SEALNET: Facial recognition software for ecological studies of harbor seals

Methods for long-term monitoring of coastal species such as harbor seals, are often costly, time-consuming, and highly invasive, underscoring the need for improved techniques for data collection and analysis. Here, we propose the use of automated facial recognition technology for identification of individual seals and demonstrate its utility in ecological and population studies. We created a software package, SealNet, that automates photo identification of seals, using a graphical user interface (GUI) software to identify, align and chip seal faces from photographs and a deep convolutional neural network (CNN) suitable for small datasets (e.g., 100 seals with five photos per seal). We piloted the SealNet technology with a population of harbor seals located within Casco Bay on the coast of Maine, USA. Across two-years of sampling, 2019 and 2020, at seven haul-out sites in Middle Bay, we processed 1529 images representing 408 individual seals and achieved 88% (93%) rank-1 accuracy in closed set (open set) seal identification. We identified four seals that were photographed in both years at neighboring haul-out sites, suggesting that some harbor seals exhibit site fidelity within local bays across years, and that there may be evidence of spatial connectivity among haul-out sites. Using capture-mark-recapture (CMR) calculations, we obtained a rough preliminary population estimate of 4386 seals in the Middle Bay area. SealNet software outperformed a similar face recognition method developed for primates, PrimNet, in identifying seals following training on our seal dataset. The ease and wealth of image data that can be processed using SealNet software contributes a vital tool for ecological and behavioral studies of marine mammals in the emerging field of conservation technology.

A simple tool for linking photo-identification with multimedia data to track mammal behaviour

Identifying individual animals is critical to describe demographic and behavioural patterns, and to investigate the ecological and evolutionary underpinnings of these patterns. The traditional non-invasive method of individual identification in mammals—comparison of photographed natural marks—has been improved by coupling other sampling methods, such as recording overhead video, audio and other multimedia data. However, aligning, linking and syncing these multimedia data streams are persistent challenges. Here, we provide computational tools to streamline the integration of multiple techniques to identify individual free-ranging mammals when tracking their behaviour in the wild. We developed an open-source R package for organizing multimedia data and for simplifying their processing a posteriori—“MAMMals: Managing Animal MultiMedia: Align, Link, Sync”. The package contains functions to (i) align and link the individual data from photographs to videos, audio recordings and other text data sources (e.g. GPS locations) from which metadata can be accessed; and (ii) synchronize and extract the useful multimedia (e.g. videos with audios) containing photo-identified individuals. To illustrate how these tools can facilitate linking photo-identification and video behavioural sampling in situ, we simultaneously collected photos and videos of bottlenose dolphins using off-the-shelf cameras and drones, then merged these data to track the foraging behaviour of individuals and groups. We hope our simple tools encourage future work that extend and generalize the links between multiple sampling platforms of free-ranging mammals, thereby improving the raw material needed for generating new insights in mammalian population and behavioural ecology.

Comparison of Three Individual Identification Algorithms for Sperm Whales (Physeter macrocephalus) after Automated Detection

Photo-identification of individual sperm whales (Physeter macrocephalus) is the primary technique for mark-recapture-based population analyses for the species The visual appearance of the fluke - with its distinct nicks and notches - often serves as the primary visual differentiator, allowing humans to make recorded sightings of specific individuals. However, the advent of digital photography and the significant increase in volume of images from multiple projects in combination with pre-existing historical catalogs has made applying the method more challenging.with the required human labor for de-duplication (reduction of Type II errors) and reconciliation of sightings between large datasets too cost- and time- prohibitive. To address this, we trained and evaluated the accuracy of PIE v2 (a triplet loss network) along with two existing fluke trailing edge-matching algorithms, CurvRank v2 and Dynamic Time Warping (DTW), as a mean to speed comparison among a high volume of photographs. Analyzed data were collected from a curated catalog of well-known sperm whales sighted across years (2005-2018) off the island of Dominica. The newly-trained PIE model outperformed the older CurvRank and DTW algorithms, and PIE provided the following top-k individual ID matching accuracy on a standard min-3/max-10 sighting training data set: Rank-1: 87.0%, Rank-5: 90.5%, and Rank-12: 92.5%. An essential aspect of PIE is that it can learn new individuals without network retraining, which can be immediately applied in the presence of (and for the resolution of) duplicate individuals in overlapping catalogs. Overall, our results recommend the use of PIE v2 and CurvRank v2 for ID reconciliation in combination due to their complementary performance.

FIN-PRINT a fully-automated multi-stage deep-learning-based framework for the individual recognition of killer whales

Biometric identification techniques such as photo-identification require an array of unique natural markings to identify individuals. From 1975 to present, Bigg’s killer whales have been photo-identified along the west coast of North America, resulting in one of the largest and longest-running cetacean photo-identification datasets. However, data maintenance and analysis are extremely time and resource consuming. This study transfers the procedure of killer whale image identification into a fully automated, multi-stage, deep learning framework, entitled FIN-PRINT. It is composed of multiple sequentially ordered sub-components. FIN-PRINT is trained and evaluated on a dataset collected over an 8-year period (2011–2018) in the coastal waters off western North America, including 121,000 human-annotated identification images of Bigg’s killer whales. At first, object detection is performed to identify unique killer whale markings, resulting in 94.4% recall, 94.1% precision, and 93.4% mean-average-precision (mAP). Second, all previously identified natural killer whale markings are extracted. The third step introduces a data enhancement mechanism by filtering between valid and invalid markings from previous processing levels, achieving 92.8% recall, 97.5%, precision, and 95.2% accuracy. The fourth and final step involves multi-class individual recognition. When evaluated on the network test set, it achieved an accuracy of 92.5% with 97.2% top-3 unweighted accuracy (TUA) for the 100 most commonly photo-identified killer whales. Additionally, the method achieved an accuracy of 84.5% and a TUA of 92.9% when applied to the entire 2018 image collection of the 100 most common killer whales. The source code of FIN-PRINT can be adapted to other species and will be publicly available.

The Obs-Services and their potentials for biodiversity data assessments with a test of the current reliability of photo-identification of Coleoptera in the field

The current reliability of species identifications by the Nature Identification API (NIA) of the app ObsIdentify is tested with a Coleoptera (Insecta) sample set from Germany. Seventy-five photographic beetle records taken with a smartphone camera under “average user” conditions are analysed in terms of correctness of the app’s identification result on various taxonomic levels, the displayed confidence level of the identification and the time until validation of the results. More than 60% of samples were identified correctly at the species level, but only 53% were validated within a month. The mechanisms by which users can upload pictures of their observations to be identified by the artificial intelligence and the validation process by experts are briefly explained. Regional specifics and further opportunities for data usage as well as currently existing problems are discussed and improvements are suggested. The expert validation of records is identified as a huge quality advantage of the Obs-Services. They are generally found to be a promising tool for lay people and professional institutions, despite still existing deficiencies such as identification failure in mutilated specimens, cryptic and rare species, doubtful species rarity ratings as well as the still insufficient capacity of validation. Experts and institutions are encouraged to volunteer as validators and collaborators.

Population Dynamics Reveal a Core Community of the Common Bottlenose Dolphin (Tursiops truncatus) in Open Waters of the South-Western Gulf of Mexico

The presence of transient and temporary individuals in capture-mark-recapture studies may violate the assumption on equal catchability, and thus yield biased estimates. We investigated the effects of residency patterns on population parameters of bottlenose dolphins inhabiting the coastal waters off the Alvarado Lagoon System (ALS), Veracruz, Mexico. We hypothesized that this population is open but there exists a “core community” that behaves as a closed population. Between 2006 and 2010, we conducted 75 photo-identification surveys and recorded 263 dolphin group encounters, in which 231 dolphins were identified. Individuals present during only one season, classified as transients (n = 85), were excluded from the study, and a standardized residency index (IH4) was computed for each dolphin that remained in the sample (n = 146). We used the K-means clustering method to split the sample into groups based on individual (seasonal, annual) IH4 values. These clusters were named as regular residents (RR, n = 55), occasional residents (OR, n = 45), and occasional visitors (OV, n = 46). The cumulative frequency of newly identified individuals displayed an asymptotic trend for the whole sample and all clusters, indicating that most of the individuals present in the study area during the study period were identified. The assumption of demographic closure was tested to define the core community, and was rejected for the whole sample and the OV cluster (p < 0.001 in both cases), indicating that the population is open. The closure assumption was not rejected for RR and OR clusters (χ2 = 6.88, DF = 13, p = 0.91, and χ2 = 17.8, DF = 16, p = 0.33, respectively), indicating that these clusters were demographically closed over the 5-year period. Thus, we defined this aggregation of individuals as the “core community”. The closed population model Mth indicated that the total abundance of this core community was 123 individuals (95% CI: 114–133). Our results provide quantitative evidence of the existence of a core community in open waters of the Gulf of Mexico, and points toward residency pattern as a main driver of population dynamics. These results highlight the importance of considering residency patterns when dealing with heterogeneity in the sample of a highly mobile species.

Salinity and Water Temperature as Predictors of Bottlenose Dolphin (Tursiops truncatus) Encounter Rates in Upper Galveston Bay, Texas

Bottlenose dolphins (Tursiops truncatus) that inhabit urban estuaries like Galveston Bay, Texas, are exposed to cumulative stressors including pollution, fisheries, shipping, freshwater inflows, and construction operations. With continuing development, it is imperative to understand the key environmental variables that make the Galveston Bay estuary suitable habitat for this protected species. The Galveston Bay Dolphin Research Program conducted monthly photo identification surveys of bottlenose dolphins in a previously understudied 186 km2 area in upper Galveston Bay (UGB). To understand occurrence patterns in this region, we calculated monthly encounter rates of dolphins (dolphins/km) for four consecutive years (2016–2019). Using multiple linear regression models, we investigated the relationship between encounter rates, and water temperature and salinity. Monthly encounter rates ranged from 0.00 to 1.23 dolphins/km with an average of 0.34 dolphins/km (SE = 0.05). Over 80% of the variance was explained by the predictor variables water temperature and salinity (R2 = 0.820). Water temperature had a positive linear effect on encounter rates at over 23.37°C (SE = 1.42). Accordingly, higher encounter rates occurred during months with warm temperatures (May–September) compared to cooler months (November–April), indicating a predictable yearly movement pattern. Moreover, salinity was a highly significant predictor variable, with encounter rates dropping linearly with decreases in salinity. Higher numbers of dolphins are found in UGB during summer, but an exodus of dolphins occurs with low salinity levels, regardless of the time of year and water temperature. These findings should be considered during infrastructure projects (i.e., flood gate system) that may alter dolphin habitat and prey availability.

An Analysis of Urban Ecological Knowledge and Behaviour in Wellington, New Zealand

<p>The majority of the worlds' population now live in urban areas, with urban areas growing at a faster rate than other land use types. Most urban residents are concentrated in neighbourhoods of low biodiversity, and there are concerns about urban residents' disconnection from nature. A lack of ecological knowledge and pro-environmental behaviour has conservation implications. I surveyed Wellington households (n=453) and investigated possible predictors of residents' ecological knowledge about birdlife, bird feeding, and tree planting connected to birdlife. Three measures of knowledge were tested, species freelisting, neighbourhood bird knowledge, and photo identification. Key predictors of higher levels of ecological knowledge were increased frequencies of visiting local and regional parks, higher levels of garden space, and higher educational qualifications. However, all models had low predictive power. Species richness and perceived access to greenspaces were not significant predictors of ecological knowledge. Residents had a lower knowledge of native birdlife compared with exotic species across all measures. Forty-two percent of respondents fed birds, 10% targeting native species, and 36% planted trees for birdlife. Knowledge of neighbourhood birds and garden size were key predictors for each. This study shows the importance of urban greenspaces for ecological knowledge and behaviour, and efforts should be made to encourage visits to greenspaces.</p>