Development and application of a machine learning algorithm for classification of elasmobranch behaviour from accelerometry data.

Discerning behaviours of free-ranging animals allows for quantification of their activity budget, providing important insight into ecology. Over recent years, accelerometers have been used to unveil the cryptic lives of animals. The increased ability of accelerometers to store large quantities of high resolution data has prompted a need for automated behavioural classification. We assessed the performance of several machine learning (ML) classifiers to discern five behaviours performed by accelerometer-equipped juvenile lemon sharks (Negaprion brevirostris) at Bimini, Bahamas (25°44'N, 79°16'W). The sharks were observed to exhibit chafing, burst swimming, headshaking, resting and swimming in a semi-captive environment and these observations were used to ground-truth data for ML training and testing. ML methods included logistic regression, an artificial neural network, two random forest models, a gradient boosting model and a voting ensemble (VE) model, which combined the predictions of all other (base) models to improve classifier performance. The macro-averaged F-measure, an indicator of classifier performance, showed that the VE model improved overall classification (F-measure 0.88) above the strongest base learner model, gradient boosting (0.86). To test whether the VE model provided biologically meaningful results when applied to accelerometer data obtained from wild sharks, we investigated headshaking behaviour, as a proxy for prey capture, in relation to the variables: time of day, tidal phase and season. All variables were significant in predicting prey capture, with predations most likely to occur during early evening and less frequently during the dry season and high tides. These findings support previous hypotheses from sporadic visual observations.

Three decades of longlining in Bimini, Bahamas, reveals long-term trends in lemon shark Negaprion brevirostris (Carcharhinidae) catch per unit effort.

In Bimini, Bahamas, the consistent employment of longlines, beginning in 1982, provided a rare opportunity to explore population trends for large resident sharks. This study assessed three shallow water longline survey periods at this location; 1982-1989, 1992-2002 and 2003-2014, with the aim of determining trends in annual catch per unit effort (CPUE) for an IUCN listed near-threatened species, the lemon shark Negaprion brevirostris. A general additive model (GAM) was used to analyse the non-linear annual CPUE values over the entire 32-year research period. The GAM displayed high variability of annual CPUE, with a peak value of 0·026 N. brevirostris per hook day (hooks day(-1) ) in 2000. The temporal pattern of CPUE indicated an abundance trend with a complete cycle, from trough to trough, occurring over a period of approximately 18 years. The 1982-1989 survey period saw the highest proportion of mature individuals (19·8%) and the smallest average pre-caudal length (LPC ; 124·8 cm). The 1992-2002 survey period had the highest average annual CPUE (0·018 hooks day(-1) ), while the 2003-2014 research period saw largest average LPC size (134·8 cm) and the lowest average CPUE values (0·009 hooks day(-1) ) of the entire research period. The long-term trend identified in this study provides a baseline for future assessment.

Using genetic inference to re-evaluate the minimum longevity of the lemon shark Negaprion brevirostris.

A combination of mark-recapture and genetic sampling was used to extend the minimum longevity of an elasmobranch species and the life span estimate of the lemon shark Negaprion brevirostris was increased conservatively from 20·2 to 37 years. This increase in longevity means higher vulnerability and a longer recovery time from exploitation.

Occurrence and habitat use of the critically endangered smalltooth sawfish Pristis pectinata in the Bahamas.

This study documents and discusses recent (2002-2015) sightings and captures of smalltooth sawfish Pristis pectinata in the Bahamas. Movement patterns and habitat preferences of five P. pectinata are examined: two tracked with acoustic telemetry in Bimini and three tagged with pop-up archival transmitting tags in Andros. Historically, P. pectinata may have been distributed throughout the Bahamas; however, since 2002 only 61 encounters were recorded including: Andros (30), Bimini (19) and a handful across other Islands (12). In Bimini, all P. pectinata were >225 cm (stretched total length, LST) suggesting that it is not used as a nursery area. Pristis pectinata in Andros ranged from c. 80 to 450 cm (LST) indicating that this island might be an important nursery and breeding habitat. Pristis pectinata tracked in both islands remained at depths <3 m, often adjacent to mangrove habitats, displaying residency from 42 days (Bimini) to 180 days (Andros). These preliminary findings confirm the Bahamas as an important habitat for P. pectinata and emphasize the urgent need for national protection and management of this population.

Genetic structure, population demography and seasonal occurrence of blacktip shark Carcharhinus limbatus in Bimini, the Bahamas.

A longline survey was conducted from 2004 to 2014 to investigate the demographic population structure and seasonal abundance of the blacktip shark Carcharhinus limbatus in the Bimini Islands, the Bahamas. All individuals sampled (n = 242) were sub-adult or adults [70·1-145·1 cm pre-caudal length (LPC) range] with no neonates or YOY recorded in Bimini. Carcharhinus limbatus abundance peaked in September, coincident with the largest ratio of female to male sharks and a peak in fresh mating wounds on females. Mitochondrial control region (mtCR) DNA sequences were obtained from C. limbatus at Bimini to test whether Bimini C. limbatus are most closely related to geographically proximate populations sampled on the south-eastern coast of the U.S.A., the closest known nursery areas for this species. Nine mtCR haplotypes were observed in 32 individuals sampled at Bimini [haplotype diversity (h) = 0·821, nucleotide diversity (π) = 0·0015]. Four haplotypes observed from Bimini matched those previously found in the northern Yucatan (Mexico)-Belize and two matched a haplotype previously found in the U.S.A. Four haplotypes were novel but were closely related to the northern Yucatan-Belizean haplotypes. Pair-wise ΦST analysis showed that Bimini was significantly differentiated from all of the populations previously sampled (U.S.A. Atlantic, U.S.A. Gulf of Mexico, northern Yucatan, Belize and Brazil). This indicates that C. limbatus sampled from Bimini are unlikely from the described, proximate U.S.A. nurseries.

The behaviour and recovery of juvenile lemon sharks Negaprion brevirostris in response to external accelerometer tag attachment.

Behavioural responses of lemon sharks Negaprion brevirostris to a fin-mounted tag package (CEFAS G6A tri-axial accelerometer with epoxied Sonotronics PT4 acoustic transmitter) were measured in a controlled captive environment (n = 10, total length, LT range 80-140 cm) and in free-ranging sharks upon release (n = 7, LT range 100-160 cm). No changes were detected in behaviour (i.e. swimming speed, tailbeat frequency, time spent resting and frequency of chafing) between control and tagged captive shark trials, suggesting that the tag package itself does not alter behaviour. In the free-ranging trials, an initial period of elevated swimming activity was found in all individuals (represented by overall dynamic body acceleration). Negaprion brevirostris, however, appeared to recover quickly, returning to a steady swimming state between 2 and 35 min after release. Post-release tracking found that all sharks swim immediately for the shoreline and remain within 100 m of shore for prolonged periods. Hence, although N. brevirostris are capable of quick adaptation to stressors and demonstrate rapid recovery in terms of activity, tracking data suggest that they may modify their spatial use patterns post release. This research is important in separating deviation in behaviour due to environmental stressors from artefacts caused by experimental techniques.