🧬 biology

National mapping of hyperdiverse insect communities beyond named species

By integrating standardized DNA metabarcoding from 196 Malaise traps with order-specific joint species distribution models, this study successfully mapped the national distributions of 11,694 insect taxa across Sweden, revealing that while unnamed "dark taxa" constitute over half the diversity and mirror broad richness patterns, distinct spatial variations exist among different insect orders.

Original authors: Jan Borgelt, Francesca Verones, Tobias Andermann

Published 2026-10-12

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jan Borgelt, Francesca Verones, Tobias Andermann

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

To understand where life is most abundant on Earth, scientists have long relied on counting species in specific places. This approach, however, hits a wall when faced with the sheer volume of life that exists in the form of insects. These tiny creatures make up the majority of animal diversity on land, yet for most of them, we do not know their names, nor do we know exactly where they live. Traditional methods of mapping biodiversity require experts to identify every single insect by hand, a process that is slow, expensive, and impossible to scale across an entire country. Consequently, maps of insect life often remain incomplete, showing only the few species that are well-known while leaving the vast majority of the insect world invisible. This gap in knowledge is a significant hurdle for conservation, as we cannot protect what we cannot see or map.

A new study by researchers in Sweden offers a way to bridge this gap without needing to name every single creature first. By combining a modern technique called DNA metabarcoding with advanced computer modeling, the team created the first national-scale map of insect diversity for Sweden that includes both named species and the many unnamed ones. The researchers set up 196 traps across the country to catch insects, then extracted and sequenced their DNA to identify thousands of distinct groups. Instead of discarding the groups they could not name, they treated them as valid units of biodiversity. Using a statistical framework that models the distributions of many species simultaneously, they predicted where these thousands of groups live across the entire nation. The result is a detailed picture of insect richness that reveals a clear pattern: insect diversity is highest in the south and declines steadily toward the north, but this broad trend hides a complex reality where different groups of insects have very different hotspots.

The study focused on 11,694 distinct insect groups, a number that includes 5,160 named species and 6,534 groups that the researchers refer to as "dark taxa." These dark taxa are molecular clusters that lack a formal species name, representing the vast portion of insect life that remains taxonomically unresolved. The researchers found that these unnamed groups make up more than half of the total insect community they mapped. Despite being detected at slightly fewer traps than the named species, the dark taxa contributed 51.9% of the total mapped richness. This means that if the researchers had ignored the unnamed insects, they would have missed more than half of the biodiversity in the country. The spatial patterns of these dark taxa closely matched those of the named species, following the same general south-to-north decline in diversity. This suggests that the lack of a scientific name does not mean an insect group is invisible to the landscape; they respond to environmental conditions in similar ways to their named cousins.

However, the study also revealed that the overall picture of insect diversity is not a simple sum of identical patterns. When the researchers looked at specific orders of insects, such as flies, beetles, or moths, they found that each group has its own unique geography. While the total richness map showed a strong gradient from south to north, this pattern was not representative of all groups. For instance, the richness of stoneflies was actually highest in the north and west, showing a distribution that was almost the opposite of the overall trend. Similarly, caddisflies showed no clear relationship to the national pattern at all. This means that a map showing the total number of insect species in an area might miss the specific high-diversity zones for certain groups. The researchers found that only a tiny fraction of the country served as a hotspot for ten or more different insect orders simultaneously. This highlights that protecting biodiversity requires looking beyond a single aggregate map and understanding the specific needs and distributions of different insect groups.

The researchers used a sophisticated modeling approach that allowed them to predict the presence of each insect group across the entire country, even in areas where no traps were set. They tested their models against the actual data from the traps and found that the predictions were highly accurate, correctly ranking the richness of sites with high precision. The models showed that the unnamed dark taxa were just as predictable as the named species, challenging the idea that we must know the name of a species before we can map its distribution. The study also noted that while the overall richness patterns were similar between named and unnamed groups, the unnamed groups tended to have slightly smaller geographic ranges on average. This difference varied among insect orders, with some groups showing larger ranges for unnamed taxa and others showing smaller ones. These nuances suggest that while the broad patterns of diversity are robust, the specific distributions of individual groups can differ in meaningful ways.

This work demonstrates a scalable path forward for biodiversity monitoring. By retaining the molecular data of unnamed groups rather than discarding them, the researchers were able to capture a much more complete picture of the insect world. The resulting maps provide a baseline for understanding where insect life is concentrated and where it is scarce, offering valuable information for conservation planning. The study confirms that national biodiversity mapping does not need to wait for taxonomists to name every species. Instead, by using standardized DNA sampling and joint species distribution models, scientists can now generate detailed, national-scale maps of hyperdiverse communities, even when much of the assemblage remains taxonomically unresolved. This approach allows for the identification of broad patterns and the detection of deviations, guiding more detailed field surveys and helping to prioritize areas for protection in a world where insect diversity is under increasing pressure.

Technical Summary: National mapping of hyperdiverse insect communities beyond named species

Problem Statement
Spatially explicit knowledge of biodiversity distribution remains profoundly uneven, particularly for hyperdiverse groups like insects. While species distribution models (SDMs) and joint species distribution models (jSDMs) offer a route to predict distributions across landscapes, their application to hyperdiverse assemblages is hindered by two primary "shortfalls": the Linnean shortfall (incomplete taxonomic knowledge) and the Wallacean shortfall (incomplete knowledge of species distributions). Conventional approaches face a trade-off between taxonomic breadth, sampling consistency, and spatial coverage. Furthermore, standard biodiversity inventories often discard "dark taxa"—molecular clusters lacking species-level assignments—thereby losing a substantial portion of the sampled assemblage and its spatial signal. There is a need for a scalable framework that can map entire communities, including taxonomically unresolved units, at a national scale without requiring complete taxonomic resolution first.

Methodology
The authors combined standardized DNA metabarcoding with joint species distribution modelling to map insect diversity across Sweden.

Data Source: The study utilized data from the Swedish Insect Biome Atlas, comprising 196 Malaise traps distributed across 98 sites representing diverse habitats (forests, wetlands, croplands, etc.). Insects were collected in 2019 and processed via DNA metabarcoding.
Taxonomic Processing: Operational Taxonomic Units (OTUs) were used. Named species were consolidated, while OTUs lacking species-level assignments were retained as "dark taxa." The final dataset included 11,694 taxa across 12 insect orders (5,160 named species and 6,534 dark taxa). Taxa were filtered to ensure presence in at least 10 traps and absence in at least 10 traps.
Environmental Predictors: Twenty environmental variables were used at a 10-m spatial resolution, including climate data (CHELSA v2.1), vegetation structure (LiDAR-derived tree height, canopy cover), soil moisture, peat depth, land cover, and anthropogenic factors (distance to roads, forest edges).
Modelling Framework: Separate jSDMs were fitted for each of the 12 insect orders using the R package sjSDM. The models employed a binomial probit formulation with a feed-forward neural network to capture non-linear taxon-environment relationships.
- Architecture: Five hidden-layer architectures were tested.
- Regularization: Elastic-net regularization (combining L1 and L2) was applied to environmental coefficients and the residual covariance component.
- Hyperparameter Tuning: 500 hyperparameter configurations were evaluated per order using 5-fold site-grouped cross-validation (CV) to prevent spatial autocorrelation bias. The top five configurations (based on a one-standard-error rule applied to mean test AUC) were selected for final prediction.
Prediction and Mapping: National predictions were generated for 2 million randomly sampled 10-m cells. Taxon-specific occurrence probabilities were converted to presence-absence using taxon-specific F1 thresholds. Richness was calculated by summing binary predictions across taxa and orders.

Key Results

National Diversity Gradient: The models revealed a pronounced south-to-north decline in predicted insect richness, ranging from 374 to 4,844 taxa per ~2-km grid cell. Richness was highest in southern Sweden and lowest in the western mountains.
Contribution of Dark Taxa: Dark taxa comprised 55.9% of the modelled assemblage (6,534 taxa) and contributed 51.9% of the mapped richness. While their detection frequency was slightly lower than named species (median 24 vs. 30 traps), their cross-validated discrimination (AUC) was nearly identical (0.772 vs. 0.768).
Spatial Concordance and Divergence: At the national scale, the richness patterns of dark taxa closely reproduced those of named species (Spearman's $\rho$ = 0.976), suggesting that taxonomic incompleteness does not preclude accurate mapping of broad geographic gradients. However, at the individual taxon level, dark taxa had modestly smaller predicted spatial footprints (6.2% smaller area to contain 90% of probability mass) compared to named species, though this varied by order.
Order-Specific Heterogeneity: Aggregate richness maps masked significant variation among orders. While Diptera, Hymenoptera, Lepidoptera, Hemiptera, and Coleoptera correlated strongly with the aggregate pattern, Trichoptera showed no correlation, and Plecoptera exhibited a strongly opposing distribution (higher richness in the north/west). Consequently, only 0.17% of the mapped area was a hotspot for at least ten of the 12 orders.
Model Performance: Cross-validated discrimination was generally high, with a median AUC of 0.770 across all taxa. 71.8% of taxa exceeded an AUC of 0.7.

Significance and Claims
The paper claims to demonstrate a scalable route for mapping hyperdiverse assemblages at a national scale, even when taxonomic knowledge is incomplete. By retaining taxon-level predictions for both named species and dark taxa, the framework allows aggregate richness patterns to remain connected to the individual distributions that generate them. The study argues that progress in spatial mapping does not require the resolution of the Linnean shortfall first; taxonomically unresolved molecular clusters can be represented in the analysis rather than discarded.

The authors position this work as a method to overcome the persistent trade-off between taxonomic breadth and spatial coverage. They suggest that such maps can complement field inventories by identifying broad patterns and highlighting taxonomic or geographical departures, thereby directing more detailed surveys. The framework is presented as transferable to other species-rich, poorly known groups (e.g., soil fungi, microbial communities) and applicable to national biodiversity monitoring and reporting, including efforts associated with the Kunming-Montreal Global Biodiversity Framework. The study emphasizes that while the maps provide model-based estimates of spatial patterns, they should be interpreted as distributions of taxa detectable under the specific sampling protocol rather than complete local inventories.

More like this