To enable reproduction of analyses in the linked journal article, we provide a table of habitats, woody linear features and elevation of all pollinator survey transect sections. This table can be linked to the publicly available data, based on square ID and section ID, to reproduce the analyses (except for one discontinued square with ID marked ‘NA’ in the publicly available data).

Journal article abstract:

1. Pollinating insects provide economic value by improving crop yield. They are also functionally and culturally important across ecosystems outside of cropland. To understand landscape-level drivers of pollinator declines, and guide policy and intervention to reverse declines, studies must cover (1) multiple insect and plant taxa and (2) a range of agricultural and semi-natural land uses. Furthermore, in an era of woodland restoration initiatives and rewilding ideologies, the contribution of woodland and woody linear features (WLFs; e.g. hedgerows) to pollinator abundance demands further investigation. 2. We demonstrate fine-scale analysis of high-quality, co-located measurements from a national environmental survey. We relate pollinator transect counts to ground-truth habitat and WLF maps across 300 1km squares in Wales, UK. We look at effects of habitat type, flower cover, WLF density and habitat diversity on summer abundance (July and August) of eight insect groups, representing three insect orders (Lepidoptera, Hymenoptera and Diptera). 3. Compared with improved grassland (the dominant habitat in Wales), pollinator abundance is consistently higher in cropland and woodland - especially broadleaved woodland. For solitary bees and two hoverfly groups, abundance is predicted to be at least 1.5× higher in woodland ecosystems than elsewhere. Furthermore, we estimate contributions of WLFs to abundance in agriculturally improved habitats to be up to 14% for honeybees and up to 21% for hoverflies. 4. The abundance of all insect groups increases with flower cover, which is a key mechanism through which woodland, cropland and grassland support pollinators. Importantly, we observe diminishing returns of increasing flower cover for abundance of non-Apis pollinator groups, expecting roughly twice the increase in abundance per % flower cover from 0-5%, as compared with 10-15%. However, the shape of the relationship was inverted for honeybees, which showed steeper increases in abundance at higher flower cover. 5. Synthesis & applications: We provide a holistic view of the drivers of pollinator abundance in Wales, in which flower cover, woodland, WLFs and cropland are critical. We propose a key role for woodland creation, hedge-laying and farmland heterogeneity within future land management incentive schemes. Finally, we suggest targeting of interventions to maximise benefits for non-Apis pollinators. Specifically, increasing floral provision in areas where existing flower cover is low – e.g. in flower-poor improved grasslands - could effectively increase pollinator abundance and diversity, while prioritising wild over managed species.

Further information on habitat and landscape feature survey methodology is available in supporting documentation for the published datasets (Maskell et al., 2020a, 2020b). This dataset represents an intersection of spatial versions of the published datasets with locations of a set of 200m pollinator transect sections. The main linked paper (Alison et al. 2021) also provides lots of additional information.

Habitat, woody linear features (WLFs) and elevation variables were extracted for each transect section using ArcGIS Desktop 10.6 (ESRI, Redlands, California). To compare transect sections in different habitats, we classified the underlying broad habitat for 200m pollinator transect sections through intersection with habitat polygons. For each section, the broad habitat accounting for the greatest proportion of its length was assigned as the dominant habitat type. During this process we sought to maximise sample size, while avoiding unrepresentative classifications related to missing or ambiguous habitat data. As such, we made no dominant habitat assignment if (1) the dominant broad habitat accounted for <100m of the section, (2) a section had incomplete (<90%) overlap with habitat survey data or (3) the dominant broad habitat was recorded as “Mosaic”. We also calculated the Shannon index of habitat diversity for each transect section, taking into account the total number of broad habitats and the dominance among them. To relate flower cover and insect counts to habitat classifications at the finest possible scale and resolution, our habitat intersections were length-based, and not area- or buffer-based. However, a Euclidean buffer was necessary to extract the total density of WLFs (m ha^-1) within 10m of each transect section, including managed WLFs (hedges), unmanaged WLFs (lines of trees), and forestry linear features (belts of trees or scrub). Outside of GMEP survey data, a 5m resolution raster of elevation was provided by Welsh Government (the Nextmap Britain DTM by Intermap Technologies). The elevation of each transect section was taken as the mean elevation of all vertices in the digitised transect section. The final modelled dataset included pollinator counts from 4,449 section-visits across 295 1km squares.

Maskell, L. C., Astbury, S., Burden, A., Emmett, B. A., Garbutt, A., Goodwin, A., Henrys, P., Jarvis, S., Norton, L. R., Owen, A., Sharps, K., Smart, S. M., Williams, B., Wood, C. M., & Wright, S. M. (2020a). Landscape and habitat area data from the Glastir Monitoring and Evaluation Programme, Wales 2013-2016. NERC Environmental Information Data Centre. (Dataset). https://doi.org/10.5285/82c63533-529e-47b9-8e78-51b27028cc7f

Maskell, L. C., Astbury, S., Burden, A., Emmett, B. A., Garbutt, A., Goodwin, A., Henrys, P., Jarvis, S., Norton, L. R., Owen, A., Sharps, K., Smart, S. M., Williams, B., Wood, C. M., & Wright, S. M. (2020b). Landscape linear feature data from the Glastir Monitoring and Evaluation Programme, Wales 2013-2016. NERC Environmental Information Data Centre. (Dataset). https://doi.org/10.5285/f481c6bf-5774-4df8-8776-c4d7bf059d40

The headings should be self-explanatory after reading the paper. Units for elevation are metres, for WLFs are metres ha^-1.