Species distribution models (SDMs) address the whereabouts of species and are central to ecology. Deep learning (DL) is poised to further elevate the already significant role of SDMs in ecology and conservation, but the potential and limitations of this transformation are still largely unassessed.
North America.
2009–2021.
Mammals, amphibians, reptiles, ants, and butterflies.
We evaluate DL SDMs for 2299 terrestrial vertebrate and invertebrate species at continental scale and 0.00833° resolution in a like-for-like comparison with latest implementation of conventional SDMs. We compare two DL methods (a multi-layer perceptron (MLP) on point covariates and a convolutional neural network (CNN) on geospatial patches) against non-DL SDMs (Maxent and Random Forest), emphasising fair comparison and induced pitfalls from information leakage, species imbalances, and location biases.
On average, DL models match, but do not surpass, the performance of non-DL methods. DL performance is moderately to substantially weaker for species with narrow geographic ranges, fewer data points, and those assessed as threatened and hence often of greatest conservation concern. Furthermore, information leakage across dataset splits substantially inflates performance metrics, especially of CNNs.
Our results indicate that biases known from conventional SDM settings are strongly amplified for DL models. Although recent advancements in DL draw promising new avenues for ecological process modelling, their benefits beyond improved numerical performance can only be met when pitfalls are accounted for. Realising the potential of DL in its entirety will thus require a closer collaboration between ecology and machine learning disciplines.
The role of each animal species in an ecosystem is largely determined both by the resources it uses and the behaviours through which these resources are obtained. Even in well-studied vertebrate groups, like birds, quantitative data on the relative use of different food resources in the context of foraging strategies are generally lacking. Most analyses in macroecology, macroevolution and conservation biology are therefore limited to simplified dietary categories, ignoring the specific foraging behaviours and substrates used to access resources. Here we present AVONICHE, a dataset quantifying proportional membership in 32 foraging niches, representing a combination of dietary categories and associated foraging strategies used by all bird species.
Species-level information on the proportional use of foraging niches, each of which is defined as a particular foraging strategy within a specific dietary category (e.g., invertebrate feeding is subdivided into 7 foraging niches based on different foraging behaviors).
Global.
Present.
All bird species (Class Aves). To allow integration with global phylogenies and other data resources published in future, we align species-level niche data with four different taxonomic treatments: BirdTree (9993 species), Clements/eBird (10,661 species), BirdLife International (10,999 species) and the new AviList taxonomy (10,981 species).
Spreadsheet (.csv).
Homogenisation (decreasing beta-diversity) among biological assemblages is often interpreted as being caused by already-widespread species increasing. The link between individual species level trends and homogenisation between assemblages, however, has not been fully addressed, with most studies focused solely on either assemblage level or species level changes. Here we aim to test the widely held hypothesis that homogenisation is driven by the decrease of localised species and increase of those already widespread using species contribution to beta-diversity.
North America, Europe, South Africa.
1970 to 2019, 1966 to 1996, 11,700 years ago to present, 2011 to 2021, 1960 to 2016.
Birds, plants, benthos and mammals.
Here, we consider individual species contributions to spatial beta-diversity and how these change over time. We focus on the relative contributions of localised and widespread species across five case studies to determine which are contributing most to homogenisation.
Species occurring in around half of sites provided the greatest contributions to beta-diversity at a given time, but not through time. The most widespread species (> 0.75 of sites) contributed little to beta-diversity change, with this most apparent in highly nested assemblages. In contrast, localised species (initially in < 0.25 of sites) contributed most to both homogenisation (when declining) and differentiation (when increasing) regardless of nestedness.
This challenges the hypothesis that widespread generalist species are the main drivers of homogenisation, underlining the importance of rare species and of nestedness to patterns of beta-diversity change. Conservation interventions to increase localised species occurrence would do more to limit homogenisation than attempts to control already-widespread species or prevent others becoming widespread, especially when assemblages are highly nested.
Specific leaf area (SLA) is a key plant functional trait linked to plant structural, physiological and resource-use strategies. Despite its importance, spatially continuous data sets capturing SLA variation across global biomes remain scarce, particularly under future climate scenarios. Here, we compile and model a comprehensive global dataset of SLA to assess its current distribution and project responses to climate change. This resource provides critical support for exploring trait–environment relationships, plant community assembly and vegetation-climate feedbacks at broad spatial and temporal scales.
The dataset consists of a single standardised table containing 24,237 SLA measurements, each linked to geographic coordinates. Observations span 5687 vascular plant species across 282 families and were compiled from peer-reviewed literature, the TRY database and new field collections.
The dataset has global coverage, with SLA predictions provided at a spatial resolution of 1 km2. Spatial layers include both present-day environmental conditions and projections under future climate scenarios.
The data represent current climatic conditions and future projections for the following intervals: 2020–2040, 2040–2060, 2060–2080 and 2080–2100 under multiple emissions scenarios.
Trait data were collected at the species level for vascular plants, with particular emphasis on grasses and trees. These SLA values were measured excluding petioles and were georeferenced to 2437 sampling sites worldwide, enabling consistent cross-biome analyses of functional trait variation.
The raw trait data are provided in comma-separated value (.csv) format. Model outputs, including global SLA predictions, are available as Geotiff (.tif) files suitable for integration into GIS platforms and Earth system models.
Plant–soil feedbacks (PSFs) are important ecological phenomena that influence plant performance and community assembly. Mycorrhizal fungi, including arbuscular mycorrhizal (AM) and ectomycorrhizal (EcM) fungi, play vital roles in modulating PSFs by mediating plant–soil interactions. However, global patterns of PSFs in relation to mycorrhizal type remain poorly understood.
Global.
1994–2022.
Woody species.
We conducted a meta-analysis of published data to investigate the difference in the global expression of PSFs between EcM and AM woody plant hosts.
Our results show that EcM hosts generally experience positive PSFs, whereas AM hosts generally experience negative PSFs, a difference that is consistent across angiosperm and gymnosperm woody plant species. The difference in the mean effect size of PSFs between EcM and AM hosts was strongest for studies conducted in North America and Asia, and was not significant for studies conducted in Europe and Oceania. AM woody species displayed more negative PSFs in wet and cool regions, whereas EcM woody species showed no latitudinal or climatic gradients in the strength of PSFs. These trends resulted in a convergence in the effects of PSFs on EcM and AM woody plants among studies conducted in the tropics, and a significant divergence in temperate and boreal studies. The meta-analysis also revealed that the expression and detection of PSFs in both AM and EcM woody species are affected by the experimental setting.
Our findings highlight the importance of mycorrhizal type in shaping PSFs and their implications for plant community assembly and dynamics in different biogeographic settings. We discuss how experimental designs influence the detection of PSFs and suggest appropriate methods for future studies.
Large herbivores play key roles in ecosystems by promoting plant diversity, dispersing seeds, regulating nutrient cycling, and shaping vegetation structure. Since the Late Pleistocene, their declines have led to profound ecosystem changes. While often viewed as problematic, alien herbivores may partly compensate for these losses; yet their spatial contributions remain poorly understood. We provide the first spatially explicit assessment of native and alien large herbivores in North America (NA), evaluating their potential to restore herbivore diversity and functions relative to a present-natural baseline.
North America (United States and Canada).
Late Pleistocene to present.
Ungulates ≥ 20 kg.
We combined species distribution models based on citizen-science records with functional trait data (body size, grazing, browsing) to estimate current and potential distributions of 12 native and 23 alien ungulate species. We compared four scenarios (current and potential natives, with and without aliens) against a present-natural baseline.
Alien ungulates are widespread, especially in southern NA, with Texas showing highest richness. If both native and alien species expanded to their climatic potential ranges, they could offset c. 40% of the deficit in richness and diet composition caused by Late Pleistocene extinctions, with full recovery possible in the Southeast and Arctic. Native species contribute about twice as much as alien species overall, but aliens disproportionately restore grazing intensity.
Alien ungulates can enhance herbivore diversity and partially restore lost ecosystem functions, though neither natives nor aliens fully compensate for extinct megafauna. Their tropical–subtropical origins limit potential ranges, and expansions must be balanced against ecological risks.
Terrestrial biodiversity is impacted by both climate and land use change. Yet, future biodiversity projections have rarely considered these two drivers in combination. In this study, we aim to assess the individual and combined impact of future climate and land use change on global terrestrial vertebrate diversity under a ‘sustainability’ (SSP1-RCP2.6) and an ‘inequality’ (SSP4-RCP6.0) scenario.
Global land, excluding Antarctica.
1995, 2080.
Amphibians, birds, and mammals.
We combined global climate-driven species distribution model (SDM) projections of 13,903 vertebrates (amphibians, birds, and mammals) with future and present land use projections from the Land Use Harmonisation 2 (LUH2) project. We refined the SDM outputs by the habitat requirements of each species using a land use filtering approach. We then analyzed future species richness changes globally, per region, and per land use category, and looked at taxon-specific effects.
Under both scenarios, decreases in future species richness dominate at low and mid-latitudes, with climate and land use change playing an equally important role. Land use change can be either an alleviating (SSP1-RCP2.6) or an exacerbating (SSP4-RCP6.0) factor of climate-induced biodiversity loss. Sub-Saharan Africa is projected to become a high-risk area for future land use-driven biodiversity loss under the SSP4-RCP6.0. Under SSP1-RCP2.6, forested and non-forested land areas increase, while SSP4-RCP6.0 leads to higher rates of deforestation and pasture expansion. Mammals experience the largest climate-driven losses, affecting 56.4% of land area under SSP4-RCP6.0, while amphibians are particularly vulnerable to land use-driven losses, especially under SSP4-RCP6.0.
Our results suggest that both climate and land use pressures on biodiversity will be highest in lower latitudes, which harbor the highest levels of biodiversity.
Lu L., Li J., Cheng F., He J., Hao M., Fan C., Zhang Chunyu, von Gadow Klaus, Zhao X., 2025 “Dispersal, Climate Anomaly and Contemporary Climate Shape the Latitudinal Patterns of β-Diversity of Angiosperm Trees.” Global Ecology and Biogeography 34 no. 12: e70178 https://doi.org/10.1111/geb.70178.
In the originally published article, the article type was incorrect. It is a Research Article, not a Review Article.
In the ‘Results’ section of the ‘Abstract’ section, the text ‘After accounting for abiotic factors, high DCMc values increased richness differences but reduced replacement, while contemporary climate continued to dominate across scales.’ was incorrect. This should have read as ‘After accounting for abiotic factors, high DCMc values decreased richness differences but increased replacement, while contemporary climate continued to dominate across scales.’
The online version of the article has been updated.
We apologise for these errors.
Li, J., S. Lin, M. Hao, et al. 2025. “Effects of Nitrogen Deposition, Biodiversity and Climate on Productivity in a Large Temperate Forest Region.” Global Ecology and Biogeography 34, no. 12: e70173. https://doi.org/10.1111/geb.70173.
In the originally published article, author Minhui Hao's name was misspelled as Minuid Hao. This has been corrected in the online version of the article.
We apologise for this error.
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：