Test species
Osmia bicornis (red mason bee) (Megachilidae) is the most abundant solitary bee species in Central Europe (Westrich 1989). This species is univoltine and polylectic. Its lifecycle begins in spring when adult males emerge from their cocoons. Males stay close to the nest to wait for females which emerge few days later. After mating, females begin to build and provision their nests. Nests are built in a wide variety of cavities, which are either naturally occurring or artificially provided. Females collect pollen and small amounts of nectar to provision the separated brood chambers. Each of the up to 30 brood chambers per nest contains one egg (Westrich 1989). Fertilized eggs develop into females and are found in the back end of the linear nest, while unfertilized eggs are laid into the front brood chambers and develop into males. Completely filled nests are sealed with mud.
The foraging range of O. bicornis is typically less than 600 m (Gathmann and Tscharntke 2002, Zurbuchen et al. 2010). The active period of foraging and nest building occurs between April and June, and lasts from 6 to 8 weeks. Larvae hatch approximately 3 days after egg laying and feed on the stored pollen and nectar. After the metamorphosis of the larvae to the adult, the bees remain in their cocoons during winter and emerge in the following spring.
For this study, 18 000 cocoons were provided by bienenhotel.de (Rostock, Germany).
Study location and design
The exposure phase of this study was conducted at two neighbouring study sites in the vicinity of Sternberg, Mecklenburg-Vorpommern in Northern Germany. Each study site covered an area of approximately 65 km2 with a diameter of 9 km. In autumn 2013 (Fig. 1), Elado®-dressed OSR seeds were drilled at all study fields at the test site (Fig. 2), whereas Elado®-free OSR seeds were drilled at the reference site (Fig. 1). For a detailed description of seed treatment, OSR fields and planting, see Heimbach et al. (2016, this issue).
Six study locations were established at the inner cores of the test site and reference sites to prevent mason bees to forage outside the study sites. Three out of the six study locations were established at the edge of OSR fields, while the other three were situated 100 m distant from the nearest OSR field. At each study location, three nesting shelters (containing up to three nesting blocks) were established in front of a hedge or forest south-east-facing to be protected against wind and rain and exposed to direct sunlight. Each nesting shelter consisted of a plastic container fixed to two wooden stakes (see Fig. 3). The opening of the shelter was covered with chicken wire to prevent predation by birds. Nesting blocks were obtained from bienenhotel.de (Rostock, Germany). Their special design allowed a non-destructive sampling of pollen as well as the later harvest of cocoons: Nesting blocks consisted of a stack of several fibre boards each drilled with semi-circular nesting tunnels, which were closed on top by the overlying board. Each board contained 10 rows of nesting tunnels arranged in parallel (8 mm in diameter). Boards were lashed together with a strap so that they could be removed and access to the brood chambers was possible (see Fig. S1 Supplementary material).
Two perforated cardboard boxes containing 1500 cocoons (sex ratio male:female 6:5) and in total eight nesting blocks were placed within the three shelters at each location, providing a total of 1600 nesting holes per study location. In each nesting shelter, one nesting block was placed facing south-east, while the two other blocks were placed in a north-east direction. This setup was chosen because of the behaviour of female mason bees when searching for nesting sites: they would be attracted by the south-east-facing blocks as these are clearly visible and then explore the more sheltered north-east-facing blocks which usually were preferred for nesting.
The exposure phase started at the beginning of OSR full flowering (63–65 on the extended BBCH-scale) on 21 April 2014 with the placement of the cocoons (Day After Placement 0). One day after the last assessment of nest occupation at the end of OSR flowering (DAP 32, 23 May 2014, BBCH 74–79), the opening of the nesting shelters was covered with gauze to avoid further nest building activities. Subsequently, on DAP 35, nesting blocks were removed from the study locations and transferred to a sheltered place (an agricultural warehouse) to avoid predation or parasitism.
Assessments during the exposure phase
Emergence success from the initially provided cocoons was assessed by counting all empty cocoons in the cardboard boxes. This was done twice during the study in order to determine the potential nest building activities.
Nest building activity was observed after sunset, when female mason bees had returned to their nesting holes for the night. Each nesting hole was illuminated with an electric torch and the presence of bees was recorded. In order to determine the start of nest building activities, the first assessment took place on DAP 3. Since there were no females present in 50 % of all study locations (reference and test site), the procedure was repeated the following days until females were found at 75 % of all study locations. After this date (DAP 5) assessments were continued on a weekly basis. In addition to the observation whether females are present, the number of nesting holes sealed with mud was also counted per nesting block.
In order to assess foraging behaviour in nesting cavities, pollen was collected from brood cells. Pollen for the composition analysis was sampled twice at every study location during OSR flowering (DAP 12–13 and 19–20). Nesting blocks were opened for the pollen sampling. 10 subsamples were taken from each nesting block from different nesting holes, these subsamples were pooled. Pollen was only sampled from yet incomplete cells, which were unsealed and did not contain any egg. This procedure was chosen to ensure that all sampled pollen were from roughly the same time and in order not to destroy any finished brood cells, which would have had a negative influence on the outcome of reproduction. Samples were taken from all nesting blocks except for two nesting blocks at study location ORB that did not yield sufficient amounts of pollen during the first sampling event. A total of 190 samples were collected (96 samples per sampling event, 8 nesting blocks per location, and 6 locations at 2 sites, minus two samples as mentioned). Pollen samples were stored deep frozen (−18 °C) until required for microscopical evaluation.
Pollen for residue analysis was sampled likewise once per study location during OSR blossom. The pollen samples for residue analysis were shipped deep frozen on dry ice to Eurofins Agroscience Services Chem GmbH, Hamburg, Germany. Analyses of pollen samples were based on the multi-residue sample preparation technique QuEChERS. The clothianidin residue content was analysed by LC-MS/MS. The Limit of Quantification (LOQ) was 1.0 µg/kg and the Limit of Detection (LOD) 0.3 µg/kg. For further details see Rolke et al. (2016, this issue).
Temperature, humidity, rain fall and wind conditions were recorded at each location. Additionally, the relative duration of sunshine exposure was recorded. A detailed methodology for weather recordings as well as results are given in Heimbach et al. (2016, this issue).
Assessments after the exposure phase
In autumn 2014, cocoons of offspring were harvested by dismantling the nesting blocks to evaluate the reproductive output, including an assessment of undeveloped larvae, eggs and parasitism. Subsequently, cocoons were stored in a refrigerator at about 2–4 °C for a duration of approximately 5 months. In spring 2015 (3 March 2015), all cocoons were set up for emergence over a period of approximately 4 weeks. 10 days before the cocoons were removed from the refrigerator, the incubating temperature was gradually raised to 12.5 °C. Cocoons were then placed in a large, black plastic box that was fitted with an “emergence trap” (eclector) (see Fig. 4 for details). Temperature during emergence was about 20 °C.
Emerged male and female bees were counted almost every day until 30 March 2015. Prior to counting, bees were anaesthetised with ether and subsequently released. The number of cocoons that remained closed after 4 weeks were counted and sliced with razor blades to check for undeveloped bees or parasites. During this assessment several bees were found to be alive in the cocoons, but these individuals were not counted as having successfully emerged because it was not clear whether these bees would have managed to emerge from their cocoons.
Statistical evaluation
Generalized linear mixed models (GLMMs) and Generalized additive mixed models (GAMMs) provide a flexible tool to analyse non-normal and normal data when measurements are not independent due to spatial or temporal grouping. Hence, GLMMs and GAMMS were applied to study the fixed effects of treatment and environmental conditions, while the study location and the individual nesting block were included as random effects. In addition Beta Regression Models were used.
Poisson lognormal GAMMs were used to model the relationship between the number of nesting females and other covariables including the DAP, distance to OSR fields, rotation of nesting block and weather covariables. Study location C at the reference site was identified to differ in the development of nesting females from other locations, because differences in the weather and the combination of a high density and great height of OSR plants at a very short distance from the nesting shelters obstructed access for the mason bees. For the pollen composition data, a beta regression model was fitted to the relative amount of OSR. Poisson GLMMs with observational level random effects and zero-inflated negative binomial GLMMs were fitted to the count data of reproduction endpoints.
Statistical evaluation was conducted with the statistical software package “R” (version 3.0.1, R Development Core Team, Vienna, Austria, 2013). GLMMs were fitted to the data by using the packages “lme4” (Bates et al. 2015) and “nlme” (Pinheiro and Bates 2015). For multiple comparisons of parameters the package “multcomp” (Hothorn et al. 2008) was applied.
The minimum detectable difference (MDD) concept has been developed as an indicator of the power of a test a posteriori for aquatic mesocosm/microcosm studies (Brock et al. 2015). However, its calculation depends on the statistical analyses (or tests) applied to analyse the data. The calculation of the MDD for the monitoring study extended the MDD concept to suite the mixed model approach. Augmented predicted confidence intervals were used as the basis for the derivation of the MDD and MDD%.