Impact of Great skuas on terrestrial ecosystem functioning and biodiversity in Shetland

We simulate acute, dramatic changes in seabird density by conducting in situ controlled manipulation experiments where turfs were reciprocally transplanted between areas of high and low skua density. We make inference about chronic, long-term changes or differences in seabird density through replicated stratified surveys using areas of contrasting great skua density. We measured a suite of soil, plant and invertebrate parameters and applied structural equation modelling (SEM) to quantify the strengths of each hypothesised mechanistic pathway shaping the impacts of changing seabird density on ecosystem functioning and biodiversity. We quantified the amount of marine nutrients that seabirds transport to a terrestrial ecosystem, and traced these through soil and plants using stable isotope analysis to confirm that soil and plant nutrients are marine-derived and strongly correlated with seabird density.
Data were collected over three great skua breeding seasons at three large great skua (Stercorarius skua) colonies in Shetland, UK, comprising a mix of blanket bog and modified bog: Lamb Hoga peninsula, Fetlar (Fetlar: centred at 60° 34' 39.77”N, 0° 53' 33.63”W), Hermaness National Nature Reserve, Unst (Hermaness: centred at 60° 49' 0.15”N, 0° 53' 32.40”W) and Isle of Noss National Nature Reserve (Noss: centred at 60° 8' 22.94”N, 1° 1' 2.40”W).
Great skua abundance across each colony is not homogeneous, but comprises a loose collection of breeding territories, each one containing the nest and at least one raised mound on which the birds stand to observe their territory, which we refer to as “high density – mounds”. In addition, there are one or more discrete areas called club sites where adult and immature non-breeding birds form dense aggregations of between 10 and 200 individuals; we refer to these as “high density – club”. Most of the land area of a colony away from nests, mounds and club sites is used infrequently by great skuas and is referred to here as the “low density - colony”. In addition, we identified and sampled from areas that were spatially distinct from the colonies, supporting no great skua breeding colony, but which shared similar land-use, soils, geology and distance and orientation to the sea. These areas are referred to as “low density – outside colony”. A total of 12 sampling sites were identified for each treatment type. Mound sampling sites were selected at random from all breeding territories identified within approximately a 1km radius of the club sites, and club sampling locations were established in a grid formation separated by at least 5m. Low density - colony sampling sites were positioned at approximately 30 m from each mound sampling site on a random bearing, ensuring there were no other nearby mound sites. Low density – outside colony sampling sites were established in a grid formation separated by at least 15m.
We undertook two broad experimental approaches, both of which were reviewed by the Animal Welfare & Ethical Review Body at University of Stirling. They comprised stratified surveys that were extensive sampled across great skua colonies and areas where great skuas were absent to quantify i) nutrient inputs from regurgitated pellets and faeces, ii) soil nutrient status, iii) plant nutrient status, iv) natural abundance stable isotope ratios of faeces and pellets, soils, plant leaves and litter, v) plant community composition, vi) above-ground invertebrate community composition, and vii) rates of litter decomposition. The stratified surveys used high density club and mound sites and low density colony and outside colony (control) sites previously identified.
A second complementary approach used a manipulation study that comprised a reciprocal transplant experiment of intact turfs from areas with high and low great skua density, to simulate the effects of either increasing or decreasing great skua populations on plant community composition and nutrient concentrations. The experiment was established at a high density club site and an adjacent low density - colony area at Hermaness, in a separate area of the colony used for the stratified surveys. In March 2016, pairs of turfs in a block design (n =12; block locations) were cut from club and intra-colony areas and placed into freely draining plastic boxes (35cm x 53cm x 15cm deep). One of the pair was reciprocally transplanted with the corresponding block in the club or intra-colony area, while the other was placed back into the hole from where it was cut, establishing four different treatments: 1) Club to club area (high to high skua density; control); 2) Club to low density colony area (high to low skua density treatment); 3) Low density colony to club area (low to high skua density treatment); 4) Low density colony to low density colony area (low to low skua density; control).

Faecal deposition was quantified in 1m2 quadrats and pellets quantified in 4m2 quadrats in June 2016. The quadrats were cleared completely of faeces and pellets and after 15 days, faeces and pellets within the quadrats were counted. Where faeces overlapped, a count was approximated based on size, shape and orientation of the outline. The chemical composition of faeces and pellets were determined in 12 samples of each collected at random from within the quadrats. All additional material, such as grass and soil, were removed from the samples by hand, before each sample was air dried, weighed, and ball-milled. Nitrogen concentrations were quantified using an automated elemental analyser (FlashSmart, Thermo Scientific). Phosphorus and calcium concentrations were determined using a nitric acid microwave digest (MARS 6, CEM) and subsequent analysis using ICP-OES (iCAP 6000, Thermo Scientific).

In July 2017 and August 2018, plant samples were collected from half of the locations used in the stratified surveys (n = 6; sample locations selected randomly), and from all turfs in the manipulation study where the species occurred. Plant samples comprised fresh, above ground vegetative tissue (i.e. not flowering stem) of four functional groups: a sedge: Eriophorum angustifolium, a grass: Anthoxanthum odoratum, a forb: Potentilla erecta, and a shrub: Calluna vulgaris. These species were selected because they are ubiquitous across the study area and have contrasting life histories. The plant samples were air dried and ball-milled before chemical analysis.

Stable isotope analysis was used to determine if the nitrogen in the soil and plant tissue was marine derived. A sub-section of the faecal and pellet samples, plant leaves and soil samples (n = 6) were analysed for the ratio of 14N and 15N using continuous-flow isotope ratio mass spectrometry (CF-IRMS, Costech ECS 4010 elemental analyser and Thermo Finnigan Delta Plus XP mass spectrometer). Results are expressed in δ15N (‰) deviations from the international standards.
Senescing graminoid plant material was collected from a club site and an adjacent area of low density colony at the North Hermaness colony. Litter was air dried, cut to approximately 2cm strips and mixed to produce two bulk samples of litter from high and low density areas. A total of 24nylon mesh litter bags, measuring 4.5cm x 4.5cm, with a 1mm aperture, were filled with0.5g of either club (12 litter bags) or control litter (12 litter bags). Litter bags were buried vertically in the top 5cm of soil in a block formationwith club litter bags buried at high density clubsites and low density colony litter buried in low-density colonyareas (n =12) and incubated 24 months. Once harvested, litter was air dried and roots and soil were removed by hand. The litter was weighed to determine mass loss prior to undertaking chemical analysis, as described previously.
The pin drop method was used to quantify the plant species abundance and community composition in both the stratified surveys and manipulation experiment. In the stratified surveys, 60cm2 quadrats were randomly placed within sampling locations during July 2017. A total of 16 pins were positioned evenly across the quadrat (15cm spacing). Each vascular plant touching the pin was identified to species level (where possible) and the number of times each species touched the pin recorded. Bryophytes, liverworts and lichens were also recorded, but not identified so are not included in the analysis. The pin touches across the quadrat were summed and used as an objective measure of relative abundance. The same method was used for the turf transplants; however, 16 pins were placed evenly in a 20cm quadrat (5cm spacing).
Invertebrate sampling was carried out using pit-fall trapping. Traps comprised plastic cups (127mm depth x 96mm diameter) buried flush with the soil surface and partially filled a 50% propylene glycol antifreeze solution (diluted with water). In a preliminary study, flooding of traps and interference by great skuas was an issue. To overcome this, two traps were set at least 1m apart within each sampling location. One trap from each site was selected randomly (unless it was the only one still intact) for sample processing. Traps were protected by a clear lid (140mm diameter) held 15mm from the ground by two U shaped pegs. Traps were established in June 2017 and were left open for a period of 4 weeks and emptied at two-week intervals. The trap contents were stored in 70% ethanol, before all individuals were identified to order, all Coleoptera were identified to family and all carabids to species.