Assessing the restoration of plant-pollinator interactions in an upland system: a network approach

Transcription

1 Assessing the restoration of plant-pollinator interactions in an upland system: a network approach Rationale All species on earth are connected to at least one other species either directly or indirectly and these interactions can be visualised as ecological networks. Ecological networks are used to identify the interactions (links) between species (nodes) within and between trophic groups and describe the structure and function of complex systems. There are two types of ecological network (i) qualitative, i.e. a simple network that uses binary links to describe which species are interacting with which (Fig.1) and (ii) quantitative, i.e. uses observed data to describe which species are interacting with each other and in what frequencies (Fig.2) (Dunne, 2006). In this study I will be focussing on a quantitative, mutualist (plant-pollinator) network where I will create flower visitor and pollen transfer networks to investigate the success of moorland restoration. Fig. 1. A qualitative ecological network for one plant species. Bottom layer = Broom, Second = Phytophagus insects that feed on broom, third = insects that feeds on or parasitize the phytophagus insects in level 2 and third = the insects that feed and parasitize on the insects in level 3. Spheres represent species within the network and the lines represent the links between species. 1

2 Fig.2. A quantitative ecological network showing flower visitors (on top) and pollen transfer (below). The width of the bars on the top and bottom of each network shows the species abundance whereas the width of the links shows the frequency of the interactions. From Forup et al (2008). UK moorland is of international importance due to its ecosystem services such as carbon sequestration, recreational use, habitat for a wide range of biodiversity including rare and threatened species and the ecosystem service that this study focuses on, pollination. Ecological restoration can be small or large-scale, nevertheless, there needs to be the proper monitoring in place to assess its success and identify any key changes that should be implemented (Hobbs and Norton, 1996; Lewis et al, 1996; Pullin, 2002). Ecosystems have many elements but can be broken down into two main features; i) structure i.e. species richness and abundance and ii) function i.e. pollination, pest 2

3 control (Bradshaw, 1997). In a damaged system you would expect that both these attributes have been negatively affected and so both need consideration when deciding the best restoration and monitoring techniques (Bradshaw, 1997). Focussing on the interactions between species, in order to determine ecosystem function, could be a superior method for assessing the success of restoration of complex ecosystems; simply analysing target communities may not be suitable as habitat modification may have detrimental effects on species interactions without significantly affecting species richness (Tylianakis, 2007). Key Questions The aim of this project is to assess the restoration of ecological interactions on restored moorland whilst answering the following questions: 1) Can plant-pollinator interactions be reinstated in restored moorland? 2) Can the function of an ecosystem be restored as well as species richness? 3) Have parasites been reinstated as the top trophic group and is this a sign of ecosystem health? 4) How long does it take for plant-pollinator and host-parasite interactions to be fully restored? Materials and methods Study sites Plant-pollinator surveys were conducted on the Bleaklow/Peaknaze area of the Peak District National Park, UK. This is an upland moorland area (c.500 c.600m elevation) that comprises mainly of blanket bog with areas of wet and dry heath and dwarf shrub plants such as ling heather (Calluna vulgaris) and bilberry (Vaccinium myrtillus), along with grasses and sphagnum mosses. Sample sites were selected according to age and the restoration techniques used. Only sites applied with all three applications (lime/seed/fertiliser, heather brash and geo-textiles) were considered to avoid any biases in the results relating to restoration 3

4 management. As the most recently restored sites ( ) had not yet received geo-textile treatments these were removed from the potential sample sites and only areas restored between 2003 and 2005 were included in the final selection. An intact site which was not severely affected by the fire in 2003 was located as a reference of potential ecological status before the moorland was damaged. Final treatments included the control/bare peat site, intact site and sites restored in 2003, 2004 and 2005 (Table 1). Six replica plots (40x40m) per treatment (n=30) were measured using a Garmin etrex Geographic Positioning System (GPS) device and marked with bamboo garden canes. In treatments where there were two sites restored in the same year the number of replicate plots were split between both sites. The correct permissions were sought from United Utilities and Natural England for working on the moorland away from paths during bird breeding season, setting up 30 sample plots including the use of bamboo canes and collecting pollinating insects and small plant samples. Table 1. OS grid coordinates for Bleaklow sample sites. A bare peat, B 2005, C 2004, D 2003, E intact. Treatment Replica OS Grid Coordinates (BNG) A 1 SK A 2 SK A 3 SK A 4 SK A 5 SK A 6 SK B 1 SK B 2 SK B 3 SK B 4 SK B 5 SK B 6 SK C 1 SK C 2 SK C 3 SK C 4 SK C 5 SK C 6 SK D 1 SK D 2 SK D 3 SK D 4 SK D 5 SK D 6 SK

5 E 1 SK E 2 SK E 3 SK E 4 SK E 5 SK E 6 SK Flower visitor surveys Each sample plot was visited once per month between mid May and late September During each visit, temperature, wind speed and weather conditions were recorded and the plot was walked in a systematic pattern for one hour. All insects seen interacting with a flower within this hour were collected either using a sweep net or directly into the killing tubes and the species of plant recorded. An insect was classed as interacting with a flower if it was seen to be actively collecting pollen or nectar, any insect remaining stationary on a flower or any other area of the plant was said to be resting thus was not collected. Any flowering plant that was not able to be identified in the field was collected in a sample bag for later identification. Killing tubes contained a numbered paper container (body bag) in contact with a paper tissue disk soaked with 8-10 drops of Ethyl Acetate to quickly euthanize the insect. Any insect caught in a sweep net was transferred to a killing tube as quickly as possible to reduce stress and decrease the chance of any pollen being deposited onto the net. Data was collected between 10am and 4pm on dry, calm days (5 or below on the Beaufort scale) to coincide with sampling protocols used in other flower visitor projects. In the event of a light rain shower, the survey was paused until 20 minutes after the shower had stopped and then data collection was resumed. If, however, the shower became persistent or turned into heavy rain the survey was discontinued and repeated in better weather conditions. Any specimens collected during these shorter surveys were discarded and not included in the final analysis. All plots were visited during the morning, mid-day and afternoon to avoid any effect of daytime on potential pollinator activity. All samples collected were frozen at the end of each day to minimize decomposition before analysis. 5

6 Specimens were kept cool with ice packs in a cool bag whilst being transported between the field site and the laboratory and placed in a -20 C freezer on arrival. All plant-pollinator interactions were collected by the same researcher to avoid sampling bias. Vegetation transects Vegetation transects were conducted at every sample site once a month from May to September in 2011, although only transects on 2 replica plots per treatment were conducted in May because of time constraints. Two parallel 20m transect lines were laid out 10m in from the top and side of the sample site, thus 20m apart from each other. A 1m quadrat was placed every 2m along both transect lines (n = 20) (Fig. 3) and vegetation species richness, percentage cover, height and floral abundance were recorded. Floral abundance was calculated by counting floral units per species, one flower head counted as one floral unit. Every floral unit was counted in the quadrats with the use of a clicker counter until August when Calluna vulgaris flowered and made counting every single floral unit unfeasible, after this point, floral units were only counted in ¼ of each quadrat and multiplied to gain an estimate of floral abundance. Floral abundance from the 20 quadrats was used to calculate the estimated average floral unit per m 2 for the whole sample site each month. 6 Fig. 3. Bamboo canes marked out to 1m were used to create a 1m 2 quadrat which was placed every 2m on the 20m transect line to record vegetation diversity, percentage cover, height and floral abundance.

7 A sample specimen of each new flowering species was collected and the anther swabbed to form a pollen reference collection for later identification of pollen grains collected from insect visitors. Any flowering species not able to be identified in the field was collected for later identification. This data not only gives information on the potential pollen and nectar recourses available to pollinators at each sample site but also allows for an observation of the success of vegetation restoration. Laboratory protocol All insects were thawed before pollen collection. The head, legs and abdomen of flower visitors collected from the study sites were swabbed using a small cube of Fuchsin-glycerin jelly as described by Beattie (1971) (Fig. 4). The gel pulls pollen grains off of the insect and is mounted on a glass slide with a cover slip. It is then melted over a gentle heat source where it stains the pollen grains and allows for identification under a microscope. Any pollen sacs found on bees were discarded as this resource would most likely not be available to plants for pollination but instead deposited back in the nest (Forup et al, 2008). Pollen grains were identified to species using the pollen reference collection gathered from the plants found in the vegetation transects on each site. Any pollen species that could be identified by this collection were done so using a large pollen reference collection already existing within the University. Insects were identified to species using a high-power microscope (Olympus SZX10 Research High-class Stereo Microscope), any insects that could not be identified were morphotyped and sent to specialist taxonomists for identification. The workers of two-banded white tail bumble bee species Bombus terrestris and B. lucorum can be difficult to distinguish between each other, because of this they may be grouped together as B. terrestris/lucorum similar to Forup et al (2008). After identification and pollen removal, all Bombus bees will be examined for external parasites (any parasites found during identification or swabbing were recorded and 7

8 kept with the bee hosts). The abdomen will be separated from the thorax and searched under a microscope for any further parasites or their larvae. The abdominal cavity will also opened and examined for any parasitoids following protocols used by Henson et al (2009). Fig. 4. Bombus lacorum specimen being swabbed for pollen with a dissecting tool holding a small cube of fucshin gel. In order to compare species richness between sites we need to make sure that the vegetation data is standardised and has not been affected by sampling effort. As the number of quadrats sampled increases, so too should the number of species recorded. This forms a sample-based accumulation curve. It is expected that the curve ascends rapidly within the first few quadrats as the most abundant species are recorded and slows down as only the rarer species are added in later quadrats until the curve plateaus off to an asymptote (Gotelli and Colwell, 2001), if the curve does not reach an asymptote by the last quadrat sampled it is not certain that full number of species in that sample plot has been recorded. Sample-based rarefaction on all of the vegetation data collected from the different sites will produce a smooth rarefaction curve that will allow direct comparisons between the data sets. EstimateS (version 7.5) will be used to compute the sample-based rarefaction with a 95% confidence limit. This dataset, however, is from a relatively species-poor habitat and it is expected that an asymptote will be reached well before the last quadrat sampled at each site. 8

9 For comparisons between bare peat, intact and 3 restoration sites two different quantitative networks will be produced; a flower visitation network and a pollen transfer network (Fig. 2). Flower visitation networks simply describe the choice of plant species made by potential pollinators. This type of network is good for identifying species diversity and how often insects encounter a pollen or nectar resource but the plant-insect interactions it describes may or may not lead to pollination and the efficiency of assumed pollinators is not taken into account (Memmott, 1999). The pollen transfer network, on the other hand, identifies which insect species is carrying which pollen species and quantifies the pollen transported (Memmott, 1999; Forup et al, 2008). These networks will be created in the bipartite (pollen transfer) and tripartite (flower visitor with parasitoids) packages in the statistical computer analysis software R (version ). Network descriptors will be calculated from the flower visitation and pollen transfer networks. These will include simple descriptors of network appearance such as Linkage Density which is the number of actual links in the network (L)/the number of species (S) and connectance which is the actual number of direct links in the network divided by the potential number of direct links: C = L/S 2 (Warren, 1994) to more informative descriptors such as presence and number of compartments within the network, the weighted nestedness which takes interaction frequencies into consideration and is thought to be more consistent than normal nestedness calculations, any specialisations present, extinction slopes for both higher and lower trophic levels which calculate the number of secondary extinctions in one trophic level following the deletion of species from the other, the robustness which is the area below the calculated extinction slope, genrality and vulnerability which is the average number of prey per predator and predator per prey respectively and mean interaction diversity of prey per predator and predator per prey. These network descriptors will also be created in the package bipartite in R. Both treatment and replicate will be taken into consideration in analyses. The pollinator importance (PI) will be calculated from the pollen transfer data to determine the key pollinators in this system. PI will be calculated following 9

10 protocols used by Gibson et al (2006) and Forup et al (2008) using the following equation: PI = [Relative abundance of pollinator] x [Pollen fidelity] where [Relative abundance of pollinator] is the proportion of all insects carrying a given pollen species that belong to the pollinator species and [Pollen fidelity] is the average proportion of individual pollen loads on the pollinator species that originate from the given plant species. Findings so far Analysis on floral abundance data showed that floral abundance has not been reinstated in restored sites when compared to the intact reference site. The fragments restored in 2004 appear to support more floral units than those restored one year previously (Fig. 5). This is the opposite of what would be expected as the 2003 site has had one extra growth and flowering season. There could be multiple explanations for the 2004 site appearing to be more established than the other restored sites including peat moisture content and climatic conditions in 2004 when plants were first introduced, however, further research needs to be conducted in the area. We know that this difference is not the result of different restoration techniques as all sites included in the study were restored using the same techniques and to similar specifications. Analysis on vegetation height, however, showed the expected trend of vegetation cover with vegetation on the oldest of the restored sites actually being taller than in the reference site (Fig. 6). Once species composition is taken into account, however, it is expected that the tall vegetation in the 2003 restoration sites will comprise mainly of grasses, whereas, the reference site is expected to comprise mainly of dwarf shrub plants, predominantly Calluna vulgaris. Analysis on plant species richness will show more information on both the floral abundance and vegetation height analyses, identifying which species are dominating in both measures. 10

11 Restoration stage Fig. 5. Barplot of mean number of floral units as a function of restoration. Bare Peat 2005 Restoration 2004 stage 2003 Reference Fig. 5. Barplot of mean number of floral units as a function of restoration. 11

12 Work left to do Floral abundance and vegetation height must be re-analysed to incorporate species composition as a function of restoration. Also, accumulation curves and samplebased rarefactions must be conducted to ascertain if the full species diversity was captured in the vegetation transects. Visitation and pollen transfer will be created once unknown insect specimens have been identified to species by taxonomists. Experts in Diptera, Lepidoptera, Coleoptera and Hymenoptera have been hired to conduct species identifications. Use of funding The money given to this project by Moors for the future has been used for accommodation and fuel whilst in the field. The next instalment will be used to may taxonomists in order to create detailed ecological networks, comprising of accurate species interactions. References Beattie, A. J. (1971). A technique for the study of insect- borne pollen. Pan-Pacific Entomologist, 47: 82. Bradshaw, A.D. (1997). What do we mean by restoration? Pp in Restoration ecology and sustainable development. Eds. Urbanska, K.M., Webb, N.R. and Edwards, P.J. Cambridge:Cambridge University Press. Dunne, J.A. (2006). The network structure of food webs. Pp in Ecological Networks Linking structure to dynamics in food webs, (Eds. Pascual, M. and Dunne, J.A.), Oxford University Press. Forup, M.L., Henson, K.S.E., Craze, P.G. and Memmott, J. (2008). The restoration of ecological interactions: plant-pollinator networks on ancient and restored heathlands. Journal of Applied Ecology, 45: Gibson, R.H., Nelson, I.L., Hopkins, G.W., Hamlett, B.J. and Memmott. J. (2006). Pollinator webs, plant communities and the conservation of rare plants: arable weeds as a case study. Journal of Applied Ecology, 43: Gotelli, N.J. and Colwell, R.K. (2001). Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters, 4:

13 Henson, K.S.E., Craze, P.G. and Memmott, J. (2009). The restoration of parasites, parasitoids, and pathogens to heathland communities. Ecology, 90: Hobbs, R.J. and Norton, D.A. (1996). Towards a conceptual framework for restoration ecology. Restoration Ecology, 4: Lewis, C.A., Lester, N.P., Bradshaw, A.D. et al (1996). Considerations of scale in habitat conservation and restoration. Workshop on the Science and Management for Habitat Conservation and Restoration Strategies (HabCARES) in the Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences, 53: Memmott, J. (1999). The structure of a plant-pollinator food web. Ecology Letters, 2: Pullin, A.S. (2002). Ecological Restoration. Pp in Conservation Biology, Cambridge:Cambridge University Press. Tylianakis, J.M., Tscharntke, T. and Lewis, O.T. (2007). Habitat modification alters the structure of tropical host parasitoid food webs. Nature, 445: Warren, P.H. (1994). Making connections in food webs. Trends in Ecology and Evoloution, 9: