Page 2 - Eagles_Circaetus_gallicus_2016

544 N. Agostini et al.

the beginning (autumn) of migration to avoid the
crossing of large bodies of water (Alerstam 2001;
Agostini et al. 2002; Mellone et al. 2011, 2016;
Panuccio et al. 2012). This research investigates the
factors affecting an eventual reverse migration in
some Afro-Palearctic raptors approaching a sea
strait. The aim of this research is to verify whether
inter-specific differences occur and which variables
can induce raptors to move in the opposite direction
of migration when still flying some kilometers far
from the coastline. The hypothesis is that larger
and heavy species should be more likely to perform
reverse migration than smaller species. Moreover, it
is expected that migrants change direction of migra-
tion more often with adverse wind conditions (strong
or head winds) or during the afternoon.

Figure 2. Location of the four watch points in the southernmost
portion of the Calabrian Apennines (white = 0–800 m; grey = 801–
Study area and methods
1600 m; dark grey: 1601–1956 m).
To test these hypotheses we analyze here directions
of raptors migrating along the southernmost portion
of the Calabrian Apennines, in the “toe” of the analysis, flocks (two or more individuals flying
Italian peninsula, oriented along their direction of together) and solitary birds were considered sam-
migration (NE–SW; Agostini et al. 2015b). In this pling units to avoid a bias of data (Hurlbert 1984).
study area, there is a flat highland west of a mountain All the observed directions of migration were cate-
ridge, while west of the highland lies the Strait of gorized as expected or reversed. As a result, this
Messina which is the narrowest water surface (mini- research focused on the short-toed snake eagle,
mum distance about 3.5 km) between southern con- which was the only species showing a substantial
tinental Italy and eastern Sicily (Figure 1; see also reverse migration (see Results section). Therefore,
Agostini et al. 2015b). Data were collected from four we verify which factors affected this behavior by
observation posts from 23 August to 10 October running a binary logistic regression analysis (here-
2011, from 12 August to 10 October 2012 and after BLRA) with binomial error distribution com-
from 11 August to 10 October 2013. Three watch paring recorded directions of birds flying toward a
points were located on the mountain ridge or close northerly (reverse migration) or a southerly
to it (minimum distance from the coast about 10, 15 (expected) direction, and using as predictors the
and 18 km; altitude 1052, 1807 and 1762 m, respec- following variables: time of day (morning: 09:00–
tively), and one about 5 km from the Tyrrhenian 11:59; midday: 12:00–14:59; afternoon: 15:00–sun-
coast, on the edge of the flat highland (altitude set), flock size, location of the observation post
987 m; Figure 2). During the fieldwork, the direc- (coastal vs. inland zone), wind direction (head com-
tions of disappearance were recorded. In the ponent: S, SW; tail component: N, NE; lateral com-
ponent: W, NW, E, SE), wind speed (km/h) and the
interaction between the last two variables. At first we
tested for autocorrelation of flock size and wind
speed variables using Spearman’s correlation test.
We selected variables using a stepwise procedure
based on the Akaike Information Criterion (AIC)
values of the different models (Akaike 1973). We
furthermore tested the significance of each variable
in the selected model using analysis of covariance
(ANCOVA). We tested the ability of the BLRA
model to distinguish between the two different beha-
viors by means of the area under the curve (AUC) of
the receiver operating characteristic (ROC) using the
pROC package in R software (Pearce & Ferrier
Figure 1. Location of the study area (the black square) in southern
continental Italy. M = Marettimo. 2000; Boyce et al. 2002; Fawcett 2006; Robin et al.