Oikos

Habitat fragmentation, driven by human activities, disrupts habitat connectivity and alters ecological processes through geometric and demographic fragmentation effects. Dispersal plays a fundamental role in shaping the distribution, abundance, and persistence of species in modified landscapes. While previous research looked at the evolution of dispersal strategies at the species level, community-level dynamics remain underexplored. Species exhibit diverse dispersal strategies to persist in modified landscapes, yet predicting how these strategies interact at the community level requires a more integrated approach. This study employed an individual-based simulation model to explore how fragmentation and other landscape characteristics influence community-level dispersal strategies. We tested the effects of varying fragmentation levels, environmental autocorrelation, habitat amount, and disturbance levels on the emerging distribution of dispersal distances within a community in modified and continuous landscapes. We hypothesised that fragmentation and other spatial patterns would significantly shape community composition, favouring particular dispersal strategies under specific environmental conditions. The findings reveal that higher disturbance levels and greater habitat amount increased the community-weighted mean of dispersal distance, while fragmentation showed only minor variation. Additionally, low autocorrelation was associated with the highest community-weighted mean of dispersal distance. These results highlight the importance of considering community-level dynamics when predicting ecosystem responses to landscape modification. By clarifying how landscape structure and disturbance shape community-level dispersal strategies, this study advances our understanding of the mechanisms underlying species persistence and community structure in modified landscapes.

The notion of "pollinator diversity" is central to most research and interpretations in animal pollination ecology. Nevertheless, when the term "diversity" is applied to pollinators its usage is often closer to the vernacular meaning (variety of kinds) than to concepts rooted in the "ecological diversity" tradition of community and statistical ecology. This paper attempts to fill a conspicuous knowledge gap in pollination ecology by presenting a comprehensive analysis of patterns of species abundance and diversity in a hyperdiverse insect pollinator assemblage from well-preserved Mediterranean montane habitats of southeastern Spain. Data on pollinator visitation to flowers of the community of entomophilous plants (288 species) were gathered over a 29-year period, and [~]95% of the pollinator individuals recorded were identified to species, totalling 46,401 individuals in 845 species. The shape of species abundance distributions (SADs) was virtually identical at regional (N = 56 sites) and local (one intensively studied site) scales, and SADs were best predicted by the log-series distribution. Pollinator diversity estimates corresponding to the first three Hill numbers (Species richness, Shannon diversity and Simpson diversity; 0D, 1D and 2D, respectively) were obtained for each plant species x site x year combinations ("sampling occasions", N = 472). Pollinator diversity measures varied widely among plant species; their frequency distributions were continuous, unimodal and strongly right-skewed; and variation was related to plant phylogeny, floral features (open vs. restrictive perianth, single flower vs. flower packet), and pollinator visitation to flowers and flowering patches. Pollinator diversity of individual plant species depended on habitat type, with those from dolomitic outcrops, rock cliffs and forest interior having the least diverse pollinators. 0D, 1D and 2D tended to vary independently of each other among habitats and years, revealing a complex spatio-temporal patterning of pollinator species richness and dominance. Estimated proportions of undetected pollinator diversity ("dark diversity") depended on insect order (highest for Diptera) and diversity measure (highest for 0D). Adoption of community ecology tools (SAD, sampling adequacy estimation, complementary diversity measures) to assess pollinator diversity will improve our ability to elucidate pollinator responses to natural and anthropogenic environmental change and permit hitherto unexplored questions in pollination ecology. "The ecologist sees in any measure of diversity an expression of the possibilities of constructing feedback systems or any sort of links, in a given assemblage of species" Margalef (1968, p. 19).

An organisms life history strategy is an attempt to optimize fitness, given environmental constraints and inherent demographic tradeoffs. As such, life history helps to shape an organisms ecological and evolutionary responses to environmental change. However, life history can also be shaped by the environment, as the organisms demographic rates respond--directly or through tradeoffs--to the new conditions. This feedback between life history and environment remains poorly understood, limiting our ability to predict the outcomes of environmental change. Here, we studied the effects of environmental change - specifically altered pollination services - on four perennial plant species. We conducted a field-based demography experiment that subjected naturally occurring populations of Delphinium nuttallianum, Hydrophyllum fendleri, Potentilla pulcherrima and Erigeron speciosus to three pollination treatments: ambient (control), reduced, or increased pollination. We estimated population growth rate ({lambda}) and 11 metrics describing life history strategy and demographic resilience from an Integral Projection Model we constructed for each species and parameterized with 4-5 years of census data. Although most life history metrics responded idiosyncratically to pollination treatment, we found consistent effects of pollination on generation time, longevity and, in three of four species, recovery time. Specifically, reduced pollination led to increased longevity, generation time, and recovery time, and increased pollination led to the opposite. These changes in life history resemble shifts along the slow-fast continuum; reduced pollination led to slower lives and increased pollination led to faster lives. This is consequential because generation time and longevity influence short- and long-term population dynamics - for example, by affecting demographic stochasticity and sensitivity to environmental stochasticity, or rates of adaptation to novel conditions. Notably, these changes occurred largely independent from changes in population growth. Altogether, our results highlight changes in life history as an important but underappreciated consequence of environmental change.

Substantial anthropogenic changes to the environment have motivated efforts to quantify temporal trends in population dynamics. While most ecological research has focused on the mean and variance of population density and reproduction, the frequency of these fluctuations through time may also be changing. We analyzed 1,563 datasets of population density and 1,456 datasets of plant reproduction (masting) across the globe. The average frequency of fluctuations increased by [~] 0.5 - 3% per decade within each time series, representing a moderate change (Cohens d {approx} 0.4) over a period of 60 years. We tested four hypothesized mediators of this trend: increased temperature, increased frequency of environmental forcing, increased intrinsic growth rate, and increased distance from a saddle at zero density. Although all hypotheses were rejected, changes in the frequency of environmental forcing and intrinsic growth rate exhibited positive correlations with changes in population fluctuation frequency as expected. Our results suggest that successive peaks in population and masting density fluctuations are becoming closer in time, which may reduce the effectiveness of predator satiation, resilience of food-webs, and the risk of critical transitions, such as population extinction. We suggest some alternative hypotheses for what may underlie this surprising global pattern.

Establishing protected areas is a promising tool to address the accelerating loss of biodiversity. However, protection levels are often low, and there is an ongoing debate over the most effective spatial configuration of reserves. This debate rarely considers trophic structure and ignores biodiversity outside protected areas. In this study, we investigate which reserve configurations best support species diversity and the persistence of high trophic levels, across systems and spatial scales, both inside and outside protected areas. Using a spatially explicit stochastic model, we assess how reserve architecture influences multiple conservation objectives across 27 empirical terrestrial, freshwater, and marine food webs. Specifically, we explore reserve architecture along three dimensions: the aggregation of protected areas, their proportion at the landscape scale, and the effectiveness level of protection measures. Our results show that having few but larger protected areas enhances all conservation metrics within reserves, while diversity within and outside reserves is relatively insensitive to reserve aggregation. Smaller and more dispersed reserves improve the overall abundance of species off-reserves through spillover effects. Reconciling all objectives inside and outside reserves becomes feasible when protection effectiveness is sufficiently high. Increasing the efficiency of protection allows for a reduction in the total amount of protected land without compromising conservation outcomes. Moreover, higher species dispersal facilitates the achievement of multiple conservation goals, supporting the implementation of architectures that enhance connectivity among reserves. These findings highlight the importance of an integrated approach combining spatial ecology and trophic functioning to optimize protected area planning under multiple objectives.

Given the accumulation of evidence that evolution can affect ecological dynamics, especially under global change scenarios, a key question is how such ecoevolutionary dynamics may change the coexistence of species and biodiversity in general. In the present article, we propose a graphical approach allowing to simultaneously discuss ecological coexistence and phenotype evolution. Our graphical approach allows tackling the two aspects in the same parameter space, allowing direct links between ecological and evolutionary perspectives. While evolution is often thought positive for the resilience of ecological systems, we first highlight it does not usually allow for better coexistence for the system as a whole. Even when focusing on the fate of the species that is evolving, evolution often leads to greater vulnerability. The graphical approach we propose is flexible and can be applied to all interaction types and covers variations in trade-off structures. Using this flexibility, we highlight how evolutionary effects can be positive or negative for coexistence, depending on these two components. Finally, we illustrate how the approach can be applied, using empirical examples derived from the literature. We thereby highlight the critical ingredients needed to inform the graphical approach, its potential use for proposing testable scenarios, but also clarify its limits.

Climate change is pushing species northward, where they will encounter novel abiotic conditions, such as novel daily light cycles and seasonal lengths, to which they will have to adapt. Despite the diversification of these adaptations being well described, the direct fitness consequences of the natural variation is rarely estimated in the field. To test the fitness effects of local adaptations to daylength and season length, we studied diapause induction and growth rate in the range expanding butterfly Lasiommata megera (wall brown). Using a common garden field-experiment (conducted near the northern range margin) where we manipulated the start of the last generation in the year that typically enter diapause, we compared populations from the southern-Swedish core range with populations from the northern-Swedish margin. Our results show differences between populations in diapause response and growth rate. In line with adaptive predictions caterpillars from the northern populations entered diapause earlier in the season compared to the southern populations. However, this difference was only present in larvae produced by the earliest individuals in the last adult generation. Consequently, the early laid eggs of the southern populations were more likely to produce an additional generation that turned out to be highly maladaptive. Additionally, caterpillars of northern origin grew faster compared to caterpillars from the southern populations, even though we found no clear evidence of prewinter larval mass affecting winter survival. The wall brown butterfly showed local evolution of seasonal timing traits, but just at a specific time of the season. This highlights the importance of local adaptations in northern-Swedish populations, during the early stages of the last annual generation. Despite that the additional generation is presently maladaptive, our fitness estimates suggests that a warmer climate is likely to favour the production of an additional generation.

Successful seagrass restoration will provide habitat for a variety of species. Here, ecological community assembly in a newly planted seagrass meadow has been modelled mathematically using a combination of numerical integration and a permanence-based method, and using real data to parametrise the models. We have studied the transient dynamics of the system: how the ecological communities assemble and change over a 100-year period. Using a trophic structure and a range of species pool sizes, we investigated how much variability there was in community size for a given sized species pool, whether it is possible to use early monitoring to predict the final community size, and to what extent monitoring gives an indication of final vs transient species. For the majority of cases modelled, the community either reached or was headed towards an endpoint community which was uniquely determined by the species pool. However, for 1.4% of cases, no unique endpoint community could be calculated. The simulated communities began to assemble within the first ten years, but 13% had still not reached their endpoint community even after 100 years. In 62% of our models, no consumer species colonised in the first two years, suggesting that monitoring should certainly be continued beyond a two-year period. We counted how many of the species that were present at any observation point in the 100 years would also be present in the endpoint community, and found that this proportion generally decreased with increasing species pool size, to an average of 86% when the species pool had 49-56 consumer species. By monitoring the community over the first ten years, it is not possible to deduce what the final community will be; however a very small number of fauna species present over the first ten years might be used to predict very small endpoint communities.

The mid-domain effect (MDE) predicts that geometric constraints drive unimodal species richness patterns within bounded gradients. However, the role of this effect in ecological networks is currently unexplored. Here we evaluate the role of the MDE in structuring interaction networks. We combine null-model simulations and empirical analyses of plant-pollinator and ant-plant networks along elevational gradients to assess whether the MDE can drive systematic variation in network structure. Our simulations demonstrated that the MDE alone can generate unimodal/U-shaped patterns in network metrics such as connectance, generality, and vulnerability. However, empirical networks only partially conformed to MDE predictions, with deviations indicating the likely influence of other ecological processes. MDE-based models best explained patterns in network-level specialization and nestedness, while only partially explaining patterns in connectance and generality. Because MDEs can shape interaction networks, MDE null models should be used when quantifying the influence of other ecological processes on network structure.

The phenology of organisms worldwide is shifting in response to changes in environmental conditions. There is growing concern that resulting timing mismatches among interacting species will negatively impact system-level properties, yet there is no general framework for evaluating community responses to changes in phenology. To address this gap, we developed a mathematical framework based on local stability analysis and used it to assess the resilience implications of phenological perturbations with a multi-year, highly time-resolved empirical dataset on subalpine plant-pollinator communities. The forecasted effects of phenological perturbations were largely independent of perturbations to species densities, indicating the potential for even small changes in phenology to disrupt the functioning of ecosystems that are otherwise highly stable.

O_LISpecies diversity potentially has a dual effect on communities: a generally positive effect on overall community biomass, reflecting the expression of species response and interaction traits, and a poorly characterised effect on mass-specific species contribution to ecosystem functions, reflecting the expression of their effect traits. Disentangling the effects of biodiversity on total biomass from those on effect trait expression would help settle a long-standing debate by clarifying how biodiversity relates to both facets of species effects on ecosystem functioning. C_LIO_LIFollowing the classical BEF approach, we calculate expected ecosystem function based on observed functioning in monoculture. We then derive a net biodiversity effect (NBE) and decompose it into four components: the classical complementarity and selection effects on total community biomass, and complementarity and selection effects on effect trait expression. The latter two reflect, respectively, a complementarity or facilitation in how effect traits influence the function, and how species with the highest potential for increasing the function become dominant in the community. C_LIO_LIWe illustrate this NBE decomposition with three ecosystem functions (nitrogen retention capacity, soil hydraulic conductivity improvement, and forage digestibility) measured in assembled communities under controlled experimental conditions of perennial grassland plants. Regarding nitrogen retention, we find a positive complementary effect via total biomass, but a negative biodiversity effect via effect trait expression. For hydraulic conductivity improvement, biodiversity effects are mostly mediated by total biomass. As for forage digestibility, we found a positive complementarity effect on trait expression, outweighed however by a negative selection effect. This analysis reveals how biodiversity may have contrasting effects on ecosystem functions via its impact on biomass and effect trait expression. C_LI SynthesisSeparating between the effect of biodiversity on plant community biomass and on effect trait expression at the community level is one important step towards understanding the pathways by which diverse plant communities drive ecosystem functioning.

Mediterranean mountain grasslands are ecosystems of high ecological and economic value. They are shaped by the dry and warm climate and land use, such as grazing, although the combined effects of both drivers remain poorly understood. In this study, we analyzed shifts in functional composition in thirty-two plant communities in Mediterranean mountain grasslands of the Pindos Range (Greece) by measuring five plant functional traits related to resource acquisition in dominant plant species. We examined the adaptive value of each trait as well as community-level responses along a well-defined two-dimensional gradient of grazing intensity and aridity, using mixed models and functional diversity analyses, and tested whether individual species trait shifts are related to aridity and grazing intensity. At the community level, aridity decreased plant height and leaf area whereas grazing only affected traits associated with tissue recovery such as high specific leaf area (SLA) and low community-weighted mean leaf dry matter content (LDMC). As aridity increased, plant height functional dispersion decreased. This convergence pattern indicates a shift towards more similar growth forms under arid conditions. Species-specific analysis indicated various responses of traits to the interaction of aridity and grazing that could not be detected using only community-level patterns. Overall, our findings demonstrate that aridity and grazing act through separate functional axes at the community level, while their combined effects emerge through species-specific trait plasticity.

Human activities are driving unprecedented environmental change, yet assessments of ecosystem resilience often overlook the rapid pace of change in the Anthropocene. Predator-prey systems are sensitive to the rate of environmental change and the whole system can collapse if predator population fail to promptly adjust to environmentally-driven shifts in resource population. Here, we investigate how different combinations of predator responsiveness and rates of environmental change influence the system vulnerability to critical transitions, explicitly addressing its interplay with magnitude of change. We found that, as predator responsiveness decreases, relatively slower rates and smaller magnitudes of environmental change leads to system collapse. Hence, even low and seemingly inoffensive total magnitudes of environmental change can be catastrophic if the rate of change is beyond a critical threshold. We propose considering predator responsiveness and current rates of environmental change as crucial factors in predicting the Anthropocenes impact on ecosystems.

Spatial heterogeneity of abiotic resources is essential for species coexistence. Ecological theory often assumes predefined heterogeneity of resources that constrains community dynamics, but the recent developments of meta-ecosystem ecology and zoogeochemistry highlight nutrient patterns could result from the interactions between the activities and movements of organisms and their abiotic environment. Here we investigate the mechanisms by which biotic-abiotic feedbacks could generate nutrient spatial heterogeneity in a simple plant-herbivore occupancy model where populations forage, recycle, and disperse in a homogenous landscape. By systematically varying organisms ranges of foraging and dispersal, and recycling levels, we found that limited dispersal of plants plays a key role on the emergence of nutrient patchiness by favoring small clusters of vegetation that shape their environment through consumption and recycling. However, herbivores could also create nutrient spatial heterogeneity when large foraging and dispersal ranges, and high recycling, allow them to efficiently track plant hot spots and to increase population persistence. Unexpectedly, strong aggregation of herbivore populations did not necessarily result in nutrient clustering. Rather than via recycling, herbivores mainly affected nutrient distribution indirectly, through their top-down impact on plant distribution. When evenly spread in the landscape, herbivore populations with large foraging ranges created areas of strong herbivory pressure unfavorable to plant colonization where nutrient can accumulate. These results can help understand the dynamical feedback between biota and abiotic resources. In a context where human activities alter both nutrient distribution and species abundances, a better understanding of this biotic-abiotic feedback will be key to anticipate the response of ecosystems to current perturbations.

The ecological and evolutionary processes determining species range limits remain poorly understood. Ultimately, range limits depend on the species abilities to persist under heterogeneous conditions, by adaptive differentiation and phenotypic plasticity, including transgenerational effects. To investigate ecological differentiation and transgenerational effects in the clonal invasive knotweed, Reynoutria japonica, in Europe, we conducted a two-phase transplant experiment: plants sampled along the entire latitudinal gradient were planted in three sites located at the northern range margin, mid-range and near the southern range margin, and then re-transplanted among all three sites after two years. Biomass production and allocation were generally not associated with latitude of origin and previous growth at the same site did not promote performance. We therefore find no evidence that adaptive differentiation or transgenerational effects contribute to the wide distribution of R. japonica in Europe. However, at the northern site, with a 25% shorter season, knotweed plants invested much less biomass below-ground, and the pattern was further strengthened in plants that had grown in the northern site in the previous generation. Overwintering below-ground rhizomes are essential for survival and spread. We further explored limiting climate conditions in a species distribution model for the European range and found that mean annual temperature and temperature annual range are the main predictors of the European distribution of R. japonica. Taken together, our study suggests that low temperatures and associated short seasons may pose a limit to the broad environmental tolerance of R. japonica and restrict its northward spread by reducing below-ground biomass accumulation.

Stingless bees are key pollinators in tropical ecosystems, yet their ecological dynamics remain poorly understood in highly seasonal environments such as the seasonally dry tropical forests of Ecuador. These ecosystems experience pronounced climatic seasonality, with sharp transitions between dry and wet periods that strongly affect floral resource availability. Understanding interspecific competition and niche partitioning in such systems is critical, particularly given the global decline of pollinators. We investigated resource use and niche dynamics in two native stingless bees, Melipona mimetica and Scaptotrigona sp., by quantifying pollen, nectar, and resin collection across seasons. Log-linear models were used to test the effects of species, season, and their interaction on resource use, while non-metric multidimensional scaling (NMDS) assessed niche overlap. Contrary to the expectation that niche overlap increases under resource scarcity, we found greater overlap during the wet season, when resources are more abundant. This suggests that both species converge on high-quality floral resources during peak availability, reflecting an adaptive response to strong environmental seasonality. Pollen use remained stable across seasons, consistent with generalist foraging behavior. In contrast, nectar collection increased significantly during the wet season, while resin exhibited a shared seasonal peak, likely associated with synchronized nest construction or maintenance. These findings reveal context-dependent competition dynamics and highlight the role of environmental seasonality in shaping pollinator interactions. Our study provides new insights into the ecology of threatened stingless bees and contributes to their conservation in tropical dry forest ecosystems.

1. Biotic insular systems differ from conventional islands because patch attributes change dynamically as patch-forming organisms develop. It therefore remains unclear whether the assembly mechanisms predicted by island biogeography theory (IBT) operate in such systems. Here, using epiphytic birds nest ferns (BNFs, Asplenium nidus) as a model biotic island system, we tested whether fungal and bacterial community diversity conform to species-area relationships predicted by IBT. With a stratified sampling scheme, we further evaluated the underlying mechanisms (passive sampling, disproportionate effects, and environmental heterogeneity) of species-area relationships, and assessed isolation effects using distance-decay patterns in community similarity. 2. We treated each BNF individual as a microbial island and categorized 24 BNFs into three size classes. Microbial and humus samples from multiple litter layers within each BNF individual were collected; microbial communities were characterized using next-generation sequencing, and humus chemical properties (pH and C:N ratio) were measured to characterize microhabitat conditions. To investigate mechanisms underlying species-area relationships, we applied a multi-scale rarefaction framework to partition diversity components. Spatial distances among BNFs were quantified to evaluate isolation effects. 3. Consistent with IBT predictions, both fungal and bacterial communities exhibited positive species-area relationships, indicating that larger BNFs harbored greater microbial richness. Diversity partitioning suggested that fungal richness increased through both disproportionate effects and environmental heterogeneity, whereas bacterial richness was primarily driven by environmental heterogeneity. Within larger ferns, greater heterogeneity in litter pH was associated with increased species turnover across litter layers, suggesting that decomposition-driven pH gradients create diverse microhabitats that promote microbial diversity. In addition, both microbial communities exhibited distance-decay patterns, indicating that isolation contributes to community assembly through dispersal limitation. 4. Synthesis. Our results demonstrate that BNFs function as a biotic insular system, in which both patch size and spatial isolation structure microbial diversity, consistent with predictions from IBT. Furthermore, we show that environmental heterogeneity generated by the growth of the habitatforming BNF mechanistically links island area to microbial diversity. Our study integrates both local habitat heterogeneity and regional spatial structure, highlighting the potential to extend IBT and metacommunity theory to organism-formed habitats.

O_LIAbundance of insect herbivores often depends on host plant suitability for their specialised immature stages. Suitability can be strongly influenced by both microclimate and the nutritional quality of the plants themselves. Where soil nitrogen is high, host plants tend to have high nutritional quality, but vigorous growth of surrounding vegetation reduces microclimatic temperatures. Thus, thermophilous insects may face a choice between host plants with optimal microclimates and those with optimal nutritional quality. C_LIO_LIWe investigated how microclimate and nitrogen content influence oviposition choices by the declining Small Copper butterfly, Lycaena phlaeas, on its host plant, Rumex acetosa. We predicted that warmer plants would have lower nitrogen content, and that butterflies would choose cooler, high-nitrogen plants during warmer ambient conditions. C_LIO_LIAlthough warmer R. acetosa plants had lower nitrogen content, L. phlaeas consistently chose to lay eggs on plants in warm microclimates, implying a trade-off between temperature and the nutritional quality of host plants. C_LIO_LIPatches of bare ground created by Talpa europaea (European Mole) near R. acetosa plants increased microclimatic temperatures and decoupled the negative correlation between nutritional quality and thermal suitability. C_LIO_LIOur results have implications for the conservation of thermophilous insect herbivores, especially close to their range margins and in the context of climate change. Rather than maximising host plant abundance or nutritional quality, management that creates suitable microclimatic conditions is likely to be critical. Our findings also suggest that, while nitrogen pollution may increase host plant nutritional quality, its negative impacts on microclimate will likely further reduce breeding habitat for L. phlaeas and other insects in grassland habitats. C_LI

Species tolerating the same environmental conditions can potentially colonize and thrive in the same habitats and eco-regions. Are any pair of those species equally probable to co-occur in the same community? Can we quantify the propensity of two species to co-occur together? Here, we focus on a simple but largely overlooked community-level pattern: the co-occurrence-occupancy curve, which relates the tendency of species to co-occur with others to their total occupancy across sites. We first define this empirical curve and then derive its expected shape under a random null model that assumes site equivalence and species independence. Building on these results, we introduce the Species Association Index (SAI), an occupancy-standardized measure that quantifies the tendency of a species to associate with others independently of its overall frequency of occurrence. The SAI enables meaningful comparisons among species with contrasting occupancies and provides a transparent benchmark against which departures from neutrality can be assessed. We illustrate the approach using two contrasting systems--tropical rain forest trees on Barro Colorado Island and organisms from Mediterranean rocky shores--highlighting both the generality of the co-occurrence-occupancy framework and its limitations.

Fishing deeply alters marine food webs structure and can drive the evolution of species traits, whether the species are directly targeted or not. Yet, studies rarely account for fisheries-induced evolution, and consequences are generally interpreted at the single-species level. Theory however predicts that eco-evolutionary dynamics within food webs can either promote biodiversity maintenance or accelerate its decline. In this study, we investigate how evolution affects the robustness of trophic networks under fishing pressure. Modifying evolution speed and the allocation of fishing effort across 458 structurally distinct allometric networks enables us to show that evolution most often enhances robustness. Network evolutionary response however becomes more variable (and possibly negative) as evolutionary rates increase and when fishing preferentially targets predators. By contrast, fishing strategies that concentrate effort on lower trophic levels, or distribute it more evenly, promote network persistence through evolutionary rescue while substantially reducing the risk of evolutionary collapse. Moreover, our results appear to be sensitive to the main forces governing ecological dynamics within the network such as competition or predation intensity. Finally, the consequences of network evolution differ across trophic levels. Evolution often drives the collapse of higher trophic levels while simultaneously promoting evolutionary rescue and enhancing diversity at lower levels through increased diversification, thereby generating a trade-off between vertical diversity (number of trophic levels) and total diversity. This highlights the importance of accounting for evolutionary dynamics and food web functioning in fisheries management, and suggests that reducing predator mortality may help prevent network evolutionary collapse.