Species-habitat networks reveal conservation implications that other community analyses do not detect

Grassland restoration is an important conservation intervention supporting declining insect pollinators in threatened calcareous grassland landscapes. While the success of restoration is often quantified using simple measures of diversity or similarity to a target communities, these measures do not capture fundamental aspect of community reconstruction. Here, we develop species-habitat networks that aim to define habitat level foraging dependencies of pollinators across restored grassland landscapes and compare their value to these more conventional measures of community restoration. We assessed this across Salisbury Plain (UK) which represents the largest area of chalk grassland in N.W. Europe, encompassing six distinct management types aimed at restoration and maintenance of species-rich calcareous grassland. Sites that were previously disturbed or reverting from arable agriculture were comparable in their pollinator abundance and species richness to those of ancient grasslands, although intensively managed grasslands supported low pollinator diversity. Applying our species-habitat network approach we found that pollinator communities in grasslands recovering from past military disturbance showed stronger modular associations with those in ancient grasslands compared to areas recovering from intensive agriculture. This highlights the importance of habitat history in shaping restoration trajectories. We propose that species-habitat networks should be part of the standard analytical toolkit assessing the effectiveness of restoration at landscape scale, particularly for mobile species such as insects.

Methods
The study took place in the Salisbury Plain Training Area, southern England. It covered six distinct grassland habitats across the area, representing a gradient of plant species richness and flower resources. These included ancient grassland, previously disturbed grassland, recently disturbed grassland, reverting grassland, intensive grazing, and ungrazed grassland.

In 2010, 36 permanently marked line transects (6 × 85 m) were established, one in the center of each habitat plot. Sampling occurred four times between May 30 and September 19, 2011, monthly. Surveys took place between 10:00 and 17:30, coinciding with major pollinator flight periods. The method followed the UK Butterfly Monitoring Scheme (BMS) and adapted for bumblebee surveys. Surveys occurred under standardized environmental conditions (wind speed <5.5 m/s, no rain, temperature >17ºC if cloudy or >13ºC if less than 40% cloud cover). Temperature, sunshine percentage, and wind speed were recorded. It took approximately two days to complete all transect counts for the 36 plots.

During each survey, all butterflies and day-flying moths were identified and counted at the species level. Foraging bumblebees were also recorded to species level, and rare species were verified with voucher specimens. Bombus terrestris and B. lucorum were recorded collectively as they couldn't be reliably distinguished in the field. Summed counts were made for honeybees, solitary bees, and hoverflies. The first flowering plant visited by each bee or butterfly was recorded to species, enabling the construction of plant-pollinator interaction webs for each grassland type.

Regarding flower resources, during each pollinator survey, eight 0.5 × 0.5 m quadrats were used, spaced equally along each transect. All flowering dicotyledonous plants were identified, and the number of flower units per species was counted. Flower units encompassed single flowers, multi-flowered stems, flower-heads, and umbels, providing estimates of timing, diversity, and abundance of floral resources available to pollinators in each grassland habitat throughout the season.