Species identification by conservation practitioners using online images: accuracy and agreement between experts

Empathy for wildlife: The importance of the individual

Because climate change and the biodiversity crisis are driven by human actions, determining psychological mechanisms underpinning support for environmental action is an urgent priority. Here, we experimentally tested for mechanisms promoting conservation-related motivation and behavior toward a flagship species, wild Tamanend's bottlenose dolphins. Following evidence that empathy increases prosocial motivations and behavior, and that the ability to identify individual humans promotes empathy, we tested whether this relationship applied to the ability to identify individual dolphins. Participants identified dolphins from their dorsal fins at above chance levels, and better individuation correlated with higher empathy for dolphins and higher willingness to pledge environmental behaviors. Pairing a narrative with an image of an injured dolphin leads to higher donations relative to a narrative alone. Our novel finding that the ability to individually identify dolphins relates to empathy and conservation-related behavior suggests pathways for strengthening environmental attitudes and behavior.

Bumble Bee Watch community science program increases scientific understanding of an important pollinator group across Canada and the USA

In a time of increasing threats to bumble bees (Hymenoptera: Apidae: Bombus), it is important to understand their ecology and distribution. As experts are limited in resources to conduct field surveys, there is potential for community scientists to help. The Bumble Bee Watch (BBW) community science program involves volunteers taking photos of bumble bees in Canada and the USA and submitting them, along with geographic and optional plant information, to a website or through an app. Taxon experts then verify the bee species identification. The Bumble Bees of North America database (BBNA) stores data (no photographs) collected and identified by more traditional scientific methods over the same range. Here we compared BBW data to BBNA data over all years and just 2010-2020 to understand the scientific contribution of community scientists to the state of the knowledge about native bumble bees. We found that BBW had similar geographic and species coverage as BBNA. It had records from all 63 provinces, states, and territories where bumble bees occur (including four more than BBNA in 2010-2020), and represented 41 of the 48 species in BBNA (with ten more species than BBNA in 2010-2020). While BBW contributed only 8.50% of records overall, it contributed 25.06% of all records over 2010-2020. BBW confirmed the persistence of species and identified new locations of species, both inside and outside of the previously known extent of occurrences. BBW also contributed a wealth of ecological information, such as unique plant genera and species data for almost all the bee species. Thus, while BBW had fewer bee records than the BBNA database overall, it helped to fill in data gaps and provided novel information, complementing the traditional methods. This community science program is valuable in helping to inform conservation management for bumble bee species.

Reliability of environmental DNA surveys to detect pond occupancy by newts at a national scale

The distribution assessment and monitoring of species is key to reliable environmental impact assessments and conservation interventions. Considerable effort is directed towards survey and monitoring of great crested newts (Triturus cristatus) in England. Surveys are increasingly undertaken using indirect methodologies, such as environmental DNA (eDNA). We used a large data set to estimate national pond occupancy rate, as well as false negative and false positive error rates, for commercial eDNA protocols. Additionally, we explored a range of habitat, landscape and climatic variables as predictors of pond occupancy. In England, 20% of ponds were estimated to be occupied by great crested newts. Pond sample collection error rates were estimated as 5.2% false negative and 1.5% false positive. Laboratory error indicated a negligible false negative rate when 12 qPCR replicates were used. Laboratory false positive error was estimated at 2% per qPCR replicate and is therefore exaggerated by high levels of laboratory replication. Including simple habitat suitability variables into the model revealed the importance of fish, plants and shading as predictors of newt presence. However, variables traditionally considered as important for newt presence may need more precise and consistent measurement if they are to be employed as reliable predictors in modelling exercises.

Linking camera-trap data to taxonomy: Identifying photographs of morphologically similar chipmunks

Remote cameras are a common method for surveying wildlife and recently have been promoted for implementing large-scale regional biodiversity monitoring programs. The use of camera-trap data depends on the correct identification of animals captured in the photographs, yet misidentification rates can be high, especially when morphologically similar species co-occur, and this can lead to faulty inferences and hinder conservation efforts. Correct identification is dependent on diagnosable taxonomic characters, photograph quality, and the experience and training of the observer. However, keys rooted in taxonomy are rarely used for the identification of camera-trap images and error rates are rarely assessed, even when morphologically similar species are present in the study area. We tested a method for ensuring high identification accuracy using two sympatric and morphologically similar chipmunk (Neotamias) species as a case study. We hypothesized that the identification accuracy would improve with use of the identification key and with observer training, resulting in higher levels of observer confidence and higher levels of agreement among observers. We developed an identification key and tested identification accuracy based on photographs of verified museum specimens. Our results supported predictions for each of these hypotheses. In addition, we validated the method in the field by comparing remote-camera data with live-trapping data. We recommend use of these methods to evaluate error rates and to exclude ambiguous records in camera-trap datasets. We urge that ensuring correct and scientifically defensible species identifications is incumbent on researchers and should be incorporated into the camera-trap workflow.

Assessment of volunteered geographic information for vegetation mapping

Vegetation mapping requires extensive field data for training and validation. Volunteered geographic information in the form of geotagged photos of identified plants has the potential to serve as a supplemental data source for vegetation mapping projects. In this study, we compare the locations of specific taxa from the iNaturalist platform to locations identified on both a fine-scale vegetation map and high-resolution ortho-imagery in open-canopy shrubland in San Clemente Island, CA. Due to positional uncertainty associated with the iNaturalist observations, as well as the presence-only nature of the data, it was not possible to perform a traditional accuracy assessment. We instead measured the distance between the location recorded by an iNaturalist observer for a given taxon and the closest mapped individual of that taxon. This distance was within 10 m for a majority of the observations (64%). When comparing the iNaturalist location to the closest individual detected through image interpretation, 87% of the observations were within 10 m. The discrepancy in agreement between the vegetation map and imagery is likely due to mapping errors. While iNaturalist data come with important limitations, the platform is an excellent resource for supporting vegetation mapping and other ecological applications.

Using Bumble Bee Watch to investigate the accuracy and perception of bumble bee (Bombus spp.) identification by community scientists

Community science programs provide an opportunity to gather scientific data to inform conservation policy and management. This study examines the accuracy of community science identifications submitted to the North American Bumble Bee Watch program on a per species level and as compared to each species’ conservation status, as well as users (members of the public) and experts (those with expertise in the field of bumble bee biology) perceived ease of species identification. Photos of bumble bees (Hymenoptera: Apidae: Bombus) are submitted to the program by users and verified (species name corrected or assigned as necessary) by an expert. Over 22,000 records from over 4,900 users were used in the analyses. Accuracy was measured in two ways: percent agreement (percent of all records submitted correctly by users) and veracity (percent of all verified records submitted correctly by the users). Users generally perceived it harder to identify species than experts. User perceptions were not significantly different from the observed percent agreement or veracity, while expert perceptions were significantly different (overly optimistic) from the observed percent agreement but not the veracity. We compared user submitted names to final expert verified names and found that, for all species combined, the average percent agreement was 53.20% while the average veracity was 55.86%. There was a wide range in percent agreement values per species, although sample size and the role of chance did affect some species agreements. As the conservation status of species increased to higher levels of extinction risk, species were increasingly more likely to have a lower percent agreement but higher levels of veracity than species of least concern. For each species name submitted, the number of different species verified by experts varied from 1 to 32. Future research may investigate which factors relate to success in user identification through community science. These findings could play a role in informing the design of community science programs in the future, including for use in long-term and national-level monitoring of wild pollinators.

Sharing for science: high-resolution trophic interactions revealed rapidly by social media

Discrete, ephemeral natural phenomena with low spatial or temporal predictability are incredibly challenging to study systematically. In ecology, species interactions, which constitute the functional backbone of ecological communities, can be notoriously difficult to characterise especially when taxa are inconspicuous and the interactions of interest (e.g., trophic events) occur infrequently, rapidly, or variably in space and time. Overcoming such issues has historically required significant time and resource investment to collect sufficient data, precluding the answering of many ecological and evolutionary questions. Here, we show the utility of social media for rapidly collecting observations of ephemeral ecological phenomena with low spatial and temporal predictability by using a Facebook group dedicated to collecting predation events involving reptiles and amphibians in sub-Saharan Africa. We collected over 1900 independent feeding observations using Facebook from 2015 to 2019 involving 83 families of predators and 129 families of prey. Feeding events by snakes were particularly well-represented with close to 1,100 feeding observations recorded. Relative to an extensive literature review spanning 226 sources and 138 years, we found that social media has provided snake dietary records faster than ever before in history with prey being identified to a finer taxonomic resolution and showing only modest concordance with the literature due to the number of novel interactions that were detected. Finally, we demonstrate that social media can outperform other citizen science image-based approaches (iNaturalist and Google Images) highlighting the versatility of social media and its ability to function as a citizen science platform.

Community science participants gain environmental awareness and contribute high quality data but improvements are needed: insights from Bumble Bee Watch

Bumble Bee Watch is a community science program where participants submit photos of bumble bees from across Canada and the United States for expert verification. The data can be used to help better understand bumble bee biology and aid in their conservation. Yet for community science programs like this to be successful and sustainable, it is important to understand the participant demographics, what motivates them, and the outcomes of their participation, as well as areas that are working well or could be improved. It is also important to understand who verifies the submissions, who uses the data and their views on the program. Of the surveyed users, most participate to contribute to scientific data collection (88%), because of a worry about bees and a desire to help save them (80%), to learn more about species in their property (63%) or region (56%), and because of a personal interest (59%). About 77% report increased awareness of species diversity, while 84% report improvement in their identification skills. We found that 81% had at least one college or university degree. There were more respondents from suburban and rural areas than urban areas, but area did not affect numbers of submissions. While half were between 45 and 64 years of age, age did not influence motivation or number of submissions. Respondents were happy with the program, particularly the website resources, the contribution to knowledge and conservation efforts, the educational values, and the ability to get identifications. Areas for improvement included app and website functionality, faster and more detailed feedback, localized resources, and more communication. Most respondents participate rarely and have submitted fewer than ten records, although about five percent are super users who participate often and submit more than fifty records. Suggested improvements to the program may increase this participation rate. Indeed, increased recruitment and retention of users in general is important, and advertising should promote the outcomes of participation. Fifteen experts responded to a separate survey and were favorable of the program although there were suggestions on how to improve the verification process and the quality of the submitted data. Suggested research questions that could be asked or answered from the data included filling knowledge gaps (species diversity, ranges, habitat, phenology, floral associations, etc.), supporting species status assessments, effecting policy and legislation, encouraging habitat restoration and management efforts, and guiding further research. However, only about half have used data from the project to date. Further promotion of Bumble Bee Watch and community science programs in general should occur amongst academia, conservationists, policy makers, and the general public. This would help to increase the number and scope of submissions, knowledge of these species, interest in conserving them, and the overall program impact.

Use of Machine Learning to Detect Wildlife Product Promotion and Sales on Twitter

Social media is an important channel for communication, information dissemination, and social interaction, but also provides opportunities to illicitly sell goods online, including the trade of wildlife products. In this study, we use the Twitter public application programming interface (API) to access Twitter messages in order to detect and classify suspicious wildlife trafficking and sale using an unsupervised machine learning topic model combined with keyword filtering and manual annotation. We choose two prohibited wildlife animals and related products: elephant ivory and pangolin, and collected tweets containing keywords and known code words related to these species. In total, we collected 138,357 tweets filtered for these keywords over a 14-day period and were able to identify 53 tweets from 38 unique users that we suspect promoted the sale of Ivory products, though no pangolin related promoted post were detected in this study. Study results show that machine learning combined with supplement analysis approaches such as those utilized in this study have the potential to detect illegal content without the use of an existing training data set. If developed further, these approaches can help technology companies, conservation groups, and law enforcement officials to expedite the process of identifying illegal online sales and stem supply for the billion-dollar criminal industry of online wildlife trafficking.

Error rates in wildlife image classification

We address the comments made by Thornton et al. (Ecology and Evolution, 2019) in response to our recent article on measuring the agreement among experts in classifying camera images of bobcats and Canada lynx.

Designing online species identification tools for biological recording: the impact on data quality and citizen science learning

In recent years, the number and scale of environmental citizen science programmes that involve lay people in scientific research have increased rapidly. Many of these initiatives are concerned with the recording and identification of species, processes which are increasingly mediated through digital interfaces. Here, we address the growing need to understand the particular role of digital identification tools, both in generating scientific data and in supporting learning by lay people engaged in citizen science activities pertaining to biological recording communities. Starting from two well-known identification tools, namely identification keys and field guides, this study focuses on the decision-making and quality of learning processes underlying species identification tasks, by comparing three digital interfaces designed to identify bumblebee species. The three interfaces varied with respect to whether species were directly compared or filtered by matching on visual features; and whether the order of filters was directed by the interface or a user-driven open choice. A concurrent mixed-methods approach was adopted to compare how these different interfaces affected the ability of participants to make correct and quick species identifications, and to better understand how participants learned through using these interfaces. We found that the accuracy of identification and quality of learning were dependent upon the interface type, the difficulty of the specimen on the image being identified and the interaction between interface type and ‘image difficulty’. Specifically, interfaces based on filtering outperformed those based on direct visual comparison across all metrics, and an open choice of filters led to higher accuracy than the interface that directed the filtering. Our results have direct implications for the design of online identification technologies for biological recording, irrespective of whether the goal is to collect higher quality citizen science data, or to support user learning and engagement in these communities of practice.

Use-inspired basic research on individual differences in face identification: implications for criminal investigation and security

This journal is dedicated to "use-inspired basic research" where a problem in the world shapes the hypotheses for study in the laboratory. This review considers the role of individual variation in face identification and the challenges and opportunities this presents in security and criminal investigations. We show how theoretical work conducted on individual variation in face identification has, in part, been stimulated by situations presented in the real world. In turn, we review the contribution of theoretical work on individual variation in face processing and how this may help shape the practical identification of faces in applied situations. We consider two cases in detail. The first case is that of security officers; gatekeepers who use facial ID to grant entry or deny access. One applied example, where much research has been conducted, is passport control officers who are asked to match a person in front of them to a photograph shown on their ID. What happens if they are poor at making such face matching decisions and can they be trained to improve their performance? Second, we outline the case of "super-recognisers", people who are excellent at face recognition. Here it is interesting to consider whether these individuals can be strategically allocated to security and criminal roles, to maximise the identification of suspects. We conclude that individual differences are one of the largest documented sources of error in face matching and face recognition but more work is needed to account for these differences within theoretical models of face processing.