Site selection by geese in a suburban landscape

Introduction

In Europe and North America, wild and feral geese frequently inhabit lakes and their surrounding parks in urban and suburban areas. These parks are appreciated by people for their recreational and aesthetic value. However, this often brings geese in conflict with people (Conover & Chasko, 1985; Hughes, Kirby & Rowcliffe, 1999; Smith, Craven & Curtis, 1999; Fox, 2019). While people often enjoy seeing small numbers of geese, when there are large flocks the soil becomes fouled and people are intimidated by the geese’s threatening behaviour (Miller et al., 2001). Geese are also known to exert pressure on small water bodies such as ponds, reducing water quality through eutrophication (Allan, Kirby & Feare, 1995; Gosser, Conover & Messmer, 1997; Smith, Craven & Curtis, 1999; Kumschick & Nentwig, 2010). They have also been suggested to be a disease risk, though the evidence is circumstantial and other domestic and wild animals pose a greater known risk (Fleming & Fraser, 2001; Clark, 2003; Bönner et al., 2004). Throughout Europe and the western Palearctic, native as well as non-native geese are increasing in numbers and distribution (Allan, Kirby & Feare, 1995; Fox et al., 2010). Several populations have developed a resident component and their year-round presence increases human-wildlife conflicts and impacts on biodiversity (Buij et al., 2017). A variety of strategies are needed to reduce these impacts (Austin et al., 2007; Gyimesi & Lensink, 2012).

In Europe, from the 18th century onwards, it has been traditional to create landscaped parks reflecting an idealised vision of the countryside. Lakes with islands, open vistas, lawns and patches of woodland are typical (Turner, 1985). Lake-side vegetation and lawns are cut regularly and the canopies of trees are kept high to ensure unimpeded views. For those goose species that are habituated to the presence of people, such landscapes are very suitable, they have abundant grazing; proximity to water and islands for undisturbed nesting sites. In addition, people often provide supplementary feeding.

In north-western Europe, four species of “geese” are the main inhabitants of urban and suburban parks: non-native Egyptian geese (Alopochen aegyptiaca), Canada geese (Branta canadensis), mixed populations of wild and feral greylag geese (Anser anser) and barnacle geese (Branta leucopsis) (Fox et al., 2010; Huysentruyt et al., 2019). All are members of the family Anatidae, but Egyptian geese are members of the subfamily Tadorninae, which are referred to as tadornine geese, whereas the others are members of subfamily Anserinae, which are referred to as anserine geese (Livezey, 1996). Egyptian geese are similar in several aspects to anserine geese, such as their large size, long neck and feeding behaviour, but they do differ in other important aspects. Anserine geese, such as Canada geese, barnacle geese, greylag geese and their hybrids, usually nest on the ground close to bodies of water and are also likely to form large flocks (Adriaens et al., 2019). Egyptian geese are also water birds, but their biology shows many characteristics of a duck, including larger clutch sizes. Although they nest on the ground, their nest site selection is highly variable and they also nest in large tree holes, on buildings, on top of willow trees or in nest boxes (Gyimesi & Lensink, 2012; Huysentruyt et al., 2020). They also differ in their social behaviour. Paired Egyptian geese defend territories near their nest site before and during nesting. Large flocks of Egyptian geese only occur after breeding during moulting (Gyimesi & Lensink, 2010).

The site selection criteria of geese are important, because their sites can bring them into conflict with people. The proximity of water, food and breeding sites are relevant to goose site selection, but there are likely to be additional influences. These habitat features may be related to predator avoidance (Conover & Kania, 1991), accessibility of feeding grounds for adults and families with chicks, nutritional quality of feed (Owen, Nugent & Davies, 1977; Fox & Kahlert, 2005), sward length (Hassall, Riddington & Helden, 2001; Feige et al., 2008; Conover, 1991; Van Gils et al., 2009; Huysentruyt & Casaer, 2010) and competition with other grazers such as other geese, livestock and rabbits (Van der Wal, Kunst & Drent, 1998). Given this, it may be possible to identify management strategies and landscape features that alter the site selection of geese and these might be used to control the geese in such a way to reduce conflict between geese and people (Conover, 1992; Owen, 1975).

Culling is often used to reduce the impact of geese (Reyns et al., 2018), but several other strategies have been used to discourage and redistribute geese, including birds scarers and chemical antifeedants (Conover, 1985), fencing of feeding grounds or landscape modification, including altered mowing regimes or landscaping solutions (Cooper, 1998; Van Daele et al., 2012). In the context of a landscaped park with large numbers of visitors, culling risks losing public support for a public garden and bird scaring might disturb people too. At the same time, a botanic garden needs to consider the impact of grazing and fouling on plantings, lawns and vegetation, without losing the recreational opportunities for wildlife watching provided by the presence of these attractive birds. Therefore, habitat modification is considered as a cost effective, sustainable solution to reduce numbers of geese on sites and to mitigate the impact (Conover, 1992). Previous studies on site occupancy of geese have concentrated on wild geese in more or less rural settings. These studies have concentrated on ways to discourage geese from feeding on crop plants (e.g., Olsson, Gunnarsson & Elmberg, 2017; Si et al., 2011). In the case of Canada geese, most studies have occurred in North America (e.g., Conover, 1992).

The aim of this study is to quantify the site selection of the different species of geese within Meise Botanic Garden (Belgium) and create models to predict their behaviour based upon the landscape of the Garden. These models can then be used to suggest strategies to reduce conflict between the geese and the visitors to the Garden without losing the opportunities they represent for wildlife watching.

Materials & Methods

Most of the research was conducted at Meise Botanic Garden (Flanders, Belgium), situated just north of Brussels, Belgium (50°55′42.4″N 4°19′37.6″E). The exception was the study on the effect of islands and those data are described below. The 92 ha Garden is a landscaped park like many such parks in northern and western Europe. It has extensive lawns, woodlands, two large lakes and one small one (Fig. 1). The Garden is subdivided into different numbered areas, divided by paths, which join various historic buildings and greenhouses with formal gardens, with approximately half the area covered by woodland. Most of the grassland is mown between two and four times a month during the growing season, though small areas are maintained as wildflower meadows and are cut once or twice a year. All geese in the Garden are considered either non-native or feral. All species breed in the Garden, though the breeding of Canada geese is, in part, controlled by egg-shaking. The birds using the Garden are part of a larger population of geese that inhabit the greater Brussels area, and birds move in and out of the Garden to the many other lakes and waterways in the neighbourhood. None of these populations are truly migratory, except for local movements (Anselin & Cooleman, 2007). Canada goose is under management in the region and flocks of geese are regularly moult captured on water bodies in neighbouring municipalities since 2010 (Reyns et al., 2018). The Garden is in almost constant use by geese except for on the rare occasions when the lakes freeze over for long periods in the winter. Geese feed on all the lawns and grasslands within the Garden, but the extent to which these areas are used varies considerably from area to area and from species to species.

Results

Do geese avoid proximity to trees?

During the study geese were rarely ever observed in woodland. Egyptian geese are occasionally found perched in trees where they nest, but rarely on the ground in woodland. It was hypothesised that this negative association with woodland would extend beyond the boundary between the woodland and lawns and be the cause of an edge effects on grazing.

Quantification of the length of geese droppings showed a clear edge effect at the border to woodland (Fig. 2). A shorter length of droppings was found close to the woodland, but this effect only extended 5–10 m from the boundary.

As a control modelling was also performed in parallel to the woodland boundary, but models either failed to converge or showed no directional trend.

Which habitat features attract geese?

Here we model the site selection of geese based upon habitat features we suspect might be important to geese. The area of the sector, barriers to flight, presence of woodland and proximity to lakes all appear relevant from observations of geese and the literature cited in the introduction. The mean individual counts of geese in the different sectors of the Garden are mapped in Fig. 3. From these maps it is clear that all species had a high affinity to the sectors containing lakes, though there are clear differences between species. The greylag geese in particular are far more wide-ranging than other species notably in the large western sectors.

The models of sector usage were evaluated with various means. The Cook’s distance was used to evaluate if particular sectors had an exaggerated influence on the model outcomes, but this does not appear to be the case (Fig. S1). Variograms of the residuals did not show evidence for spatial autocorrelation that was not accounted for in the model parameters (Figs. S2–S5). A plot of residuals versus fitted values indicates that there may be some non-linearity between the predictors and the abundance of geese, but this was not clear (Fig. S6). The Q–Q plot shows that the residuals were quite normally distributed for all models (Fig. S7). The Scale-Location plot showed that some heteroscedasticity was evident in all models, however we consider that only the model for B. leucopsis was so heteroscedastic that it might impact our interpretation of the results. Given that no real-world model will perfectly match our assumptions and some of the reasons for deviation from these assumptions are suggested in the discussion.

A summary of the minimum adequate models is given in Table 1. The simplest minimum adequate model selected was for Anser anser. Only the area of the sector and the presence of woodland were significantly correlated to their distribution in the Garden. Note, that these models do not include areas of the Garden containing a lake. For B. canadensis the area was also positively correlated with the number of geese, but not significantly in the model. However, in contrast to Anser anser, distance from a lake was a significant factor for B. canadensis, but also barriers to direct flight and their interacting term. For Alopochen aegyptiaca, area and barriers are significant as single factors, and they reoccur in interacting terms. Distance from the lake was not a significant term, but it did occur in an interaction term with area. In the case of B. leucopsis, area was a significant correlate, the other terms are more difficult to interpret, but both distance from a lake and the presence of barriers remained in the model due to their interactions and their interaction with area.

Table 1:

A summary of the minimum adequate models results for the distribution of geese in the Garden.

A ‘+’ indicates a positive association of geese numbers with the independent variables and ‘–’ indicates a negative association. The independent variables are the area of that sector of the garden, the distance from a lake, the presence of woodland on the garden sector and barriers to direct flight out of a sector. The number of asterisks indicate the degree of significance (^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001). Details of the models are presented in Tables S2–S5.

	Alopochen aegyptiaca	Anser anser	Branta canadensis	Branta leucopsis
Area	+*	+***	+	+**
Distance from a lake	+		–***	+
Woodland	–	–**	–
Barriers to direct flight	–*		–***	+
Area:distance	–*			–*
Area:barriers	+*			–
Distance:barriers			+***	–
Area:distance:barriers				+*

DOI: 10.7717/peerj.9846/table-1

Goose abundance was negatively correlated with woodland for all except B. leucopsis, but this variable is not ideal as all those areas of woodland are also surrounded by trees as barriers to flight, So, there are no areas of woodland without barriers. Therefore, some of the variance stemming from the presence of woodland may be being accounted for in the barrier variable.

Therefore, for all species the area of the sector was positively correlated with goose abundance and the area was part of the significant interactions included in the models for Alopochen aegyptiaca and Branta leucopsis. The distance from the lake remained in models for all species, except Anser anser. This is also evident in Fig. 3, where Anser anser can be seen to range more widely than other geese. All other predicted habitat determinants were included in one or more of the models.

For Canada and greylag geese there was a negative influence of barriers on site usage, particularly for Canada geese. In the case of Egyptian and barnacle geese, barriers were not a clear determinant of site selection, but did remain in minimum adequate models as interactions with distance and area.

Do islands in lakes attract geese?

Lakes with islands attract more Canada, greylag and barnacle geese in the summer (Fig. 4). These results indicate that a lake without an island had 35%–60% fewer anserine geese than a lake of an equivalent size with an island (p < .05). However, islands made no difference to the number of Egyptian geese. All goose numbers showed a positive relationship with lake size (p < 0.05), although this is not significant in the case of barnacle geese.

Discussion

The modelling results, edge effects and impact of islands demonstrated the complicated relationship between habitat choice and the landscape for suburban geese (Figs. 1 and 2, Table 1). A casual observer could assume that there is a rather passive relationship between geese and their landscape, but as with any other animal, geese are clearly actively selecting and using particular landscapes and landscape features suited to their preferences.

Edge effects are relevant to the usage of geese on lawns because they reduce the active area of use for the geese. Our methodology did not distinguish whether there are species differences, however, the effect was so distinct that we speculate that all species are influenced. While there may be many potential causes of an edge effect, such as predator avoidance and poorer grazing, an area of lawn less than 20 m in diameter is likely to be undesirable to geese. However, with increasing ratio of area to circumference means that the relevance of this effect will diminish with increasing area. In ornamental parks individual specimen trees might extend the influence of this edge effect.

Sector area was the most consistent predictor of goose abundance (Table 1). This was anticipated, as more space can contain more geese. Yet in addition to the edge effects there are reasons to expect a more sophisticated relationship between goose number and area. Firstly, anserine geese are social species forming large flocks and they may only select areas with sufficient capacity to hold the whole flock. Secondly, if an area is surrounded by tall trees the flight angle needed to enter and leave it from the air becomes progressively steeper the smaller the area becomes. Mature trees stand 15–20 m tall, but average vertical and horizontal airspeeds of geese are approximately 0.5 m s⁻¹ and 16 m s⁻¹ respectively (Hedenström & Alerstam, 1992). Therefore, to enter and escape a small area surrounded by trees they must either considerably steepen their descent or climb rate, or circle while gaining or losing height. Both of these strategies would be more energetically expensive (Norberg, 1996). For these reasons, it is not surprising that the area of the sector also appears in interacting terms in the models with barriers. Barriers particularly restrict movement of geese when flight is not an option, such as, when raising young or moulting. However, the negative influence of barriers was scarcely significant for Alopochen aegyptiaca. This may be a result of their behaviour of nesting in tree holes. Though they do not inhabit densely forested areas, their preferred habitat is open grassland with some trees in proximity to freshwater (Cramp et al., 1984; Carboneras, 1992; Gyimesi & Lensink, 2012). They defend territories around nest sites and therefore must be in proximity to trees (Sutherland & Allport, 1991).

Distance from lakes was not as important to site selection as had been assumed, and the interactions with area and the presence of barriers suggests that the ease of access to grazing is more important to site selection than the linear distance. This perhaps indicates that careful usage of landscape features could guide geese to use particular feeding sites, irrespective of their distance from the lake.

The results show a strong preference of anserine geese for lakes with islands during the summer (Fig. 4). Islands are used by geese year-round, as they provide protection from disturbance where geese can rest and nest. The lack of a similar preference for Egyptian geese is consistent with the territorial breeding behaviour of Egyptian geese and their use of nest holes in trees. Although anserine geese prefer lakes with islands in the summer, the reasons are probably many and this preference may not be true in winter. Island breeders are presumably more protected from predators, particularly foxes (Vulpes vulpes) (Wright & Giles, 1988), stone marten (Martes foina), brown rat (Rattus norvegicus) and carrion crow (Corvus corone) (Huysentruyt et al., 2020). However, when breeding success on islands has been examined it is not always better than on the mainland (Gosser & Conover, 1999; Petersen, 1990). Other studies on the influence of islands on goose nest site selection vary. Fox et al. (1989) showed no influence for greylag goose, whereas others report an effect for Canada Goose (Lokemoen & Woodward, 1992; Bromley & Hood, 2013). Huysentruyt et al. (2020), in their study of 200 breeding pairs of barnacle goose in Flanders, also note that barnacle goose mainly breeds on small islands in lakes and ponds in the region.

Based on the results of this study we suggest that landscape adaptations could indeed reduce the number of geese in suburban parks, which could be an alternative to lethal control and prevent conflict with people. Unfortunately, many of the landscape adaptations that would reduce the presence of geese are in opposition to popular landscape design features, such as ponds and lakes, islands, open vistas and extensive lawns. Other sorts of landscape and garden design with more enclosed and higher vegetation are more suitable where geese are a problem. Woodlands, shrubberies, coppice, hedges, tall grass meadows, prairie planting, hard landscaping features, shallow water and moving-water features would all deter geese from using an area (Allan, Kirby & Feare, 1995; Gosser, Conover & Messmer, 1997; Allan, 1999; Baxter, Hart & Hutton, 2010).

If artificial islands were eliminated from suburban lakes it might be argued that native birds would also suffer from the lack of island breeding sites, however, islands in suburban parks are mostly unsuitable for island nesters of conservation concern, such as common terns (Sterna hirundo) which do breed well on artificial rafts in bigger lakes and lagoons (Coccon et al., 2018; Dunlop, Blokpoel & Jarvie, 1991). Islands could perhaps be made less attractive if they were connected to the mainland by constructing bridges or an isthmus. They can also be modified with banks that deter access from the water, rather than from the air. However, making feeding areas inaccessible is controversial as chicks can then starve (Allan, 1999). Modifications or removal of islands should however consider the trade-off with ongoing management. For example, when practicing egg shaking or egg oiling for fertility reduction, the success of this measure depends on sustained effort and a high percentage of treated nests (Klok et al., 2010; Beston et al., 2016). Hence, having all geese nest on the same island makes it easier to perform this management.

There is also a need to educate the public to the benefits of geese. In the Botanic Garden their selective grazing of grasses has created an exceptional species rich grassland that is unlikely to be maintained with mowing alone (Ronse, 2011). An adaptive management approach, whereby vegetation and goose numbers in the Garden are thoroughly monitored and objectives are clearly stipulated, could be a good way to learn more about the behaviour and impacts of geese.

Conclusions

Landscape features have a powerful influence on the distribution of geese, though these influences differ between species. For example, we show that…

• Lakes with islands attract more than twice as many anserine geese

• Flight barriers between grazing areas and lakes deter geese

• Small grazing areas surrounded by trees reduces their attractiveness to geese

• Proximity of a lake is most important to Canada geese, and least to greylag geese

Landscape modifications cannot completely remove geese from a suburban landscape and an integrated management strategy may be necessary (Allan, Kirby & Feare, 1995). Retroactively modifying landscapes to reduce their attractiveness to geese is difficult, so designing landscapes for wildlife usage should be among the primary design criteria.

Supplemental Information