Linear infrastructure habitats increase landscape-scale diversity of plants but not of flower-visiting insects

Semi-natural grasslands, traditionally managed by mowing or grazing without agrochemical inputs, have declined drastically in Sweden (about 90% loss over the last 80–90 years). This has driven declines in associated species and altered community dominance patterns, potentially affecting ecosystem functioning. Linear infrastructure habitats (road verges, railways, power line corridors) are novel, anthropogenically managed open habitats that can resemble semi-natural grasslands under suitable abiotic conditions and can facilitate movement and spread of species. Despite local high diversity reported in such habitats, it remains unclear whether they increase biodiversity at the landscape (gamma) scale. Biodiversity assessment should consider not only species richness but also evenness and phylogenetic diversity. The study aimed to test whether greater amounts of linear infrastructure habitats and semi-natural grasslands in 2×2 km landscapes increase species richness, community evenness, and phylogenetic diversity of plants, butterflies, and bumblebees. The authors predicted positive relationships between these diversity aspects and the area/connectivity provided by linear infrastructure habitats.

Previous work documents substantial loss of semi-natural grasslands in Sweden and associated biodiversity declines. Linear features can harbor rare and threatened species and may match or exceed local diversity of natural counterparts but can also have negative impacts (e.g., traffic mortality, pollution, barriers to movement). Species-area relationships suggest more habitat increases alpha and gamma diversity, and increased connectivity may further enhance diversity, particularly for disturbance-driven grassland species. Biodiversity facets beyond richness, such as evenness and phylogenetic diversity, are important predictors of ecosystem functioning. If linear habitats increase area and connectivity, they may raise the probability of occurrence of rare and evolutionarily distinct species, potentially elevating phylogenetic diversity. However, evidence is mixed on whether area/connectivity effects at patch scales translate to landscape scales, and road verges versus power line corridors differ notably in management intensity and disturbance regimes, which may differentially influence biodiversity.

Study design: Thirty-two 2×2 km (4 km²) landscapes in east-central Sweden were selected in a crossed design: presence (n=16) vs absence (n=16) of a power line corridor traversing forest, and high (n=16) vs low (n=16) road density (proxy for road verge area). Landscapes were forest-dominated (45–81% forest) and spanned a gradient of semi-natural grassland area (0.8–9.9%). Locations are shown in Figure 5. Sampling habitats: In each landscape, one site of each habitat was surveyed: a grazed (or previously grazed) semi-natural grassland, a road verge along a larger paved road, and a road verge along a smaller gravel road. In the 16 landscapes with power lines, one power line corridor site was also surveyed (managed on an ~8-year cycle; widths 28–95 m, mean 39.9 m). Transects and taxa: For each site, a 200 m transect was divided into four 50 m sections. Butterflies (including day-flying burnet moths) were surveyed using Pollard walks, counting individuals within 5 m (2.5 m each side, 5 m ahead/up); narrower verges used full width. Bumblebees were netted within 1 m on each side and identified in the field or lab (queens not collected). Butterflies and bumblebees were surveyed four times between 1 June and 23 August 2016, 10:00–16:00, under suitable weather and randomized daily site order. Plants: In each transect section, a 1 m² quadrat at the center recorded presence-absence from a predefined list of 170 grassland plant species. Additionally, indicator plant species (55 species indicating long-term grassland management/soil conditions) were surveyed along the entire transect once between 13 July and 24 August 2016. Land use quantification: Using ArcMap 10.7 and Swedish databases (GSD Terrain Map; TUVA), the areas of arable land, semi-natural pastures (high and low nature value), road verges (public/private roads), power line corridors (only forest-crossing segments), water, urban land, and railways were estimated. Verge areas used mean verge widths (public 5.6 m; private 5.9 m); power line corridor area used mean width 39.9 m multiplied by length. Analytical approach: To avoid sampling bias (half the landscapes lacked power lines), analyses at the landscape scale excluded species recorded in power line corridor transects; only semi-natural grasslands and road verges were pooled within each landscape. For each taxon (butterflies, bumblebees, plants, and indicator plants), gamma-scale species richness was modeled using general linear models (indicator plants via Poisson GLM) with predictors: amount of semi-natural grasslands, presence/absence of power line corridor in the landscape, and road verge density (high/low). All two-way interactions were tested; model selection used AICc (ΔAICc<2 rule favoring simpler models unless an interaction model was within 2 AICc). Community composition differences among the four landscape categories were assessed via NMDS ordination. Evenness: Shannon’s evenness (E_shannon) was calculated per landscape for butterflies, bumblebees, and plants, then modeled against the same predictors and interactions. Diversity profiles (Hill numbers for q=0,1,2,∞ corresponding to richness, Shannon, Simpson, Berger-Parker) were generated to compare abundance distributions. Phylogenetic diversity: Phylogenetic trees for each group were assembled from the Open Tree of Life via rotl; two plants (Rosa dumalis, Potentilla tabernaemontani) were removed due to taxonomy mismatch. Branch lengths were assigned using Grafen’s method (ape). Faith’s PD was computed (Picante) and standardized using 1000 randomizations to obtain standardized effect size (SES-PD). SES-PD was modeled with the same predictors. Correlations between SES-PD and species richness were assessed with Pearson correlation.

• Sampling totals: 2,704 butterfly individuals (51 species), 1,316 bumblebee individuals (19 species), and 128 plant species across the three habitats within the 32 landscapes. Most abundant butterfly: Aphantopus hyperantus (974), followed by Coenonympha arcania (281), Thymelicus lineola (178), Pieris napi (167). Most abundant bumblebees: Bombus pascuorum (347), B. lucorum (170), B. terrestris (157). Common plants included Achillea millefolium, Agrostis capillaris, Dactylis glomerata, Taraxacum vulgare.
• Species richness: Landscapes with power line corridors had on average ~6 (±2.36) more plant species than landscapes without (t=2.6, p=0.01). No relationship between linear infrastructure amount and butterfly or bumblebee species richness, nor indicator plant richness. No relationship between semi-natural grassland area and species richness for any group.
• Species composition: NMDS showed no clear clustering among the four landscape categories for butterflies (stress 0.23), bumblebees (0.24), or plants (0.24), indicating similar species identities across categories.
• Evenness: Diversity profiles indicated similar abundance distributions across landscape categories for all taxa. Butterfly communities had low evenness; bumblebees also low but less extreme; plant communities showed higher evenness overall. Mean plant E_shannon across landscapes was 0.93 ± 0.01 (range 0.90–0.96). A significant interaction for plants: without power lines, increasing road verge proportion decreased plant evenness; with power lines, increasing road verge proportion increased evenness (estimate 0.023 ± 0.01, t=2.1, p=0.04). No association between semi-natural grassland amount and evenness in any group.
• Phylogenetic diversity: No relationship between SES-PD and the amount of linear infrastructure or semi-natural grassland for butterflies, bumblebees, or plants. The simplest additive models were generally best; for butterflies and plants, interaction models were within ΔAICc<2 but not superior. SES-PD was not correlated with species richness (butterflies r=-0.23, p=0.2; bumblebees r=0.24, p=0.19; plants r=-0.09, p=0.63).

Despite high local diversity reported for linear infrastructure habitats, their contribution to landscape-scale biodiversity was limited overall. Power line corridors were associated with a modest but significant increase in the number of plant species and higher plant evenness at the landscape scale, but similar effects were absent for butterflies and bumblebees, and no increases in phylogenetic diversity were detected. Community composition did not differ among landscape categories, suggesting the plant richness increase was not due to unique species sets. Two non-mutually exclusive explanations are discussed: (1) landscapes with power line corridors may have larger effective grassland area, which could elevate plant richness; however, this would also be expected to affect insects and indicator plants, which was not observed. (2) Extinction debt: additional area and connectivity provided by power line corridors may be slowing plant species loss in landscapes experiencing long-term grassland decline; insects with shorter generation times may have already responded (i.e., paid the debt). Road verges did not show positive effects, potentially due to different management (frequent mowing 1–3 times/year) and disturbances (pollution, traffic) compared to power line corridors (8-year management cycle), and because roads can increase mortality and act as movement barriers. No relationship between semi-natural grassland area (0.8–9.9%) and diversity suggests that the habitat proportion/variation may have been too low to detect species–area effects at the landscape scale or that such relationships are context-dependent. Butterfly and bumblebee evenness was low across categories, driven by dominant species (Aphantopus hyperantus; Bombus pascuorum), independent of linear infrastructure effects. While theory suggests habitat area and connectivity can influence evenness, their effects can be opposing and were not separable here. Lack of phylogenetic responses implies any ongoing species losses are not strongly phylogenetically clustered for these groups at this scale, and for bumblebees, limited phylogenetic spread may constrain detectable variation.

Power line corridors can make a positive contribution to landscape-scale plant biodiversity (richness and evenness), whereas effects on flower-visiting insects (butterflies, bumblebees) were not detected, and road verge effects were unclear. Linear infrastructure habitats did not increase phylogenetic diversity at the landscape scale. Future research should determine whether these habitats function as sources or sinks for insect populations, expand focus to other pollinators (e.g., solitary bees, moths), and estimate demographic parameters to elucidate mechanisms. Incorporating local edaphic conditions and historical habitat amount/connectivity using historical maps and aerial imagery could reveal extinction debt dynamics, particularly in landscapes with power line corridors. Conservation assessments are needed in regions with near-complete loss of semi-natural grasslands, as the value of linear infrastructure habitats may depend on remaining habitat context.

• Analyses excluded species recorded in power line corridor transects to avoid sampling bias, potentially underestimating total contributions of corridors to landscape gamma diversity.
• The proportion and variation of semi-natural grassland area in landscapes were low (0.8–9.9%), possibly limiting power to detect species–area relationships.
• Plant abundance was proxied by presence in 1 m² quadrats (frequency), not true abundance, which can inflate evenness metrics and complicate comparisons with insect evenness.
• Road verge density was used as a proxy for verge area and may not fully capture habitat quality or management intensity differences.
• Inability to disentangle effects of habitat area versus connectivity on community evenness.
• Phylogenetic analyses used Grafen branch lengths rather than time-calibrated trees; limited phylogenetic scope (especially in bumblebees) may constrain detection of SES-PD responses.
• Single-season sampling (2016) and focus on butterflies and bumblebees; other pollinator groups were not assessed.
• Study landscapes were forest-dominated in Sweden; generalizability to other regions/landscape contexts may be limited.