cOMBINED BOTTOM-UP AND TOP-DOWN PRESSURES DRIVE A CATASTROPHIC POPULATION DECLINE OF ARCTIC SKUAS IN SCOTLAND Allan Perkins, Norman Ratcliffe, Dave Suddaby, Brian Ribbands, Claire Smith, Pete Ellis, Eric Meek and Mark Bolton Journal of Animal Ecology (2018) Contact details: allan.perkins@rspb.org.uk Data collation and processing: Data sources comprised a skua/tern monitoring programme in Orkney and Shetland throughout the 1990s (RSPB unpublished data); skua censuses in 1992 and 2010 (see paper for references); ‘Seabird 2000’ census (see paper for reference); and the UK Seabird Monitoring Programme database (SMP) (http://jncc.defra.gov.uk/smp/). Note that the latter database contains the majority of the data collated, as it also includes data from the other sources listed. Counts of Apparently Occupied Territories (AOTs) and data on annual productivity (chicks fledged per pair) were collated for all Scottish Arctic skua colonies. Colonies with at least 5 years of AOT counts during 1992-2015 were retained for analysis. For great skua, black-legged kittiwake, common guillemot, Atlantic puffin and Arctic tern, data were collated for all colonies in northern Scotland for 1992–2015. Those within 20 km of at least one selected Arctic skua colony were retained for analysing each species’ population and productivity trends, and further subset for specific analyses of localised pressures on Arctic skuas. For analysing effects of spatial variation in host and great skua densities on Arctic skua trends and productivity, we used digital mapping (MapInfo Professional v6) of Seabird 2000 and Orkney/Shetland skua census data. Each Arctic skua colony was categorised by abundance of cliff-nesting hosts (black-legged kittiwake, common guillemot and Atlantic puffin) breeding within foraging range (5 km) of its colony centroid. Thus, type 1 colonies had > 10 000 pairs of cliff-nesting hosts; type 2, 1000–10 000 pairs; type 3, < 1000 pairs. For Arctic tern, whose distribution fluctuated greatly, we instead used counts only from the same year as the Arctic skua AOT count. Great skua AOTs were estimated within the likely range of each Arctic skua colony centroid over which competitive and predatory interactions between the two species occur (1 km). Great skua surveys were often frequent, allowing use of AOT counts from the same years as Arctic skua AOT counts. For years without great skua counts, we assumed linear rates of change between successive counts to impute missing great skua AOT values. For quantifying effects of food availability on Arctic skua productivity, we estimated local host breeding success as an index. Host productivity inside 5 km Arctic skua colony buffers was often unknown, so for black-legged kittiwake and common guillemot, data from annually monitored sites up to 30 km and 110 km away were used. Small samples were pooled to give island or sub-region estimates. For Arctic tern, whose foraging range is much smaller, we took a similar approach but restricted the radius to 20 km. For each Arctic skua colony, we then calculated annual indices of cliff-nesting host productivity (mean number of black-legged kittiwakes and common guillemots fledged per pair within that island or sub-region) and Arctic tern productivity (estimated Arctic terns fledged per Arctic skua AOT within the 5 km buffer). The latter involved multiplying Arctic tern AOTs within the 5 km buffer in that year by the local productivity estimate, then dividing by Arctic skua AOTs. This finer scale approach was not possible for cliff-nesting hosts due to sparsity and infrequency of AOT counts (hence categorisation of colony types by cliff-nester abundance during Seabird 2000). Statistical analyses: All statistical analyses were conducted using SAS for Windows, version 9.4 (https://www.sas.com/en_gb/software/stat.html). GLMMs were fitted using the GLIMMIX procedure. For details, please refer to the Statistical Analysis section of the paper. Contents of this data package: 5 data files (xls), 4 SAS programme code files (txt), and 4 results files (xls). These files are numbered sequentially, following the order in which the analyses and results appear in the paper. Please see the comments boxes appended to variable names within the xls spreadsheets for a full description of their meaning. Data files: Perkins_ArcticSkua_0001_RawData_22July2018.xls This file contains 10 spreadsheets: One sheet for each of the six species included in analyses, containing raw data on population size and productivity per site per year, collated from the data sources described above. An 'Arctic skua sites' sheet giving abundance estimates for great skuas and the four host species within 1km and 5km buffers around each Arctic skua colony (obtained from the digital mapping exercise). A 'host productivity by area' sheet, summarising the overall annual breeding success of each host species recorded in each sub-area within our overall study area. A 'sites within 1km and 5km buffers' sheet, showing which great skua and host sites fall within 1km and 5km of each Arctic skua colony (derived from the digital mapping exercise). A 'sites within each area' sheet, listing which great skua and host sites fall within each of the sub-areas mentioned in point 3. Perkins_ArcticSkua_001_DataPopulationTrends_22July2018.xls This file contains 6 spreadsheets: One sheet for each of the six species included in analyses of population trends, containing cleaned data for the period 1992-2015 and including site names, codes, location and coordinates, colony type, number of AOTs in each year, and number of AOTs during Seabird 2000. These spreadsheets were imported directly into SAS for analysis (see SAS code file Perkins_ArcticSkua_005_CodePopulationTrends_22July2018.txt). Perkins_ArcticSkua_002_DataProductivityTrends_22July2018.xls This file contains 6 spreadsheets: One sheet for each of the six species included in analyses of productivity trends, containing cleaned data for the period 1992-2015 and including site names, codes, location and coordinates, colony type, and number of pairs sampled and chicks fledged in each year. These spreadsheets were imported directly into SAS for analysis (see SAS code file Perkins_ArcticSkua_006_CodeProductivityTrends_22July2018.txt). Perkins_ArcticSkua_003_DataPopulationTrendAssociations_22July2018.xls This file contains 2 spreadsheets: An 'astrendvprod' sheet, combining cleaned data on Arctic skua population size and productivity, including additional variables that show the number of chicks fledged four years previously, population size one year previously, and the number of great skua AOTs within 1km of each Arctic skua colony in each year. The 'astrend' sheet (copied from 'Perkins_ArcticSkua_001_DataPopulationTrends_22July2018.xls'). These data were used to analyse effects of Arctic skua breeding success and great skua densities on Arctic skua population trends, and the spreadsheets were imported directly into SAS for analysis (see SAS code file Perkins_ArcticSkua_007_CodePopulationTrendAssociations_22July2018.txt). Perkins_ArcticSkua_004_DataProductivityAssociations_22July2018.xls This file contains 1 spreadsheet: An 'asprodassocs' sheet, combining cleaned data on Arctic skua productivity, great skua densities, and the breeding success of hosts around each Arctic skua colony in each year. These data were used to analyse effects of host productivity and great skua densities on Arctic skua breeding success, and the spreadsheet was imported directly into SAS for analysis (see SAS code file Perkins_ArcticSkua_008_CodeProductivityAssociations_22July2018.txt). SAS programme code files: Perkins_ArcticSkua_005_CodePopulationTrends_22July2018.txt This file contains the SAS code used to run GLMMs for modelling population trends for each of the six species analysed, and relates to Figures 2a, 3, S1 and S2 in the paper. Explanatory notes for each section of code are given within the asterisks. Note that the 'data=work.names' in the first line of each section of code are the names given to the relevant spreadsheet in the corresponding xls file (_001). Perkins_ArcticSkua_006_CodeProductivityTrends_22July2018.txt This file contains the SAS code used to run GLMMs for modelling productivity trends for each of the six species analysed, and relates to Table 1, Figures 2b, 3 and S1 in the paper. Explanatory notes for each section of code are given within the asterisks. Note that the 'data=work.names' in the first line of each section of code are the names given to the relevant spreadsheet in the corresponding xls file (_002). Perkins_ArcticSkua_007_CodePopulationTrendAssociations_22July2018.txt This file contains the SAS code used to run GLMMs for modelling effects of Arctic skua breeding success and great skua densities on Arctic skua population trends, and relates to the 'Associations with Arctic skua trends' results section in the paper. Explanatory notes for each section of code are given within the asterisks. Note that the 'data=work.names' in the first line of each section of code are the names given to the relevant spreadsheet in the corresponding xls file (_003). Perkins_ArcticSkua_008_CodeProductivityAssociations_22July2018.txt This file contains the SAS code used to run GLMMs for modelling effects of host productivity and great skua densities on Arctic skua breeding success, and relates to Table 2 and Figures 5 and 6 in the paper. Explanatory notes for each section of code are given within the asterisks. Note that the 'data=work.names' in the first line of each section of code are the names given to the relevant spreadsheet in the corresponding xls file (_004). Results files: Perkins_ArcticSkua_009_ResultsPopulationTrends_22July2018.xls This file contains 10 spreadsheets: One sheet for the overall population trend for each of the six species analysed. Each of these sheets contains the model output (copied/pasted directly from SAS) from three GLMMs (overall trend - year fitted as a covariate; annual means - year fitted as a categorical variable; variation with latitude - year and latitude fitted as covariates). Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. These sheets also contain population trend graphs used in the paper (Figures 2a and 3), with calculations from the SAS outputs shown - note that some cells contain formulas which link to the relevant values of SAS output used in the calculations. One sheet for population trend by colony type for the four species for which there was sufficient data to run these models. Each of these sheets contains the model output (copied/pasted directly from SAS) from two GLMMs (overall trend - year fitted as a covariate, colonty type as a categorical variable; annual means - year and colony type fitted as categorical variables). Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. These sheets also contain population trend graphs used in the paper (Figures S1 and S2), with calculations from the SAS outputs shown - note that some cells contain formulas which link to the relevant values of SAS output used in the calculations. The corresponding data files and SAS code files for this results file are numbered _001 and _005 respectively. Perkins_ArcticSkua_010_ResultsProductivityTrends_22July2018.xls This file contains 11 spreadsheets: One sheet for the overall productivity trend for each of the six species analysed. Each of these sheets contains the model output (copied/pasted directly from SAS) from two GLMMs (annual means - year fitted as a categorical variable; variation with latitude - year and latitude fitted as covariates). Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. These sheets also contain productivity trend graphs used in the paper (Figures 2b and 3), with calculations from the SAS outputs shown - note that some cells contain formulas which link to the relevant values of SAS output used in the calculations. One sheet for productivity trend by colony type for the four species for which there was sufficient data to run these models. Each of these sheets contains the model output (copied/pasted directly from SAS) from two GLMMs (overall trend - year fitted as a covariate, colonty type as a categorical variable; annual means - year and colony type fitted as categorical variables). Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. These sheets also contain productivity trend information and graphs used in the paper (Table 1 and Figure S1), with calculations from the SAS outputs shown - note that some cells contain formulas which link to the relevant values of SAS output used in the calculations. A 'correlation matrix' sheet containing a correlation matrix of annual productivity for each of the six species (Table S2 in the paper), and the dataset used for this. The corresponding data files and SAS code files for this results file are numbered _002 and _006 respectively. Perkins_ArcticSkua_011_ResultsPopulationTrendAssociations_22July2018.xls This file contains 4 spreadsheets: A 'Model outputs' sheet containing the model output (copied/pasted directly from SAS) from four GLMMs modelling effects of Arctic skua breeding success and great skua densities on Arctic skua population trends. Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. A 'Model AICC comparisons' sheet containing the model output (copied/pasted directly from SAS) from three of these GLMMs refitted with Laplace approximation, and a null model, to compare the deviance explained, and goodness of fit from AICc values. See results file _012 ('comparing model fit' sheet) for a worked example of how this was done. An 'Intercorrelations' sheet containing a correlation matrix of the three explanatory variables used in these models, showing that the two great skua variables were inter-correlated. A 'Figure 4 data and scatterplot' sheet containing the dataset used to construct a scatterplot of overall change in Arctic skua population size and mean annual Arctic skua productivity for 20 colonies with sufficient data (Figure 4 in the paper). The corresponding data files and SAS code files for this results file are numbered _003 and _007 respectively. Perkins_ArcticSkua_012_ResultsProductivityAssociations_22July2018.xls This file contains 5 spreadsheets: An 'Intercorrelations' sheet containing a correlation matrix of the explanatory variables used in these models, showing that the two host productivity variables and the great skua and cliffnester variables were inter-correlated. A 'Host prod & great skua models' sheet containing the model output (copied/pasted directly from SAS) from four GLMMs modelling effects of host productivity and great skua densities on Arctic skua breeding success. Each set of model output includes a 'Conditional Pearson Residuals' panel for assessing model validation. These sheets also contain graphs and results used in the paper (Table 2, Figures 5 and 6), with calculations from the SAS outputs shown - note that some cells contain formulas which link to the relevant values of SAS output used in the calculations. An 'AICC values from Laplace models' sheet containing the model output (copied/pasted directly from SAS) from the four GLMMs refitted with Laplace approximation, and a year model and a null model, to compare the deviance explained and goodness of fit from AICc values. A 'Comparing model fit' sheet explaining how the model deviance and AICc values were used to compare between models (presented in Table 2 in the paper) The corresponding data files and SAS code files for this results file are numbered _004 and _008 respectively.