Species distribution models (SDMs) commonly produce probabilistic occurrence predictions that must be converted into binary presence-absence maps for ecological inference and conservation planning. However, this binarization step is typically heuristic and can substantially distort estimates of species prevalence and community composition. We present MaxExp, a decision-driven binarization framework that selects the most probable species assemblage by directly maximizing a chosen evaluation metric. MaxExp requires no calibration data and is flexible across several scores. We also introduce the Set Size Expectation (SSE) method, a computationally efficient alternative that predicts assemblages based on expected species richness. Using three case studies spanning diverse taxa, species counts, and performance metrics, we show that MaxExp consistently matches or surpasses widely used thresholding and calibration methods, especially under strong class imbalance and high rarity. SSE offers a simpler yet competitive option. Together, these methods provide robust, reproducible tools for multispecies SDM binarization.

Human activities are rapidly eroding the biodiversity of most ecosystems, threatening the myriad contributions they provide to nature and people. Protected areas are often seen as key management tools for their conservation. However, the lack of historical baselines hinders our ability to fully assess these declines and the extent to which protected areas can compensate for decades of human-mediated degradation. Using a Bayesian framework, we modelled 22 fish community contributions across 2,800 tropical reefs and predicted their levels under counterfactual scenarios to compare the relative benefits of marine protected areas (MPAs) and anthropogenic impacts on unprotected reefs. We show that human activities have significantly reduced fish biodiversity- and biomass-related contributions with, for example, a 120% decline in piscivore biomass, corresponding to a net loss of 19 kg per hectare of reef. In contrast, the benefits of MPAs appear comparatively low, with conservation efforts potentially offsetting only 5% of this decline. Ultimately, only old and fully protected areas provide marked benefits to nature and people. This suggests that even if we drastically increase our protection efforts across the ocean (30% coverage by 2030), we cannot expect short-term socio-ecological benefits to counterbalance a long history of human footprint. A desirable future for nature and people thus requires a paradigm shift in our relationship with ecosystems and their biodiversity, beyond MPA establishment.

ABSTRACT Habitat configuration governs the movement of organisms across landscapes, thereby shaping both population structure and community assembly. While theoretical and empirical studies have assessed how habitat connectivity simultaneously influences intra‐ and interspecific diversity, direct comparisons across contrasting biogeographic regions remain limited. Here, we investigate patterns of genetic and species β‐diversity in tropical reef fishes across two ocean basins with distinct spatial configurations: the Caribbean Sea and the Western Indian Ocean. Using a comparative framework based on species occurrence data from five fish families and single nucleotide polymorphism (SNP) data from 19 species, we detected significant isolation by distance at both population and community levels in the Western Indian Ocean, but only at the community level in the Caribbean Sea. Additionally, genetic and species β‐diversity were positively correlated among species in the Western Indian Ocean, but not in the Caribbean Sea. Together, these results suggest that the shorter inter‐reef distances of the Caribbean Sea promote higher connectivity, leading to a decoupling of intra‐ and interspecific β‐diversity patterns.

ABSTRACT Describing and understanding diversity patterns from populations to communities remains a fundamental challenge in ecology and evolutionary biology. The main barriers to address this challenge are linked to the difficulty of concurrently assessing diversity from intra‐ to interspecific level and of obtaining genetic data for multiple species, particularly for hyperdiverse taxa like tropical reef fishes. Here, we propose environmental DNA (eDNA) metabarcoding with level‐specific primers (the conserved 12S and the hypervariable D‐loop) as a standardized approach to bridge this gap. Using 21 eDNA samples from the Caribbean Sea, we estimated species diversity across all teleosts (378 Molecular Operational Taxonomic Units, MOTUs) and intraspecific haplotype diversity for several grunt species (Haemulon 1149 Amplicon Sequence Variants, ASVs). Our results revealed no covariation between the haplotype diversity of Haemulon and the overall species richness detected. However, we detected isolation‐by‐distance patterns at the interspecific level and species‐dependent isolation‐by‐distance at the intraspecific level. Notably, the effect of distance led to a positive covariation between fish species dissimilarity and Haemulon plumierii haplotype dissimilarity across samples. By enabling simultaneous, standardized monitoring of biodiversity across scales, eDNA opens new perspectives on unifying biodiversity assessments and understanding the eco‐evolutionary processes that shape diversity patterns from genes to communities.

This study examines the adaptive capacity of French Mediterranean fisheries as perceived by the fishers themselves. Recognizing the role of fisheries for food security and economic livelihoods, this research addresses the dual challenges of maintaining sustainable fisheries while adapting to environmental and socio-economic changes. The Mediterranean Sea, one of the world's most overexploited marine regions, faces significant pressures from overfishing, climate change, and socio-economic fluctuations among others. This study employs participatory workshops and semi-structured interviews with 48 fishers across the French Mediterranean coast to understand the adaptive responses of local fisheries to these challenges. Based on an existing framework, we decompose adaptive capacity into five domains: assets, flexibility, social organization, learning, and agency. The results show that flexibility is the most mobilized domain, indicating that the ability of fishers to switch between opportunities and adaptation options is crucial for sustainability. Short-term crises such as fuel price hikes and the COVID-19 pandemic highlighted the reliance on public subsidies and social organization, whereas long-term environmental changes emphasized the need for continuous learning and flexibility in fishing practices. By integrating local knowledge with scientific assessments, we aim to provide a comprehensive view of fisheries' adaptive capacity, emphasizing the value of interview methods in capturing the nuanced realities faced by fishers. These insights are critical for developing co-constructed adaptive strategies that align with both local and broader-scale management goals. The results of our interviews highlight empirical adaptation measures, based on fisher's lived experiences, that policy makers need to consider to promote sustainable fisheries in the Mediterranean. We provide a robust framework to address the complex challenges facing Mediterranean fisheries, which could be replicated in other fisheries of the world and contribute to their long-term resilience and sustainability.

Abstract The metabolic processes sustaining coral reefs, from carbonate and primary production to secondary production, remain poorly integrated and rarely quantified simultaneously at global scales. This hampers our ability to predict global responses to accelerating human pressures and manage coral reef functioning. Using metabolic scaling and bioenergetic models applied to surveys from 1,100 reefs worldwide, we provide a global, standardized quantification of 14 ecosystem functions spanning benthic (corals and algae) and fish communities. Our analysis reveals a continuous functional spectrum of global coral reefs organized along four dominant axes: 1) primary production, 2) calcification and habitat structure, 3) secondary biomass production and consumption, and 4) biomass turnover. Functions mediated by fish and benthic communities show weak associations at the global scale rather than tight coupling. Climate stressors reduced calcification and local human impacts lowered secondary production. Yet these directional effects unfolded against a backdrop of substantial natural variability in reef functional configurations, such that heavily and minimally impacted reefs overlap substantially in the global functional space. Temporal analyses across three representative reef systems further revealed that functional trajectories following disturbance are context-dependent, with no universal pattern of recovery across locations. This continuous and context-dependent functional spectrum challenges the notion of universal functional benchmarks and supports locally tailored conservation strategies.

Equilibrium concepts and the expectation of compensatory density dependence remain fundamental to fisheries science, but stock collapses and an increasing appreciation of environmental factors have raised questions about their real-world applicability. To explore the demographic variability of harvested marine fishes, we have calculated metrics commonly used in conservation biology to describe the demographics for 77 assessed stocks from the North Atlantic and Northeast Pacific Oceans using life-tables. We found that median annual population growth rates ([Formula: see text]) were centered around 1, and surprisingly, they were only slightly higher when the effect of fishing was excluded. For most stocks, as abundance declined, [Formula: see text] tended to increase and become more variable as would be expected from compensatory dynamics. The population growth of several stocks was sustained by a limited number of years with exceptionally high rates. However, the ability of a stock to increase from low abundance appeared largely independent of life history characteristics and exhibited stronger geographical differences among stocks of the same species (notably Atlantic cod). Life history characteristics alone were poor predictors of annual population growth or future recovery potential, whereas regional factors appeared to be more influential. Overall, recovery potential remained relatively high, with simulations indicating that 62 of the stocks would be highly likely to double in size within 20 years in the absence of fishing. Low recovery potential was exclusively observed in stocks with a low median [Formula: see text] and low variability in [Formula: see text]. These results suggest that understanding stock-specific (rather than species-specific) demographic parameters is necessary to promote sustainable management or develop rebuilding plans for collapsed stocks.

Abstract Understanding how climate change and exploitation affect marine species distribution is key for biodiversity conservation. While warming is typically expected to drive species poleward and into deeper waters, unexpected patterns such as equatorward shifts and shallowing also occur. Fishing pressure may further contract species’ ranges. Yet, the mechanisms underpinning these 3-dimensional range shifts remain largely unresolved. Using 92,961 bottom trawl samples across the European continental shelf, covering 607 fish species over four decades, we show that deepening is predominant, with 60% of populations experiencing an increase in their average depth. Contrary to expectations, latitudinal shifts are balanced, with 49% of populations shifting poleward and 51% equatorward. Here, we quantify how population relocations are driven by the interplay between environmental factors and species traits as species track their optimal ecological niches. We also highlight that the recent reduction in fishing pressure has likely facilitated spatial expansions and increases in populations abundance.

Human pressure on global ecosystems is strongly mediated by how accessible they are, yet we know little about accessibility changes worldwide. Here, we quantify the changing accessibility of global coral reefs, an entire ecosystem, to the nearest human settlement and nearest markets between 2001 and 2023. Our study reveals that the small changes in average coral reef accessibility mask the substantial shifts occurring at the national and regional scales. Decreasing accessibility is related to coastal vegetation restoration blocking access, while increasing accessibility is predominantly due to the expansion of rural settlements. We also highlight that coral reefs in marine protected areas tend to be slightly less accessible, but as travel time in surrounding areas increases, protected areas may face growing pressure. Our results provide insights into how the accessibility of the world’s coral reefs is changing and highlight what that may mean for their protection.

The Convention on Biological Diversity’s ambitious target to “effectively conserve” 30% of the world’s oceans in protected areas by 2030 risks being undermined by a primary focus on total area protected rather than realized conservation outcomes. Area-based assessments can overstate success because few marine protected areas (MPAs) possess strong restrictions and high compliance. Here, we estimate the extent that conservation outcomes are realized across the global MPA network for shallow reefs using standardized fish survey data. We first compare observed fish biomass from 552 surveyed MPA zones in temperate and tropical seas to modeled counterfactuals in the absence of protection, then extend these findings across the global network using key MPA attributes and predictive models. We estimate that at best, a quarter of the global MPA network is likely delivering meaningful conservation outcomes, with effectiveness strongly contingent upon the level of compliance with fishing restrictions. If shallow reef responses are representative of other marine ecosystems, then <2% of the ocean is effectively protected, compared to 9.8% presently reported. The unsupported assumption that MPA designation and management attributes directly translates to tangible biodiversity outcomes has apparently resulted in pervasive overestimation of conservation achievement. Critically, long-term success of MPAs will require further research, informed governance, and improved strategic policies that focus on the intricate mechanisms underlying compliance with mandated area restrictions.

Urbanization is rapidly expanding along the world's coasts, transforming natural ecosystems into increasingly uniform artificial environments. Yet, the patterns and processes shaping marine biodiversity under urban pressure remain poorly understood. We investigated whether coastal urbanization drives marine communities toward greater similarity, a process well documented in terrestrial and freshwater systems as urban biotic homogenization but rarely examined in the sea. We analyzed fish communities using 222 seawater environmental DNA samples collected across Mediterranean seaports and less urbanized coastal sites. Seaports exhibited higher local species richness than less urbanized sites, yet their communities were markedly more similar to one another, revealing a local-regional decoupling paradox in which gains in local diversity coincided with biotic homogenization at the regional scale. This pattern emerged in the absence of non-indigenous species, which are commonly invoked as main drivers of homogenization. Instead, homogenization was primarily driven by increased occupancy (i.e., the proportion of sites occupied) of disturbance-tolerant native taxa, half of which showed occupancy levels approximately 2.6 times higher in seaports than elsewhere. Biotic homogenization was evident at both taxonomic and phylogenetic levels, indicating that disturbance-tolerant taxa are phylogenetically clustered. Overall, urban pressure emerged as the dominant driver of community similarity, exceeding spatial and environmental predictors. Together, these results show that coastal urbanization emerges as a fundamental and underappreciated driver of marine biodiversity reorganization.

ABSTRACT Video surveys from drones or planes equipped with cameras have become invaluable tools for monitoring populations of rare and elusive megafauna over increasingly large spatial scales. To analyze hours of such video surveys, automated species detection approaches based on deep learning are now widely used, but simply tallying detections on the successive frames of videos leads to double counting of individuals, significantly biasing population size estimates. Here, we leverage a multiple object tracking approach that links detections across video frames to derive counts of unique individuals. Our tracking‐by‐detection framework integrates the full pipeline from aerial video surveys to automated individual count and estimation of minimum population abundance. Applied to aerial video surveys of dugongs ( Dugong dugon ) on the west coast of New Caledonia in winter, our framework automatically counted individuals with a slight negative bias of −5.7% (−15 individuals over a total of 265 ground‐truth individuals) compared to a baseline method relying on the sum of detections that led to a 48 times overestimation. Our tracking‐by‐detection framework directly estimated the number of unique individuals per flight, leading to a mean seasonal index of abundance of 42 individuals (±28 SD) in the surveyed area. Overall, 35 individuals were missed across flights, representing 13.2% of all ground‐truth individuals. Among missed individuals, two‐thirds were calves superposed to their mothers, stressing the challenge of occlusion when counting group‐living animals. Despite being applied to entire videos with very low prevalence of dugongs, our framework generated just 8% of false positives (20 instances) across all flights, mostly due to misidentifications with other species. Illustrated through the case of dugong, our tracking‐by‐detection framework more generally provides a scalable method to support digital population surveys, benefitting long‐term monitoring and conservation programs of vulnerable megafauna.

Predicting marine species distribution and abundance is essential for effective conservation and management. Yet, it remains challenging in data-limited regions where traditional biodiversity surveys are logistically or financially constrained. Combining underwater visual census and eDNA fish sampling across the northwestern Mediterranean Sea, we tested a novel modelling framework that uses eDNA metabarcoding sequences to complement socio-environmental covariates in predicting species-specific local abundances. The eDNA-derived community proxies revealed ecological gradients complementary to visual census data, helping to distinguish sites dominated by coastal demersal fishes versus offshore predators, territorial reef fishes versus mobile dispersers and benthic versus pelagic species. Using joint Species Distribution Models (jSDMs), within the Hierarchical Modelling of Species Communities (HMSC) framework, we compared the predictive performance of models based solely on socio-environmental covariates with those that additionally incorporated eDNA-based information. Including eDNA information significantly improved model fit for 16 out of 26 species, including the endangered dusky grouper (Epinephelus marginatus), and contributed to one-third of explained variance in species local abundances on average. Synthesis and applications. This study demonstrates that integrating eDNA metabarcoding data into species distribution models can improve fish abundance predictions, especially for site-attached and reef-associated species. This approach provides a scalable and cost-effective tool for monitoring, impact assessment, spatial planning and adaptive management of marine resources.Read the free Plain Language Summary for this article on the journal's .

Species Distribution Models based on Convolutional Neural Networks (CNN-SDMs) have recently emerged, demonstrating greater effectiveness than traditional SDMs in several contexts. A limited number of studies, however, have focused on species abundance patterns, as the datasets available for this purpose are generally too small to effectively learn a deep learning model with millions of parameters. Our study demonstrated that CNN-SDMs can circumvent the small sample size of species abundance datasets through the combined use of a large presence-only species dataset and transfer learning to significantly improve the performance of abundance-based CNN-SDMs. Applied to Mediterranean coastal fishes, our approach significantly improves the abundance prediction performance of CNN-SDMs, with average gains of 35% (D-squared regression score). This allows CNN-SDMs to perform better than classical SDMs in abundance prediction, with average gains of 10%. These gains are stemming from enhanced abundance predictions for rare species and where widespread species are locally rare.

Monitoring optical properties of coastal and open ocean waters is crucial to assessing the health of marine ecosystems. Deep learning offers a promising approach to address these ecosystem dynamics, especially in scenarios where gap-free ground-truth data is lacking, which poses a challenge for designing effective training frameworks. Using an advanced neural variational data assimilation scheme (called 4DVarNet), we introduce a comprehensive training framework designed to effectively train directly on gappy data sets. Using the Mediterranean Sea as a case study, our experiments not only highlight the high performance of the chosen neural network in reconstructing gap-free images from gappy datasets but also demonstrate its superior performance over state-of-the-art algorithms such as DInEOF and Direct Inversion, whether using CNN or UNet architectures.

In a context of rapidly declining biodiversity, knowing the distribution of endangered species is critical to ensure the protection of the areas they occupy. To achieve this, species distribution models (SDMs) typically use a range of variables to determine the suitability of an area to a given species, which enables scientists to produce species distribution maps. Most SDMs are statistically limited in the number of predictors they can take into account, which leads to using summary variables such as temperature yearly average or bathymetry extrema. Deep-learning-based SDMs have been proposed to tackle this limitation by bringing powerful implicit feature extractors. Here, we describe a new model to advance the adaptation of such Deep-SDMs to marine environments and take advantage of the knowledge of the environmental seascape around surveyed points. We used data from the Reef Life Survey data set to predict the presence of 1,796 fish species around all of Australia, in a wide range of climates. The environmental seascape around each point was encoded as a raster image populated with 15 environmental variables, 4 human activity variables, and completed by a 10-year time series of temperature anomaly. The model performance was highly dependent on the species, with a 0.95 F1 score for the best performing species, Notolabrus parilus , but rapidly decreased, with the 100th best species having a 0.38 F1 score. ### Competing Interest Statement The authors have declared no competing interest. Agence Nationale de la Recherche, https://ror.org/00rbzpz17, ANR-21-AAFI-0001-01

Even though environmental DNA metabarcoding is revolutionizing biomonitoring, many critical steps remain unstandardized, leading to arbitrary choices, particularly regarding the selection of metabarcode, clustering method and similarity threshold, among others. Additionally, these studies were hindered by biases resulting from the presence of mislabeled sequences in international databases such as GenBank and the lack of explicit definitions for taxonomic resolution. To address these issues, we developed a robust framework to compare the performance of 22 metabarcodes derived from the same mitogenomes (all available for Actinopterygians in NCBI) against a standardized taxonomic baseline based on COI Barcode Index Numbers (BINs). This framework allows for the separate quantification of over-splitting (splitting the same taxon/BIN) and over-merging (merging different taxon/BIN). Comparison of OTUs obtained with multiple de novo clustering methods to BINs confirmed the metabarcode ranking based on error sums. Although each metabarcode exhibited varying sensitivities to over-merging or over-splitting errors, the clustering threshold emerged as the most important factor influencing biodiversity estimates whatever the clustering method. This led us to propose optimal thresholds for each metabarcode to delineate taxonomic levels (metabarcode gaps). Additionally, we found that taxonomic resolution varied significantly among genes, orders and community diversity, but independently of metabarcode length. Overall, the choice of metabarcode and clustering threshold should aim to minimize over-merging or over-splitting while ensuring accurate lower taxonomic delineations. A set of documented R functions makes this evaluation of taxonomic resolution easily applicable to any other taxonomic group for which a representative set of full genes or mitogenomes is available.

Coastal ecosystems are affected by numerous direct anthropogenic pressures (urbanization, fisheries, etc.) and global changes (climate change, biological invasions, etc.). Long-term monitoring of their biodiversity is crucial for (1) diagnosing disturbances as early as possible, (2) evaluating conservation (Marine Protected Areas) and restoration actions, and (3) better predicting the dynamics of their functioning. However, national and international programs are often limited in time (1-4 years) and offer few opportunities for standardized long-term monitoring in partnership with local stakeholders. To address this gap, French organizations, including the UMR Marbec and CEFE research labs, the Agence de l’Eau Rhône Méditerranée Corse, and the companies Spygen and EDF have collaborated to set up the ‘Sentinel Marine Areas’ network as part of the VigiLife international initiative. Since 2023, 13 Sentinel Marine Areas (SMA) distributed across the different maritime fronts of mainland France (Mediterranean, Atlantic, and English Channel) have been monitored annually using eDNA metabarcoding technology to identify the fish and crustacean species present. This non-destructive, highly efficient monitoring technology is ideal for assessing coastal biodiversity. Each SMA is centered on an area of interest (marine reserve, wind farm, thermal discharge from a nuclear power plant, artificial reef, entry point for non-native species), and several water filtrations are carried out by boat both inside and outside this area of interest. Each transect is conducted over 30 minutes during which 30L of seawater are collected and filtered 1m below the surface, sometimes complemented by filtration in the mesophotic zone (30-150m depth). Samples are then sent to the laboratory and processed using a standardized bioinformatic workflow to obtain a list of the species present. The objectives of this monitoring and research program are (i) to assess the influence of each area of interest on coastal biodiversity and its evolution over time, and (ii) to establish a network of sentinel sites to monitor the effects of climate change and the arrival of new species, as well as the effects of direct human pressures. During the first campaign in 2023, DNA copies of more than 140 fish species and 200 crustacean taxa were detected. These initial results have helped to update certain distribution ranges, with some species, such as the salema porgy Sarpa salpa, detected more than 500 km north of their former known northern limit. In connection with these northward shifts, we have also identified unprecedented species co-occurrences, the consequences of which on the functioning of coastal ecosystems are still unknown. The ‘Sentinel Marine Areas’ network is intended to become a permanent monitoring program, in particular by making local partners autonomous in carrying out the sampling, and to be extended to other areas of interest in France, including overseas, as well as to other territories abroad.

Environmental DNA (eDNA) metabarcoding is changing the way biodiversity is surveyed in many types of ecosystems. eDNA surveys are now commonly performed and integrated into biodiversity monitoring programs and public databases. Although it is widely recognized that eDNA records require interpretation in light of taxonomy and biogeography, there remains a range of perceptions about how thoroughly records should be evaluated and which ones should be reported. Here, we present a modular procedure, available as an R script, that uses a set of five steps to assess the confidence of species-level eDNA records by assigning them a score from 0 to 5. This procedure includes evaluations of the known geographic distribution of each taxon, the taxonomic resolution of the marker used, the regional completeness of the reference database, the diversification rate, and the range map of each taxon. We tested the procedure on a large-scale marine fish eDNA dataset (572 samples) covering 15 ecoregions worldwide, from the poles to the tropics, using the teleo marker on the mitochondrial 12S ribosomal gene. Our analysis revealed broad variation in the average confidence score of eDNA records among regions, with the highest scores occurring along the European and Eastern Atlantic coasts. Generalized linear models applied to record covariates highlighted the significant influences of latitude and species richness on low confidence scores (< 2.5). The polar regions notably displayed high proportions of low confidence scores, probably due to the limited completeness of the regional reference databases and the taxonomic resolution of the teleo marker. We conclude that only records with high confidence scores (> 2.5) should be integrated into biodiversity databases. The medium (2.5) to relatively low-confidence (< 2.5) records correspond to species that require further investigation and may be integrated after inspection to ensure high-quality species records.