Optimal management of a goose flyway: migrant management at minimum cost

Authors

  • Marcel Klaassen,

    Corresponding author
    1. Centre for Limnology, Netherlands Institute of Ecology, PO Box 1299, 3600 BG Maarssen, The Netherlands;
      *Correspondence author. E-mail: m.klaassen@nioo.knaw.nl
    Search for more papers by this author
  • Silke Bauer,

    1. Centre for Limnology, Netherlands Institute of Ecology, PO Box 1299, 3600 BG Maarssen, The Netherlands;
    Search for more papers by this author
  • Jesper Madsen,

    1. Department of Arctic Environment, National Environmental Research Institute, University of Aarhus, PO Box 358, 4000 Roskilde, Denmark; and
    Search for more papers by this author
  • Hugh Possingham

    1. The Ecology Centre, University of Queensland, School of Integrative Biology, St Lucia, Queensland 4072, Australia
    Search for more papers by this author

*Correspondence author. E-mail: m.klaassen@nioo.knaw.nl

Summary

  • 1We adopt a ‘whole flyway’ approach to modelling scenarios for protecting migratory birds, aiming at efficient and cost-effective conservation of flyway habitat.
  • 2We developed a model to minimize flyway management costs while safeguarding a migrating bird population. The model assumes that the intensity of the birds’ use of sites can be manipulated by varying management regimes (with concomitant costs) and that the birds make optimal use of the conditions created along their flyway.
  • 3We used dynamic programming to find the sequence of migratory decisions that maximizes the fitness of the migrants given a range of management scenarios, followed by a management cost estimate of all these scenarios and selection of those scenarios yielding an optimal solution from both an economic and the migrants’ perspective.
  • 4Using the population of pink-footed geese Anser brachyrhynchus that breed in Svalbard as an example, we calculated that the cheapest management scenario given current compensation payment rates at the various goose stopover sites yielded a 35% cost saving over current management. This cheapest scenario provides a migration itinerary that is very similar to the current itinerary used by the geese. This is fortuitous since changing environmental conditions may put the migrants at risk.
  • 5Synthesis and application. Given the global threats to migratory birds, developing a framework for efficient and effective conservation of flyway habitat is an urgent need. Such a framework may likewise be used to assist in controlling migrants causing conflict with agriculture, such as several goose species, in an economic and responsible fashion. Our suggested exemplified framework identified large unexplainable differences in management costs between regions. Differences in management costs between staging sites for birds make big differences to the optimal management of a flyway. Hence, to achieve efficient and effective management of migratory birds, we firstly need an objective assessment of the cost of management in different locations, followed by a modelling approach as here advocated, and followed up by a collaborative action of managers along the entire flyway.

Introduction

The allocation of conservation effort to the management of populations is largely focused on sedentary species. Very few papers consider how resources should be spent optimally managing habitat for migratory species within an economic framework, and the only studies we know of which do this focus on the management of a habitat, not an entire flyway (Amano et al. 2007; Martin et al. 2007). Given the global threats to migratory birds (BirdLife International 2000; Gaston, Blackburn & Goldewijk 2003; Pimm et al. 2006), developing a framework for efficient and cost-effective conservation of flyway habitat is an urgent need – this study is the first to develop such a framework.

One of the predictions of the time-minimization hypothesis for migratory birds is that stopover site use is interdependent (Alerstam & Lindström 1990; Lindström & Alerstam 1992). Accordingly, various theoretical migration models and notably (stochastic) dynamic programming (SDP) models, have shown that for the effective protection of migratory birds, a flyway approach needs to be adopted (Weber, Houston & Ens 1999; Klaassen et al. 2006). SDP models for migratory birds (Farmer & Wiens 1998, 1999; Weber, Ens & Houston 1998; Clark & Butler 1999; Weber et al. 1999; Beekman, Nolet & Klaassen 2002; Klaassen et al. 2006) provide information on the population dynamic consequences of habitat loss or improvement (in terms of size or quality) along a species’ flyway. These models can thus potentially be used to identify key sites for protection and nature management. However, halting habitat deterioration and implementing a management strategy for migratory birds invariably comes at a cost. In this study, we investigate how to optimize the allocation of financial resources in order to maximize the protection of targeted migratory bird species.

Many migratory goose populations are actively managed with often elaborate hunting, scaring, crop-damage compensation and feeding programmes in place along their entire flyway (van Roomen & Madsen 1992). This also applies to Svalbard-breeding pink-footed geese Anser brachyrhynchus, a relatively small population of approximately 56 000 geese (winter 2006/2007; J. Madsen unpublished data). The Dutch, Danish and Norwegian governments annually pay a few hundred thousand Euros in refuge management and goose-damage compensation to their farming communities for this population. Management of goose and other bird populations as well as their staging sites need not be an all-or-nothing undertaking, where protection measures are either in place or are not implemented. By carefully choosing hunting or scaring seasons or the area set aside for geese to forage, the quality of stopover sites for geese can be manipulated and management costs regulated. The scheme recently introduced in The Netherlands serves as an example (van der Zee & Verhoeven 2007). Here specific areas have been designated where geese are tolerated and farmers receive compensation, whereas in other areas, it is the farmers’ own responsibility to minimize goose damage by scaring geese off their fields. These farmers are not entitled to compensation. By fine-tuning the size of the protected area, the desired level of food provisioning to the geese could be attained. A similar scheme has recently been adopted by the Norwegian government to manage the agricultural conflict caused by spring-staging pink-footed geese (Jensen, Wisz & Madsen in press).

Building on a previously published SDP model for the spring migration of this population from Denmark to Svalbard via Norway (Klaassen et al. 2006), we estimated the management and goose population fitness costs for a range of flyway management scenarios in which we manipulated food availability at all staging sites independently. From these results, we extracted those scenarios that minimized flyway-management costs while maintaining the reproductive potential of the population.

Methods

study system and management costs

The Svalbard-breeding population of pink-footed geese winters in Denmark, The Netherlands and Belgium. During March and April, the population congregates in western Jutland, Denmark, before migrating via Norwegian stopover sites to the breeding grounds in Svalbard. In Norway, the geese have traditionally stopped in Vesterålen in northern Norway, but since the late 1980s, increasing numbers of geese have also stopped in Trøndelag, in central Norway. During spring in west Jutland, the geese predominantly feed on pastures, winter green cereals and newly sown cereal fields. Damage to crops is alleviated by feeding geese with grain at designated sites (Jepsen 1991; Madsen 1996). In Trøndelag, the geese feed on pastures, stubble grain fields, and newly sown cereal fields (Madsen et al. 1997). In Vesterålen, the geese primarily feed on pastures (Tombre et al. 2005). These three major stopover sites for pink-footed geese will be referred to as D (Denmark), T (Trøndelag) and V (Vesterålen) in the remainder of this study. In the past, the heavy use of these Danish and Norwegian agricultural areas by geese has led to considerable conflicts with agriculture, culminating in organized scaring campaigns in Norway. In Denmark, a ‘lure crop’ feeding programme covering the major sites has been implemented since the early 1990s (Jepsen 1991; Madsen 1996), and in Norway since 2005, farmers have been compensated for the damage the geese inflict to their crops. We used data from 2006, a year in which a systematic goose monitoring programme was carried out in all three areas and the government agreed to instigate a longer term commitment to a compensation system in Norway. Knowing the total amount of money used to alleviate the conflict in the three areas and the total number of ‘goosedays’ (i.e. the number of days times the number of geese) for each site, allowed us to calculate the site-specific daily costs that a visiting goose incurs (ci, € gooseday−1). Assuming that management costs are indeed directly proportional to the site use of the geese (and thus goose damage), we used the site-specific daily costs that a visiting goose incurs to calculate the total costs for a range of management scenarios (see below). Number of goosedays was calculated based on goose monitoring schemes carried out in Denmark, Trøndelag and Vesterålen from 21 March to the end of May during the spring migration from Norway to Svalbard (J. Madsen and I. Tombre unpublished data). In addition, reading of neckbands on individually marked pink-footed geese was carried out in the three sites as part of the monitoring programme, enabling estimation of the length of stay in the three regions (http://pinkfoot.dmu.dk; Bauer et al. 2008).

model

Klaassen et al. (2006) developed a stochastic dynamic programming model to find the sequence of migratory decisions that would maximize the fitness of a female pink-footed goose under various environmental conditions during spring migration. They used this model to evaluate the impact of environmental change, notably the effect of management-related changes in food availability along the migratory flyway of the geese in terms of their behaviour (i.e. Ti, staging duration in days at the various stopover sites) and fitness (F, expected number of young produced per female over its entire lifetime). Bauer et al. (2008) used the model to generate spring migration itineraries for the same population of pink-footed geese in relation to observed climatic changes along their flyway between 1990 and 2004. They found that model predictions generally agreed well with empirically observed migration patterns over the same period, lending credit to this modelling approach.

In this study, we incorporated management costs into Klaassen et al.'s (2006) model by first calculating the optimal solution of how to organize the migratory journey from the goose perspective (i.e. exactly following Klaassen et al. 2006), given a range of management scenarios for all staging sites along the migrants’ flyway, subsequently calculating total management costs for all these scenarios, and finally selecting those scenarios yielding a (near) optimal solution from both an economic and goose perspective. In our model, we assumed that the geese are omniscient and have full knowledge of the feeding conditions along their route and behave optimally with regard to these conditions. We furthermore assumed that as a result of management, food availability at the three sites can theoretically vary from zero, [e.g. by intensive scaring campaigns, to what is on offer in terms of maximum intake rate (MJ day−1)], without scaring the geese off the fields or impairing their foraging in other ways. For the calculations outlined below, we divided the range of food available on each site into 11 evenly spaced levels from zero to the site-specific maximum and ran all possible combinations, reflecting the full range of possible management scenarios (i.e. 1331 = 113 combinations, 11 possibilities at each of three sites) using the model developed by Klaassen et al. (2006). All other parameters were given their default values (Bauer, Madsen & Klaassen 2006; Klaassen et al. 2006), except the time onwards from which food becomes available in the two Norwegian sites (i.e. onset of spring). The latter parameters were given an average value over the period 1990–2004 (i.e. 15 April and 3 May for Trøndelag and Vesterålen, respectively), where values for the individual years resulted from the date at which satellite-derived NDVI values passed a long-term threshold (Høgda, Karlsen & Tømmervik 2007). For all model simulations, we calculated population average staging times for each site, Ti, as well as population mean expected fitness F. Next, we converted the average staging times, Ti, for all 1331 combinations of management to management costs. The total cost of the entire stay of an individual goose at a specific site (Ci, € goose−1) is:

Ci = Ti × ci,

where ci is the cost per day of a goose at site i. The total costs over all sites along the migratory flyway (CT, € goose−1) is:

CT = ∑Ci (where i€{D,T,V}).

Consequently, for each of the 1331 management (i.e. food availability) scenarios, we have a measure of both the total economic cost, CT, and the expected fitness consequences F for the geese.

Results

For spring 2006 (i.e. from 21 March until 1 June), the number of goosedays for each site as well as the total compensation payments to the farming communities for each site are presented in Table 1. The calculated gooseday costs vary greatly between areas with a ratio of 1:3:9 as we consider sites D, T and V, respectively. However, in this ratio, no correction is being made for differences in food consumption among the three areas, which mainly varies as a result of differences in day length, which greatly determines access to food (Klaassen et al. 2006; Madsen & Klaassen 2006). After correction for differences in average energy intake per goose per day, which amount to 1·98, 2·80 and 3·93 for the three sites, respectively (Klaassen et al. 2006), the ratio in costs among the three sites amounts to 2:4:9. The spring management costs per pink-footed goose along its entire flyway amounted to 3·84 € goose−1.

Table 1.  The total number of goosedays, total compensation payments to the agricultural community (€) and daily costs that a visiting goose incurs (€ gooseday−1), for the three major staging sites along the spring flyway of Svalbard-breeding pink-footed geese for spring 2006 (21 March–31 May, which is the season with crop damage management). Sources: goose data, J. Madsen and I. Tombre unpublished; compensation, Forest and Nature Agency, Denmark; Ministry of Agriculture, Norway
 DenmarkTrøndelagVesterålen
  • *

    including 76 000 barnacle goose Branta leucopsis goosedays.

Total no. of goosedays1 680 0001 165 000386 000*
Total compensation payment (€)41 00086 000 86 000
ci (€ goose day−1)0·02440·07380·2228

For the 1331 simulated management scenarios, the expected reproductive success for omniscient Svalbard-breeding pink-footed geese as a function of total management costs along their spring flyway indicates that fitness levels vary greatly (Fig. 1). Even at low levels of investment, relatively high goose fitness can be attained. On the other hand, high investments do not guarantee healthy goose populations (e.g. when spending large sums of money in Vesterålen and none in Denmark and Trøndelag, thereby turning the latter in poor staging sites for geese). In the model, we used a residual fitness value (i.e. the expected fitness after the current breeding attempt) of 1·80 for non-breeders (Klaassen et al. 2006). A fitness higher than 1·80 thus indicates that a goose was able to breed. Thirty-four per cent of all simulated management strategies yielded a relatively high average expected fitness, ranging between 1·80 and 2·62 young per female while varying costs between 1·35 and 4·52 € goose−1. The bimodal distribution of fitness values (Fig. 1), above and below 1·80 young per female, is due to the fact that the remaining 66% of the management scenarios do not allow for breeding. These cases lead to an enormous fitness drop, many geese even failing in their attempt to make it to the breeding grounds.

Figure 1.

Expected lifetime reproductive success (young per female goose) for 1331 management scenarios as a function of total management costs for pink-footed geese during the spring migratory season (in € goose−1) from Denmark via Trøndelag and Vesterålen to Svalbard. The graph illustrates that high goose fitness can be attained at relatively low levels of investment and that large investments need not necessarily result in high goose fitness levels.

For the scenarios with mean expected fitnesses greater than 1·80 (i.e. relatively favourable management scenarios approximating an almost stable or growing goose population), the average time geese stayed in Denmark varied between 30 and 56 days, for Trøndelag between 0 and 24 days and for Vesterålen between 0 and 12 days. Over the period 1990–2004, the average staging durations for marked individuals within this population were between 29 and 43 days, 11 and 26 days, and 1 and 5 days, for the three sites, respectively (Fig. 2). Thus, the predicted staging durations overlap with empirical data. Narrowing down the selection to expected fitness levels exceeding 2·40 young per female only marginally reduced the predicted time-windows for the geese at each site (Fig. 2), but the minimum management cost increased from 1·35 to 2·43 € goose−1. When we require very high goose fitness (> 2·6 young per female), the management options are very limited. In this case, expected stopover-site uses ranged between 32 and 36 days, 14 and 20 days, and 6 and 9 days, for Denmark, Trøndelag and Vesterålen, respectively, with management costs in the range of 3·60 to 4·06 € goose−1.

Figure 2.

The relationship between calculated staging durations (in days) of Svalbard-breeding pink-footed geese at the three major spring-migration stopover sites (D, Denmark; T, Trøndelag; V, Vesterålen) for a range of management scenarios. The different symbols indicate the total management flyway costs in € goose−1. Only cases where expected lifetime reproductive success exceeds 1·8 young per female goose are included. The light-grey shaded areas indicate management scenarios yielding an expected fitness exceeding 2·4 young per female, whereas dark-grey, black-encircled shading identifies the management scenarios with an expected fitness exceeding 2·6 young per female. For reference, the actual average staging durations obtained in a project on marked individuals of Svalbard-breeding pink-footed geese (cf. Klaassen et al. 2006) over the period 1990–2004 are indicated (field data). The circle identifies a set of management scenarios yielding high fitness at low cost while allowing geese to largely retain current migration behaviour.

Discussion

The actual average site use by pink-footed geese in the period 1990–2004 overlaps with predicted site use for geese with expected fitness greater than 1·80. But remarkably, the empirical data do not coincide with the predicted site-use patterns for those geese with the highest fitness values. We would expect more scatter in the empirical data than in the predictions based on omniscient and identical geese since the latter are based on long-term average foraging conditions on the three sites and thus do not reflect annual variations in conditions. Furthermore, we should consider that empirical data on site use are based on the differences between the last and the first sighting in an area and are therefore minimal staging-duration estimates. Taking this into account, the empirical data still show less use of Vesterålen than we predict for birds maximizing fitness. Possibly, this relates to the scaring campaigns that have intermittently taken place in this area in the periods 1993–1995, 1999–2002. Because of this scaring, the geese may have perceived Vesterålen as a less profitable place than the 70–100% food availability that was assumed for Vesterålen in the set of management scenarios yielding a fitness > 2·6 young per female goose, and thus, reduced their staging time there. Alternatively, as Vesterålen is the most northerly of the sites, the geese might avoid the potentially significant costs of arriving too early and experiencing bad weather.

Population stability is safeguarded with a lifetime reproductive success of 2 young per female. However, adopting a reasonable level of prudence with respect to our modelling results, we selected management scenarios yielding expected fitness values higher than 1·8 young per female goose (i.e. management scenarios where the average female goose reaches Svalbard successfully and breeds). Quite remarkably, a wide range of management scenarios existed with fitness values exceeding 2·4 young per female. The potential for pink-footed geese to successfully cope with a wide range of environmental conditions thus seems to be large. However, a major assumption in our modelling approach is that the geese are omniscient and have full knowledge of the feeding conditions along their route and behave optimally with regard to these conditions. Klaassen et al. (2006) compared two alternative modelling strategies assuming birds to have full knowledge of their environment as in this study and, alternatively, birds having dated environmental information. That comparison highlighted that a sudden change in environmental conditions may have tremendous negative impacts on the geese's fitness. Nevertheless, the Svalbard-breeding population of pink-footed geese has shown some remarkable and apparently adaptive changes in migratory behaviour over the past 40 years, both in relation to climate change and land use (Madsen 2001; Fox et al. 2005), which may be associated with their considerable longevity and their reliance on experience (i.e. exchange of memes through high sociality and strong and long-lasting family bond). Still, it may be wise to only allow for gradual transitions between management scenarios. Furthermore, any implementation of a management plan should be accompanied by a monitoring programme not only to identify potentially unforeseen population responses but also to allow for the refinement and (cost-) effectiveness of the management.

Inspection of Fig. 1 indicates that there are a range of high fitness (> 2·6 young per female goose), high cost (3·60–4·06 € goose−1) scenarios. Intriguingly, there are several management scenarios that yield only a slightly lower fitness (2·4–2·5 young per female goose) at a much lower cost (2·44–2·70 € goose−1). The trade-off between cost and goose fitness is quite favourable, in that we can reduce costs dramatically with a minimal impact on the geese. Furthermore, these scenarios fall close to the range of actual pattern of site use between 1990 and 2004 (Fig. 2). When repeating our analysis under the assumption of dealing with naïve geese rather than omniscient geese, this would result in a large fitness penalty to the geese for management scenarios that would require a drastic change in site use in order to behave optimally. Thus, departing from both naïve and omniscient geese in our modelling exercise favours the same set of scenarios falling close to the range of actual site use. This set of management scenarios with food availability ranging between 70% and 100%, 60% and 100%, and 0% and 40% in Denmark, Trøndelag and Vesterålen, respectively, would result in a reduction of the total management cost by 35%, mainly because of the reduced use of the expensive site, Vesterålen. For this particular case, a large difference in the costs of maintaining geese among the three sites exists, which results in ample potential for optimizing management costs on a flyway scale. The reason for these cost differences among sites is partly because geese forage on different crops of different vulnerability to goose grazing, and the goose grazing pressure varies. However, the figures presented in Table 1 suggest that the level of compensation is only partially based on actual goose damage. We suggest that the alleviation of crop damage in Denmark by luring geese with grain at designated sites (Madsen 1996) mainly serves a psychological effect within the local farming community and has only taken the heat off the problem. For all sites considered, there are no estimates available on actual goose damage in terms of loss of crop yields and loss in income to the farmers. Thus, it seems likely that the compensation payments are largely based on perceptions and political pressure rather than true estimation of the economic loss.

This subjectivity in the allocation of compensation may also result in considerable instability in the level of compensation. Variations in the level of compensation result in varying optimal management decisions. As mentioned above, sudden changes in the management of staging sites may incur considerable fitness costs, and changes in management regimes should thus be avoided as much as possible. To avoid this problem, we call for the implementation of objective goose-damage compensations and long-term commitments to compensation schemes and management schemes in general.

Once objective goose-damage compensation has been established, and we know exactly the site-specific costs associated with each visiting goose, the approach presented in this study could be applied to calculate the optimal management regime (i.e. the amount of forage made available to the geese to safeguard their fitness for minimal costs). That regime might well be very different from the currently calculated optimum. For instance, if we assume equal management costs within Norway, this is equivalent to a situation where management costs are entirely neglected (and therefore comparable to the exercise conducted by Klaassen et al. 2006) and geese should be allowed to stay anywhere within Norway from an economic perspective. Nevertheless, management investments should preferentially be made in Vesterålen, since 1 day of foraging in Vesterålen incurs greater positive fitness consequences than foraging in Trøndelag (Bauer et al. 2006).

management implications

We have devised a method that allows for the protection and management of a migrant's flyway for minimal cost. Given the global threats to migratory birds, developing such a methodology for efficient and effective conservation of flyway habitat is urgently needed. In the case of migratory geese, which increasingly cause conflicts with agriculture at their wintering and migratory staging grounds, the method may also allow management scenarios to be designed that target population regulation in an economic fashion.

Besides pink-footed geese, the migration behaviour of a variety of other waterfowl and wader species has been modelled using dynamic programming (Farmer & Wiens 1998, 1999; Weber et al. 1998, 1999; Clark & Butler 1999; Beekman et al. 2002; Klaassen et al. 2006), highlighting the general applicability of the methodology in finding the sequence of migratory decisions that maximizes the fitness of migrants under a range of environmental conditions and/or management scenarios. These models, which are often programmed in such a flexible fashion that application to other species and migration systems is easily achieved, can be extended without difficulty to include our approach to optimal management. However, for such purposes, management costs are required, which may not always be readily available. Nevertheless, tentative estimates of these costs, which locally may vary enormously across seven orders of magnitude (Balmford et al. 2003), may still be made using approaches such as those advocated by Balmford et al. (2003).

When applying our approach to the flyway management of Svalbard-breeding pink-footed geese, there was indeed considerable space for improving the management of this migratory species across its entire flyway. Large differences in management costs existed among regions, which not only suggested sub-optimal management but also considerable subjectivity and concomitant instability in management scenarios. The fact that flyway management usually involves several nations (in the present spring management case only two, but the entire flyway includes four countries), it necessitates internationally agreed management and conservation objectives, including an understanding of shared economic burdens (Madsen & Jepsen 1992). To increase management stability not only for geese, but also for other migratory birds, we therefore call for an objective assessment of regional costs after which our modelling approach can assist in further optimizing the management on a flyway scale.

Acknowledgements

This work was made possible by ALW-NWO grant 816·01·007. The goose monitoring in Norway was carried out under the AGRIGOOSE project funded by the Norwegian Research Councils and Norwegian Directorate for Nature Management. We are grateful for the insightful comments of T. Amano and an anonymous referee on our draft manuscript. This is publication 4352 of The Netherlands Institute of Ecology (NIOO-KNAW).

Ancillary