Insect decline in forests depends on species’ traits and may be mitigated by management

The study addresses whether and why insect populations are declining in forests, a major global biome that has received less attention compared to agricultural landscapes. Prior work indicated heterogeneous temporal trends among insect taxa and ecosystems, with land use implicated as a driver in farmlands. In forests, multiple processes can concurrently influence insect populations: natural disturbance-succession dynamics, forest management practices that alter habitat structure and resources (e.g., canopy, deadwood), and broader drivers such as climate change, nitrogen deposition, and landscape configuration. Beyond aggregate metrics (richness, abundance, biomass), species’ life-history traits (e.g., body size, trophic level, dispersal ability) may modulate responses and reveal mechanisms. Using long-term standardized sampling of beetles and true bugs across 140 German forest sites spanning a management gradient, the study aims to identify which local (site-level) and landscape-scale environmental variables are associated with temporal changes in insect communities, and how species’ traits predict species-level trends. The authors hypothesize that site-level declines relate to management-induced changes in canopy and forest structure; that larger-bodied species will decline more; and that higher trophic levels will exhibit stronger declines than herbivores.

Previous studies and a global meta-analysis reported terrestrial insect declines but with wide variation across taxa, regions, and ecosystems. Agricultural land-use intensification is frequently implicated, sparking public debate. Forests, despite harboring high biodiversity and often serving as refuges, show mixed evidence: some taxa (e.g., carabids, some moths) exhibited stability or increases, while others declined across taxonomic groups. Trait-based studies suggest species with narrow dietary niches, specialized habitat requirements, larger body size, lower dispersal ability, and higher trophic position are more vulnerable. Larger insects require more resources, tend to be rarer, and often show stronger negative trends. Higher trophic levels may be more sensitive due to smaller population sizes and cascading effects of environmental perturbations. Landscape context and forest structural attributes (e.g., tree diversity, deadwood, canopy openness, vertical layering) have been linked to insect diversity, yet disentangling local versus landscape drivers and management effects on temporal trends remains a gap. Prior work in the same regions found declines in species richness and biomass but did not identify clear drivers, motivating the present analysis that explicitly tests site- and landscape-level predictors and species traits.

Study system: 140 forest sites (1 ha each; 100 m × 100 m) across three German regions within the Biodiversity Exploratories project, spanning unmanaged broadleaved to intensively managed conifer stands, with varying proportions of non-native trees. Sites were selected by stratified random sampling to avoid initial diversity bias and monitored to avoid land-cover change. Sampling: Insects were sampled mainly from 2010–2017, with additional intermittent sampling in 2008, 2011, and 2017 for some sites. Two flight-interception (window) traps per site (approximately 1.5 m height) operated during the main flight season (harmonized to May–July for annual comparisons). Adults of Coleoptera and Heteroptera were identified to species (>99.7% success). Biomass was estimated from body length via an allometric function (y[kg] = 0.00348 × BL^2.1600). Diet/trophic group (herbivore, myceto-detritivore, omnivore, carnivore), vertical stratum, and dispersal ability were compiled from literature and expert curation. Response variables: Community responses per site-year included species richness, total abundance, and biomass for all insects and by trophic group. Species-level responses were the number of individuals per species per region (excluding single occurrences), summarized over years. Temporal trend metric: Pearson’s correlation coefficient (r) between sampling year and the response variable quantified direction and strength of temporal change. Site-level r values were computed for richness, abundance, and biomass for all insects and trophic groups. Species-level r was computed per species per region for species-level analyses. Environmental predictors (site scale): proportion of non-native trees and its change (inventory 1 vs. 2), harvesting intensity prior to sampling start (proportion of basal area harvested in the 1–5 years before 2008/2010), change in harvesting (2008–2017), deadwood volume and its change, canopy openness and its change, tree diversity. Environmental predictors (landscape scale): metrics derived from remote sensing within buffers (e.g., 1000 m), including structural heterogeneity (Sentinel-1 radar backscatter principal components), forest cover, disturbance intensity (percentage forest area lost adjacent to canopy-change, 2008–2017), vertical vegetation structure (effective number of layers from terrestrial LiDAR). Statistical analysis: Linear mixed-effects models (R, lme4) related site-level correlation coefficients (r) to environmental predictors. Models considered fixed effects at site and landscape scales; random effects accounted for regional/site structure as appropriate. Sensitivity analyses included models conditioning on baseline (first-year) conditions and models accounting for abundance when modeling species richness correlations. Species-level mixed models tested effects of log-transformed body length, total incidence (regional abundance/occurrence), trophic group, and dispersal ability, with species identity and region as crossed random effects. Multiple comparisons were adjusted via Bonferroni-Holm where relevant. Model diagnostics verified normality and homoscedasticity; collinearity was inspected (VIF < 2.5 when PCA variables were excluded).

- Across 140 sites and 10 sampling years (2008–2017 combined series), site-level correlations (Pearson’s r between year and response) were on average negative for total insect species richness (mean r = -0.182, 95% CI -0.257 to -0.106) and biomass (mean r = -0.152, 95% CI -0.237 to -0.066), but not for total abundance (mean r = 0.011, 95% CI -0.079 to 0.100). - A higher proportion of non-native trees was associated with more negative site-level correlations for: species richness (estimate = -0.106 ± 0.048 SE, p = 0.032), abundance (-0.177 ± 0.050, p = 0.001), and biomass (-0.137 ± 0.050, p = 0.016). Similar negative relationships were observed across trophic groups (e.g., herbivore abundance, myceto-detritivore abundance and biomass, omnivore richness and abundance, carnivore abundance). - Species richness correlations were additionally negatively related to: change in deadwood volume (-0.073 ± 0.036, p = 0.044), change in the proportion of non-native trees (-0.115 ± 0.042, p = 0.008), change in canopy openness (-0.130 ± 0.040, p = 0.002), and PC1 of landscape heterogeneity (Sentinel-1, 1000 m radius; -0.130 ± 0.040, p = 0.002). - Total abundance correlations decreased with harvesting intensity prior to sampling (-0.089 ± 0.044, p = 0.045); were negatively related to the effective number of vertical layers (-0.084 ± 0.042, p = 0.046) and to change in canopy openness (-0.104 ± 0.042, p = 0.041); and were positively related to tree diversity (0.103 ± 0.043, p = 0.017). - Biomass correlations were negatively related to the change in the proportion of non-native trees (-0.107 ± 0.044, p = 0.029). - Trophic groups differed in trends: herbivores showed positive average correlations for species richness (mean r = 0.306, 95% CI 0.232 to 0.379) and abundance (0.389, 0.304 to 0.468) while herbivore biomass was stable (-0.023, -0.112 to 0.067). Myceto-detritivores and omnivores showed negative average correlations; carnivore abundance had the strongest negative mean correlation (r = -0.299, 95% CI -0.371 to -0.223). - Harvesting intensity prior to sampling was linked to more negative correlations across groups (significant for herbivore abundance and omnivore richness and abundance). For herbivores, positive abundance correlations at low non-native tree proportion and low prior harvesting diminished towards zero or negative with increasing non-native tree proportion and harvesting. - Species-level analyses (1,050 species; 1,874 species×region combinations) showed more negative correlations over time for larger-bodied species (estimate = -0.026 ± 0.011, p = 0.016), more abundant/widespread species (total incidence: -0.039 ± 0.009, p < 0.001), and species at higher trophic levels; herbivores tended to increase relative to other groups. - Landscape-scale predictors generally had fewer and weaker associations with temporal trends than site-level forest properties. Positive associations were observed with tree diversity (abundance) and structural heterogeneity (species richness).

The results indicate that forest management legacies and composition, particularly high proportions of non-native trees and intensive timber harvesting prior to monitoring, are consistently associated with more negative temporal trends in forest insect communities. These patterns are compatible with a resource-driven mechanism: non-native tree dominance and harvesting can reduce the quantity and quality of resources and alter microclimate and habitat structure, disproportionally affecting higher trophic levels and larger species. Positive associations of trends with tree diversity and heterogeneous forest structure suggest that diverse resources across scales may buffer declines. Contrasting responses among trophic groups imply potential shifts in food web structure. Herbivores increased in species richness and abundance, possibly benefiting from increased understory vegetation, disturbed landscapes, associations with native host trees (e.g., European beech), and warming-enhanced plant–herbivore dynamics. In contrast, myceto-detritivores and carnivores showed declines, consistent with higher sensitivity of upper trophic levels and decomposers to habitat and resource changes, which could impair ecosystem functions such as nutrient cycling and top-down control. Deadwood, often emphasized for saproxylic taxa, showed limited explanatory power for temporal trends at the site scale in this dataset, potentially reflecting generally low deadwood amounts relative to old-growth conditions or complex time lags. Landscape variables exerted comparatively modest effects, highlighting the primacy of local forest conditions and management history for short- to medium-term trends in these systems. The species-level trait patterns—stronger declines in larger-bodied, more abundant, and higher-trophic species—underscore that not only rare, specialized species are at risk; common species also contribute substantially to overall declines. This has implications for ecosystem functioning given the disproportionate role of common species. Overall, while effect sizes were moderate and multiple drivers likely act simultaneously (the “death by a thousand cuts”), the convergence of evidence across community- and species-level analyses points to actionable management levers to mitigate declines: reduce dominance of non-native trees, moderate harvesting, and enhance tree diversity and structural heterogeneity.

Most insect communities in German forests exhibited declining trends in species richness and biomass over a decade, with declines linked to high proportions of non-native trees and prior intensive harvesting. Species’ traits predicted vulnerability: larger-bodied, more abundant/widespread species and higher trophic levels declined more, whereas herbivores tended to increase. These patterns suggest food web reorganization with potential consequences for ecosystem functions. Management can help mitigate declines by promoting natural tree species composition, reducing harvesting intensity (especially avoiding high pre-monitoring extraction), and fostering tree diversity and structural heterogeneity at local and landscape scales. Future research should employ manipulative experiments to establish causality of specific forest attributes (e.g., tree composition, harvesting regimes, canopy structure) and examine interactions with climate variability (including post-2017 droughts), time lags, and resource pathways across trophic levels.

- Observational design limits causal inference; evidence is correlative and circumstantial. - Time series end in 2017 and do not include extreme drought years post-2018 that could alter trends. - Some methodological complexities (e.g., intermittent sampling at a subset of sites, sensitivity of correlations to baseline conditions) may influence effect estimates, though sensitivity analyses were conducted. - Limited explanatory power of some landscape metrics and deadwood variables may reflect measurement constraints, low variance, or time lags in responses. - Analyses focus on Coleoptera and Heteroptera captured by window traps, which may not represent all insect taxa or strata equally.