New records of the Japanese seahorse Hippocampus mohnikei from Cambodia, Malaysia, Thailand and Vietnam, along with recently published studies from India and Singapore, have greatly expanded the known range of H. mohnikei within Southeast Asia. These new records reveal novel habitat preferences and threats to H. mohnikei in the region. Although the global conservation status of H. mohnikei is classified as Data Deficient according to the IUCN Red List of Threatened Species, new sightings indicate that this species is found in similar habitats and faces similar threats as other Hippocampus species that are considered Vulnerable.
Life-history variables for three incidentally captured species of seahorse (Kellogg's seahorse Hippocampus kelloggi, the hedgehog seahorse Hippocampus spinosissimus and the three-spot seahorse Hippocampus trimaculatus) were established using specimens obtained from 33 fisheries landing sites in Peninsular Malaysia. When samples were pooled by species across the peninsula, sex ratios were not significantly different from unity, and height and mass relationships were significant for all species. For two of these species, height at physical maturity (HM ) was smaller than the height at which reproductive activity (HR ) commenced: H. spinosissimus (HM = 99·6 mm, HR = 123·2 mm) and H. trimaculatus (HM = 90·5 mm, HR = 121·8 mm). For H. kelloggi, HM could not be estimated as all individuals were physically mature, while HR = 167·4 mm. It appears that all three Hippocampus spp. were, on average, caught before reproducing; height at 50% capture (HC ) was ≥HM but ≤HR . The results from this study probe the effectiveness of assessment techniques for data-poor fisheries that rely heavily on estimates of length at maturity, especially if maturity is poorly defined. Findings also question the sustainability of H. trimaculatus catches in the south-west region of Peninsular Malaysia, where landed specimens had a notably smaller mean height (86·2 mm) and markedly skewed sex ratio (6% males) compared with samples from the south-east and north-west of the peninsula.
Along with propagating the input toward making a prediction, Bayesian neural networks also propagate uncertainty. This has the potential to guide the training process by rejecting predictions of low confidence, and recent variational Bayesian methods can do so without Monte Carlo sampling of weights. Here, we apply sample-free methods for wildlife call detection on recordings made via passive acoustic monitoring equipment in the animals' natural habitats. We further propose uncertainty-aware label smoothing, where the smoothing probability is dependent on sample-free predictive uncertainty, in order to downweigh data samples that should contribute less to the loss value. We introduce a bioacoustic dataset recorded in Malaysian Borneo, containing overlapping calls from 30 species. On that dataset, our proposed method achieves an absolute percentage improvement of around 1.5 points on area under the receiver operating characteristic (AU-ROC), 13 points in F1, and 19.5 points in expected calibration error (ECE) compared to the point-estimate network baseline averaged across all target classes.