The Earth’s climate has always had a powerful and ubiquitous influence on biodiversity, hence the great concern about the anthropogenic climate change that has been set in motion in the last century. There is an increasing need to understand the vulnerability of species to climate change but this is not a trivial task (Williams et al., 2008). Statistical modelling of the relationship between climate and the distribution and abundance of species is now a widespread tool for studying present-day climate–organism interactions as well as the future consequences of climate change (Elith & Leathwick, 2009). It first involves the collation of geo-referenced records of species’ occurrences and the querying of spatially interpolated climate data at these locations. These data can then be used to develop models that predict probability of occurrence, or abundance, as a function of climate. Such a model can be used to infer distribution and abundance across the full extent of spatial climatic data, now, in the future, or in the past.
This kind of species distribution modelling approach is a powerful predictive tool because of the intimate connections between organisms and climate. Statistical species distribution models (SDMs) are often difficult to interpret, however, because processes linking organisms to climate are extremely diverse and often indirect. For example, air temperature may be a strong predictor of geographic range through its direct effect on the physiology of the organism (e.g. development or growth rate). Alternatively, air temperature may have a strong impact on a key prey species, or on the competitive ability of the focal species, or its interaction with a disease or a predator, and so forth. Statistical, or correlative, modelling approaches capture all such processes but only implicitly – they are hidden in a ‘black box’. One cannot easily open and unpack this box to understand processes driving it. In this issue, however, Oswald et al. (2011) have attempted just that in their study of climatic constraints on the distribution of Northern Hemisphere seabirds.
Seabirds are tricky from a distribution modelling point of view in that they are extremely vagile; many of them forage over vast areas of ocean. For example, the European herring gull, Larus argentatus, forages on average 120 km from its breeding ground (Oswald et al., 2011). However, the distribution of the breeding sites of seabirds is more easily definable and this is where Oswald and colleagues focus their attention. They reason that statistical, climate-based models of the distribution of seabird nesting sites should reflect the direct physiological impact of climate on the birds at those sites, rather than indirect impacts of climate on food resources, because birds forage far from nesting sites.
What are the likely direct effects of climate on seabirds? From a heat regulation perspective, high-latitude seabirds must satisfy a diverse range of conflicting requirements across their lives. They maintain a constant core body temperature of around 40 °C and, like any other endotherm, they respond to cold environments by expending excess energy to produce metabolic heat. The environmental temperature at which an increase in metabolic rate is required depends strongly on the shape, size and insulation of the animal (Porter & Kearney, 2009). The birds considered by Oswald and colleagues range in size from 57 to 1700 g. Whereas large size and spherical shape act to reduce this lower critical temperature, morphological shifts in these directions are relatively constrained in birds because of aerodynamic requirements for efficient flight. While foraging, these birds must have sufficient thermal insulation to withstand substantial heat loss imposed by the cold, windy environments in which they fly; rapid loss of heat to water when they dive for food; and strong evaporative cooling when they return to the wing after a dive. When nesting on land, however, the birds are vulnerable to overheating because they are often fully exposed to solar radiation and are immersed in the hot boundary layer of air near the ground. They are behaviourally constrained because they must stay on the nest to protect the eggs from predation. Under warm conditions the birds must cool themselves through panting, and they or their chicks may run a serious risk of desiccation or death from overheating (Salzman, 1982; McKechnie & Wolf, 2010). There are thus strong, direct linkages between climate and seabird breeding sites.
Oswald and colleagues developed SDMs for 13 European seabirds, deliberately choosing climatic predictor variables that are among those important for processes of heat exchange (air temperature, solar radiation and wind speed). They then assessed performance of their models using a statistic called the ‘area under the receiver operating characteristic curve’ (AUC), which ranges from 0 (model gets all predictions wrong) to 1 (all predictions correct). Based on this statistic, the models fitted well on average (AUC = 0.8), supporting the idea that location of seabird breeding sites is influenced by thermal conditions at those sites. More interestingly, they found that model fit increased as a species’ foraging range increased (AUC ranging from 0.65 to 0.90). This pattern was independent of confounding effects such as body size, breeding habitat availability and phylogeny. It supports their initial assertion that direct (heat balance) and indirect (food) climatic drivers of geographic range become decoupled as foraging becomes more spatially segregated from other parts of the life cycle. The positive relationship between AUC and foraging range cannot unambiguously exclude a role for biotic interactions such as nest predation and parasitism, which may well be climate sensitive. However, the authors argue that such biotic interactions tend to become weaker with latitude, whereas they found that AUC increased with latitude.
Oswald and colleagues’ study brings into focus a number of important issues about understanding climatic constraints on species’ ranges. First, it emphasizes the need to focus on the whole life cycle, because different stages are likely to have different vulnerabilities and sensitivities. This is important whether one is applying a correlative or a mechanistic approach to developing SDMs. The potential range may look different on the basis of distribution points or physiological limits of the adult, juvenile or egg stages. In this sense, the title of Oswald and colleagues’ piece is somewhat perplexingly focused generally on endotherms. Their approach and results would seem most generally applicable to those organisms, endotherms and ectotherms alike, with a strong spatiotemporal separation between different stages of their life-cycle or daily activities; marine turtles or laticaudid sea snakes instantly come to mind.
A second important message of Oswald et al.’s paper is that endotherms can be just as sensitive to the direct impacts of climate as ectotherms. Whereas the body temperature and thus physiological performance of an endotherm remains relatively constant, this homeostatic control may come at significant costs in terms of water or energy, and the requirements for behavioural thermoregulation may also constrain activity, including foraging (Goldstein, 1984; Bozinovic et al., 2000). Finally, Oswald et al.’s study highlights our relatively poor understanding of the direct physiological impacts of climate on endotherms. Their approach for disentangling direct versus indirect impacts of climate on range is itself quite indirect but is a valuable first step that could be combined with mechanistic models of climatic impacts (e.g. Fort et al., 2009) and empirical studies at the breeding sites (Salzman, 1982). To the extent that we can disentangle and understand the diversity of processes by which climate affects organisms, we will be better equipped to predict the consequences of future climate change for biodiversity.