Generative models are being used to address data scarcity, modality gaps, and semantic grounding in wearable Human Activity Recognition (HAR) and Health Monitoring (HM), but deployment evidence remains limited. This focused evidence-map review examines Scopus-indexed Q1/Q2 journal literature on generative Artificial Intelligence (AI) for wearable information fusion, focusing on datasets, validation practices, and deployment claims. A PRISMA-guided Scopus search, followed by Q1/Q2 journal filtering and backward reference checking, yielded a frozen 2023–2025 corpus of 22 studies: 17 primary, 4 reviews/surveys, and 1 framework paper. In this corpus, Generative Adversarial Networks (GANs) are the most frequent generative family for augmentation, class balancing, and biosignal synthesis. Large Language Models (LLMs) appear mainly as semantic or interface layers for activity description, pseudo-label assessment, knowledge retrieval, and decision-use prototypes, rather than as clinically validated systems. The primary studies show a consistent reality gap: downstream classification or recognition gains on benchmark tasks are commonly reported, but physiological plausibility, privacy evaluation, clinical or contextual grounding, and deployment-level generalization remain sparse or indirect. We map each study to information-fusion roles, assess dataset representativeness and alignment with intended deployment settings, and summarize validation coverage through a qualitative matrix. The corpus supports technical progress, but the evidence is narrower than deployment-oriented language sometimes implies. Stronger deployment claims require richer longitudinal wearable datasets, validation beyond benchmark utility, explicit privacy reporting, and clinical claims matched to the available evidence.
Trioza erytreae is a vector of Huanglongbing (HLB), a highly damaging citrus disease. Lemon plants (Citrus ×limon) are the preferred host for T. erytreae, although the underlying mechanisms behind this remain to be fully elucidated. A comparative proteomic analysis of T. erytreae nymphs in their fourth and fifth instars that were fed either lemon or sweet orange (SwO) was carried out to investigate the interaction with its hosts. A 24-hour sucrose feeding assay was conducted to understand proteomic responses to a nutrient-poor diet. Proteomic profiling using nanoscale liquid chromatography coupled to tandem mass spectrometry (nanoLC-MS/MS) identified a total of 1,477 psyllid proteins with high confidence. Oviposition and nymphal development were also evaluated across citrus hosts, revealing higher numbers of nymphs developing on lemon than on SwO. Feeding on SwO enriched pathways related to “transmission across chemical synapses” and “metabolism of proteins”. Responses observed under a 24-hour sucrose-only diet enriched the biological processes “response to external stimulus”, “response to stress” and “cytoskeleton organization”. In contrast, these enrichments were absent on lemon host, suggesting that lemon provides a more favourable environment for psyllid development. In addition, nymphs developed on lemon exhibited enhanced energy metabolism and an increase in translation initiation factors. Overall, the results demonstrate that development strongly depends on host plant species, with SwO impairing optimal growth and lemon promoting successful nymphal development.
The distribution and migration of fin whales (Balaenoptera physalus) remain poorly understood in parts of the North Atlantic. One of these regions is the Algarve in southern Portugal, which is shaped by seasonal upwelling and the exchange of Atlantic and Mediterranean waters. This area represents a potentially critical yet poorly understood habitat for this species. This study presents the first comprehensive assessment of fin whale abundance and site fidelity in the Faro area, located approximately between 36–37° N and 8–7° W. Data was collected from 2020 to 2024 through photo-identification and GPS vessel tracks, in collaboration with a local whale-watching company. A regional catalog of 86 individuals was established, and an open spatially explicit capture-recapture model estimated a population of 1.476 individuals. Sightings peaked in spring, likely linked to phytoplankton blooms and foraging opportunities. While re-sighting rates were low, some individuals returned across years, indicating possible seasonal site fidelity. Our results demonstrate the value of data collected from platforms of opportunity in an under-surveyed region, providing a foundation for ongoing monitoring efforts. Results also highlight the Faro area’s ecological relevance as a seasonal feeding ground for fin whales and offer insights to support future conservation in the eastern North Atlantic.
The ocean is undergoing significant changes, including warming, acidification, and deoxygenation, which pose great challenges to marine biodiversity. However, most models projecting the impacts of climate change on marine species overlook predictor variables critically meaningful for species' ecologies such as pH and dissolved oxygen. The recent release of high-resolution projections of different future climate-change scenarios offers the opportunity to explore species redistribution under multiple threats beyond ocean warming. Accordingly, we conducted a global comparative analysis to study the impact of incorporating predictor variables describing pH and dissolved oxygen into marine species distribution models. We used models trained for 268 cold-water coral species to project potential future distributions for different climate and dispersal scenarios over different time periods. We found that, irrespective of scenario or period, models using pH and dissolved oxygen projected 11.5-21.4% higher impacts of climate change than those without them. For instance, by the end of the century under a high emission scenario, models including pH and oxygen projected an average range contraction of 48.2% for cold-water corals under a no-dispersal scenario, compared with a 26.8% contraction projected by models excluding these two predictors. Given the substantial differences in the predicted distribution patterns and the biological importance of these variables, we highlight that researchers should consider more diverse sets of predictor variables when predicting future range shifts for marine biodiversity assessments under climate change.
Species distribution models (SDMs) have been increasingly combined with thermal performance data to enhance their transferability and to provide a physiological explanation for the predicted geographic patterns. Yet, while it is widely acknowledged that thermal sensitivities may vary among biological traits, it remains largely unexplored to what extent predictions obtained using hybrid models depend on the choice of trait, and, hence, the type of thermal performance data considered. In this study, we examine the thermal performance of three fitness-related traits, namely spore germination rate, maximum quantum yield of photosystem II (Fv/Fm), and growth rate, in the brown seaweed Dictyota dichotoma, a common macroalga along European coastlines. To predict the species' current and future distributions, we constructed a traditional correlative SDM, fitting a comprehensive dataset of occurrence records and environmental data, as well as three distinct hybrid SDMs, incorporating the thermal responses of each fitness-related trait. Despite considerable differences in thermal performance among traits, with Fv/Fm and growth displaying the broadest and narrowest thermal performance curves, respectively, predictions made by the distinct hybrid SDMs were largely congruent under both current and future conditions. This seems to be linked to the models' ability to detect even small differences in trait performance, which are then used to fine-tune suitability predictions and achieve the best fit. In addition, predictions made by hybrid SDMs were similar to those made by the correlative SDM, demonstrating that when occurrence data are representative for the species' distribution, environmental predictors are relevant, and model responses are ecologically constrained, correlative and hybrid approaches can converge. Importantly, all models predicted range contractions for D. dichotoma at the warm-edge limit in the Mediterranean Sea under future climate change scenarios. This suggests that these populations may be particularly vulnerable, although cooler local microhabitats could mitigate some of these effects.
