Dr. Philipp Brun

Citizen science has become key to biodiversity monitoring but critically depends on accurate quality control that is scalable and tailored to the focal region. We developed FlorID, a free-to-use identification service for all native and many non-native plants of Switzerland. FlorID can identify >3000 species, using vision transformers trained on 1.5M photos, and ecological predictions from multilayer perceptrons, trained on 6.7M occurrence observations and 20 high-resolution environmental variables. Embedded in a free-to-use application programming interface, FlorID can be accessed directly, via webservice, and via FlorApp smartphone application. If multiple images and spatiotemporal location are available, FlorID correctly identifies 93% of field observations and has a top-5 accuracy of 99%. Ecological predictions boost identification success especially for native species with distinct distributions. By evaluating information on appearance and fine-grained ecology, FlorID is a blueprint for similar solutions targeting different taxa or regions, and a basis for developments like automated community inventories.

Persistent multiyear drought (MYD) events pose a growing threat to nature and humans in a changing climate. We identified and inventoried global MYDs by detecting spatiotemporally contiguous climatic anomalies, showing that MYDs have become drier, hotter, and led to increasingly diminished vegetation greenness. The global terrestrial land affected by MYDs has increased at a rate of 49,279 ± 14,771 square kilometers per year from 1980 to 2018. Temperate grasslands have exhibited the greatest declines in vegetation greenness during MYDs, whereas boreal and tropical forests have had comparably minor responses. With MYDs becoming more common, this global quantitative inventory of the occurrence, severity, trend, and impact of MYDs provides an important benchmark for facilitating more effective and collaborative preparedness toward mitigation of and adaptation to such extreme events.

Climate projections for continental Europe indicate drier summers, increased annual precipitation, and less snowy winters, which are expected to cause shifts in species' distributions. Yet, most regions/countries currently lack comprehensive climate-driven biodiversity projections across taxonomic groups, challenging effective conservation efforts. To address this gap, our study evaluated the potential effects of climate change on the biodiversity of an alpine country of Europe, Switzerland. We used a state-of-the art species distribution modeling approach and species occurrence data that covered the climatic conditions encountered across the full species' ranges to help limiting niche truncation. We quantified the relationship between baseline climate and the spatial distribution of 7291 species from 12 main taxonomic groups and projected future climate suitability for three 30-year periods and two greenhouse gas concentration scenarios (RCP4.5 and 8.5). Our results indicated important effects of projected climate changes on species' climate suitability, with responses varying by the taxonomic and conservation status group. The percentage of species facing major changes in climate suitability was higher under RCP8.5 (68%) compared to RCP4.5 (66%). By the end of the century, decreases in climate suitability were projected for 3000 species under RCP8.5 and 1758 species under RCP4.5. The most affected groups under RCP8.5 were molluscs, algae, and amphibians, while it was molluscs, birds, and vascular plants under RCP4.5. Spatially, by 2070–2099, we projected an overall decrease in climate suitability for 39% of the cells in the study area under RCP8.5 and 10% under RCP4.5, while projecting an increase for 50% of the cells under RCP8.5 and 73% under RCP4.5. The most consistent geographical shifts were upward, southward, and eastward. We found that the coverage of high climate suitability cells by protected areas was expected to increase. Our models and maps provide guidance for spatial conservation planning by pointing out future climate-suitable areas for biodiversity.

In the age of big data, scientific progress is fundamentally limited by our capacity to extract critical information. Here, we map fine-grained spatiotemporal distributions for thousands of species, using deep neural networks (DNNs) and ubiquitous citizen science data. Based on 6.7 M observations, we jointly model the distributions of 2477 plant species and species aggregates across Switzerland with an ensemble of DNNs built with different cost functions. We find that, compared to commonly-used approaches, multispecies DNNs predict species distributions and especially community composition more accurately. Moreover, their design allows investigation of understudied aspects of ecology. Including seasonal variations of observation probability explicitly allows approximating flowering phenology; reweighting predictions to mirror cover-abundance allows mapping potentially canopy-dominant tree species nationwide; and projecting DNNs into the future allows assessing how distributions, phenology, and dominance may change. Given their skill and their versatility, multispecies DNNs can refine our understanding of the distribution of plants and well-sampled taxa in general.

We propose a deep learning approach for high-resolution species distribution modelling (SDM) at large scale combining point-wise, crowd-sourced species observation data and environmental data with Sentinel-2 satellite imagery. What makes this task challenging is the great variety of controlling factors for species distribution, such as habitat conditions, human intervention, competition, disturbances, and evolutionary history. Experts either incorporate these factors into complex mechanistic models based on presence-absence data collected in field campaigns or train machine learning models to learn the relationship between environmental data and presence-only species occurrence. We extend the latter approach here and learn deep SDMs end-to-end based on point-wise, crowd-sourced presence-only data in combination with satellite imagery. Our method, dubbed Sat-SINR, jointly models the spatial distributions of 5.6k plant species across Europe and increases the spatial resolution by a factor of 100 compared to the current state of the art. We exhaustively test and ablate multiple variations of combining geo-referenced point data with satellite imagery and show that our deep learning-based SDM method consistently shows an improvement of up to 3 percentage points across three metrics. We make all code publicly available at https://github.com/ecovision-uzh/sat-sinr.

The Arctic ecosystems and their species are exposed to amplified climate warming and, in some regions, to rapidly developing economic activities. This study assesses, models, and maps the geographic patterns of community-level plant species richness in the Western Siberian Arctic and estimates the relative impact of environmental and anthropogenic factors driving these patterns. With our study, we aim at contributing toward conservation efforts for Arctic plant diversity in the Western Siberian Arctic. Western Siberian Arctic, Russia. We investigated the relative importance of environmental and anthropogenic predictors of community-level plant species richness in the Western Siberian Arctic using macroecological models trained with an extensive geobotanical dataset. We included vascular plants, mosses and lichens in our analysis, as non-vascular plants substantially contribute to species richness and ecosystem functions in the Arctic. We found that the mean community-level plant species richness in this vast Arctic region does not decrease with increasing latitude. Instead, we identified an increase in species richness from South-West to North-East, which can be well explained by environmental factors. We found that paleoclimatic factors exhibit higher explained deviance compared to contemporary climate predictors, potentially indicating a lasting impact of ancient climate on tundra plant species richness. We also show that the existing protected areas cover only a small fraction of the regions with highest species richness. Our results reveal complex spatial patterns of community-level species richness in the Western Siberian Arctic. We show that climatic factors such as temperature (including paleotemperature) and precipitation are the main drivers of plant species richness in this area, and the role of relief is clearly secondary. We suggest that while community-level plant species richness is mostly driven by environmental factors, an improved spatial sampling will be needed to robustly and more precisely assess the impact of human activities on community-level species richness patterns. Our approach and results can be used to design conservation strategies and to investigate drivers of plant species richness in other arctic regions.

The emergence of alternative stable states in forest systems has significant implications for the functioning and structure of the terrestrial biosphere, yet empirical evidence remains scarce. Here, we combine global forest biodiversity observations and simulations to test for alternative stable states in the presence of evergreen and deciduous forest types. We reveal a bimodal distribution of forest leaf types across temperate regions of the Northern Hemisphere that cannot be explained by the environment alone, suggesting signatures of alternative forest states. Moreover, we empirically demonstrate the existence of positive feedbacks in tree growth, recruitment and mortality, with trees having 4–43% higher growth rates, 14–17% higher survival rates and 4–7 times higher recruitment rates when they are surrounded by trees of their own leaf type. Simulations show that the observed positive feedbacks are necessary and sufficient to generate alternative forest states, which also lead to dependency on history (hysteresis) during ecosystem transition from evergreen to deciduous forests and vice versa. We identify hotspots of bistable forest types in evergreen-deciduous ecotones, which are likely driven by soil-related positive feedbacks. These findings are integral to predicting the distribution of forest biomes, and aid to our understanding of biodiversity, carbon turnover, and terrestrial climate feedbacks.

1. Species distribution models (SDMs) are widely used to infer species–environment relationships, predict spatial distributions and characterise species' environmental niches. While the importance of space and spatial scales is widely acknowledged in SDM applications, temporal components of the niche are rarely addressed.
2. We discuss how phenology and demographic stages affect model inference in plant SDMs. Ignoring conspicuousness and timing of phenological stages may bias niche estimates through increased observer bias, while ignoring stand age may bias niche estimates through temporal mismatches with environmental variables, especially during times of rapid global warming.
3. We present different methods to consider phenology and demographic stages in plant SDMs, including the selection of causal, spatiotemporally explicit predictors and the calibration of stage-specific SDMs. Based on a case study with citizen science data, we illustrate how spatiotemporal SDMs provide deeper insights on the coincidence of range and phenological shifts under climate change.
4. The proliferation of digitally available biodiversity and citizen science data increasingly allows considering time explicitly in SDMs. This offers a more mechanistic understanding of plant distributions, and more robust predictions under global change, especially if the reporting of phenological stages and age is facilitated and promoted by relevant data portals.

The conservation and restoration of forest ecosystems require detailed knowledge of the native plant compositions. Here, we map global forest tree composition and assess the impacts of historical forest cover loss and climate change on trees. The global occupancy of 10,590 tree species reveals complex taxonomic and phylogenetic gradients determining a local signature of tree lineage assembly. Species occupancy analyses indicate that historical forest loss has significantly restricted the potential suitable range of tree species in all forest biomes. Nevertheless, tropical moist and boreal forest biomes display the lowest level of range restriction and harbor extremely large ranged tree species, albeit with a stark contrast in richness and composition. Climate change simulations indicate that forest biomes are projected to differ in their response to climate change, with the highest predicted species loss in tropical dry and Mediterranean ecoregions. Our findings highlight the need for preserving the remaining large forest biomes while regenerating degraded forests in a way that provides resilience against climate change.

Predicting contemporary and future species distributions is relevant for science and decision making, yet the development of high-resolution spatial predictions for numerous taxonomic groups and regions is limited by the scalability of available modelling tools. Uniting species distribution modelling (SDM) techniques into one high-performance computing (HPC) pipeline, we developed N-SDM, an SDM platform aimed at delivering reproducible outputs for standard biodiversity assessments. N-SDM was built around a spatially-nested framework, intended at facilitating the combined use of species occurrence data retrieved from multiple sources and at various spatial scales. N-SDM allows combining two models fitted with species and covariate data retrieved from global to regional scales, which is useful for addressing the issue of spatial niche truncation. The set of state-of-the-art SDM features embodied in N-SDM includes a newly devised covariate selection procedure, five modelling algorithms, an algorithm-specific hyperparameter grid search, and the ensemble of small-models approach. N-SDM is designed to be run on HPC environments, allowing the parallel processing of thousands of species at the same time. All the information required for installing and running N-SDM is openly available on the GitHub repository https://github.com/N-SDM/N-SDM.

1. Predicting community-level trait indices, such as community-weighted mean or functional dispersion, in space or time, is crucial, as they are markers of ecosystem functions, proxies for ecosystem services, and are now integral part of conservation planning. On one hand, since members of a species share similar traits, a natural way is to model species distributions as a function of the environment first (i.e. with species distribution models), and then to reconstruct trait indices for any locality of interest (predict-first). On the other hand, community-level trait indices can be seen as the direct result of environmental filtering, and thus their distributions may directly be modelled in response to the environment (assemble-first). Although relatively different, these two approaches have been used interchangeably in trait-based ecology with unknown consequences on their usability.
2. Here, using plant community (4463 plots) and trait data (LNC, leaf nitrogen content; PLH, plant height; SLA, specific leaf area, for >800 species) covering the French Alps, we compared the two approaches to predict community mean and functional dispersion indices, accounting or not for abundance (CM, community mean; CWM, community weighted mean; FDis, functional dispersion; uFDis, unweighted functional dispersion). We tested both interpolation versus extrapolation capabilities of the two approaches. To support the empirical findings, we also run the same comparative analysis on simulated community data.
3. While both approaches produced similar and skillful CM/CWM predictions (R² > 0.63 for CWM PLH) for interpolation, the assemble-first outperformed the predict-first approach for extrapolation to unobserved environmental conditions (R² = 0.60 against 0.54 for CWM PLH). Functional dispersion was generally less well predicted, although the assemble-first approach again provided better predictions for both interpolation (R² = 0.31 against 0.27 for FDis PLH) and extrapolation (R² = 0.28 against 0.18 for FDis PLH). The predict-first approach systematically overestimated FDis. Results were confirmed by the simulation experiments. We conclude that the direct modelling of community-level indices provides more robust spatial predictions over space and time and should thus be preferred. This approach also does not suffer from having generally too many species predicted by the predict-first approach.

The entry of microplastics (MPs) into marine food webs is a major environmental concern. We investigated how the behavior of planktonic copepods influences the risk of MPs to enter marine food webs by applying a trait-based approach and by combining experiments (bottle incubations and video observations) with biogeographical analyses. We aimed to evaluate which type of feeding behavior is most risky in terms of MP ingestion and which marine geographical areas are more susceptible to MP ingestion by planktonic copepods. We used different species as models of the main foraging behaviors in planktonic copepods: feeding-current, cruising, ambush, and mixed behavior feeding. All behaviors showed a similarly low risk of MP ingestion, up to 1 order of magnitude lower than for similar-sized microalgae. We did not observe any influence of the prey type or MP size (8 and 20 μm) on MP ingestion for any of the behaviors. By mapping the global distribution of feeding behaviors, we showed that feeding-current feeding is the most common behavior, but the risk of MP ingestion remains equally low across the global ocean, independently of the predominant behavior. Overall, our results suggest a low risk of MP ingestion by planktonic copepods and therefore a minimal risk of trophic transfer of MPs via marine pelagic copepods in marine ecosystems.

A multitude of physical and biological processes on which ecosystems and human societies depend are governed by the climate, and understanding how these processes are altered by climate change is central to mitigation efforts. We developed a set of climate-related variables at as yet unprecedented spatiotemporal detail as a basis for environmental and ecological analyses. We downscaled time series of near-surface relative humidity (hurs) and cloud area fraction (clt) under the consideration of orography and wind as well as near-surface wind speed (sfcWind) using the delta-change method. Combining these grids with mechanistically downscaled information on temperature, precipitation, and solar radiation, we then calculated vapor pressure deficit (vpd), surface downwelling shortwave radiation (rsds), potential evapotranspiration (pet), the climate moisture index (cmi), and site water balance (swb) at a monthly temporal and 30 arcsec spatial resolution globally from 1980 until 2018 (time-series variables). At the same spatial resolution, we further estimated climatological normals of frost change frequency (fcf), snow cover days (scd), potential net primary productivity (npp), growing degree days (gdd), and growing season characteristics for the periods 1981–2010, 2011–2040, 2041–2070, and 2071–2100, considering three shared socioeconomic pathways (SSP126, SSP370, SSP585) and five Earth system models (projected variables). Time-series variables showed high accuracy when validated against observations from meteorological stations and when compared to alternative products. Projected variables were also highly correlated with observations, although some variables showed notable biases, e.g., snow cover days. Together, the CHELSA-BIOCLIM+ dataset presented here (https://doi.org/10.16904/envidat.332, Brun et al., 2022) allows improvement to our understanding of patterns and processes that are governed by climate, including the impact of recent and future climate changes on the world's ecosystems and the associated services on societies.

Outside controlled experimental plots, the impact of community attributes on primary productivity has rarely been compared to that of individual species. Here, we identified plant species of high importance for productivity (key species) in >29,000 diverse grassland communities in the European Alps, and compared their effects with those of community-level measures of functional composition (weighted means, variances, skewness and kurtosis). After accounting for the environment, the five most important key species jointly explained more deviance of productivity than any measure of functional composition alone. Key species were generally tall with high specific leaf areas. By dividing the observations according to distinct habitats, the explanatory power of key species and functional composition increased and key-species plant types and functional composition-productivity relationships varied systematically, presumably because of changing interactions and trade-offs between traits. Our results advocate for a careful consideration of species’ individual effects on ecosystem functioning in complement to community-level measures.

Monitoring biodiversity is paramount to manage and protect natural resources. Collecting images of organisms over large temporal or spatial scales is a promising practice to monitor the biodiversity of natural ecosystems, providing large amounts of data with minimal interference with the environment. Deep learning models are currently used to automate classification of organisms into taxonomic units. However, imprecision in these classifiers introduces a measurement noise that is difficult to control and can significantly hinder the analysis and interpretation of data. We overcome this limitation through ensembles of Data-efficient image Transformers (DeiTs), which not only are easy to train and implement, but also significantly outperform the previous state of the art (SOTA). We validate our results on ten ecological imaging datasets of diverse origin, ranging from plankton to birds. On all the datasets, we achieve a new SOTA, with a reduction of the error with respect to the previous SOTA ranging from 29.35% to 100.00%, and often achieving performances very close to perfect classification. Ensembles of DeiTs perform better not because of superior single-model performances but rather due to smaller overlaps in the predictions by independent models and lower top-1 probabilities. This increases the benefit of ensembling, especially when using geometric averages to combine individual learners. While we only test our approach on biodiversity image datasets, our approach is generic and can be applied to any kind of images.

• The documentation of biodiversity distribution through species range identification is crucial for macroecology, biogeography, conservation, and restoration. However, for plants, species range maps remain scarce and often inaccurate.
• We present a novel approach to map species ranges at a global scale, integrating polygon mapping and species distribution modelling (SDM). We develop a polygon mapping algorithm by considering distances and nestedness of occurrences. We further apply an SDM approach considering multiple modelling algorithms, complexity levels, and pseudo-absence selections to map the species at a high spatial resolution and intersect it with the generated polygons.
• We use this approach to construct range maps for all 1957 species of Fagales and Pinales with data compilated from multiple sources. We construct high-resolution global species richness maps of these important plant clades, and document diversity hotspots for both clades in southern and south-western China, Central America, and Borneo. We validate the approach with two representative genera, Quercus and Pinus, using previously published coarser range maps, and find good agreement.
• By efficiently producing high-resolution range maps, our mapping approach offers a new tool in the field of macroecology for studying global species distribution patterns and supporting ongoing conservation efforts.

Central Europe has been experiencing unprecedented droughts during the last decades, stressing the decrease in tree water availability. However, the assessment of physiological drought stress is challenging, and feedback between soil and vegetation is often omitted because of scarce belowground data. Here we aimed to model Swiss forests' water availability during the 2015 and 2018 droughts by implementing the mechanistic soil-vegetation-atmosphere-transport (SVAT) model LWF-Brook90 taking advantage of regionalized depth-resolved soil information. We calibrated the model against soil matric potential data measured from 2014 to 2018 at 44 sites along a Swiss climatic and edaphic drought gradient. Swiss forest soils' storage capacity of plant-available water ranged from 53 mm to 341 mm, with a median of 137 ± 42 mm down to the mean potential rooting depth of 1.2 m. Topsoil was the primary water source. However, trees switched to deeper soil water sources during drought. This effect was less pronounced for coniferous trees with a shallower rooting system than for deciduous trees, which resulted in a higher reduction of actual transpiration (transpiration deficit) in coniferous trees. Across Switzerland, forest trees reduced the transpiration by 23% (compared to potential transpiration) in 2015 and 2018, maintaining annual actual transpiration comparable to other years. Together with lower evaporative fluxes, the Swiss forests did not amplify the blue water deficit. The 2018 drought, characterized by a higher and more persistent transpiration deficit than in 2015, triggered widespread early wilting across Swiss forests that was better predicted by the SVAT-derived mean soil matric potential in the rooting zone than by climatic predictors. Such feedback-driven quantification of ecosystem water fluxes in the soil–plant-atmosphere continuum will be crucial to predicting physiological drought stress under future climate extremes.

Planktonic copepods are the most abundant animals in the ocean and key players in global biochemical processes. Recent modeling suggests that zooplankton ingestion of microplastics (MPs) can disrupt the biological carbon pump and accelerate a global loss of oceanic oxygen. Here we investigate the behavioral responses and ingestion rates of a model feeding-current generating copepod when exposed to microplastics of different characteristics by small-scale video observations and bottle incubations. We found that copepods rejected 80% of the microplastics after touching them with their mouth parts, in essence exhibiting a kind of taste discrimination. High rejection rates of microplastics were independent of polymer type, shape, presence of biofilms, or sorbed pollutant (pyrene), indicating that microplastics are unpalatable for feeding-current feeding copepods and that post-capture taste discrimination is a main sensorial mechanism in the rejection of microplastics. In an ecological context, taking into account the behaviors of planktonic copepods and the concentrations of microplastics found in marine waters, our results suggest a low risk of microplastic ingestion by zooplankton and a low impact of microplastics on the vertical exportation of fecal pellets.

Although the importance of edaphic factors and habitat structure for plant growth and survival is known, both are often neglected in favor of climatic drivers when investigating the spatial patterns of plant species and diversity. Yet, especially in mountain ecosystems with complex topography, missing edaphic and habitat components may be detrimental for a sound understanding of biodiversity distribution. Here, we compare the relative importance of climate, soil and land cover variables when predicting the distributions of 2,616 vascular plant species in the European Alps, representing approximately two-thirds of all European flora. Using presence-only data, we built point-process models (PPMs) to relate species observations to different combinations of covariates. We evaluated the PPMs through block cross-validations and assessed the independent contributions of climate, soil, and land cover covariates to predict plant species distributions using an innovative predictive partitioning approach. We found climate to be the most influential driver of spatial patterns in plant species with a relative influence of ~58.5% across all species, with decreasing importance from low to high elevations. Soil (~20.1%) and land cover (~21.4%), overall, were less influential than climate, but increased in importance along the elevation gradient. Furthermore, land cover showed strong local effects in lowlands, while the contribution of soil stabilized at mid-elevations. The decreasing influence of climate with elevation is explained by increasing endemism, and the fact that climate becomes more homogeneous as habitat diversity declines at higher altitudes. In contrast, soil predictors were found to follow the opposite trend. Additionally, at low elevations, human-mediated land cover effects appear to reduce the importance of climate predictors. We conclude that soil and land cover are, like climate, principal drivers of plant species distribution in the European Alps. While disentangling their effects remains a challenge, future studies can benefit markedly by including soil and land cover effects when predicting species distributions.