Considerable attention has been given to the impact of climate change on avian populations over the last decade. In this paper we examine two issues with respect to coastal bird populations in the UK: (1) is there any evidence that current populations are declining due to climate change, and (2) how might we predict the response of populations in the future? We review the cause of population decline in two species associated with saltmarsh habitats. The abundance of Common Redshank Tringa totanus breeding on saltmarsh declined by about 23% between the mid-1980s and mid-1990s, but the decline appears to have been caused by an increase in grazing pressure. The number of Twite Carduelis flavirostris wintering on the coast of East Anglia has declined dramatically over recent decades; there is evidence linking this decline with habitat loss but a causal role for climate change is unclear. These examples illustrate that climate change could be having population-level impacts now, but also show that it is dangerous to become too narrowly focused on single issues affecting coastal birds. Making predictions about how populations might respond to future climate change depends on an adequate understanding of important ecological processes at an appropriate spatial scale. We illustrate this with recent work conducted on the Icelandic population of Black-tailed Godwits Limosa limosa islandica that shows large-scale regulatory processes. Most predictive models to date have focused on local populations (single estuary or a group of neighbouring estuaries). We discuss the role such models might play in risk assessment, and the need for them to be linked to larger-scale ecological processes. We argue that future work needs to focus on spatial scale issues and on linking physical models of coastal environments with important ecological processes.
Coastal waterbird populations are considered vulnerable to climate change, primarily because of sea-level rise. Climate models predict that global sea-levels might be 5–32 cm higher by the year 2050 as a result of the thermal expansion of the ocean, and melting of polar ice sheets (IPCC 2001). However, sea-levels in northern Europe have been rising since the last ice age due to the process of glacial rebound (see Lambeck et al. 1998a, 1998b). Tide gauge data for the UK show that sea-levels are currently rising by up to 3 mm/yr on the coast of England (Woodworth et al. 1999). Sea-level rise is an important political issue because of its implications for coastal protection. As sea-levels rise, coastal habitats such as saltmarsh are lost to erosion, removing important natural protection from wave action. This causes an increase in the wave energy experienced by sea defences, resulting in a massive increase in their maintenance costs. There is already evidence of the widespread erosion of saltmarshes in southeast England (Harmsworth & Long 1986, Burd 1992), although sea-level rise is unlikely to be the only causal factor (Crooks 2004).
Conservationists are concerned that habitat loss resulting from sea-level rise might threaten important waterbird populations by causing a decline in abundance. However, there is considerable uncertainty as to whether this potential risk will be realized. This raises two important questions. First, because coastal habitats are currently being lost to erosion, is there any evidence that these habitat changes are causing particular populations to decline in abundance now? Secondly, even if the status of a particular population is not currently a cause for concern (i.e. it is either stable or increasing in abundance), how can we predict if/when habitat loss driven by climate change is likely to have an adverse impact in the future? These two questions form the primary focus of this paper.
Two species associated with saltmarsh habitats have declined over recent decades. First, saltmarsh habitats in the UK are important breeding grounds for Common Redshank Tringa totanus. Brindley et al. (1998) showed that in 1985 the saltmarsh Common Redshank population numbered approximately 21 000 pairs, but this number had declined by 23% in 1996 to 16 000 pairs. Secondly, saltmarsh habitats on the southeast coast of England are important wintering grounds for Twite Carduelis flavirostris, which feed on the seeds produced by saltmarsh plants. Atkinson (1998) showed that there was evidence of widespread declines in the abundance of birds wintering on a number of estuaries in East Anglia. To what extent can these declines be related to habitat loss caused by sea-level rise and climate change?
There seems little evidence to suggest that the Common Redshank breeding population has declined because of sea-level rise. Population changes have occurred throughout Britain (Brindley et al. 1998), rather than being concentrated in southeast England where sea-levels have been rising most rapidly, as might be expected if habitat loss driven by sea-level rise caused the decline. Instead, the most likely explanation for the decline is an increase in the intensity of grazing. Norris et al. (1998) showed that the density of breeding Common Redshank recorded on a saltmarsh site was significantly correlated with the intensity of grazing. Sites that were heavily grazed had the lowest density of breeding Common Redshank, and the highest densities were recorded on sites with relatively light grazing levels. There was also evidence that the intensity of grazing had increased significantly on saltmarsh sites included in the census between 1985 and 1996, and that Common Redshank declines had been most pronounced on sites that had seen an increase in grazing intensity from ungrazed/light grazing to moderate/heavy grazing. These changes in Common Redshank density in relation to grazing intensity were sufficient to explain the observed decline in abundance at a UK level (Norris et al. 1998). The cause of the increase in grazing intensity remains unknown, although heavily grazed sites had saltmarsh habitats characteristic of being heavily grazed by sheep.
The Twite population wintering on estuaries in East Anglia and the south coast of England has shown dramatic recent declines of > 50% at certain sites. In the context of sea-level rise, this decline is interesting because the estuaries used by the birds have also seen extensive loss of saltmarsh habitats over the last 25 years, particularly low marsh habitats that are extremely important in providing food resources for Twite (see Burd 1992). Atkinson (1998) developed a spatial depletion model of the Twite system, and showed that the predicted decline in bird numbers caused by the loss of their food supplies closely matched the actual decline in bird numbers. Is this evidence of an adverse impact of sea-level rise? Crooks (2004) points out that coastal habitats around Britain are currently experiencing readjustments to past land claim in addition to any effects of sea-level rise and climate change. This means that Twite may have declined directly because of habitat loss, but it is unclear whether the loss of saltmarsh has been caused by sea-level rise, at least acting in isolation. Atkinson's model is also valuable because it allows predictions to be made about how the Twite population is likely to decline in response to future habitat changes driven by sea-level rise and other components of climate change such as storm frequency. The application of such small-scale or local models to predicting the impact of climate change on bird populations is presented in further detail below.
MAKING PREDICTIONS ABOUT THE FUTURE
Although coastal habitats in the UK are important for a number of breeding bird populations, the main conservation interest lies in the internationally important populations of waterbirds (primarily wildfowl and waders) that overwinter on many estuaries (Kershaw & Cranswick 2003, Rehfisch et al. 2003). Applied ecological research has therefore primarily focused on attempting to understand how habitat loss or change might impact on the abundance of these birds (e.g. Goss-Custard & Durell 1990, Goss-Custard et al. 1996, 2002, Percival et al. 1996, 1998, Pettifor et al. 2000a, Stillman et al. 2001). To do this, models have been used to build a link between prey availability and demography or abundance. Understanding how climate change might impact on the demography and abundance of coastal birds requires an understanding of how prey availability might be affected by climate change, how bird populations respond to these changes, particularly in terms of their dispersion patterns, and how changes in dispersion patterns might affect demography and abundance.
Here we review a number of issues that are relevant to the development and application of models to predicting how coastal bird populations may respond to habitat loss (i.e. a reduction in prey availability), including the impact of climate change, by considering the link between prey availability and the demography and abundance of coastal birds. First, it is important to consider the issue of spatial scale at which important ecological processes operate. This is because predictions are required for biogeographical populations whose range covers 1000s of kilometres. Secondly, we examine ‘local’ models that describe important ecological processes acting within a single site (= estuary) or several neighbouring sites, and consider how they might be applied to climate change issues. Finally, we discuss important gaps in our knowledge.
The importance of spatial scale
At which spatial scale(s) do we need to make predictions? To illustrate how we may need to tackle this question, we briefly review recent work on the population of Black-tailed Godwits Limosa limosa islandica wintering in estuaries on the south and east coasts of England. This population has shown a dramatic increase in abundance over the last 25 years (Rehfisch et al. 2003). Prior to this increase, birds primarily wintered on estuaries in southern England, but as the overall population has increased in size, progressively more birds have started wintering on the east coast estuaries, whereas there has been very little change in the abundance of birds wintering on the more traditional south coast sites (Fig. 1). This dispersion pattern is interesting because it implies that there is a buffer effect operating. That is, as the population grows in size, progressively more individuals are forced to occupy relatively poorer quality habitats. This hypothesis was tested by Gill et al. (2001a), who showed that in estuaries with negligible population increases (i.e. those that have supported significant wintering populations throughout the time series) birds had higher survival rates, higher food intake rates in spring and arrived earlier in Iceland than birds occupying sites in which bird numbers have only recently increased (Fig. 2).
This work has important implications for the development of predictive models for two reasons. First, it shows that important ecological processes that need to be incorporated in realistic predictive models can operate at relatively large spatial scales (see also Moser 1988). Secondly, it also shows that the demographic consequences of ecological processes operating at one stage in the annual cycle (winter) can be realized 1000s of kilometres away at a different (breeding) stage of the annual cycle (for a recent review see Webster et al. 2002). There is evidence from some migratory bird populations that early arrival on the breeding grounds correlates positively with breeding success (e.g. Møller 1994, Kokko 1999). Constructing ecologically realistic models of processes at an appropriate spatial scale that might be applied to climate change issues is therefore a substantial challenge, but one that is beginning to be addressed (see Pettifor et al. 2000a).
Local population models
Making predictions at a local level has been done using process-based or behaviour-based models. This type of model describes the processes, such as prey depletion and interference, that occur between individual birds competing for food resources in a population, and we derive from these processes demographic parameters such as population size or mortality rates (for reviews see Sutherland 1996, Goss-Custard & Sutherland 1997, Pettifor et al. 2000b, Norris & Stillman 2002). Although the application of process-based models is not limited to a local scale, most models of actual waterbird populations produced to date are local in that they apply to only a single estuary or group of neighbouring estuaries (e.g. Percival et al. 1996, 1998, Stillman et al. 2001, but see Clark & Butler 1999). There are models of global populations of waterbirds (e.g. Goss-Custard et al. 1996, Pettifor et al. 2000a), but these are only process-based for part (usually winter) of the annual cycle, incorporating other life-history stages using a traditional demographic modelling approach. How do process-based local models work?
The ecological processes usually included in local models are interference (contest) or scramble (depletion) competition between individuals. Here we illustrate a local model developed for Black-tailed Godwits by Gill et al. (2001b) because it is interesting to examine ecological processes acting at a local scale against the background of the large-scale processes described in the previous section. However, having outlined this example, we then discuss more general issues concerning process-based models.
Black-tailed Godwits wintering on estuaries in Suffolk and Essex feed primarily on soft-shelled clams. These prey items are heavily depleted by the birds over winter. Gill et al. (2001b) used a spatial depletion model developed by Sutherland and Anderson (1993) to simulate the depletion process, and to make predictions about the number of bird-days that the food supply could support over the winter at different spatial scales. The model uses (1) parameters from the functional response, (2) the time spent feeding each day by an individual bird and (3) the density of prey below which patches are abandoned because they contain too few prey for a bird to survive to calculate how many bird-days would have been required to cause the observed levels of prey depletion. They then compared these predictions with actual census data, and showed that the model's predictions were reasonably correlated with observed bird numbers.
This approach is encouraging because it illustrates that with an understanding of a relatively simple ecological process (depletion), reasonable predictions can be made about the population sizes of birds a particular food supply is capable of supporting over a given period of time. This could be valuable in terms of predicting the impacts of climate change because if we were able to predict how climate change might affect food availability, we could use the model to examine the potential risk to the Godwit population. Although this prospect seems appealing, it is important to note that the modelling approach applied to the Godwits implicitly assumes that there are no individual differences between birds that influence food intake rates. As Goss-Custard et al. (2002) point out, when individual birds vary in either foraging efficiency or susceptibility to interference, significant emigration or mortality may occur at prey densities above the threshold density incorporated in simple depletion models. This occurs because inefficient foragers or poor competitors are unable to maintain intake rates high enough to survive within a particular area of habitat. This has important applied implications because if such processes are important but not included in a model then it is possible to conclude that a specific environmental change that reduces food availability might have a negligible impact whereas in fact the opposite might be the case. The crucially important issue here is that, although process-based models provide great potential to examine the impacts of environmental change, a particular model must capture important ecological processes adequately to make reliable predictions. In the Godwit example, a simple depletion model seems adequate. This is because of their basic ecology. As Goss-Custard et al. (2002) point out, interference competition is unlikely because of the small prey items being consumed and their lack of mobility, and because the functional response for soft-shelled clams is steep, i.e. intake rates rise quickly with increasing prey densities before reaching an asymptote (see Gill et al. 2001b). This means that even if differences in foraging efficiency between individuals were quite marked, the population would still be vulnerable to a dramatic decline in survival as food availability declined across the threshold because birds would need to balance their daily energy budgets (see figure 4 in Goss-Custard et al. 2002).
How should local models be linked with large-scale processes? We agree with Goss-Custard et al. (2002), who argue that process-based models should be used to generate predictions concerning vital rates such as mortality, and that vital rates should be the currency by which the impact of environmental change is assessed. However, the Godwit case study shows that to do this adequately may require an understanding of large-scale processes. In this particular case, although a local model may identify important ecological constraints on numbers overwintering in particular sites, understanding the impact of environmental change will involve, at least in part, assessing the demographic consequences of emigration. For example, if we were to predict that habitat loss or change driven by climate change reduced the number of Godwits that the east coast estuaries could support, we would still not understand the demographic consequences resulting from the habitat loss. This is because the birds could die or more likely disperse to other estuaries. If the latter occurs, the work of Gill et al. (2001a) implies that the birds would be forced to occupy poorer quality sites, which is likely to reduce their survival rate and cause them to arrive later in Iceland. Therefore, although local models can provide some useful predictions, they may be limited if they work at spatial scales that are too small to understand the demographic impact of habitat loss.
Process-based models provide a potentially valuable ecological tool to address the potential impact of climate change provided models could be constructed and/or interpreted at an appropriate scale. Such an approach has been applied to climate change issues. For example, Percival et al. (1996, 1998) developed a simple prey depletion model for the population of Light-bellied Brent Geese Branta bernicla hrota and Eurasian Wigeon Anas penelope overwintering at Lindisfarne, on the northeast coast of England. These birds feed on intertidal algae, and this food supply could be influenced by climate change if, for example, sea-level rise alters the abundance of algae by changing the shore profile, or because an increase in storms might remove a greater fraction of the algae. The model showed that current numbers of birds was well below the maximum that could (potentially) be supported by the available food supply, implying that the population was not at risk from relatively minor habitat loss.
This example at least illustrates how a local model might be used as an aid to risk assessment, but it also raises an important general issue. The model implies that there are more than sufficient food supplies available to support the population, and hence it is not ‘at risk’ from small-scale habitat loss. Is this conclusion reasonable? Recent work on Eurasian Oystercatchers Haematopus ostralegus has shown that significant mortality can occur even in the presence of relatively abundant food supplies (Goss-Custard et al. 2001). This is because interference competition reduced food intake rates as population density increased, even though there was no significant prey depletion. There is some evidence from goose populations that interference competition can also impact negatively on intake rates, by forcing subdominant individuals to forage in areas already depleted heavily by feeding birds (Rowcliffe et al. 2003). It would be interesting and important to see whether such a process could cause significant mortality among subdominant birds even in the face of relatively abundant food supplies. If so, it would be essential to examine the potential impact of habitat loss driven by climate change by incorporating such processes within a process-based model, otherwise risks to the population may be considered minimal whereas in reality the opposite would be true.
This review of past population declines and approaches to generating predictions about the future raise a number of important issues. First, populations of coastal waterbirds that have experienced a dramatic decline over recent decades are relatively few. However, those that are declining may have very different causes for these declines. This is not news to a conservation biologist, but does emphasize that in the politics surrounding coastal conservation issues it is unwise to focus efforts too closely on a single issue to the exclusion of others. The work on Common Redshank shows that anthropogenic effects other than climate change can cause population declines. In a conservation sense, it is important that policy initiatives tackle all relevant problems, rather than becoming totally immersed in a politically potent issue like climate change/sea-level rise.
Secondly, the Black-tailed Godwit study illustrates that the ecological processes we need to model in order to make reasonable predictions about future population changes may operate over large spatial scales. This is a substantial challenge given the paucity of our knowledge of processes acting at such scales, but important if we are to make predictions at the global rather than local population level.
Thirdly, models of local populations based on ecological processes such as scramble and context competition do have a role in risk assessment at the local level. A number of different models have now been constructed for different systems that provide the possibility for us to look at climate change scenarios. Of central importance to this application is our ability to predict how climate change may influence the distribution and abundance of resources (usually food) in the environment (see below), and how birds respond to these changes. Climate change may also alter the energetic costs of living in different areas of the coast due to changes in temperature, wind speeds and solar radiation. Process-based models provide a framework for addressing how these changes may influence demography and abundance. However, the predictions of local models are limited because any habitat loss that causes individuals to disperse to other sites cannot be adequately assessed in terms of its demographic impacts.
So, how can some of these outstanding problems be addressed? First, we need increasingly to think about large-scale processes. This will be a substantial challenge, but the work by Gill et al. (2001a) shows that it is possible to gain insights into processes operating at large spatial scales, and Pettifor et al. (2000a) illustrate how important processes can be incorporated into a model that operates at large spatial scales. However, more empirical studies are required, and, in our view, detailed studies of particular ‘model’ systems such as Black-tailed Godwits offer a potentially profitable way forward. How can a ‘model’ system be identified? In our view there are four basic requisites: (1) long-term census data covering large spatial scales (e.g. the Wetland Bird Survey in the UK), (2) sufficient knowledge of the birds’ foraging ecology to allow large-scale estimates of resource availability to be made, (3) a sufficient number of marked individuals in the population to allow the fitness consequences of dispersion patterns for individuals to be estimated and (4) preferably some perturbation of the population so that distribution patterns at different total population sizes can be examined (as in Gill et al. 2001a). Secondly, there is an urgent need to bring together models of physical coastal processes with ecological models. This is particularly important in terms of predicting how invertebrate communities might change in response to climate change. At present there is massive uncertainty in the climate change predictions. However, even if uncertainty decreases in future, we will only be in a position to examine, with any degree of realism, climate change scenarios and their potential impact on important coastal waterbird populations if both spatial scale issues and linkages with the physical coastal environment are addressed.
We would like to thank John Goss-Custard for his detailed comments on previous versions of our paper, and for forcing us to clarify our thinking.