Summary of analysis of trends in uk seabird abundance and breeding success 1969-2007




Дата канвертавання27.04.2016
Памер61.63 Kb.
Summary of analysis of trends in UK seabird abundance and breeding success 1969-2007 (unpublished report)

Matt Parsons and Ian Mitchell (JNCC), Adam Butler (Biomathematics and Statistics Scotland), Norman Ratcliffe (RSPB, now BAS) and Morten Frederiksen (CEH, now NERI, Aarhus University)



  1. Introduction

This report presents a summary of analyses of trends in abundance (undertaken by Adam Butler of BioSS) and breeding success (undertaken by Morten Frederiksen, Norman Ratcliffe and Matt Parsons) of seabirds in the UK over the period 1969-2007. As well as presenting updates on previous reports’ findings (i.e. Mavor et al. 2008) we use sophisticated and up-to-date analytical techniques not previously applied to these data (see Methods). Full results of the analyses, including regional assessments and recent updates, will be presented on the JNCC website jncc.gov.uk/seabirds.

2. Methods



    1. Abundance

2.1.1 Spatial sampling

The data comprised counts from two distinct sources within the Seabird Monitoring Programme: whole colony counts and sample plot counts. Whole colony counts are generated for all species by a complete survey of a colony. Counts of representative sub-sections of the colony – ‘plots’ - are for some species/colonies conducted instead of (and sometimes in addition to) whole colony counts. Plot counts were available for seven species – black-legged kittiwake, northern fulmar (Fulmarus glacialis), European shag, Arctic skua (Stercorarius parasiticus), great skua (Stercorarius skua), razorbill and common guillemot.

For each species, indices were calculated that expressed abundance in a given year relative to the abundance in the baseline year, which in this case was 1986 – the first year in the data set for most of the species. The amount of uncertainty in the estimation of each index was represented by 95% confidence intervals. This uncertainty originated primarily from the fact that not all colonies had been surveyed in all years.

2.1.2 Statistical methods

The data constituted whole colony and/or plot counts taken at each colony in each year from 1986 (or 1969, for tern species) to 2007. These counts were sometimes available, and sometimes missing.

Two methods of analysis were used: i) a form of ‘imputation’ (Thomas, 1993), that makes use of counts for years and colonies in which data are available to draw inferences about the values of missing counts; and ii) a specific form of Bayesian hierarchical modeling, as proposed by Parsons et al. (2006), that assumes each colony has a distinct smooth but non-linear trend, and which provides a natural framework for incorporating whole colony count and plot count data into a single analysis. Details of these methods can be found in ICES (2008).

Annual indices of abundance were estimated for each species (for which data was available) at two geographical scales: UK and each regional sea. The estimated indices for each species at each scale were compared against the associated indices derived from the two complete censuses undertaken in 1985-88 and 1998-2002 (see 3.3). While the estimated indices were derived from annual counts from samples of colonies, the complete censuses provided a direct estimate of the total national (and regional) populations for all species. Thus, the deviation of the estimated indices from the census results provides some indication of how well the different methods of analysis estimate the actual changes in abundance of each species between 1985-88 and 1998-2002 at both national and regional scales.

2.1.3 Selection of species

The estimated indices derived using the imputation method generally performed well, in the sense that the estimated indices were relatively similar to the census-based indices: more precisely, the census-based indices were typically well within the 95% confidence intervals associated with the imputed indices.

The Bayesian hierarchical modelling approach had mixed performance, with severe lack of convergence being encountered for certain species (e.g. European shag, herring gull (Larus argentatus) and great black-backed gull (Larus marinus). Since the method explicitly assumes that plot count data are synchronous with corresponding trends at the whole-colony level, the corresponding 95% credible intervals are typically narrower than the 95% confidence intervals associated with the estimates obtained using imputation, and the method therefore yielded more precise estimates of trend.

UK indices of abundance were estimated for the 16 species shown in Table 3.1. The Bayesian hierarchical modelling approach was adopted for Arctic skua and razorbill, with the imputation method adopted for the remaining 14 species. For all of these species the estimated indices were in close or fairly close agreement with those obtained from the census, and the 95% confidence/credible intervals were not excessively wide.

Four species were omitted: Atlantic puffin,, black guillemot, common gull (Larus canus), and great black-backed gull. The estimated indices for these species had extremely wide confidence intervals and/or there were large differences between the imputed indices and those derived from the censuses.



Table 3.1: Seabird species selected for UK trend analysis of abundance data

Species

Method

Comparison of estimated indices & census

Width of 95%confidence /credible interval

Arctic skua

Bayesian

Quite close

Moderate

Arctic tern

Imputation

Close

Narrow in later years

Black-headed gull

Imputation

Close

Moderate

Common guillemot

Imputation

Close

Moderate

Common tern

Imputation

Fairly close

Narrow

Northern fulmar

Imputation

Close

Moderate

Great cormorant

Imputation

Close

Narrow

Great skua

Imputation

Fairly close

Moderate

Herring gull

Imputation

Fairly close

Narrow

Black-legged kittiwake

Imputation

Close

Narrow

Lesser black-backed gull

Imputation

Close

Moderate

Razorbill

Bayesian

Moderate

Wide in later years

Sandwich tern

Imputation

Close

Very narrow

European shag

Imputation

Quite close

Narrow

Roseate tern

Imputation

Close

Narrow, especially in later years

Little tern

Imputation

Close

Narrow

2.1.4 UK Multi-species index

A multi-species index of abundance was produced by taking the geometric mean of the annual indices of abundance for the 16 species shown in Table 3.1.


    1. Analysis of trends in breeding success of seabirds

2.2.1 Data source and statistical analysis

Data describing the mean number of chicks fledged per breeding pair (breeding success) and sample sizes (number of breeding pairs that were investigated per colony) were extracted from the SMP database for all seabird species breeding in the UK between 1989 and 2007. Where sample sizes were missing for a particular colony-year, the average sample size recorded in other years was calculated and used for further analysis.

Data were analysed at the UK and regional scale (using seven ‘Regional Seas’ areas), though only trends at a UK scale are shown here. The number of chicks fledged in each colony-year was calculated as the product of the sample size and mean success, rounded to the nearest integer. This variable was used as the response variable in generalized linear mixed models (GLMM), using the Genstat statistical software package. Colony was included as a random effect in all models to allow for the fact that repeated measures of productivity from individual sites are not independent. Two classes of models were employed, and applied to species depending on their maximum clutch size. For those species that lay a single egg (northern fulmar, Manx shearwater, northern gannet, common guillemot, razorbill, Atlantic puffin) the sample size was declared as a binomial denominator and the modelling proceeded with a binomial error distribution and logit link, which constrains the predicted values to lie between zero and one. For species that lay more than one egg (great cormorant, European shag, both skua spp., all tern and gull spp., and black guillemot) the sample size was declared as an offset and modelling proceeded with a Poisson error distribution and log link, which constrains predicted values to be greater than zero. In both types of model, overdispersion was corrected by scaling the deviance by the residual Pearson Chi-square divided by the residual degrees of freedom.

Five models were tested where data allowed: a full interactive model of time and region effects, additive effects of time and region, time only, region only and constant productivity. Note that only year-dependent UK trends (and constant productivity) are shown in this report. Model fit was tested using F-ratio statistics and a backward elimination approach was used to arrive at the minimum adequate model. The parameter estimates were extracted from the minimum adequate model and back transformed to produce estimates of breeding success.

Seabird abundance varies greatly among the regional groupings and so, for species where success was region-dependent post hoc weightings were used to produce national indices. Abundance within the required regional groupings was extracted from the most recent census of breeding seabirds, ‘Seabird 2000’ (Mitchell et al., 2004). Success for each year was then estimated using analysis of variance, with success for each region-year being weighted by the number of pairs in the region. The least-square means from the analysis were then used to calculate the national year-dependent trends.

2.2.2 Selection of species and calculation of multi-species indices

Sufficient data were available to model productivity of 22 species: Manx shearwater, Arctic skua, Arctic tern, Atlantic puffin, black guillemot, black-headed gull, common guillemot, common tern, northern fulmar, northern gannet, great black-backed gull, great cormorant, great skua, herring gull, black-legged kittiwake, lesser black-backed gull, mew gull, razorbill, Sandwich tern, European shag, little tern and roseate tern. Very few data existed for Leach’s storm petrel Oceanodroma leucorrhoa, European storm petrel Hydrobates pelagicus and Mediterranean gull Larus melanocephalus, which were excluded from the analyses. Significant year-dependence at the UK scale was detected for all species, except Manx shearwater, black guillemot, great cormorant and northern gannet, for which a model of constant success was the most parsimonious description of the data.

To express relative annual variation of species’ breeding success within the UK, standardized indices for each species were calculated by dividing the breeding success value in each year by the value of the earliest year in which data were available for all species: 1989 for UK. Multi-species indices of breeding success for UK were produced by taking the geometric mean of the annual indices of abundance for all modelled species.

3. Results

3.1 Breeding abundance

The abundance of the 16 seabird species included in the UK multi-species index has declined overall by 8.3% between 1986 and 2007 (Figure 1). Figure 2 shows the percentage change in abundance of individual species between the late 1960s (terns only; other species since 1986) and 2007. The abundance index of eight species (northern fulmar, European shag, black-legged kittiwake, Roseate tern, Arctic skua (Stercorarius parasiticus), little tern (Sternula albifrons), common tern (Sterna hirundo) and herring gull) declined overall by 10% or more between 1986 and 2007; five (great cormorant (Phalacrocorax carbo), lesser black-backed gull (Larus fuscus), great black-backed gull, Sandwich tern, and Arctic tern (Sterna paradisaea)) were relatively stable overall, having experienced changes of less than 10% over the same period; four species (great skua, black-headed gull (Larus ridibundus), common guillemot and razorbill) showed overall increases since the mid 1980s.

Roseate tern declined the most (by 91% between 1969 and 2007 and by 73% between 1986 and 2007), though the population since 1991 has been low and relatively stable, greatest declines having occurred during the 1970s and 1980s. Herring gull numbers declined by 29% 1986-2007, despite a substantial rise during the 1980s and 1990s in the number nesting in urban sites (Mitchell et al. 2004). Arctic skua has shown severe recent declines- by 35% between 1986-2007. Similarly, northern fulmar has suffered a recent decline, by 25% between 1986-2007.

Notable increases include that of great skua (Stercorarius skua), whose index increased by 71% between 1986-2007, though there are signs since 2003 that numbers are decreasing. The index for common guillemot showed an increase of 30% between 1986-2007, though since around 2003 has showed signs of levelling off or decline. Recent increases have occured in the population index of razorbill – by 80% between 1986-2007, though the confidence intervals here are wide in recent years.

3.2 Breeding success

The mean breeding success of the 22seabird species included in the UK multi-species index has declined from 2000, reaching a low in 2005, with a temporary increase in 2006 then falling in 2007 (Figure 3).

Breeding success in most species showed greater annual variation than did abundance (Figure 4). Breeding success more immediately reflects feeding conditions at sea and the impacts of stoachastic events such as storms. Persistent trends in productivity within a species over a number of years would not generally be expected to occur given the variety of factors that affect productivity and their individual variability. However, downward trends in annual breeding success were apparent at a UK scale for some species: northern fulmar, Atlantic puffin and, particularly, in black-legged kittiwake, common guillemot and razorbill. Mean breeding success of black-legged kittiwake nearly halved over the period 1989-2007. Common guillemot success was until recently fairly stable at around 0.75 chicks per pair. However, from 2003onwards productivity has fallen markedly, with a UK mean of just 0.2 chicks per pair raised in 2007.

4. References

ICES (2008) , , . Matt Parsons. 12 pp.

Mavor, R.A, Heubeck, M., Schmitt, S. And Parsons, M. 2008. Seabird numbers and breeding success in Britain and Ireland, 2006. Peterborough, JNCC. (UK Nature Conservation, No. 31).

Mitchell, P.I., Newton, S.F., Ratcliffe, N. & Dunn, T.E. 2004. Seabird Populations of Britain and Ireland. T. & A.D. Poyser, London.

Parsons, M., Mitchell, P.I., Butler, A., Mavor, R.A., Ratcliffe, N. & Foster, S. (2006) Natural heritage trends: abundance of breeding seabirds in Scotland. Scottish Natural Heritage Commissioned Report No. 222 (ROAME No. FO5NB01).

Thomas, G.E. (1993) Estimating annual total heron population counts. Appl. Statistics, 42, 473-486.

5. Acknowledgements

The Seabird Monitoring Programme relies on the generous help of many surveyors – often volunteers - to supply information. The following organisations work on the SMP in partnership with JNCC: Scottish Natural Heritage, Countryside Council for Wales, Natural England, Northern Ireland Environment Agency, Royal Society for the Protection of Birds, Centre for Ecology and Hydrology, The Seabird Group, Shetland Oil Terminal Environmental Advisory Group, BirdWatch Ireland, British Trust for Ornithology, National Trust for Scotland, National Trust, National Parks and Wildlife Service (Dept. of Environment, Heritage and Local Government – Republic of Ireland).

Figure 1. UK multi-species index of abundance of 16 species breeding seabird species, 1986-2007.

Figure 2. UK annual variation in abundance of seabird species between 1969-2007 (terns) or between 1986-2007 (other species). Ther middle line indicates the modelled index (Thomas method); the upper and lower lines represent the 95% confidence limits.
























Figure 2 (cont.)





















Figure 2 (cont.)











Figure 3. UK multi-species index of breeding success of 22 species breeding seabird species, 1989-2007.

Figure 4. Annual UK variation in breeding success for seabird species 1989-2007. * denotes significance of fixed effect of year <0.05; *** <0.001. Straight lines indicate results of model of constant productivity (where fixed effect of year was not significant in the GLMM).




















Figure 4 (cont.)


















Figure 4 (cont.)















База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка