Bats are ecologically important mammals whose monitoring increasingly relies on acoustic data. However, many tools for bat call identification remain subscription‐based, closed‐source, region‐specific or limited in scalability, creating barriers to global data integration and method development. We present BSG‐BATS, an open‐access annotation portal and convolutional neural network (CNN)‐based classifier for bat calls. The portal enables researchers and practitioners to annotate calls at species or phonic group level and contribute directly to the iterative improvement of the classifier. As a proof of concept, we trained the first BSG‐BATS model using over 4000 annotated recordings of 21 European species. It gave promising results and outperformed many commonly used commercial and academic tools by providing higher AUC (area under the receiver operating characteristic curve) scores for test data. All annotations are transparently annotated in the portal ( https://bsg.laji.fi/bats/identification/instructions ) and the trained model, code and documentation are available at www.zenodo.org/records/15495676 . BSG‐BATS offers a foundation for community‐driven development of bat sound identification tools. By integrating annotation and model retraining on an open‐access platform, it enables collective improvement and adaptation to new regions, species and sound types. We invite all bat researchers and experts to join this collaborative effort, whether by contributing data ( bsg-bat@helsinki.fi ), annotating sounds ( https://bsg.laji.fi/bats/identification/instructions ) or testing the model in their own work ( https://www.zenodo.org/records/15495676 ).

How microbes influence insects, each other and vice versa, is a topical question in insect science. With this contribution, we highlight joint species distribution models (JSDMs) as a statistical framework particularly well suited for resolving it. While JSDM has been widely applied to micro-biota of organisms other than insects, only a handful of studies have applied them thus far to insect microbiota. To encourage insect microbiota researchers to catch up with the benefits of JSDM, we introduce the statistical and ecological basics of JSDM and give flesh to the modelling framework by reviewing how it has so far been applied to insect microbiota. We highlight the power of JSDMs in separating host, environmental and microbial drivers of community assembly, and in generating hypotheses about how different microbes influence each other. To stimulate a broad adoption of JSDMs in studies of insect microbiota, we propose a predictive framework for identifying and characterizing five ecological categories of insect-associated microbes: (1) obligate nutritional symbionts; (2) facultative endosymbionts; (3) specialized gut associates; (4) transient microbes; and (5) pathogens and parasites. In particular, we hypothesize how the ecological characteristics of these different microbial categories are reflected in the statistical signatures recovered by JSDMs. We further suggest how JSDM can be applied to address the phylogenetic and ecological scales at which microbiome patterns are structured and to uncover how global environmental changes may reshape insect microbiota.

DNA metabarcoding is fast becoming a method of choice for the inventory and monitoring of community‐level arthropod biodiversity, with growing interest in the recovery of haplotype‐level variation within species from metabarcoding reads. Denoising tools have been established as the standard for metabarcode processing, with the promise to provide accurate sequence data at the haplotype level. Despite their widespread use, the accuracy of denoising methods relative to the elimination of non‐target sequences and retention of target sequences has received limited attention. Recent studies based on whole organism community DNA (wocDNA) metabarcoding suggest that misleading estimates of species richness and diversity may be obtained after denoising. Understanding and evaluating how denoising pipelines perform and how they can be improved is crucial for broadening the applicability of metabarcoding to community ecology, metaphylogeography and biodiversity inventory and monitoring. To evaluate the performance of the most popularly used denoising tools, we use (i) complex communities from malaise traps subjected to wocDNA metabarcoding (representing the information yielded by metabarcoding) and (ii) additional sequencing of each individual specimen (representing the ground truth of sample contents). To these data, we apply the metabarcoding evaluation rationale described in Andújar et al. (2021), complemented by an additional layer of clustering and filtering with LULU and metaMATE. This study confirms that denoising tools alone, but also in combination with clustering, retain a high number of non‐target amplicon sequence variants (ASVs) and operational taxonomic units (OTUs). Both LULU and metaMATE removed large amounts of non‐target sequences and noise. This, however, came at the cost of the collateral removal of target sequences. We recommend a workflow including one of the better performing denoising tools: DADA2 or UNOISE3, in combination with LULU or metaMATE as a customizable option for evaluating pipeline performance.

ABSTRACT A basic question in ecological research and biodiversity monitoring concerns the estimation of species abundances from trap catches. As a case in point, a Malaise trap can yield thousands of arthropod individuals, but how this count should be converted to numbers of individuals per unit area has remained an open question. Here, we supplement observational data with an experimental approach targeted at quantifying catchability. We released marked insects in a boreal forest and examined their capture rate by a grid of Malaise traps around the release location. We estimated insect movement rates, mortality rates, and Malaise trapping capture rates by fitting a joint species movement model to these data. As a methodological novelty, we show how to convert the movement model parameters to the expected number of captured individuals, given their actual population density. Our results show that multiplying the sample content by 30 000 yields a rough estimate of the number of individuals per hectare. This conversion factor depends on the species, generally decreasing with increasing body size. We apply the estimated conversion factors to conclude that typical boreal forest contains some four million insect individuals per hectare, out of which around half belong to Diptera. SIGNIFICANCE STATEMENT Traditionally, the Malaise trap method has been used for assessing the state of the local communities and to estimate population abundances. However, a topical question is: how does the number of individuals observed in a sample relate to the true density of individuals in the surrounding community? To answer this question, we implement a movement model parametrized by a carefully designed mark-recapture experiment, in which we are able to obtain taxon-specific conversion factors for different groups of insects. We found that different insect groups come with different conversion factors, causing a mismatch between sample contents and true community composition. Thus, treating the sample contents as a direct representation of the reference community will be misleading.

Abstract Understanding global biodiversity patterns and their drivers is a prerequisite for countering the biodiversity crisis. In this paper, we introduce a novel generalized linear model, Hubbell regression, to estimate a key biodiversity descriptor, the fundamental biodiversity number. This can be converted into a set of biodiversity descriptors, including Shannon and Simpson indices, and more. Hence, quantifying the impact of environmental conditions on the fundamental biodiversity number allows us to predict the general properties of local biodiversity in any setting. In addition to having a strong mathematical foundation, Hubbell regression consistently outperformed current state-of-the-art models in predicting global biodiversity. We apply the method to arthropods, which account for the majority of terrestrial biodiversity. By parameterizing the models using samples of 1.78 million arthropods from 2415 samples collected at 135 sites spanning all continents, we pinpoint the drivers of arthropod biodiversity and its features at the global scale. We find that actual evapotranspiration is the single largest predictor of arthropod diversity and explains nearly 30% of the variation in richness. Moreover, we infer that high human activity has led to a 21.3 % and 29.2% decrease in potential insect richness in tropical and dry zones, respectively, but increased insect richness in polar regions. These insights bring a new foundation for biodiversity research and action.

Abstract DNA metabarcoding—high‐throughput sequencing of barcode regions from bulk samples—has become a key tool for insect biodiversity assessment. Yet, how methodological choices affect the accuracy of metabarcoding data remains insufficiently explored. In this paper, we ask: (1) How does the lysis method (non‐destructive lysis vs. destructive homogenization) affect community recovery? (2) How comprehensively does metabarcoding capture species richness? (3) To what extent can spike‐ins improve abundance estimates? (4) How accurately can species abundances be estimated? We evaluated the accuracy of insect metabarcoding using 4749 bulk samples from a large‐scale biodiversity survey subjected to mild lysis. Of these samples, 856 were also homogenized, allowing a systematic comparison of the effect of alternative treatments. To potentially improve abundance estimates, we added six biological spike‐ins (i.e. foreign insects) to all samples, and two synthetic spike‐ins (artificial DNA fragments) to the homogenization treatment. In addition, we established the contents of 15 samples by individually barcoding all specimens, enabling direct assessment of occurrence and abundance estimates. Our results revealed consistent differences between destructive and non‐destructive treatments. While both methods reliably detected the majority of species, small and soft‐bodied taxa were more often recovered after mild lysis than after homogenization, while the reverse was true for heavily sclerotized, hairy and large taxa. Using biological spike‐ins for calibration reduced the variance in read numbers per specimen considerably, especially in homogenized samples, while synthetic spike‐ins were less effective. In a Bayesian analysis, where species data were matched to the best‐fitting spike‐in calibration curve, accurate abundance estimates (±1 individual) were obtained for 72.9% of species occurrences. Our results show that it is possible to obtain reasonably accurate abundance estimates from metabarcoding data and that mild lysis and homogenization result in different taxon‐specific biases in terms of occurrence data, with neither method outperforming the other. Abundance accuracy is improved by homogenization rather than mild lysis of samples, and by the use of biological rather than synthetic spike‐ins. Together, these findings provide a major step towards robust, quantitative biodiversity monitoring using DNA‐metabarcoding.

Changes in land use and climate pose major threats to biodiversity 1–5 . However, variation in species responses to climate and land use across space, time, and taxa remains poorly understood 3,6–10 , hindering our ability to predict and mitigate biodiversity change. Here, we evaluate the relative importance of concurrent changes in climate and land use in driving the occurrence and abundance patterns of 503 terrestrial animal species of various taxa over 20 years in Finland. Habitat composition proved to be the main driver of biodiversity patterns but how much and with what uncertainty it explained species distributions depended highly on the context. Specifically, habitat was the dominant driver for butterflies, birds, and small mammals, while habitat and climate were equally important for large mammals and moths. Additionally, species patterns between biogeographical regions were mainly explained by climate, while habitat was the main driver within regions. Traits, such as, body size, pace of life, habitat and diet specialization modulated the relative importance of both drivers’ impacts. Climate and habitat impact on most species were also partially correlated, highlighting the tight connection between the drivers. Our findings emphasize that land use is a major force in shaping terrestrial biodiversity, while highlighting its tight connection with climate. Considering functional and spatial contexts is thus essential for building effective management and conservation strategies for biodiversity under rampant global change.

The mechanisms linking productivity to patterns of species richness, species prevalence, and beta diversity remain contested and may be scale-dependent. We address productivity-diversity relationships in arthropod communities across two subarctic landscapes. Using the normalized difference vegetation index (NDVI) as a proxy for plant productivity, we targeted three different spatial scales: the local scale (0-10 km), the landscape scale (10-50 km), and the regional scale (>300 km). At each scale, we examined variation in species richness (alpha diversity), in species prevalence patterns, and in spatial turnover (beta diversity) along gradients of productivity. We hypothesized that alpha diversity will increase towards highly productive areas regardless of the scale assessed, and that this increase will be associated with either (a) higher, (b) equal, or (c) lower mean species prevalence towards increasing productivity. We expected to encounter the biggest difference in species turnover between low- and high-productivity sites at the landscape scale, owing to maximized effects of both environmental filtering and dispersal constraints at these two ends of the gradient. We found a positive relationship between alpha diversity and productivity across all scales and arthropod community types. Species-specific prevalence did not differ between the low and high ends of the productivity gradient, and similar proportions of the species pools were shared among sites under both conditions. As a net outcome, the increase in alpha diversity did not translate into higher community dissimilarity at the highly productive sites. Our findings also suggest that while higher productivity can sustain a larger species pool, this is not reflected in higher turnover among sites. Rather, the majority of species are widely represented along the productivity gradient, with a specific subset of species present in high- and low-productive sites. The same patterns prevail across scales and across flying and ground-dwelling arthropods. We conclude that patterns of alpha and beta diversity observed here are consistent with scenario (b) advanced a priori, that is, with equal mean species prevalence across the productivity gradient. A fraction of the local species pool is restricted to conditions of either low or high productivity, causing variation in species richness but not in species turnover.

Citizen science provides large amounts of biodiversity data. Key challenges in unlocking its full potential include engaging citizens with limited species identification skills and accelerating the transition from data collection to research and monitoring outputs. Here we use a large dataset from Finland to show how even citizens who cannot identify birds themselves can contribute to real-time predictions of avian distributions. This is achieved through a digital twin that combines smartphone-based citizen science with long-term knowledge in a continuously updating model. The app submits raw audio to a backend that classifies birds with machine learning, reducing variation in data quality and enabling validation and reclassification by continuously improving classifiers. We counteracted spatiotemporal sampling biases by interval recordings and permanent point count networks. Over 2 years, the app generated 15 million bird detections. Independent test data show that the digital-twin-informed models are more accurate at predicting bird spatiotemporal distributions. Because our approach is highly scalable and has the potential to generate biomonitoring data even in understudied areas, it could accelerate the flow of reliable biodiversity information and increase inclusivity in citizen science projects. Citizen science data are increasingly used in biodiversity monitoring. This study applies a digital twin approach to biodiversity monitoring using a large citizen science dataset on birds from Finland, demonstrating its potential for ecological forecasting.

Malaise traps offer an efficient method for monitoring insect communities. Nonetheless, insect taxa may differ in diel activity, whereas longer-term trap catches will integrate over such variation. Furthermore, fluctuations in weather may affect variation in taxon-specific activity, thus affecting the relation between the community observed in trap catches and the total community present. In this study, we asked how diel- and weather-related variation in insect activity influences Malaise trap catches. To this aim, we emptied 24 Malaise traps in a boreal forest in central Sweden with short time intervals across five consecutive days, while simultaneously recording the prevailing weather conditions. We found major differences in diel patterns across insect taxa. Significantly different communities were observed during different parts of the day, with the lowest abundances observed during mornings (6:00-12:00), lowest taxon richness during nights (22:00-06:00), and highest abundances and taxon richness during afternoons (12:00-18:00). Temperature affected both the occurrences and abundances of insect taxa in trap catches. Most taxa exhibited lower abundances with increased wind speed and cloud cover, and all taxa exhibited lower abundances with a rise in dew point. In conclusion, insect communities observed in Malaise trap samples vary substantially across the day in terms of both abundances and taxon richness. Before converting summary catches from longer-term Malaise traps into metrics of insects available as, for example, prey or pollinators, we should thus consider what fraction of these communities are available for these ecosystem services during different parts of the day.

Abstract Long-term monitoring data have enabled detection of phenological change, yet it remains poorly understood how its temporal dimensions— duration and choice of start and end years— influence the inferences drawn. To examine which phenological signals emerge at different temporal scales, we analyzed the longest continuous dataset on high-arctic plant and arthropod phenology, collected from 1996 to 2024 in the Zackenberg valley, northeast Greenland. These data have been used to suggest both rapid advancement of spring in the High Arctic (2007) and little directional change but decadal regime shifts (2023). To reconcile these differing conclusions, we quantified how trend estimates varied across moving time-windows and determined the minimum time-series length required to achieve a high probability of agreement with long-term trends. We find that while trend directionality shifts with temporal windows, confidence in trend estimates increases with time-series length. Using the full time-series, we show dampened signals of warming trends, with annual increases in spring and summer air temperatures by 0.04 [-0.05, 0.13] and 0.05 [-0.01, 0.11] °C per year, respectively, alongside a 0.82% [-1.85, 0.19] decline in spring snow cover. We also see modest advancement in the seasonal activity of most arthropod taxa (by ∼0.1 days/year), whereas flowering phenology shows no consistent directional change. Shorter time-series revealed cyclical patterns in abiotic drivers yet variable biotic responses, indicating that a single pattern of “climate change” will translate into varied responses within communities. Finally, almost two decades of data were needed to reliably capture long-term trends. Ecologically, these suggest that 1) phenological shifts in the High Arctic are more moderate than early assessments implied and 2) reflect a dynamic balance between species’ life histories and ongoing climate variability. These may alter interaction potentials within communities, with consequences for ecosystem functioning.

ABSTRACT Aim To test whether spatial trait–temperature relationships predicted by ecogeographic rules (Bergmann's rule and the thermal melanism hypothesis) can be used as space‐for‐time substitutions to anticipate temporal changes in community morphology under climate warming. Location Finland. Time Period 1993–2021. Major Taxa Studied Geometrid moths (Lepidoptera: Geometridae). Methods We analysed 29 years of standardised moth monitoring data spanning a 1200 km latitudinal gradient in Finland. For each moth community, we quantified mean wingspan and pigmentation as thermal morphological traits. We examined spatial relationships between community‐mean traits, latitude and temperature and assessed temporal trends in these traits in relation to long‐term changes in temperature. Results Across space, site‐level community‐mean trait composition over the study period showed that moth communities in colder northern regions were dominated by larger and darker species, whereas warmer southern communities comprised smaller and lighter‐coloured species, consistent with both ecogeographic rules. Through time, communities shifted towards smaller mean wingspan over the study period, consistent with the spatial expectations described by Bergmann's ecogeographic rule, while mean pigmentation exhibited no consistent directional change, contrary to expectations from the thermal melanism hypothesis. Main Conclusions Our results provide mixed support for space‐for‐time substitution in predicting climate‐driven changes in community morphology. While spatial patterns in wingspan successfully anticipated temporal shifts consistent with Bergmann's rule, spatial gradients in pigmentation did not translate into temporal change. These findings indicate that the predictive power of ecogeographic rules depends on the trait considered, and highlight both the promise and limitations of space‐for‐time approaches for forecasting climate‐driven reassembly of insect communities, particularly at high latitudes.

Resolving the ecological and evolutionary processes affecting biodiversity requires long-term community data. By separating short-term variability from directional change and by revealing lags spanning years to decades, these records expose the relative roles of dispersal, environmental filtering, species interactions, and drift, as well as the relevant timescales involved. Across realms, such records show that biodiversity change will frequently concern composition rather than species richness. Across time, the records resolve dynamics such as cycles and evolutionary change invisible to short studies. Long-term records also provide rare windows into eco-evolutionary change, from climate-driven selection on phenology to trait shifts feeding back to coexistence. We integrate insights from flagship sites, monitoring networks, and global databases, highlighting statistical advances that strengthen inference despite imperfect designs. However, the full potential of long-term community data remains underused. Comparative analyses across taxa and regions, together with harmonized sampling and the initiation of new long-term monitoring, are essential to overcome existing biases.

Pollinators, both wild and managed, form diverse associations with plants and microbes which affect the wellbeing of the plants and the pollinators. The method by which these associations are sampled impacts our understanding of the system. The common ways to understand pollinator-plant or pollinator-plant-microbe associations are to observe flower visits of insects, or to collect foraging individuals and identify the pollen and microbes they carry. Honey bees offer a test case for methods of sampling these associations. Hives of managed honey bees host thousands of pollinator individuals together with jointly-collected nectar which is turned into honey. Previous studies have used DNA preserved in honey to infer honey bee associations with plants and microbes. Here, we sampled honey, individual bees while they were foraging, and groups of bees from inside the hive. We identified plants and microbes on the surface of the bees or in the honey using DNA metabarcoding - expecting that bees sampled singly or in groups would reveal a subset of the associations recorded in the communal honey deposits. However, we found that each sample type revealed different aspects of the richness and community composition of plants and microbes encountered by bees. Both honey samples and hive bees had more plant and microbial taxa per sample than samples of individual bees. Though individual bees are subsets of the larger colony, pollen and microbe associations recovered from individual bees did not represent a subsample of associations recovered from groups of hive bees or from honey. Thus, while each sampling technique provides information about honey bee ecology, they are not equivalent. DNA in honey represents time-integrated associations between bees and the surrounding ecosystem; the hive bees provide a snapshot of current colony-level associations, and individual foraging bees capture intraspecific variation in foraging preferences and microbe exposure.

Soil eukaryotes, including fungi, protists, plants, and animals, are central to biosphere functioning and resilience. The Global Standardised Soil Eukaryome Dataset (GloSED) is the first dataset encompassing the entire spectrum of soil eukaryotes, covering 4,063 sampling sites in 121 countries on all continents, revealing nearly one million operational taxonomic units. All samples were collected and analysed using a standardised protocol minimizing technical biases. Long-read sequencing of full-length ITS and 18S-V9 regions provide broad taxonomic coverage and high-resolution identification supported by specialist curation of "dark taxa". A rigorous bioinformatic processing ensures against homopolymer errors, PCR-mediated chimeras, and index switching providing high data quality. The dataset is supported by raw sequences and an open-source containerised workflow for reproducible analyses. The samples are accompanied by land-cover description and directly measured soil pH, δ13C, δ15N, as well as P, K, Ca, Mg, and total C and N contents. GloSED is the first database that enables ecological and biogeographic studies of entire soil eukaryotic communities from local to global scales.

ABSTRACT As global temperatures rise, ecological communities are increasingly dominated by warm-affiliated species, a process known as community warming or thermophilisation. Yet, why different taxa exhibit different rates of community warming remains unclear. Habitat composition and structure are likely drivers of this variation, as the ecological consequences of warming are filtered by local environmental conditions. Using over 40 years of monitoring data spanning terrestrial (birds, insects, plants) and freshwater (phytoplankton) communities, we show that habitat structure determines how strongly communities track warming. Forest cover systematically slows thermophilisation by reducing communities’ sensitivity to temperature change, whereas habitat heterogeneity has weak and variable effects that differ among ecosystems. Together, these results demonstrate that uneven thermophilisation arises from habitat-mediated differences in how communities respond to a shared climatic signal. Incorporating these effects is essential for improving predictions of biodiversity change under ongoing climate warming.

Abstract DNA barcoding and metabarcoding have emerged as cost‐efficient, standardized methods for characterizing local biodiversity. Based on the sequencing of a small targeted gene fragment, it is theoretically possible to identify a wide diversity of taxa by comparing them with reference sequence databases. However, a key challenge for accurate taxonomic classification is the incompleteness of such databases, leading to most query sequences lacking species‐level matches. Where species‐level matches are missing, it may be possible to classify query sequences to a higher taxonomic rank, such as genus or family, based on the similarity of related reference taxa. The challenge then lies in confidently recognizing whether the sequence belongs to an unobserved (here, ‘novel’) taxon on a given rank. We evaluate the performance and utility of several methods for taxonomic classification. Methods were assessed based on the classification accuracy of both observed and novel taxa, accuracy of prediction confidence estimates and computational resource use. We focus on two widely studied cases: the COI barcode for arthropods, and the ITS barcode for fungi, with the latter representing an instance with substantially greater sequence length variation within classes. To benchmark the classification of novel taxa, we used curated datasets with partially distinct taxonomic distributions between the training and test sets. Novel taxa occurred at all evaluated taxonomic ranks, such as novel species in observed genera and novel genera in observed families. We further assessed the effect on performance when shifting from full‐length barcodes to shorter sequences as generated through metabarcoding in the test dataset. This study sheds light on the strengths and limitations of different classification algorithms across taxonomic groups and barcode characteristics. It demonstrates the supreme performance of phylogenetic placement methods (e.g., EPA‐ng) for classification of arthropod COI barcodes and composition‐based classifiers (e.g., SINTAX, RDP‐NBC, IDTAXA) for fungal ITS. Differences likely reflect barcode properties: COI is alignable and evolutionary constrained, favouring phylogenetic placement, whereas ITS is too variable for reliable alignment but rich in short‐motif (k‐mer) signal, favouring composition‐based classifiers. Across most algorithms, shorter sub‐regions performed comparably to full‐length barcodes, supporting the use of short reads for high‐throughput metabarcoding.

The Arctic is a hotspot of environmental change, as demonstrated by various monitoring programs and studies north of the Polar Circle. As these activities primarily focus on detecting shifts in biodiversity and phenology, the functional dynamics of the terrestrial ecological community of Arctic systems remain comparatively understudied. Current research coverage of actual species interactions exhibits considerable temporal and spatial heterogeneity. Nevertheless, there have been numerous attempts to synthesize existing knowledge into conceptual frameworks, including for well-studied regions such as Svalbard. However, these schemes often do not incorporate the idea of interaction strength. In this work, we aim to integrate existing knowledge on interaction strengths into a conceptual model of the Svalbard Ecological Network. In doing so, we also highlight current knowledge gaps and the challenges of establishing a robust baseline of species interactions in the region. Such current challenges cannot be overcome without coordinated efforts among multiple research groups.

How communities are structured into functional groups and trophic layers is key to understanding ecosystem functioning. Nonetheless, we lack insights about spatiotemporal variation in guild composition of communities and its causes. To investigate spatial and temporal patterns and drivers of variation in insect feeding guilds, we combined data from a nationwide survey of Swedish insects using Malaise traps and DNA metabarcoding with a comprehensive trait database. We assigned species into one of three feeding guilds (phytophages, saprophages, predators) or into one of three associated parasitoid guilds. We then analysed patterns in species richness for each guild. Species richness declined with latitude in all guilds. Beyond this gradient, local variation in species richness matched between hosts and their parasitoids. Yet, hosts and their parasitoids responded differently to habitat. The phenological peak of parasitoid species richness appeared later than the peak of their hosts, but the length of time lags varied among guilds. Spatiotemporal patterns were driven by guild-specific responses to temperature, though much variation remained between seasons and locations even when controlling for temperature. Overall, these patterns suggest that shifts in both climate and land use may alter the synchrony of insect trophic layers, with unknown consequences.