Recent arthropod declines in agricultural landscapes can threaten biodiversity and the provision of ecosystem services. Wildflower strips (WFS) represent popular measures to support biodiversity. Most studies investigated their effects on pollinators, whereas ground-dwelling arthropods are less studied. In addition, time after WFS establishment is a relevant factor, but most experiments are performed in a single season. Here, we evaluated the effects of WFS on ground-dwelling arthropods (carabids, spiders, and myriapods) and ecosystem services (pest and weed seed predation) across three years. Using a standardised experimental design with paired control and WFS margins in 12 fields, we sampled before and up to the first two years after WFS establishment while also assessing spillover into adjacent fields. Our analyses revealed that spiders, as well as total pest predation and predation by rodents, were enhanced in two-year-old WFS, whereas similar but weaker patterns were observed for carabid and myriapod species richness. Spillover patterns were weak, and only carabid species richness was enhanced in parts of the field neighbouring WFS. These results reinforce the important role of perennial WFS in supporting beneficial ground-dwelling arthropod groups while also highlighting that taxon-specific responses can hinder the design of general measures to support biodiversity on arable land.

Study sites and experimental design

The study was performed between 2020 and 2022 in the northeast of Prague, Czech Republic (50°7′–50°10′N, 14°31′–14°34′W; mean altitude 252 m a.s.l), where we selected 12 conventionally managed arable fields (areas ranging from 3.5 to 33.9 ha; Table S1; Figure S1a). Fields were managed conventionally by the VIN AGRO company and sown with cereals, sugar beet, oilseed rape, and mustard. Following a BACI approach (i.e., before and after design), within each field, two margins separated by at least 75 m (mean 397 m) were chosen for sampling arthropod and ecosystem services before and after sowing flower strips. In addition to the field margins (12 m from the original boundary), we also sampled the interior of the fields adjacent to both margins, at 36 and 60 m (see Venturo et al., 2024 for details), making a total of six transects per field. In 2020 (hereafter, before WFS establishment), both field margins were part of the arable fields, whereas in 2021, one of the margins was selected as a control margin, and farmers managed it as the rest of the field, whereas the second margin was used for sowing experimental wildflower strips. Wildflower strips were sown after the crop harvest in 2020. Half of them were established in early September 2020, and the second half were established in early April 2021, depending on the sown crops. WFS were 24 m in width, and their length varied according to the size and shape of individual fields (see details in Table S1). The seed mixture used for wildflower strips was composed of 24 plant species and included three species of flowering cover crops, 17 species of other flowering dicots (mainly Fabaceae, Asteraceae, Apiaceae), and 4 species of grasses (for the detailed seed mixture composition see Table S2). WFS were sown using 25 kg of seeds per ha and managed by mowing and removing biomass. In the first year after establishment, WFS were mown once during the peak summer, whereas in the second year, they were mown twice (late June and early October). 

Arthropod and ecosystem services sampling

During the three study years, we sampled three groups of common ground-dwelling arthropods in arable fields, which are also relevant for ecosystem service provision: carabids, myriapods (diplopods and chilopods), and spiders. A 20 m transect, parallel to the field margin, was established at each transect. Arthropods were sampled using three pitfall traps per transect, separated by 10 m exposed for approximately 30 days (from mid-May to mid-June during the three years). Traps were made of a 0.5 l plastic cup (diameter 10 cm) covered by an aluminium roof (25 × 30 cm) and filled with 33% propylene glycol (diluted with water) as preservation fluid. After the exposure period, trap contents were filtered, and the material collected was stored in plastic bags in a freezer at -20°C until sorting. For each arthropod taxon, specialists identified specimens to the species level, and species richness and activity-density were calculated for each trap.

To estimate pest predation rates, we used the artificial sentinel prey method (Lövei & Ferrante, 2017). Artificial caterpillars (25 mm long and 3 mm diameter) made of light green-coloured plasticine (® Smeedi) and glued to a small piece of reed were used as sentinel prey. Twenty caterpillars were placed at the ground level on each transect, separated by 1 m, and exposed for 48 hours in two sampling periods per year, late spring (May, hereafter spring) and early summer (June, hereafter summer), making a total of 8,640 exposed caterpillars. Caterpillars were taken to the lab in plastic tubes and visually inspected for predation marks using an optical stereo microscope. Predation marks were classified into three categories: insects, small mammals, and birds.

To estimate weed seed predation rates, we employed the seed card method. Seed cards were made of rectangular pieces of sandpaper (5 x 10 cm) with 30 seeds of Taraxacum spp. glued with a tasteless and odourless glue (Westerman et al., 2003). On each transect, six seed cards separated by 4 m were exposed for six to nine days (according to weather conditions) at the ground level, fixed with two nails (2,592 seed cards in total). Each year, seed cards were exposed during two sampling periods, late spring (May, hereafter spring) and early summer (June, hereafter summer). After exposure, seed cards were taken to the laboratory in separate envelopes and dried. Undamaged seeds were counted to calculate predation rates as the number of predated or missing seeds divided by the number of exposed seeds on each transect. To correct for variations in exposure times, daily predation rates were calculated as predation rates divided by the number of exposed days.