of playfully moving across space reminded Michelle of exploring open-world video games that are designed to elicit curiosity by strategically placing game-entities.

IMG_8300

A BioBlitz participant picturing a plant species for identification. Image taken by Michelle Westerlaken

This is just a small sample of reflections and images from the BioBlitz. It reveals how the activity of mapping biodiversity locally is closely connected to people's personal interests and experiences. Rather than a collection of observations that resembles scientific accuracy, the mapping activity provides a more detailed reflection on how participants are drawn to different species. Furthermore, the results don't reflect species occurrences, but instead mainly demonstrate which species are suitable for automated identification with the application's image recognition software.

In the weeks following the BioBlitz, more Ecodorp inhabitants joined as registered users of ObsIdentify and shared experiences and perspectives. People used the application to ask each other questions on which local plants are edible, and they enjoyed quizzing themselves and others on identifying species. Inhabitants are still adding species observations to the platform and more reflections may be added as the fieldwork unfolds.

Naturalis Biodiversity Center, Leiden, Nederlands

Biodiversity Data Standardisation Practices

During the ARISE-day 2023 workshop, I (Michelle) had the opportunity to attend a data management workshop hosted by the Dutch node of GBIF, where I learned more about data standardisation principles for biodiversity data. Here we worked with the Dutch monumental tree dataset to explore how to create a database that can be uploaded to GBIF. What struck me during this workshop were the various ways in which data of individual trees that have historic, cultural, or societal value are transformed into a database that only includes species names, GPS coordinates, plant years, images, and the name of the photographer. In other words, the standardisation process works to create singular categories that work for all GBIF occurrences, thereby eliminating those types of data that make these occurrences unique.

Furthermore, while the Darwin Core protocol is open-source and described as 'flexible', GBIF's standardisation practices align with a more western type of scientific data management that, for instance, prioritises species names, timestamps, location, and ownership data. When such data standardisations are implemented globally, it risks eliminating the possibilities for other types of data to become included in global frameworks. One of the organisers remarked that while FAIR data principles are important for Dutch GBIF data, CARE principles are not really relevant because "these are for Indigenous people and data sovereignty ". Such understandings demonstrate how people involved in biodiversity data management in the Global North can view different data practices as separate initiatives across the world with distinct ethical protocols, while at the same time GBIF is working towards establishing global standards. These experiences materialise the conflicting needs for both data standardisation and unique ways of understanding biodiversity.

Screenshot 2023-06-06 at 13.08.15

A partial view of the Dutch monumental tree database used in the NLBIF data management workshop. Screenshot taken by Michelle Westerlaken from an open dataset, retrieved on March 30 via https://bomen.meetnetportaal.nl/

Species Identification Apps and Biodiversity Database Infrastructures

With the increasing use of mobile smartphones, applications that assist biodiversity monitoring are now widespread and have millions of users (predominantly in the Global North). These apps focus specifically on recognising plant, animal, and fungi species from image- or sound recordings. Previously, all this data had to be manually evaluated by taxonomists or citizen scientists to identify species. Now, the majority of the most used applications incorporate automated functionality to identify species based on image- or sound-recognition techniques.

Examples of apps with large user bases include iNaturalist, eBird,Merlin (all three developed in the U.S.), the French Pl@ntNet, the Dutch ObsIdentify, and national nodes of iNaturalist such as the UK portal iRecord. Each of these apps collects thousands of observations per day. Depending on the app 's infrastructure, besides species identification data, each observation contains metadata about the time, location, as well as user statistics. Some apps, including iNaturalist, rely predominantly on automated observation and citizen science, while others, such as ObsIdentify, selects specific observations to be evaluated by volunteer 'validators' or 'experts'. Some of these technologies can also be directly connected to hardware such a binoculars or cameras to directly record more detailed and high-resolution images.

Screenshot 2023-06-06 at 12.15.02

This table shows the different validation categories from ObsIdentify to give an indication of how automated species recognition is evaluated by a group of volunteer-experts. Screenshot taken by Michelle Westerlaken, retrieved on June 6, 2023 via https://observation.org/pages/validation/

Rather than developing new taxonomic datasets, these applications are usually connected to larger national and international digital species databases. eBird, for instance, works with the 'Working Group Avian Checklists' to create a single global taxonomy for avian species. ObsIdentify is an app of Observations.org, which is connected to Naturalis, a Dutch national biodiversity institute. Most of the (validated) data in these applications, including iNaturalist, are finally indexed as open data in GBIF, the Global Biodiversity Information Facility.

GBIF contains hundreds of millions of species occurrence records from sources such as museum collections, eDNA, research databases, and citizen data. The intergovernmental initiative started in 1999, has a secretariat in Copenhagen, involves 42 voting-participant countries, and is government -funded. One of its aims is to create protocols for data standards, specifically with the use of the Darwin Core protocol.

/
smart forests radio