Study Areas. We studied eagle-owls at two study areas along the west-central coast of Korea: (1) wetlands located along the SRA (ca. 360 km2; 37°4′–37°17′N, 126°34′–126°50′E) and (2) non-wetland areas located on the coastal area from Paju City and Gimpo City to Ganghwa District (PGG; ca. 810 km2; 37°36′–37°49′N, 126°21′–126°51′E; Fig. 1). For the SRA, the marshes and open areas created after 1987 were widely distributed throughout the habitat (K-water 2009); elevation ranges from 28–146 masl, with gentle slopes (Shin et al. 2013). The PGG was composed of different landscapes: Paju and Gimpo were cultivated and urbanized, with intense human activity, whereas Ganghwa was characterized by hills (range = 42–468 masl; Shin et al. 2013).

Figure 1. 

Study areas in west-central Korea, where landscape, diet, and reproductive rate of Eurasian Eagle-Owls were studied, September 2010–August 2012. ▪ PGG (Paju City, Gimpo City and Ganghwa District); ▴ SRA (Sihwa Reclaimed Area).

Landscape Assessment. The landscapes around nest sites were classified following Shin et al. (2013). Based on Arirang 2 satellite images from the Ministry of Environment of Korea (Ministry of Environment of Korea 2009), we plotted a circle with a 1-km radius centered on each nest, and extracted the proportional areas of seven land-cover categories (fallow lands, agricultural lands, forests, water bodies, wetlands, human settlements, and grasslands) using ArcGIS 9.3 (ESRI 2008). We also measured: (1) nearest neighbor distance between nests (NND); (2) altitude above sea level (ASL); (3) distance to a paved road; (4) distance to an occupied residence; and (5) distance to an open area (DOA).

Diet and Reproduction. We examined 44 eagle-owl territories: 19 in SRA and 25 in PGG (Paju: 7, Gimpo: 11, Ganghwa: 7; Fig. 1). In the period from September 2010 to August 2012, we searched for nests using a combination of methods: passive listening for vocalizations, playing of conspecific calls and listening for responses, observing individuals at dusk, and visiting areas around historical nests (for more details, see Shin et al. 2013). To minimize disturbances, we visited nests once per week during the breeding season, starting with incubation.

We examined laying date and clutch size. We estimated hatching date by back-calculating based on the feather development of nestlings (Penteriani et al. 2005), and estimated laying date by subtracting 35 d, the mean incubation period, from the hatching date (Cramp 1985, Penteriani 1996, Dalbeck and Heg 2006). We expressed laying dates in days since 15 December (1 = 15 December), because that was the earliest laying date. To record clutch size, we climbed to the nests during incubation. When replacement clutches (two each for SRA and PGG) were recorded, we used the date of the first clutch as the laying date.

We calculated reproductive breeding success at the two study areas using multiple parameters, i.e., the number of fledglings (1) per territorial pair, (2) per breeding pair, (3) per hatchling, (4) per egg laid, and (5) per successful pair. We defined a territorial pair as one that established and defended a territory. Pairs that moved away >300 m from their nest sites in the next breeding season were also considered to maintain a separate territory. We defined a breeding pair as one that laid at least one egg and a successful pair as one that fledged at least one young. An occupied nest site was defined as one where a territorial pair was present during the breeding season, and nesting success was calculated as the percentage of territorial pairs that successfully fledged at least one young. Because eagle-owl young may leave the nests on the ground when they are ca. 1 mo old, we visited all nest sites as well as areas around the nests until the young were 8 wk old to record brood size at fledging.

To assess diet and its effects on reproductive output, we collected pellets and prey remains (1035 in SRA and 1977 in PGG) at nests, feeding sites, and roosting places, once or twice per month during the entire study period and every time we visited a nest during the breeding season. We identified prey items at the species level using published literature (Takada and Kanochi 2004, Yoon et al. 2004, Abe 2007, Won and Kim 2012). We determined the minimum number of prey items of each type for each collection event (Marti 1987, Bosè and Guidali 2001, Marchesi et al. 2002, Penteriani et al. 2005, Shin et al. 2013). We calculated the percent biomass and percent frequency of prey items for each territory by pooling all the years together (Smith et al. 1999). Biomass data for each prey species were based on mean body weight in Won (1981), Yoon et al. (2004), Song and Song (2005), Hume (2006), and in our measurements.

The various prey items that made up >3% of the diet pooled over the entire study area by number or by mass were classified at the family level: Anatidae, Ardeidae, Columbidae, Corvidae, Phasianidae, Leporidae, Muridae, and other prey (insects, crustaceans, reptiles, and actinopterygians). Other avian and mammalian prey items with three were grouped together as other birds and other mammals, respectively. Trace prey items made up 0.01% (Jaksić and Marti 1984).

The differences in diets between the two habitats were compared using diet composition, MWP, and Shannon–Wiener diversity index (H′). Because diet composition may vary seasonally for raptors (Bosè and Guidali 2001, Forsman et al. 2001, Marchesi et al. 2002), we used only diet data collected during the breeding season (from 15 December to 31 August in our study area) to describe diet and to assess the relationship of diet with reproductive rates.

Statistical Analyses. We used discriminant function analysis (DFA; Sokal and Rohlf 1995) to ascertain whether nest sites could be correctly classified into their original habitat types (wetland or non-wetland habitats). The difference in diet composition between the two habitats was compared by χ2-test, based on the occurrence frequency of the categorized prey items. To avoid potential biases caused by the effect of sample size on diet indices (Smith et al. 1999), we excluded territories with fewer than 20 identified prey items from statistical analyses. In addition, territories affected by human disturbances such as amateur photographers or herb pickers were excluded from our analyses of the factors (landscape and diet) that may affect the reproductive output. Prior to analyses, percentage data were arcsine-transformed for normalization and variables related to MWP in the nonbreeding period were standardized as Z-scores to remove variation between years. Univariate comparisons of land-cover categories, nest location characteristics, dietary parameters and breeding estimates between the two different habitats were analyzed using t-tests, or Mann–Whitney U-tests when the data did not meet normality requirements or variances were not homogeneous after transformation. Associations of the independent variables with breeding success were analyzed using Pearson correlation. To examine factors affecting the reproductive success of eagle-owls, we used the landscape, nest location, onset of egg-laying, and diet variables as independent variable factors in a stepwise multiple linear regression (SMLR), with entry and removal criteria set at probability of F ≤ 0.05 and ≥0.10, respectively. To avoid pseudo-replication, we averaged the value of reproductive success for the two study years for each territory (Marchesi et al. 2002). Significance was determined at P < 0.05. PASW Statistics 18.0 for Windows (SPSS Korea Datasolution Inc. 2009) was used for statistical analyses. Data are expressed as means ± SD.

Landscape and Nest Location Variables. We found significant differences between the two habitat types for seven variables (DFA; χ2 = 81.851, canonical correlation = 0.942, n = 43, P < 0.001). All the nest sites in both SRA and PGG were accurately classified. The percentage of wetlands was higher in SRA, whereas the percentages of agricultural lands, forests, human settlements, and grasslands were higher in PGG (Table 1). However, the percentages of fallow lands and water bodies did not differ significantly between the two habitats (Table 1). Nests of the SRA, compared to those of the PGG, were farther from occupied houses and paved roads, nearer to open areas and located at a lower elevation (Table 1). NND did not differ between the two habitats.

Table 1.

Characteristics of landscape surrounding 43 Eurasian Eagle-Owl nest sites in wetland (Sihwa Reclaimed Area) and non-wetland populations (Paju City, Gimpo City, and Ganghwa District) of west-central Korea during the period of September 2010–August 2012. Percentage data of land-cover types were analyzed within a radius of 1 km around the nest (see methods for details) and were arcsine-transformed to compare the means between the two study areas. At one nest site in the wetland study area that was first discovered after fledging, we measured only nearest neighbor distance. n in parenthesis is number of nest sites, and P < 0.05 was considered significant. Bold type indicates statistical significance. Values are represented as mean ± SD.

Nest Sites and Reproductive Success. In 2011, 34 nests were occupied (n = 12 for SRA, n = 22 for PGG), whereas in 2012, six of those were not occupied, another seven nest sites were found, and two pairs moved away ca. 400 m and 1600 m from their previous nest sites within their own territories (n = 17 for SRA, n = 20 for PGG in 2012). Of the 43 occupied nests, eight (n = 3 for SRA, n = 5 for PGG) had pairs that did not lay any eggs during the two breeding seasons. Clutch size, brood size at hatching, and the number of hatchlings per egg did not differ between the two study areas (Table 2). Additionally, laying date ranged from mid-December to early March and did not significantly differ between SRA and PGG (Table 2).

Table 2.

Breeding parameters for Eurasian Eagle-Owls in wetland (Sihwa Reclaimed Area) and non-wetland populations (Paju City, Gimpo City, and Ganghwa District) of west-central Korea, September 2010 to August 2012. Percentage data were arcsine-transformed. P-value indicates the significance of the difference between the two study areas. n in parenthesis refers to number of samples.

Wetland-dependent eagle-owls had higher reproductive rates than non-wetland eagle-owls, using several measures of reproduction. Nesting success was 1.7 times higher in wetland habitat, and most measures of reproductive success varied between the two habitats (Table 3). SRA owls fledged more hatchlings than PGG owls (Table 3); the number of fledglings per egg was 1.6 times higher in SRA (Table 3). The number of fledglings per breeding pair of SRA was 1.5 times higher than that of PGG (Table 3). However, the number of fledglings per successful pair did not differ.

Table 3.

Mean reproductive success of Eurasian Eagle-Owls in wetland (Sihwa Reclaimed Area) and non-wetland populations (Paju City, Gimpo City, and Ganghwa District) of west-central Korea during the period of September 2010–August 2012. Percentage data were arcsine-transformed. Territories affected by human disturbances (e.g., amateur photographer, herb picker) were excluded from the analysis (see methods for details). P-value indicates the significance of the difference of breeding outputs between the two study areas, and significant difference is in bold type. n in parenthesis refers to numbers of sample.

Diet. Pellets and prey remains contributed 76% and 24%, respectively, to the total sample of prey items (Table 4). For pellets, mammalian prey were more frequent than avian prey; however, among prey remains, avian prey were overwhelmingly more frequent than mammalian prey (Table 4). During the breeding period, diet composition significantly varied between the two habitats (χ2 = 308.524, P < 0.001; Table 5). For the SRA eagle-owls, birds were the most important prey group in terms of both number and mass (Table 5). The most important bird family groups were Anatidae, Columbidae, and Phasianidae, making up >52% by number and 75% by mass. Waterbirds such as Chinese Spot-billed Ducks (Anas zonorhyncha), Mallards (A. platyrhynchos), egrets (Egretta spp. and Bubulcus ibis), and Common Coots (Fulica atra) contributed greatly to the biomass of the owl diet at SRA (47%). Muridae, including brown rats (Rattus norvegicus), black-striped field mice (Apodemus agrarius), and voles (Clethrionomys spp.), were detected quite frequently in the SRA, but their frequency of occurrence and biomass (31% and 16%, respectively) were lower than those of birds. In contrast, both birds and mammals were important components of the owl diet at the PGG. Mammals were 1.4 times more frequent than birds (Table 4), but avian prey contributed twice as much to the overall biomass (Table 5).

Table 4.

Diet composition of the Eurasian Eagle-Owl in west-central Korea (September 2010–August 2012), analyzed by pellets and prey remains. Taxonomic groups with <3% of number or biomass of prey items were included in the categories “Other mammals” and “Other birds,” respectively. tr = trace (<0.01%). Number of each species (n), frequency of each species (F), and biomass of prey (B) in the diet are shown. n in parenthesis refers to sample size.

Table 5.

Spatial variations of the Eurasian Eagle-Owl diet in the breeding and nonbreeding periods in wetland (Sihwa Reclaimed Area) and non-wetland populations (Paju City, Gimpo City, and Ganghwa District) of west-central Korea, September 2010–August 2012. Number of each species (n), frequency of each species (F), and biomass of prey (B) in the diet are shown. Taxonomic groups with <3% of number or biomass of prey items were included in the categories “Other mammals” and “Other birds,” respectively. tr = trace (<0.01%). Territories with n < 20 prey items were excluded from analyses of the difference between the two habitats. n in parenthesis refers to sample size.

MWP was 1.7 times greater at SRA than at PGG during the breeding period (SRA: 473.8 ± 498.9 g, PGG: 276.6 ± 365.9 g; Mann–Whitney U-test, z = −11.036, n = 2513, P < 0.001). H′ was also significantly higher in SRA (SRA: 2.1 ± 0.4, PGG: 1.7 ± 0.5; t-test, t = −3.555, n = 35, P = 0.001). Eagle-owls in the SRA wetland area captured more diverse prey, but MWP of avian prey that they hunted was 2.6 times greater than that of mammalian prey (Fig. 2).

Figure 2. 

Biomass contribution of avian and mammalian prey in the Eurasian Eagle-Owl diet in the breeding period in the wetland population (Shihwa Reclaimed Area) of west-central Korea (September 2010–August 2012). Aves: X̄ = 608.2 ± 446.7 g, n = 537, Mammalia: X̄ = 237.1 ± 529.4 g, n = 253 (Mann–Whitney U-test: z = −18.895, P < 0.001). The box includes the middle 50% of the distribution of prey weights, with the bold central line within the box indicating the median of the data. The top and bottom of the box represent the 75th and 25th percentile of the data, respectively. The vertical lines outside the box show the upper and lower deciles. Circles and asterisks represent outliers and extreme values, respectively.

Association of Landscape Variables, Nest Location, and Diet with Reproductive Rate. Percent wetland was the only landscape variable positively associated with the reproductive success of eagle-owls (Table 6, Fig. 3A). As the proportion of wetland area increased, there were also increases in the number of fledglings per hatchling and per egg. Of nest location characteristic variables, NND was negatively correlated with the number of fledglings per breeding pair and hatchling, and DOA was negatively associated with the number of fledglings per hatchling (Table 6). MWP and H′ were positively associated with both the number of fledglings per breeding pair (Fig. 3B) and per egg (Fig. 3C, Table 6). The percentage of Anatidae (by number) in the diet was positively correlated with the number of fledglings per hatchling; similarly, the percentage of Phasianidae (by number) in the diet was positively correlated with the number of fledglings per breeding pair and per egg (Table 6). The number of fledglings per successful pair was not significantly associated with any of the independent variables.

Table 6.

Correlation of the landscape, nest location, and diet variables with the reproductive success of Eurasian Eagle-Owls in west-central Korea, September 2010–August 2012, analyzed by Pearson correlation analysis. Dietary data in the breeding period were pooled among years, and breeding data were averaged across the two study years for each territory. Percentage data were arcsine-transformed. Territories with n < 20 prey items and affected by human disturbances (e.g., amateur photographer, herb picker) were excluded from the analysis (see Methods for details). n in parenthesis refers to sample size.

Figure 3. 

Association of percent wetlands (A), mean weight of prey (B), and diet diversity index (C) in the breeding period with reproductive success, based on 43 Eurasian Eagle-Owl territories in west-central Korea, September 2010–August 2012. Dietary data were pooled among years and breeding data were averaged across the two study years for each territory. Percentage data were arcsine-transformed. n in parenthesis is sample size. The equations of 3A, 3B, and 3C are y = 1.070x + 0.845 (r² = 0.234, n = 29, P = 0.010), y = 0.003x + 0.153 (r² = 0.218, n = 29, P = 0.011), and y = 1.186x − 0.960 (r² = 0.267, n = 29, P = 0.004), respectively.

A SMLR revealed that H′, percent biomass for “other family” and NND were significant determinants of the number of fledglings per breeding pair, whereas percent wetland was the only significant determinant of the number of fledglings per hatchling (Table 7). The percent of Phasianidae in the diet (by number) and date of the onset of egg-laying were significant determinants of the number of fledglings per egg (Table 7). A significant predictor of the number of fledglings per successful pair was not detected in the pooled samples. Overall, the predictive power of a multiple linear regression was highest for reproductive rate in terms of the number of fledglings per breeding pair, accounting for >47% of the variation.

Table 7.

Multivariate determinants influencing reproductive success of Eurasian Eagle-Owls in west-central Korea, September 2010–August 2012, analyzed by a stepwise multiple linear regression. Landscape and diet variables were entered as independent variables and each of the four measures of reproductive rate as dependent variable (only the final model is shown). Dietary data in the breeding period were pooled among years and breeding data were averaged across the two study years for each territory. Percentage data were arcsine-transformed. Territories with n < 20 prey items and affected by human disturbances (e.g., amateur photographers, herb pickers) were excluded from the analysis.