Could an event of extreme drought (2019-2020) affect the feeding ecology of Bubo magellanicus (Gmelin 1788) (Strigiformes: Strigidae) in a Mediterranean region of Chile?

Study area

The study was conducted in Las Tórtolas (33°9′41.50″S, 70°42′23.25″W), located in the Metropolitan Region of Chile (Fig. 1). The dominant vegetation of the area corresponds to Mediterranean scrub, specifically an Andean Mediterranean thorny forest of Acacia caven and Baccharis paniculata (Luebert & Pliscoff, 2017). During the study, the area was mostly dry with scarce vegetation associated and few isolated trees of the species Prosopis chilensis (Molina) Stunz emend Burkart. The climate of the study area is warm temperate, characterised by short periods of rain during winter (July–September) and a six to eight months (October–May) dry season (sensu Köppen, 1948). But, since 2010 to date, central Chile has been affected by an uninterrupted sequence of dry years (12 years), with annual rainfall deficits from 25 to 45% (Garreaud et al., 2017). This so-called Mega-Drought (MD) is the longest event on record in the last millennia in this Mediterranean-like region (CR2, 2015; Garreaud et al., 2020). The MD has co-occurred with the warmest decade on record for central Chile (Vuille et al., 2015), together with other factors increasing its severity, such as: the decrease of the Andean snowpack that affect rivers flow, diminution of reservoir volumes and groundwaters levels; the increase of evaporation reservoirs (Dai, 2013; Garreaud et al., 2017); and the reduction of vegetation productivity (Zambrano et al., 2020). During the study, the annual rainfall deficit reached 72.6% on average, with monthly deficits over 80% for January (100%), May (81.1%), July (82.2%), August (90.5%) and December (87.5%) (Fig. 2).

Pellet collection and analysis

We collected and analysed the composition of pellets of the Magellanic Horned Owl, Bubo magellanicus (Gmelin 1788), in the Mediterranean region of Chile, which is considered a hotspot of biodiversity (Myers et al., 2000) that has been suffering a prolonged drought period for more than a decade (Núñez et al., 2011; Boisier et al., 2016; CR2, 2015; Garreaud et al., 2017; Garreaud et al., 2020), reaching critical levels of rainfall deficit between the years 2018 and 2019 (Zambrano et al., 2020).

Between April 2019 and January 2020, pellets from B. magellanicus were collected at the piedmont, using a carob tree as a perch. The pellets were removed in their entirety systematically every three months, sampling the different seasons of the year. The collected pellets were deposited in hermetic bags and labelled with the date and time for further analysis in the laboratory. Only the entire pellets were weighed and measured using a calliper (0.1 mm) and a digital weight (0.1 g) precision.

Pellets were manually disaggregated using warm water to dissolve the outer mucous coat, and all the remains of invertebrate, hairs, feathers, bones and other organic fragments were separated (Marti, 1987; McDowell & Medlin, 2009). For the identification, we used keys and specific reference for small mammals (Reise, 1973; Pearson, 1995; Fernández et al., 2011; Udrizar Sauthier et al., 2020), birds (Araya et al., 1995), arthropods (Peña, 1986; Artigas, 1994a; Artigas, 1994b; Aguilera & Casanueva, 2005; Vidal & Guerrero, 2007). In addition, the reference collection of the Zoology Museum of the University of Concepción, Chile (MZUC-UCCC) was also reviewed.

Prey quantification was estimated using the Minimum Number of Individuals (MNI) for diagnostic structures. For vertebrates, it is based on the number of skulls and pair of jaws (Pearson, 1995), whereas for arthropods it is based on the presence of cranial capsules, telson and mandibles (Martínez, 2018). Chitinous remains in good condition were taxonomically determined using specialised references and keys (mentioned in the previous paragraph) and compared with the museum collections mentioned above. This method is thought to minimise estimation biases by aggregation (McDowell & Medlin, 2009). The field collections were authorised under permits from Servicio Agrícola Ganadero, SAG, of the Chilean government (No 4141/2018 and 1646/2019).

Morphometry and hypervolume

To evaluate morphology pellet differences between the seasons, first we develop a one-way analysis of variance (ANOVA), verifying the assumptions of normal distribution and homogeneity of variance of the data (Zar, 1999). When a significant result was given, a Tukey’s test (HSD) for multiple pairwise comparisons were performed. Second, we used a hypervolume estimation (Blonder et al., 2014; Blonder et al., 2017) to describe pellet characteristics and overlaps (shape, size and weight). This method allows the description of the shape, volume and overlap of features, attributes or variables using a Gaussian Kernel density estimation (Blonder et al., 2017). To determine the superposition of two hypervolumes (autumn vs winter, and spring vs summer) similarity indices of Jaccard and Sorensen were calculated, as well as the unique fractions of the hypervolumes (Blonder et al., 2017). The morphometric variables of the pellets were standardised by a log-transformation. The analysis was performed using the Hypervolume package (Blonder et al., 2017) in the R software (R Core Team, 2021).

Diet composition

To described the diet composition we used two estimators; frequency of occurrence of each prey unit (%F), and the percentage of biomass (%B). Formulas in Supplementary Materials S1.

Characterization of the diet and trophic niche

The diet composition of B. magellanicus was described at two different time scales: annual and seasonal. The specific richness or the number of species in the diet (S) was determined as species count per site, and the diversity of the diet was estimated following the Shannon-Wiener index (H’), that establishes the relationship between the species richness and the abundance of each species. This index normally fluctuates between 0.5 and 5.0, where values over 3.0 are considered high levels of diversity. Additionally, we used the Pielou index (J′), which measures the evenness of the observed diversity. The values range from 0 to 1, with 1 corresponding to a situation where all species are equally abundant and 0 indicating the absence of uniformity (Magurran, 1998).

The trophic niche breadth was evaluated using the Levins $\left(B\right)$ and standardised Levins (B_STA) indices (Levin, 1968). Levins’ index is the inverse of Simpson’s index. It reaches its maximum when the frequencies are the same for each prey unit consumed, and its minimum when only one resource is consumed, i.e., being a strict specialist (Krebs, 1989). All formulas are available in Supplementary Materials S1.

Surrounding small mammals and diet selectivity

The estimation of the richness and relative abundance of small mammals in the surrounding area was quantified through a catch per season developed from 2018 to 2020, as a complementary study that has been carried out in the surroundings of the perch (pellets point), with an annual installation of 2,800 Sherman traps (700 traps per season) and a low capture efficiency of 4.5%. The estimation of the richness and relative abundance of small mammals in the surrounding area was developed from 2018 to 2020 in a complementary study carried out in the same place, with an annual installation of 2,800 Sherman traps (700 traps per season). However, for the present study we only considered the traps installed in a surrounding radius of 700 m2 of the perch (60 traps per season, adding up to 240 traps per year with a total sampling effort equivalent to 960 traps, as each trap was active during two consecutive nights), as a potential feeding area of the owl. Each captured individual was identified at the species level, following Reise (1973), Muñoz Pedreros, Rau & Yánez (2004) and Muñoz Pedreros & Gil (2009). In the case of arthropods, study was carried out in summer (2018) and autumn (2020) using Barber soil traps, so it was not considered for the diet selectivity analysis, due to the lack of information for the remaining seasons (winter-spring).

The diet selectivity analysis was defined by a Chi-square goodness of fit test (Sokal & Rohlf, 1969), where the observed values correspond to the absolute frequencies of prey species occurrence in the pellets, whereas the expected values were calculated as the relative frequency of each prey species in the trapping study, matched up with the grand total of prey individuals detected in the pellets (Jaksic & Yáñez, 1980).

Comparative diet analysis

The collection of pellets was done during a climatic period defined as Mega-Drought (MD) in central Chile (Garreaud et al., 2020), in which the year 2019-2020 was characterised as one of the driest on record with a precipitation deficit of 72.6% with respect to a normal year, only precipitating at winter time a total of 29.90 mm (information record from the environmental Huechún station, Metropolitan Region) (Fig. 2). Under this extreme environmental scenario, a comparison with other studies on diet and trophic ecology of this species in the Mediterranean zone of Chile was assembled. The diet comparison was carried out using the following studies: Muñoz Pedreros et al. (2017) (annual study) in the Lago Peñuelas National Reserve, Valparaíso region; Jaksic & Yáñez (1980) (spring study) developed in La Dehesa sector, north of Santiago, Metropolitan region; and Yáñez, Rau & Jaksic (1978) (summer study) carried out in the city of San Fernando, O’Higgins region.

To test the null hypothesis that there are no differences in the relative frequencies of PU between the study periods, a multivariate analysis of variance, based on permutations (PERMANOVA) of three factors (Anderson, 2001) was applied. The data was based on Bray-Curtis dissimilarity, squared root transformed and 999 permutations were performed. The analyses were carried out through the vegan package (Oksanen et al., 2020) in R software (R Core Team, 2021). Data and script are available in supplementary material and through the Figshare repository (https://doi.org/10.6084/m9.figshare.20481486.v2).

Morphometry and hypervolume pellets

A total of 120 pellets were analysed during a year of study (2019-2020), establishing morphometric variations between the different seasons (Table 1), with significant differences in weight (ANOVA F = 17.12; p < 0.05) and length (ANOVA F = 2.78; p = 0.04, Table 2). The Tukey’s post hoc test determined difference of weight between autumn-winter, winter-spring and winter-summer for lengh, and difference of length between autumn-winter for length, while the width of the pellet did not show significant differences between the different periods (ANOVA F = 0.39; p = 0.75) (Table 2). The hypervolumes exhibited the highest volume in winter (0.75), followed by spring (0.55); whereas in summer (0.48) and autumn (0.24) the hypervolumes showed smaller sizes (Fig. 3). Regarding the superposition of the hypervolumes between seasons, a high similarity was established in terms of morphology and weight between the spring-summer and autumn-winter (Table 3).

Characterization of the diet and trophic niche

The specific richness of diet consisted in 22 species (prey units) with a marked seasonal fluctuation along the year, registering the highest value in winter (S = 14) and the lowest in spring (S = 6). The Shannon-Wiener index showed an average of 0.71 ± 0.33 bit/ind., with lower values in the spring (0.35 bit/ind.) and summer (0.53 bit/ind.), contrasting what was exhibited in autumn (1,086 bit/ind.) and winter (0.89 bit/ind.). The Pielou index showed the same trend as the specific diversity, with greater heterogeneity of diet in autumn (0.44), and lower evenness in the rest of the seasons (winter = 0.17; spring = 0.19; summer = 0.23) due to the dominance of a few PU (Table 4). The annual estimation of the trophic niche breadth was B = 0.37, while the standardised Levin’s index was B_STA = 0.01.

We quantified 1,071 annual PU and consumed biomass of 15,988 (g per year). The higher PU percentages identified in the pellets occurred in summer (34.30%; PU = 367) and spring (28.70%; PU = 307), adding up to 63% (PU = 674) of the total PU. The percentages of consumed biomass showed the same tendency, with higher values in summer (32.36%; 5,174 g) and spring (26.85%; 4,293 g). Most of the annual PU consumed (97.75%) were arthropods (87.86% Arachnida and 9.89% Insecta); and the remaining 2.24% were vertebrates. Grammostola rosea (from the class Arachnida) was the most consumed species with an annual PU percentage of 84.97% (PU = 910), plus it exhibited the highest percentage of occurrence throughout the year (77.48 ± 13.22% PU), with maximum values in spring (92.73% of consumed biomass) and summer (88.63% PU). The scorpion of the genus Brachistosternus was scarcely consumed (2.89% annual PU), showing a low percentage of occurrence throughout the year (0.14 ± 0.09%) with maximum values of consumed biomass in autumn (0.27%). Among Insecta, Tenebrionidae family was the most consumed with 8.40% annual PU and with an annual biomass contribution of 0.23 ± 0.11%.

Vertebrates were poorly represented in the diet, with only 2.24% of the annual PU and a marked seasonal occurrence percentage, with a maximum in autumn (33.33% PU) and a minimum in spring (3.33% PU). The annual average of biomass consumed of this group was 21.67 ± 15.74%, with the lowest values recorded in spring (5.27%) The rodents Abrocoma bennetti and Abrothrix longipilis were the most consumed prey with 52.15% of vertebrate annual PU, although it only represents 1.12% of the total annual PU. The marsupial Thylamys elegans and the bird Turdus falcklandii were scarcely consumed with only one record in autumn and spring, respectively (Table 4).

Surrounding small mammals and diet selectivity

A low richness and relative abundance of small mammals was recorded during the different sampling seasons. Phyllotis darwini was the most captured rodent with a mean relative abundance of 3.33 ± 1.36%, while the rest of mammals presented relative abundances less than or equal to 1% (Table 5). The selectivity of the diet established that B. magellanicus did not consume mammals in the same proportion as they are present in the surrounding area (X² = 7.81; p < 0.05). In decreasing order, the most consumed rodents were A. bennetti. A. longipilis and P. darwini, whereas Octodon degus and Spalacopus cyanus were the least consumed. The marsupial T. elegans and the lagomorph Oryctolagus cuniculus were extremely rare in the diet (See Table 4).

Comparative diet analysis

In comparison to other studies on dietary ecology of B. magellanicus performed in the Mediterranean region of Chile, this study determined the lowest trophic niche breadth record for the species, with values of Levin’s index B = 1.38 and Levins standardised index B_STA = 0.01 (Lago Peñuelas B = 6.23; B_STA = 0.37. La Dehesa B = 6.32; B_STA = 0.66. San Fernando B = 2.93; B_STA = 0.18). The Shannon-Wiener and Pielou indices followed the same trend, with the lowest values estimated for the species, whereas the dominance index and prey richness exhibited the highest values compared to the other trophic studies (Table 6).

Trophic characterization showed a shift in the composition and biomass of the diet for B. magellanicus during the period of extreme drought, exhibiting preferences of arthropods over vertebrates. In general, most of the biomass consumed (78%) corresponded to arthropods, with vertebrates contributing the resting 22%, thus contrasting the results of Lago Peñuelas National Reserve (Muñoz Pedreros et al., 2017), that described high preference for small mammals (90.17 ± 10.33% diet biomass) and a marginal percentage consumption in arthropods (<6.0% PU). Also, the results of La Dehesa study, carried out only in spring (Jaksic & Yáñez, 1980) described consumption only for vertebrates (small mammals 88.7% and birds 11.3%) compared to 66.6% consumption of arthropods biomass (0.2% Insecta and 66.4% Arachnida) and only the remainning 33.4% of vertebrates biomass, in the same season. Nevertheless, the research developed in summer in San Fernando (Yáñez, Rau & Jaksic, 1978) showed a greater consumption of arthropods than the rest of the comparative studies (17.70%), but less than the percentages of consumption registered in our study (88.8%). However, the present study exhibited significant differences of diet composition with all the trophic ecology studies previously documented for the species in the Mediterranean region of Chile (PERMANOVA F = 2.59; p < 0.05).

The arachnid G. rosea was the species that most contributed annually to the biomass (77.50%), followed by the rodent A. bennetti (9.01%) in this study, contrasting with Jaksic & Yáñez (1980) and Muñoz Pedreros et al. (2017) which determined an average contribution of 50.76 ± 11.76% by O. cuniculus and 23.55 ± 0.78% by A. bennetti, respectively. The study of Yáñez, Rau & Jaksic (1978) described an important contribution from the genus Grammostola sp. (15.80%) and the exotic rodent Rattus norvegicus (14.50%).