Study areas

The study was conducted in two areas located 22 km from each other, in the Baixo Alentejo province, Southern Portugal: São Marcos da Atabueira (37°42′ N, 7°54′ W) and Cabeça da Serra (37°37′ N, 8°09′ W). São Marcos da Atabueira comprises 16.8 km² of a treeless pseudo-steppe, used for pastures or cereal cultivation. Stone piles are frequent all over the area and they are used both as nesting sites (Tomé et al. 2004) and as hunting perches by Little Owls. Density of Little Owls is 2.5 pairs/km² (Tomé et al. 2008). Cabeça da Serra covers 6.1 km² and is mostly occupied by open (16.5 trees/ha) to very open (0.25 trees/ha) woodland/parkland of Holm Oaks Quercus ilex rotundifolia (Tomé et al. 2008). The area is also used for pastures or cereal cultivation and a small part is covered by a young forestation of Stone Pine Pinus pinea. The density of Little Owls in this area is high, reaching 7 pairs/km² (Tomé et al. 2008). Both study areas are included in the proposed network of Vital Sign monitoring areas for evaluating population trends and key habitat components for Little Owls (Van Nieuwenhuyse et al. 2008).

In both areas, agriculture and cattle farming are carried out in traditional ways, maintaining a rotation every 4–5 years between cereal fields and fallows used as pastures (Suárez et al. 1997, Moreira 1999). During the period of the study, 86% of the Little Owl territories in the pseudo-steppe were mostly occupied by fallows. In the woodland fallow was the dominant habitat in 67% of the territories, while the remainder was mostly young forestations or cereal crops.

In the region, the climate is dry, with an average annual precipitation of less than 600 mm (DRAA 2000). The temperature is mild in winter (monthly average c. 11°C) and summers are hot (monthly average over 25°C, with maximum values exceeding 40°C; DRAA 2000).

Behavioural observations

We searched for Little Owls during daytime (approximately 1 h after sunrise until sunset) by monitoring the study areas several days a week between January and early August 1999. We made car transects that fully covered both areas, stopping periodically at strategic points, from which we surveyed the area to locate Little Owls, using binoculars and telescope. When we detected an owl that was actively hunting, we started a 40-min period of continuous observation (‘focal animal sampling’; Altmann 1974, Martin & Bateson 1993). Observations were made from a minimum distance of 30 m to a maximum of several hundred meters. Each focal period began after the owl perched following an attack (irrespectively of its success). When the owl abandoned a perch due to disturbance (caused by raptors or people) or changed its behaviour (roosting, preening, singing), the sampling period was not considered in the analyses.

Little Owl territories were thoroughly mapped during a parallel study using playbacks (Zuberogoitia & Campos 1998, Centili 2001), individual continuous observations and colour-ringing (Tomé et al. 2008). Focal observation periods were obtained in fourteen and fifteen territories in the pseudo-steppe and in the woodland areas, respectively. Most (66%) of these periods were collected during the breeding season (April–June). The prey taken during observations were invertebrates, which are the main prey of Little Owls (by number and mass), in our study areas (Tomé et al. 2008).

During the observation periods, we measured detection and giving-up times (Bye et al. 1992) to the nearest second. Detection times corresponded to the period spent on a perch before an attack, and giving-up times corresponded to the period spent on a perch before abandoning it and moving to another. The success of each attack was also recorded. We assumed that the sampled periods were dedicated to hunting, as owls were exposed and they actively scrutinized the ground.

After each observation period, we measured the heights of used perches (Hays et al. 1981) and distances flown by the owls to the nearest cm. Perches were classified as detection perches (from which an attack was launched; sensu Carlson 1985, Bye et al. 1992) or giving-up perches (perches abandoned for other perches, without an attack; sensu Andersson 1981, Moreno 1984, Bye et al. 1992). We considered any structure used by the owls as an individual perch (i.e. different branches of the same tree were considered as different perches), except when mentioned differently.

Habitat variables

For each focal observation period we also determined: the average percentage of vegetation coverage, estimated from three 1 m² squares randomly distributed throughout the area where attacks took place; and the average vegetation height, estimated from measurements of the plants nearest to the corners of the same squares (Thiollay & Clobert 1988). An Index of Vegetation Cover (hereafter IVC; Thiollay & Clobert 1988) was determined for each sample plot by multiplying the % vegetation cover by the average vegetation height.

To estimate perch availability and the average available perch height in each area, we randomly selected ten Little Owl territories and measured the height of all perches ≥50 cm (considering each structure, tree or stone pile, as one perch) present in a 100 m radius area around the nest or territory centre (Tomé et al. 2004). In order to compare differences in vegetation between areas, we sampled vegetation features in the same twenty territories. Measurements were made in late June/early July from within eight sample-plots, located at 20 m and 40 m from the nest or territory centre in the four cardinal directions (Tomé et al. 2008). Vegetation cover and height were estimated from a 50 cm × 50 cm sample quadrate within each sample plot, together with IVC (see above).

The same territories were sampled in June to estimate invertebrate availability, following procedures described by Sutherland (1996). Four pitfall traps were buried 20 m from the nest (or main roost) to north, south, west and east directions. The traps were kept open during three nights and an additional continuous 48 h period. Sweep-netting was used to sample invertebrates in standing vegetation: a series of ten sweeping movements was performed while walking from the pitfall trap location toward north, south, west and east.

Data analysis

Habitat variables (perch and vegetation characteristics and prey availability) in the territories in both study areas were compared by Mann—Whitney U-tests. When analysing the availability of invertebrates, the samples obtained in each pitfall trap and associated sweepnetting were combined. Hence, four samples of invertebrate prey were used to calculate an average for each territory. In the present study we only analyzed data for Coleoptera (beetles) and Orthoptera (grasshoppers and locusts), the main invertebrate taxa consumed by Little Owls in both study areas, representing more than 60% of consumed prey items (Tomé et al. 2008).

In comparisons involving focal sampling we used average values for each variable calculated from all measurements made during the same focal period per territory. The comparisons of the use of different perch types by the owls with those available in random territories were made with Chi-square and likelihood ratio tests. Differences in the number and height of perches and in variables associated with hunting success between areas or between samples from focal observations and random territories were tested by Mann— Whitney U-test or t-test.

For comparisons between detection vs. giving-up variables within the same habitat we used paired tests per territory (Wilcoxon test), whereas Mann-Whitney tests were used in inter-area comparisons. In these comparisons we only analyzed territories where both attack and giving-up behaviours were observed during focal periods (n = 13 in the pseudo-steppe and n = 14 in the woodland). Spearman correlations were used to test for relationships between behavioural variables (perch height and attack distance). Multiple linear regressions were used to investigate the relationships between vegetation characteristics (vegetation height, vegetation cover and IVC) and behavioural variables.

Table 1.

Perch abundance and height, vegetation characteristics and prey availability in randomly selected territories in pseudosteppe (n = 10) and woodland (n = 10) areas (mean ± SE).

Table 2.

Comparison between detection and giving-up hunting behaviour in pseudo-steppe (number of territories = 13) and woodland (n = 14) (mean ± SE).

To analyze how the behaviour of owls varied in detection and giving-up perches, we calculated the ‘survivorship’ plots of perch times (Moreno 1984, Bye et al. 1992), which represent the variation in the proportion of birds that remain on a perch as search time increases. Means are presented with standard errors. Statistical tests were two-tailed and all analyses were performed using R statistical software (R Development Core Team 2009).

Habitat characteristics

In the pseudo-steppe, most of the available perches were stone piles (51%) and fence poles (49%), whereas in the Holm Oak woodland trees were dominant (80%), followed by fence poles (15%) and stones (3%). As a consequence, the maximum height of available perches was five times higher in the woodland than in the pseudo-steppe (Table 1). Despite this difference in the type of perches, the density of perches was similar in both habitats (Table 1).

Figure 1.

Distribution of perch heights utilized by Little Owls during focal sampling (pseudo-steppe: n = 14; woodland: n = 15 territories) and available in randomly selected territories in the pseudo-steppe (n = 10) and woodland (n = 10).

Pseudo-steppe and woodland differed in both the vegetation (which was higher and denser in the woodland) and the availability of invertebrates. Coleoptera (beetles) were more abundant in woodland, and Orthoptera (grasshoppers and locusts) in pseudosteppe (Table 1).

Hunting behaviour by habitat

ATTACK AND GIVING-UP DISTANCES

We did not find any differences in attack or giving-up distances between habitats (Table 2). In both habitats, the average giving-up distance between structures (considering trees or stone piles as a unit) was almost twice as long as the average attack distance.

Figure 2.

‘Survivorship’ plots of perch times in the pseudosteppe and woodland. Dark lines: pseudo-steppe; grey lines: woodland. Solid lines: detection times; dashed lines: giving-up times. Detection and giving-up times were pooled from all focal sampling periods (pseudo-steppe: n = 78 and 68, respectively; woodland: n = 76 and 94, respectively). The y-axis is represented on a logarithmic scale.

Figure 3.

Relationship between perch height and attack distance in the pseudo-steppe (n = 13) and woodland (n = 14) areas; pseudo-steppe: r_s = 0.65, P < 0.05; woodland: r_s = 0.43, ns.

Table 3.

Hunting success of Little Owls in pseudo-steppe (n = 13) and woodland (n = 14) areas (mean ± SE).

There was a positive correlation between perch height and attack distance in the pseudo-steppe, but not in the woodland (Fig. 3). No correlation was found between perch height and giving-up distance in pseudo-steppe (r_s = 0.14, P > 0.05) and in woodland (rs = -0.14, P > 0.05).

Hunting success

We found no between-habitat differences in the number of prey capture attempts, the prey capture rate and, consequently, in hunting success (expressed as the percentage of successful capture attempts; Table 3).

Hunting behaviour and vegetation

In the pseudo-steppe, we found that perch height decreased with increasing vegetation height. In the woodland, vegetation cover had a negative effect on attack distance (Table 4).

Similar hunting success in two distinct habitats

Although the number of available and utilized perches was similar in territories of Little Owls in both habitats, in the woodland owls benefited from a wider range of perch heights. In this habitat, owls utilized generally higher perches than in the pseudo-steppe, but also foraged from perches at different heights, including big stones on the ground (Fig. 1). In spite of the maximum available perch height (consisting of the highest branches of Holm Oaks at approximately 5 m), Little Owls selected lower perches (average 2.76 m). This suggests the existence of an optimal foraging height above which hunting is less rewarding. As experimentally shown by Andersson et al. (2009), in open habitats where grass is much taller than prey, prey are only visible from almost directly above. This should have been the case especially in the woodland, where ground vegetation was higher and denser. The fact that we found no correlation between perch height and attack distance in woodland (Fig. 3) and that in territories with more developed ground vegetation, Little Owls produced attacks closer to their perches (Table 4), corroborate the hypothesis that in this habitat vegetation was reducing the available visual field and prey accessibility (Thiollay & Clobert 1988, Andersson et al. 2009).

In the pseudo-steppe the use of the highest available hunting perches (tops of stone piles) could pay, since ground vegetation is lower and less dense (Table 1). In fact, and similarly to that found in other studies (Greig-Smith 1983, Moreno 1984, Thiollay & Clobert 1988, Sonerud 1992), in the pseudo-steppe there was a positive correlation between perch height and distance to prey attacked (Fig. 3). In this respect, differences between the habitats may also be tied to the range in variation of perch height available: according to Andersson et al. (2009), a variation between 0.5 m and c. 1.5 m in perch height (as in the pseudo-steppe) may account for a 100% increase in prey visibility, while a variation of between c. 1 and 4.5 m (as in the woodland) should involve a smaller effect. In the pseudosteppe, detection perches were also higher than giving-up perches, suggesting that Little Owls may benefit from choosing the highest available perches to forage. The existence of a wider availability of perch heights in the woodland, which allowed the owls to select an optimal perch height, may explain why no difference was found between the height of detection and giving-up perches in this habitat (Bye et al. 1992, Sonerud 1992).

Figure 4.

Diagram illustrating the two hypotheses explaining the similar prey capture rate observed in the pseudo-steppe and in the woodland (see text).

In both habitats, giving-up distances were practically twice as long as attack distances (Table 2). This corresponds well to some of the assumptions underlying optimal foraging models (Fitzpatrick 1981): Little Owls do not search for prey in an area that they covered before, and they do not waste energy by flying further than is needed to find an unsearched area.

Interestingly in the pseudo-steppe Little Owls attacked from lower perches when the vegetation was higher (Table 4). This result suggests that in this habitat vegetation may also influence the visual detection of prey (Andersson et al. 2009), though to a smaller extent than in the woodland (probably due to the reduced range of variation in height of vegetation in the pseudo-steppe). In these cases (small prey imbedded in dense vegetation and lack of high perches), Little Owls perhaps compensate for losses in visual detectability by increasing the efficiency of acoustic detection through foraging from lower heights (Rice 1983, Bye et al. 1992, Andersson et al. 2009).

While our work was not designed to investigate this question, the use of acoustic cues in prey detection by Little Owls is also suggested by the lack of correlation between perch height and giving-up distance in both areas (see Andersson 1981, Fitzpatrick 1981, Moreno 1984), and by the longer search times we observed in the pseudo-steppe (see Bye et al. 1992), in spite of the use of lower perches in this area. Nevertheless, longer times could also reflect differences in the behaviour (and, therefore, the conspicuity) of the most abundant prey in the pseudo-steppe (Orthoptera) and in the woodland (Coleoptera) (Table 1). Clearly, additional, more focused, investigation is needed to examine the role of acoustic cues and of prey behavioural specificities on the hunting behaviour of Little Owls.