Study Area

The study was carried out in a flooded area at the confluence of the Warta and the Odra rivers in western Poland (Ujście Warty National Park; 52°34′N, 14°53′E). The reservoir covers an area of about 5,000 ha and is bordered by dikes. During the study period (spring 1994-1996), the water level was high and the area resembled a shallow lake with islands of willow bushes and willow trees Salix spp. Only in June 1995 the water level was low enough to change the dominant landscape to meadows and pastures (for more details see Osiejuk et al. 1999). The dominant vegetation surrounding goose nests were willow bushes (diameter: 10-40 m, height: 2-4 m), old willow trees (up to 15 m high) and rows of young willow trees (height: 5-8 m). The dominant feature of the area is the unpredictable variation in water level which could modify nesting conditions both between and within seasons.

Potential predators of greylag goose nests were hooded crow Corvus corone, magpie Pica pica, red fox Vulpes vulpes, wild pig Sus scrofa, pine marten Martes martes, stone marten Martes foina and American mink Mustela vison (Majewski 1986, Osiejuk 1998, Bartoszewicz & Zalewski 2003).

Measurements of Flushing Distance

The study was conducted during the breeding seasons of 1994-1996. Between one and five visits were made per nest each year. Successive visits were quantified.

Greylag goose nests were approached by two people in a boat with an electric or a silent gas engine. When the female left the nest, the distance between the nest and the boat (DIST) was estimated to the nearest 1 m. The distribution of flushing distance was highly skewed and significantly different from a normal distribution (Kolmogorov-Smirnov: Z = 2.55, P < 0.001). Thus, we transformed DIST, and for all tests we used the following index of flushing distance: LOGDIST = log(DIST). LOGDIST is symmetrically distributed and does not deviate from the normal distribution (Kolmogorov-Smirnov: Z = 1.30, P > 0.05). All statistical tests were performed using LOGDIST, but to make its values meaningful, corresponding values of flushing distance (± SE) were also given in text and figures. There was no year effect on flushing distance (F_2,134 = 0.953, P = 0.388), so all data were pooled before analysis.

Factors Affecting Flushing Distance

The number of eggs in the nests was counted at each visit. For analysis of variance, clutch sizes were pooled in three classes (variable CLUTCH-SIZE: 2-4, 5-6, and > 6 eggs).

Actual hatching dates were known in 40 of 97 cases. In the remaining 57 cases, we estimated the remaining incubation time and hatching date on the basis of egg density and water test (Westerkov 1950, Osiejuk 1998). All eggs were measured and numbered with a marker when first found, and they were weighed and tested in water at each visit. The methods were developed as a part of a more extensive study in the Słońsk Reserve (1994-1996) during which 206 breeding pairs were investigated. The accuracy (mean deviation in days) of our clutch ageing method was 1.6 ± 1.9 days (Osiejuk 1998).

We noted three variables describing nest type and nest surroundings: 1) OVER-WATER: nest height above the water level (0 vs > 0 m); 2) VEGETATION: vegetation height above the nest (< 4.5 m vs ≥ 4.5 m); and 3) TO-OPEN-WATER: distance from the nest to open water (< 3 m vs ≥ 3 m). Median values were used as criteria for the above divisions. A nesting attempt was considered successful if at least one of the goslings survived to the time of leaving the nest.

During the nest surveys, we also measured temperature, time of day and wind speed. Preliminary investigations showed no significant influence of the variables on flushing distance, so they are not presented in this paper. Similar results were obtained by Gunness & Weatherhead (2002) who studied three duck species.

Statistical Approach

General Linear Models (GLM) and model notation were applied mainly according to Grafen & Hails (2002). Statistics were calculated using MiniTab 13 (Ryan & Joiner 2001) and SPSS 10 software (Norušis 2000), and P values are two-tailed, unless stated otherwise.

Material Collected

We collected data from 97 greylag goose nests (29 nests in 1994, 27 nests in 1995 and 41 nests in 1996). Altogether, we had 135 flushing distance observations (37 in 1994, 36 in 1995 and 62 in 1996) and the females were flushed once from 49 nests, twice from 32 nests, three times from six nests and four times from a single nest.

The number of visits we made was positively correlated with number of days of incubation (r = 0.54, P < 0.001, N = 135). This is not surprising as the nests that were incubated longer had a greater chance of being visited more times. However, this result leads to difficulties in testing the influence of number of visits on flushing distance for the incubating female. Therefore, we used two statistical approaches to analyse our data. First, we analysed only the earliest records, i.e. first-flushing distance measured for each nest (female), to test effects of clutch size, breeding stage and nest site. Then we used also data from repeated visits to analyse how incubation progress and humans affect females' decisions. This was a very conservative approach as LOGDIST was found to be independent of the number of visits to the nest (F_3,134 =0.58, P = 0.627).

Analysis of the First Flushing Attempts

First-flushing distance averaged 16.5 ± 1.60 m (range: 1-100 m, N = 96). We used a GLM approach to check for significant relations among a set of predictors. LOGDIST was significantly correlated with different measures of breeding stage, e.g. number of days after the first egg was laid (r = -0.32, N = 96, P = 0.001), and number of days since incubation was started (r = -0.40, N = 96, P < 0.001). We found, however, the strongest correlation between LOGDIST and number of days left to hatching (r = 0.45, N = 96, P < 0.001), so this variable (in the following termed TO-HATCH) was used in the models. We found that first-flushing distance depended mostly on time left to hatching, and the longer the clutch was incubated, the longer the females stayed at the nest (Fig. 1, Table 1). We also found that large clutches (> 6 eggs) were left earlier (first-flushing: 24.1 ± 4.6 m) than medium (16.7 ± 2.5 m; Tukey adjusted: P = 0.044) and small clutches (14.0 ± 2.4 m; Tukey adjusted: P = 0.030, Fig. 2). First-flushing distance was significantly shorter in nests located in shorter vegetation, i.e. within willow bushes, than in groups of higher trees (see Table 1 and Fig. 3).

Figure 1

Relationship between first-flushing distance (in m) and time left to hatching (number of days) for incubating female greylag geese.

Table 1

Results of the General Linear Model analysis on the effects of variables associated with the reproductive value of clutch and nest location variables on first-flushing distance of female greylag geese. Effect of CLUTCH SIZE was analysed using Tukey simultaneous tests. All pairwise comparisons among levels of CLUTCH SIZE were 1) 2-4, 2) 5-6, 3) > 6 eggs, and CLUTCH SIZE classes were 1 and 2 (P = 0.994), 1 and 3 (P = 0.030) and 2 and 3 (P = 0.044). The variable TO-HATCH is a covariate (t = 5.08, P < 0.001).

Figure 2

Relationship between mean (± SE) first-flushing distance (in m) and clutch size for incubating female greylag geese with different clutch size.

Figure 3

Relationship between mean (± SE) first-flushing distance (in m) and vegetation height (in m) for in incubating female greylag geese.

Analysis of Repeated Flushing Attempts

Flushing distance was measured at least twice in 31 nests, and it was found to be significantly longer during the first visit (23.0 ± 3.85 m) than during the second visit (15.8 ± 3.9 m; repeated measures GLM: Wilks' Lambda = 0.828, F_1,30 = 6.24, P = 0.018). In the next repeated measures model, we included also number of days to hatching at the first visit (TO-HATCH) and number of days that passed between the first and second visit (TSHIFT). In this case, the statistically significant factors were TO-HATCH (F_1,28 = 12.31, P = 0.002) and TSHIFT (F_1,28 = 5.94, P = 0.021).

We found that the longer the time between the first and second flushing attempt was, the more shortened was the second flushing distance in comparison to the first one (Spearman's rho: r_s = -0.36, N = 31, P = 0.048). However, this relationship was insignificant when controlled for clutch age at the first flushing attempt (r_s = -0.32, N = 28, P = 0.081).

Flushing Distance and Nesting Success

First-flushing distance was close to being significantly shorter in unsuccessful females, whose nests were later destroyed, than in successful females (F_1,62 = 3.733, P = 0.058). Mean first-flushing distance was 16.1 ± 2.4 m for unsuccessful females and 23.9 ± 4.0 m for successful ones. The difference was insignificant when TO-HATCH and CLUTCH-SIZE were included into the model.

Effect of Clutch Size

The parental investment theory predicts that females should invest more for offspring of a higher reproductive value (Trivers 1972, Montgomerie & Weatherhead 1988). Regarding clutch size, one may therefore expect that females with larger clutches should exhibit shorter flushing distances than females at the same breeding stage but with smaller clutches. We found that females incubating the largest clutches flushed from significantly longer distances than females having smaller clutches. The analysis involved full clutches with time of incubation and nest location included in the model. So, the observed behaviour seems to contrast the parental investment theory and some other studies on precocial species (Forslund & Larsson 1992, Forbes et al. 1994, Sjöberg 1994). This unexpected increase in risk-taking with decreasing clutch sizes might result from many reasons, e.g. unknown population age structure and female condition or experience. Waterfowl have been reported to respond to manipulation of clutch size and appear to be able to judge the value of their clutch (Armstrong & Robertson 1988, Sjöberg 1994). Therefore, individual clutch value, adjusted for several possible factors such as age, female condition and food availability, may be a better measure of parental investment for testing a nest defence hypothesis than the simple clutch size variable.

How can this ‘experience adjustment’ work? For example, older geese usually lay larger clutches and/or larger eggs than younger geese (e.g. Hamann & Cooke 1987, Robertson et al. 1994, Woog 2002). In general, reproduction is known to be age-specific in geese and ducks (Finney & Cooke 1978, Raveling 1981, Blums et al. 1997, Kampp & Preuss 2005). Therefore, we may expect that older females are more experienced and that vigilance in such birds is more effective than in younger females. So, older females may faster recognise a potential danger, may faster cover the eggs with down feathers, and therefore may be able to leave the nest earlier than inexperienced females. Such a behaviour could be advantageous, as the covering of eggs protects them against cold for quite a long time, while simultaneously the longer flushing distance hinders nest localisation by a potential predator and enables females to avoid risking their own lives. The lack of experience in nest defence is, for example, a reason for lower reproductive success rates in first-time breeders of snow geese Anas caerulescens (Cooke et al. 1995). In general, many researchers stress the importance of age-related skills for reproductive success in geese (e.g. Forslund & Larsson 1992, Rockwell et al. 1993). Moreover, a recent study on turnstones Arenaria interpres demonstrated experimentally that birds may change their response depending on their individual state independently of the strength of the disturbance by humans (Beale & Monaghan 2004a).

Stage of Incubation

The best predictor of flushing distance in greylag geese was the time left to hatching. Our observations indicate that during the first half of incubation females tend to leave the nest quite early and change their behaviour at the final stage of incubation. However, it is noteworthy that the shortest flushing distances observed are still long enough to enable a safe escape. We know of only a single example of a female goose bitten to death on the nest by a fox during the three years of our field research (Osiejuk 1998). However, many times we found completely destroyed nests, proving that a relatively strong and potentially dangerous predator was active in the area. This part of our results is in agreement with the parental investment theory, as females were found to invest more in eggs with a higher probability of hatching (Osiejuk 1998), i.e. offspring with a higher reproductive value (Trivers 1972, Montgomerie & Weatherhead 1988). Previous studies on various duck and goose species have shown results consistent with our research, i.e. females tend to wait longer before flushing as incubation progresses (Mallory & Wheatherhead 1993, Forbes et al. 1994, Sjöberg 1994, Albrecht & Klvaňa 2004). Moreover, Albrecht & Klvaňa (2004) have shown that incubating mallards Anas platyrhynchos were extremely attentive to the nest during the last part of the incubation stage, similarly to our goose species. Escape behaviour was found to be affected by offspring value also in other bird species (Dale et al. 1996) and animals from other groups, e.g. fish (Pressley 1981) or insects (Tallamy 1982).

Nest Location

Nests of greylag geese in the Słońsk Reserve were located in few microhabitats which can be divided into two main classes: 1) nests in willow bushes and 2) nests in trees. Nests in bushes were usually placed at the water level and were hidden inside the dense vegetation, invisible from a longer distance, especially when leaves appear. Nests in trees were placed above the water level, usually not concealed by dense vegetation, and consequently they were visible from a distance throughout the breeding season. Earlier research showed that these two nest types differed in overall breeding success, but this was mainly because nests in trees did not suffer from floods, typical for our study area (Osiejuk 1998). In our study we found that flushing distance was shorter for nests well concealed in bushes. This result suggests that female behaviour is adjusted to nest site location. Females appear to distinguish properly between visible and hidden nests and delay flushing when they ‘expect’ to remain unobserved. This also supports our earlier hypothesis that female experience may play an important role in shaping variation of flushing distance. Corresponding results were reported also by Albrecht & Klvaňa (2004), who suggested that vegetation may act as a protective cover for incubating females in mallards. Consequently, that ability to delay flushes in cryptically coloured birds may be regarded as an antipredator strategy to reduce nest advertising. Albrecht & Klvaňa (2004) also suggested that such behaviour may depend strongly on predator types presented (e.g. visually vs chemically oriented) and in some cases may be maladaptive. In our case, we found that step-like differences in nest concealment (within a bush vs in a tree) may affect a female's nest defence strategy. Similarly to the mallard, an escape decision seems to be balanced by cryptic colouration (Albrecht & Klvaňa 2004), rather than by reduced locomotory abilities (e.g. Cuadrado et al. 2001, Cooper 2003). Anyway, many studies confirm that geese in particular are able to adjust their behaviour to the current predator pressure. For example, Forslund (1993) demonstrated that vigilance behaviour of barnacle geese Branta leucopsis depends not only on brood size, but also is adjusted to predator abundance. Kahlert (2003) showed that the choice of moult site in wing-moulting greylag geese is mainly affected by predation risk and level of disturbance, which also indicates a high predation-risk perception in this species.

Effect of Repeated Visits

We found that geese escaped later closer to hatching, regardless of the number of visits, and hence no positive reinforcement or habituation effect of repeated flushes was observed. Sjöberg (1994) obtained similar results for Canada geese Branta canadensis, and Forbes et al. (1994) did so for different duck species. Thus, the increased defence intensity during incubation was attributed exclusively to an increased reproductive value of the clutch (Andersson et al. 1980, Redondo 1989, Sjöberg 1994).