Analyzing Range Shifts: From Global Change to Local Experiments
Anticipating specific biogeographical consequences of climate change requires understanding how species borders depend on climatic conditions. Our current understanding of this relationship is limited by the âscale gapâ (Root and Schneider 1995). An average increase in global mean temperature is a landscape-scale prediction based on climate models using 500 Ă 500 km grids (Schneider 1993). These climate model predictions apply to âthe big pictureâ of a speciesâ range: large-scale patterns of abundance and distribution. Analyses of distributions at this scale show that combinations of temperature and precipitation criteria are frequently good predictors of speciesâ ranges (e.g., Pigott 1975, Hengeveld 1985, Caughley et al. 1987, Root 1988, Cammell and Knight 1992). This association suggests that a change in environmental conditions may lead to a corresponding change in speciesâ ranges.
Unfortunately, large-scale correspondence cannot differentiate essential and coincidental environmental associations. There are a large number of potential environmental variables that may yield spurious associations with species ranges. Many environmental factors are also correlated with each other in todayâs climate, obfuscating their relative importance for organisms (Dennis 1993). However, paleontological evidence shows that temperature, precipitation, and seasonality all vary independently during climate change (Brubaker 1988, Overpeck et al. 1992). Climates have existed in the past that do not occur today; thus it is likely that future regimes will differ from todayâs. If âno-analogâ conditions result, present ranges do not necessarily predict a speciesâ future distribution. Furthermore, speciesâ response times to climate change can vary by orders of magnitude, leading to strong transient effects of rapid climate change (Davis 1976, Huntley 1991, Lawton 1995, Huntley et al. 1997). Response times are difficult to predict from static distributions due to the important but stochastic role of long-distance dispersal (Pitelka and Plant Migration Workshop 1997, Clark et al. 1998). Therefore, current large-scale associations alone are insufficient for predicting biogeographical consequences of climate change.
Ecological experiments are designed to identify causal mechanisms underlying broader patterns. For logistical reasons most ecological experiments occur on a relatively small scale, studying a single population or locality, usually for a short period of time (Kareiva and Anderson 1988). Unfortunately, experimental results cannot necessarily be extrapolated to larger scales (Levin 1992, Root and Schneider 1995). There are three potential problems with scaling up. First, larger-scale constraints may not be apparent at a smaller spatial or temporal scale or level of organization. For example, a short-term experiment would be unlikely to detect a range limit set by occasional extreme events, or subtle differences in extinction probability. Second, constraints appearing at a small scale may not be very important at a larger scale. We know, for example, that the key factor limiting population growth in one part of a speciesâ range may differ from that in another location (e.g., Pollard 1979, Shaw 1981, Dempster 1983, Thrush et al. 2000). Although there is increasing interest in this problem (e.g., Connolly and Roughgarden 1999, Thrush et al. 2000), it is not yet clear how to predict when certain classes of factors will dominate either within a species or across species and thus how small-scale processes may generate large-scale patterns.
Third, all levels of organization from individual, to population, to species, to multiple speciesâand a corresponding range of spatial and temporal scalesâinteract to shape a speciesâ distribution. For example, physiological limits to tolerance of environmental stress certainly exist and correlate broadly with geographic distribution. However, it is not necessarily clear whether this relationship demonstrates a fundamental physiological constraint in the species, or is contingent on population dynamics that inhibit local adaptation. Genetic variation in stress tolerance exists within natural populations (Hoffmann and Parsons 1991), suggesting that species could adapt to different climatic regimes. Intraspecific comparisons of populations from climatically different regions (e.g., Lamb 1977) complemented by many artificial selection experiments demonstrate that physiological tolerance can evolve rapidly (e.g., White et al. 1970, Tucic 1979, Huey and Kingsolver 1989). However, a comparison of the environmental correlates of the ranges of closely related species suggests that these niches tend to be conserved in evolutionary time (Peterson et al. 1999).
We currently know relatively little about range edge populations and how they may differ from central populations (Hoffmann and Parsons 1991, Hoffmann and Blows 1993). Little is known, for example, about the extent to which marginal populations are locally adapted, whether they are more likely to go extinct than equal-sized central populations, or what percentage of edge populations are demographic sinks, dependent on migration from central populations for persistence. There may be trade-offs between stress tolerance and fitness (Mongold et al. 1996, Bradshaw et al. 1998) such that migration from central-range populations opposes local adaptation (Stearns and Sage 1980, Kirkpatrick and Barton 1997). Thus population-level and species-level processes could interact with a physiological process to define a range limit.
One solution to the scaling problem is to alternate between scales interactively (Root 1991, Root and Schneider 1995). For example, an initial comparative analysis may reveal a pattern consistent with a particular hypothesized mechanistic relationship between factors. This hypothesis should then be explicitly tested at a local level to explore its mechanistic basis (e.g., Gilbert 1980, Muth 1980). Once the mechanism is understood, additional experiments in different localities or with different species can be targeted to more efficiently detect that mechanism at a larger scale and to predict under what circumstances it should be most important (e.g., Root 1991).
The approach I present in this chapter fits into this paradigm at the mechanistic level. After presenting some background information on what we already know about the relationship between butterflies and climate, I will focus on the mechanism controlling the leading edge of an advancing range in the sachem skipper. Understanding the direct impact of an environmental gradient in a natural population along the range edge is a first step toward integrating the individual- and population-level dynamics. Studying a population that may already be responding to rapid environmental change will further provide insight into the transient effects of rapid climate change, and hence the population- and species-level phenomenon of range change. Further experiments involving other locations and species will add to the results of this study and build a better bridge across the scale gap.
Butterfly Ranges and Climate
Research at many scales clearly shows that climate is important for butterfly abundance and distribution. The basic association between climate and insect distribution and abundance has been studied for at least 70 years (e.g., Uvarov 1931, Andrewartha and Birch 1954, Birch 1957, Dennis 1993). The challenge now is to focus on experiments that will integrate our mechanistic understanding of individual and population processes with distribution limits and change.
Large-scale patterns in butterfly distribution are generally consistent with the hypothesis that environmental conditions may constrain the northern boundaries of many speciesâ ranges, although alternative hypotheses have been proposed (see discussion in Dennis 1993). As with many taxa, butterfly species richness characteristically declines along latitudinal and elevational gradients (Scriber 1973, Gieger 1987, Dennis et al. 1991, Sanchez-Rodriguez 1995). Dennis (1993) conducted a multifactorial analysis of species richness within Great Britain based on a systematic 10-km grid database on butterfly abundance (from the Butterfly Monitoring Program, Pollard 1977). Dennis (1993) recorded that butterfly species richness correlates with July temperature (correlation coefficient, r = 0.85), but also with January temperature (r = 0.59), the number of frost-free days (r = 0.58), and precipitation (r =â0.49). This analysis demonstrates a strong correspondence with environmental factors, particularly summer temperature.
If most speciesâ ranges are constrained by temperature, then during a century of warming the prevailing direction of range shifts should be toward higher latitude and elevation...
