Data analyses

The materials collected from 1 June to 31 August were referred to the eagles’ breeding season and those from 1 November to 1 March to their autumn-winter diet. The body mass of the specimens of the various species was determined by Petrov (1964), Simeonov and Petrov (1980), Boev (1986), Simeonov et al. (1990), Böhme (1993), Kunstmüller (2000), Dunning (2008) and Aulagner et al. (2009). An average body mass was given, calculated on the basis of the average mass of individual specimens. When the material was identified up to genus level, the average values for the presented species of the genus were given. The carrion biomass was not taken into account.

In order to identify the diet differences between regions and seasons, the prey items were grouped into the following main categories, based on their specific ecological requirements: Lizards & Snakes (Squamata), Tortoises (Testudines), Water birds (Anatidae, Ardeidae), Poultry (Gallusgallusf.domestica, Anseranserf.domestica, Meleagrisgallopavof.domestica, Pavocristatusf.domestica), Phasianids (Phasianidae), Gulls (Laridae), Doves (Columbidae, Feral Pigeon), Song birds (Non-Corvidae Passerines), Corvids (Corvidae), Stork (Ciconiaciconia), Raptors & Owls (Accipitridae, Falconidae, Strigidae, Tytonidae), Hedgehog, Hare, Souslik, Rodents (Rodentia excl. European Souslik), Carnivores (Carnivora), Carrion (Artiodactyla, Perissodactyla) and Other Animals.

To understand differences in diets amongst regions, we used Generalised Linear Mixed Models (GLMM) with Poisson distribution and Log link function. Our response variable was region and our predictors were food categories that showed high Likelihood Score (p ≤ 0.07) in the likelihood estimation (Table ​(Table2).2). We ran our “Global model” in two scales: first with prey items and second with biomass adjusted data as a predictor variable. We merged territories from SG (n = 2) and ER (n = 1) into a group of high mountain regions (HM) due to the small sample sizes and the similar habitat conditions (Demerdzhiev 2011). “Eagle territory” was included in the models as a random effect. To determine which diet factors would affect the region differences, we used Akaike Information Criterion, corrected for small sample sizes (AICc), for model selection and chose the models with the lowest AIC_c value from the set of our candidate models. All models with an AIC_c value < 2 from the model with the lowest AIC_c (AIC_cmin) were considered best models (∆AIC_c = AIC_i– AIC_cmin) (Burnham and Anderson 2002).

Likelihood estimation of different food categories was used to describe the regional differences in the EIE diet. Categories included in GLMM’s are given in bold.

The relative importance of each model was estimated through the weight of AICc (w), so that all the weights for all models added up to 1. We also used explanatory parameter estimates with Lower (95%) and Upper CL (95%) and a probability value (p) of the explanatory factors.

To find out the diet differences between seasons, we used the non-parametric Mann-Whitney U Test with continuity correction.

All data were analysed using Statistica for Windows, Release 12 (StatSoft Inc 2013), R v.2.15.2 (R Core Team 2012) and Past Version 3.14 (Hammer et al. 2001). Results with p ≤ 0.05 were considered significant. Values were provided as means ± SE.