Modelling conservation management options for a southern range-margin population of Golden Plover Pluvialis apricaria vulnerable to climate change




There is considerable interest in understanding how management may help species and populations cope with climate change (climate change adaptation). I used a population model describing the demography of a southern range-margin European Golden Plover Pluvialis apricaria population vulnerable to climate change to assess the potential benefits associated with site-based adaptation management. Two forms of management were simulated: (1) counteracting management to reduce the severity of the negative climate change impacts, simulated by increasing tipulid (cranefly) abundance, and (2) compensatory management to increase populations through an alternative mechanism, simulated by manipulating nest and chick predation rates. A 1 °C rise was estimated to require a doubling of cranefly abundance, or a 35% increase in nest and chick survival rates, to maintain a stable population. For a 2 °C rise, a four-fold increase in craneflies or an 80% increase in survival rates would be required for population stability. A model based on likely realistic estimates of the magnitude of benefit associated with both adaptation management options showed that combined, they may significantly reduce the severity of population decline and risk of extinction associated with a relatively large increase in temperature of 5.8 °C above 1960–90 levels. Site-based adaptation management may therefore increase the resistance of Golden Plovers to some degree of future climate change. This model framework for informing climate change adaptation decisions should be developed for other species and habitats.

There is increasing evidence that bird populations are already being affected by climate change (Jiguet et al. 2007, Green et al. 2008, Gregory et al. 2009), which increases confidence in projections of future population and range changes (Thomas et al. 2004, Jetz et al. 2007, Huntley et al. 2008). Given the magnitude of change anticipated by these projections, there is increasing interest in the potential for management to help species and populations cope with climate change (climate change adaptation). Much of this interest has focused on the potential for increasing connectivity to facilitate the movement of species to track their changing climate (Opdam & Wascher 2004, Vos et al. 2008, Heller & Zavaleta 2009). However, there is considerable uncertainty in the likely efficacy of this approach (e.g. Bailey 2007), which has led to a re-thinking of conservation priorities in the face of climate change (Hodgson et al. 2009, Green & Pearce-Higgins 2010).

One area of adaptation management that has been largely neglected in the literature is that of the potential for management to increase the resistance to climate change of populations at individual sites (Pearce-Higgins et al. 2011). Whilst adaptation to increase the connectivity of populations should facilitate the expansion of a species’ range along the advancing range margin in response to climate change (Opdam & Wascher 2004, Huntley et al. 2007), site management (also termed intensive management or building resistance in the literature; Heller & Zavaleta 2009) is aimed at preventing or delaying the loss of species at the retreating range margin. If such adaptation can increase the persistence of species in an increasingly unfavourable climate, then the loss of species distributions will be less severe than predicted (Pearce-Higgins et al. 2011). To be successful, site-based adaptation management requires both an understanding of the mechanisms by which climate change impacts upon a species, or at least an understanding of the main demographic drivers of population change, and research into the potential for management to reduce the severity of any such negative effects. Thus, there are strong parallels between this proposed framework for site-based adaptation management and the framework for effective conservation of threatened species (Norris 2004). Adaptation management can be separated into two forms: counteracting and compensatory (Green & Pearce-Higgins 2010). Counteracting management reduces the severity of the negative climate change impact, whilst compensatory management does not address the mechanism by which climate change impacts upon a species but attempts to increase productivity or survival rates by an alternative mechanism. This latter approach is increasingly being applied to other conservation issues (e.g. Wilcox & Donlan 2007) and may therefore have merit for climate change adaptation. Both options may be useful in different circumstances, and may potentially give rise to different population responses to climate change, as tested in this paper.

I use the example of the European Golden Plover Pluvialis apricaria (hereafter Golden Plover) to illustrate how such site-based adaptation management may operate in practice. This species has a wide distribution across the tundras, heaths and peatlands of the northern Palaearctic from the UK and Ireland to central Siberia. Recent research on a population in the South Pennines, UK, located close to the southern range margin of the global distribution, has highlighted the potential for increasing summer temperatures to impact detrimentally upon Golden Plover populations through effects on their tipulid (cranefly) prey. During the breeding season, the growth and survival of young chicks is positively correlated with the abundance of emerged craneflies (Pearce-Higgins & Yalden 2004). The number of craneflies that emerge in any year is strongly negatively correlated with the previous summer temperature (Pearce-Higgins et al. 2010), as during a hot summer, the desiccation of the peat surface results in a high mortality of early cranefly larval instars (Coulson 1962). Accordingly, the abundance of the Golden Plovers is negatively correlated with August temperature with a 2-year lag. Thus, a hot summer will result in reduced cranefly emergence in the following year, and hence low Golden Plover productivity, resulting in few recruits and a population decline in the year after (Pearce-Higgins et al. 2010). Whilst climate change is likely to have a wide range of impacts upon particular populations (Mustin et al. 2007), impacts mediated through changing prey populations are likely to be among the most important and widespread (e.g. Sillett et al. 2000, Croxall et al. 2002, Frederiksen et al. 2006). Due to the projected detrimental impacts of climate change and understanding of the likely mechanism underpinning any such effects (Huntley et al. 2007, Pearce-Higgins et al. 2010), Golden Plovers in the UK form a good model system in which to explore the potential for climate change adaptation.

In this paper, I use an existing productivity model to assess how increasing adaptation management intensity may increase the resistance of the population to future climate warming, and what the potential limits to such management may be. Specifically, first, I modelled the potential for different levels of counteracting and compensatory management to alter the relationship between temperature and Golden Plover population growth, comparing the likely population responses to the two managements for a given temperature. This indicates the likely temperature limits to the effectiveness of different levels of both management options. Secondly, I reviewed the existing literature to assess the likely magnitude of counteracting and compensatory management that realistically may be achieved for Golden Plover, based on current knowledge. Thirdly, I modelled the likely population-level consequences of implementing such an adaptation management strategy under a high climate change scenario to assess whether adaptation management may indeed potentially increase the resilience of vulnerable populations to climate change.


Modelling Golden Plover productivity

The demographic model of Pearce-Higgins et al. (2010) was used to examine the potential effects of adaptation management upon Golden Plover productivity in the South Pennines population at Snake Summit (53°26′N, 1°52′W). This model was originally developed to look at phenological mismatch (Pearce-Higgins et al. 2005), and predicts the daily survival of young (< 9 days old) Golden Plover chicks as a function of daily cranefly abundance. These predictions were combined with estimates of nest survival and the daily survival of older Golden Plover chicks to predict the number of fledglings produced per 100 pairs in each year. Predicted productivity correlates with annual variation in Golden Plover population growth (Pearce-Higgins et al. 2010), enabling the model to predict non-linear changes in population growth as a function of August temperature with a 2-year lag (Fig. 1). Although the model also has been used to examine the consequences of phenological mismatch upon productivity (Pearce-Higgins et al. 2005), phenological mismatch was previously found to be unrelated to observed fluctuations in the Golden Plover population (Pearce-Higgins et al. 2010). Furthermore, as the magnitude of predicted declines in productivity as a result of phenological mismatch was much less than that caused by reductions in cranefly abundance (cf. Pearce-Higgins et al. 2005, 2010), I followed Pearce-Higgins et al. (2010) and focused only on this single mechanism. Similarly, I did not account for potential effects of increasing winter temperature on Golden Plover survival rates (Yalden & Pearce-Higgins 1997), as this did not limit the population in the analysis of Pearce-Higgins et al. (2010) and, given predicted warming trends (UKCP09 2010), is unlikely to do so in the future.

Figure 1.

 Schematic of the productivity model used to estimate annual variation in the Snake Summit Golden Plover population (adapted from Pearce-Higgins et al. 2010). In A–C, the thick and thin solid lines represent different levels of cranefly abundance and thick and thin dashed lines represent the frequency distribution of Golden Plover hatch dates for first and replacement clutches respectively. (A) Date is the mean peak emergence date estimated for Snake Summit from 1972 to 2006 from a modelled relationship between emergence date and May temperature. The model used 17 years of data spanning six sites with an r2 of 0.59 (Pearce-Higgins et al. 2005, 2010). (B) First laying dates estimated for Snake Summit from 1972 to 2006 from a modelled relationship between lay-date and March and April temperature and March rainfall. The model used 20 years of data spanning six sites with an r2 of 0.87 (Pearce-Higgins et al. 2005, 2010). Daily nest survival rates equal 0.986, with a mean brood size at hatching of 3.69. Half of failed pairs re-lay 23 days later, with equivalent nest survival rates, but a smaller brood size at hatching of 2.76 (Pearce-Higgins & Yalden 2003). (C) Total cranefly abundance = exp(16.10-0.70T−1) where T−1 is mean daily maximum August temperature in the previous year, with an r2 of 0.69 (Pearce-Higgins et al. 2010). (D) Daily survival (Sd) of young (< 9 days) chicks is a function of daily cranefly abundance (Cd) inline image (Pearce-Higgins & Yalden 2004, Pearce-Higgins et al. 2005, 2010). (E) From Pearce-Higgins and Yalden (2003). (F) The log-ratio of population change between consecutive years = −0.42 − 0.019n−1 + 0.0093P where n−1 is the population in the previous year, to account for density dependence (Yalden & Pearce-Higgins 1997), and P is estimated productivity per 100 pairs. The r2 for this model is 0.33, or 0.49 with a single outlier removed (Pearce-Higgins et al. 2010).

The productivity model was deterministic and did not specifically incorporate uncertainty in the terms within. However, error in the relationship between estimated productivity and population change (Fig. 1, step F) accounted for both uncertainty in the productivity model (Fig. 1, steps A–E) and uncertainty in the relationship between productivity and population change. Similarly, error in the estimate of the magnitude of density dependence was also incorporated in the model predictions at step F, which also included survey error (Freckleton et al. 2006). Model uncertainty may result from either process variation or measurement error, but these were not distinguished.

Modelling adaptation management

Simulations using the productivity model were sequentially run for a range of temperatures from 15 to 24 °C, representing the observed range from 1971 to 2005 of 15–21 °C, plus an additional 3 °C of increase. To simulate varying scenarios of counteracting management, I multiplied the original estimates of cranefly abundance for a given temperature at Snake Summit by 4, 2, 1, 0.5 or 0.25 to mimic 16-fold variation in habitat quality (Fig. 1C). Thus, only absolute changes in cranefly density as a function of habitat condition were modelled; the relationship between cranefly abundance and temperature did not vary in response to management as this has yet to be parameterized. Simulations of compensatory management were achieved by multiplying observed nest and chick survival rates of Pearce-Higgins and Yalden (2003) by 2, 1.5, 1, 0.75 or 0.5 to produce scenarios with a four-fold variation in predation rates (Fig. 1B and 1E). Projected annual changes in population size were plotted for each scenario as a function of the full temperature range, to model the relative effectiveness of both management approaches in influencing the relationship between Golden Plover population change and temperature. By also modelling the impacts of reducing cranefly abundance and survival rates, the results can be applied to other, poorer quality sites to indicate how robust they are likely to be to future climate change.

Estimating realistic magnitudes of management

Literature searches were conducted to indicate what a realistic likely maximum effectiveness of both counteracting management to maximize habitat quality for craneflies and compensatory management to reduce predation rates might be. The former was crudely assessed by examining the magnitude of change in cranefly abundance in relation to habitat condition, whilst the latter was based on variation in observed rates of Golden Plover nest and chick losses at different sites.

Only literature describing variation in cranefly abundance in the Pennines was used, to maximize the relevance of the vegetation data to the South Pennines Golden Plover population modelled. Data were available from four locations, Moor House (54°40′N, 2°20′W), Chapel Fell (54°43′N, 2°12′W) and Widdy Bank (54°40′N, 2°16′W) in the North Pennines and Snake Summit in the South Pennines. At Moor House, the abundance of Tipula subnodicornis was estimated using both core samples to estimate densities of final instar larvae at 13 locations (Coulson 1962) and sticky traps to estimate the abundance of emerged adults in two habitats (Coulson 1959). Weekly counts of emerged craneflies in 1994 and 1995 provided estimates of abundance from Chapel Fell and Widdy Bank (Whittingham et al. 2001). Estimates of cranefly abundance at Snake Summit were derived from two methods: fortnightly counts within pitfall traps in 1996 and weekly sweep net samples in 1998 (Pearce-Higgins & Yalden 2004). Different habitats were sampled in each study, but these were converted into six broad habitat definitions for the purposes of analysis (Table 1). To account for the different methods used to assess cranefly abundance in each study, abundance data were converted into ratios of abundance relative to the maximum recorded in each study, thus producing directly comparable data across all locations. This eliminated the need to specifically correct for differences in abundance between studies, and allowed the results to be simply expressed as relative increases or decreases in abundance between habitats. Corrected abundance was normally distributed between 1 and 0 (Kolmogorov–Smirnov test = 0.13, P = 0.12), and therefore was modelled as a function of habitat in a linear mixed model (proc mixed in sas v9.0; Littell et al. 1996), with the interaction between study and habitat as a random effect, to account for potential methodological differences between studies that might affect the relationship between cranefly abundance and habitat.

Table 1.   How descriptions of varying habitat/dominant plant taxa employed by different studies of variation in cranefly abundance have been combined for the single analysis.
Habitat definitionCoulson (1959)Coulson (1962)Whittingham et al. (2001)Pearce-Higgins and Yalden (2004)
RushJuncus squarrosusJuncus squarrosusJuncus effusus flush 
Sphagnum Sphagnum spp.  
Bog EriophorumEriophorumEriophorum
Bog/heath Calluna – Sphagnum
Calluna – Eriophorum
Sphagnum – Calluna
 Eriophorum – Empetrum
Heath Calluna – CladoniaCallunaEmpetrum – Vaccinium
GrassAlluvial grasslandAlluvial grasslandAcid grassland 

Literature describing daily Golden Plover nest and chick survival rates was collated. Due to a general lack of data to separate daily survival during laying and incubation periods, despite potentially greater rates of predation during laying (Pearce-Higgins & Yalden 2003, MacDonald & Bolton 2008), combined survival rates from laying to hatching are presented. Due to potential biases, data from studies that simply present the percentage of recorded nesting attempts that successfully hatched were excluded. Because the survival of young chicks was explicitly modelled by the productivity model, only variation in the survival of older chicks was examined. However, as the only study presenting such information was that used to construct this model (Pearce-Higgins & Yalden 2003), it was not possible realistically to parameterize the range of survival rates experienced by older chicks. Therefore, only an assessment of the likely effects of predator control upon nest predation rates was made. As studies of predator control on the related Lapwing Vanellus vanellus demonstrate strong effects of predator control on nest survival but equivocal effects on chick survival (Bolton et al. 2007), and because the observed survival rates of older chicks was high (Pearce-Higgins & Yalden 2003), estimates of likely productivity under a maximal predator control scenario are likely to be biologically realistic.

Will realistic magnitudes of management be enough?

Future projections of Golden Plover population trends were made by applying the Golden Plover productivity model to a scenario of future climate change, based on a simple extrapolation of current temperature trends with associated variance calculated from 1972 to 2005 (Pearce-Higgins et al. 2010). The model was used to predict annual abundance of Golden Plovers from 2006 to 2105, from a starting population of 25 pairs. This model was replicated 200 times, with random values of August temperature, the coefficient for density-dependence and predicted productivity. These were selected for each year from a normal distribution with appropriate variance for each variable, to estimate the mean and 95% confidence intervals of these projections, capturing both uncertainty in the model and stochasticity in August temperature. A predicted population that fell below 0.5 pairs was regarded as extinct. The proportion of replicates where extinction occurred provided a measure of likely extinction risk. Two scenarios were modelled. The first was based simply on August temperature increasing by a constant rate of 0.55 °C per decade with no adaptation, and thus replicates the high climate-change scenario of Pearce-Higgins et al. (2010), whilst the second examined the same temperature increase with a scenario of maximal adaptation management intervention outlined below. A scenario of no climate change has already been described by Pearce-Higgins et al. (2010).

For the purposes of this paper, adaptation management was modelled as resulting in an instantaneous increase in both nest survival and cranefly abundance. Whilst the former is reasonably realistic, the latter, mediated through habitat condition, is not. To eliminate this artefact of the assumptions made, the proportional change in populations between the two scenarios was compared after excluding the first year of change (2005–2006). This is equivalent to comparing the projected decline of the Golden Plover population at Snake Summit under an increasing temperature scenario with that of a hypothetical second site with maximal habitat quality and minimal nest predation.

The maximal likely benefit of counteracting adaptation management was estimated from the potential maximum improvement in cranefly abundance that could be gained from changing the vegetation composition at Snake Summit. Although supporting a high breeding density of Golden Plovers, Snake Summit currently comprises 50% bog (Eriophorum spp.), 48% heath (Calluna vulgaris, Empetrum nigrum, Vaccinium myrtillus) and 2% acid grassland (Finney et al. 2005), and therefore does not support maximal cranefly densities. I used the relative cranefly densities in Figure 4 to calculate the proportional improvement in cranefly abundance if habitat at Snake Summit was managed optimally for craneflies, i.e. if it were 100% bog (see below). Estimates of the maximum likely benefit of compensatory management were simply derived from using the lowest observed nest predation rate to simulate a high intensity of predator control (Table 2).

Figure 4.

 Variation in the relative abundance of craneflies in the Pennines in relation to habitat (see Table 1). Habitats are listed in order of decreasing soil moisture based on Hill et al. (2004).

Table 2.   Published estimates of Golden Plover daily nest survival rates.
StudyDaily nest survival rate
Pearce-Higgins and Yalden (2004)0.986
Whittingham et al. (2002)0.958
Byrkjedal (1987)0.945
Pulliainen and Saari (1993)0.996
Crick (1992) 1943–730.990
Crick (1992) 1974–810.984
Crick (1992) 1982–890.992


Modelling adaptation management

Counteracting management was predicted to have little positive impact on population trajectories following a cool August of 15 °C, as the model predicted a high survival rate of young Golden Plover chicks for all scenarios (Fig. 2). This is because cranefly abundance was no longer limiting once it exceeded a particular threshold. Based upon the abundance of craneflies at Snake Summit, the temperature at which the population will be stable was projected to be 18 °C. However, if cranefly abundance is increased four-fold beyond that expected from the models of Pearce-Higgins et al. (2010), then the temperature at which the Golden Plover population was projected to be stable increased to 20 °C, but declined to 16 °C were cranefly abundance a quarter of that expected. At the highest temperatures above 21 °C, there was again a convergence of outcomes between different scenarios of counteracting management because cranefly abundance was low under all management scenarios. Each sequential doubling of the cranefly population was projected to increase the resistance of the population to a 1 °C increase in temperature.

Figure 2.

 Modelled variation in the log-ratio of annual change in a Golden Plover population as a function of mean daily maximum August temperature with a 2-year lag in relation to varying scenarios of cranefly density. The central continuous line is based on the cranefly densities of Pearce-Higgins et al. (2010). Additional lines represent sequential doubling (going up) or halving (going down) of cranefly densities.

The model predicted that compensatory management affects the relationship between Golden Plover population trajectories and temperature in a markedly different way from that of counteracting management (Fig. 3). Thus, by altering levels of predation, compensatory management may have a much larger impact than counteracting management upon projected population trajectories when temperatures are low. Accordingly, a halving of predation rates from that observed at Snake Summit meant that the population was projected to increase or be stable up to August temperatures of 20.4 °C, an increase of 2.4 °C from the baseline. However, any marked increase in the observed predation rate was likely to result in population decline, irrespective of temperature, as predation becomes limiting. Indeed, the magnitude of variation in population change between compensatory management scenarios was much greater than between counteracting management scenarios over such low temperatures. At high temperatures, the model predicted that there was less benefit to be gained from compensatory management altering levels of predation, as the main factor limiting productivity was low cranefly abundance. The magnitude of proportional increases in nest survival rate required to compensate for increases in temperature increased with temperature. Thus a 1 °C rise in temperature from the baseline was projected to require about a 35% increase in nest and chick survival rates, whilst a 2 °C rise was projected to require an 80% increase.

Figure 3.

 Modelled variation in the log-ratio of annual change in a Golden Plover population as a function of mean daily maximum August temperature with a 2-year lag in relation to varying scenarios of nest and chick predation. The central continuous line is based on the predation rates of Pearce-Higgins et al. (2010). Additional lines represent sequential increases or reductions in rates of both nest and chick survival by 1.5 and 2 times.

Estimating realistic magnitudes of management

Across studies, cranefly abundance differed significantly between habitats (F5,8.85 = 7.18, P = 0.006). Pairwise contrasts indicated that there was little consistent difference between rush and bog habitats, but that these supported significantly more craneflies than bog/heath, heath, sphagnum and grass habitats, which did not differ significantly from each other (Fig. 4). The best habitats (rush and bog) therefore supported on average 3.5 times the abundance of craneflies than the worst (bog/heath, heath, Sphagnum and grass). This model suggested that converting the Snake Summit study area to 100% bog vegetation would result in a 39% increase in cranefly abundance, which is taken as the likely maximum magnitude of counteracting management possible at Snake Summit (see Methods).

Published daily survival rates of nests varied from 0.945 to 0.996 (Table 2). Assuming a 5-day laying period, and 30-day incubation (Pearce-Higgins & Yalden 2003), these are equivalent to a range from 14 to 87% hatching success, respectively. The estimate of 0.996 is therefore taken to represent the likely maximum effectiveness of compensatory management in reducing nest predation rates.

Will realistic magnitudes of management be enough?

As expected, the adaptation management scenario, combining both compensatory and counteracting management, contributed to a rapid increase in the Golden Plover population from 2005 to 2006 from 25 to 44.7 pairs (95% CI 13.7–120.7), whilst in the absence of management, the population was projected to remain stable at 25.5 pairs (95% CI 12.3–51.9). These two population trajectories were therefore compared from 2006 in terms of proportional change (Fig. 5). In the absence of adaptation management, the population was projected to decline with a 98% probability of extinction by 2105. The population under climate change adaptation management was largely stable to 2035, but was then projected to decline. By 2105, the two projected trends differed significantly, with only a 6.5% probability of extinction of the population under adaptation management, which averaged 3.0 pairs (95% CI 0.0–10.1).

Figure 5.

 Predicted differences in population trends from a climate change scenario with an increase in August temperature of 5.2 °C by 2105 (solid) and the same scenario, with maximal realistic adaptation management implemented (dotted). Mean values of predictions from 200 randomizations of climate change for each scenario (thick line), standardized to a value of 1 for 2006, are presented along with the 95% confidence intervals (thin lines). Thus a 0.5 proportional change represents a population halving and 2, a population doubling.


Research to assess the potential for management to increase the resistance of vulnerable populations to climate change is urgently required to inform current conservation practice (Pearce-Higgins et al. 2011). This paper makes one of the first attempts to do this. By simulating relatively simple consequences of the manipulation of habitat condition upon insect abundance, the theoretical principle of such management being able to increase the resistance of an insectivorous population of birds to increasing temperature has been supported. This suggests that counteracting management directly to address the mechanism by which climate change is likely to have a detrimental effect on a particular population may successfully reduce the severity of change. Thus, management to increase cranefly abundance is likely to benefit Golden Plover populations experiencing increases in August temperature. Similarly, the contention that varying levels of predator abundance and predator control management may increase the resistance of a population to climate change through compensatory management is also supported. In particular, such management was predicted to have a very large impact upon projected population trajectories at low temperatures because in such circumstances predation is a much more important limiter on Golden Plover productivity than cranefly abundance.

Combined, these two results suggest that in the short term, undertaking compensatory management through predator control is likely to result in the greatest immediate increase in Golden Plover abundance (Fletcher et al. 2010). However, as food availability becomes increasingly limiting as a result of temperature increases, habitat management to maximize food availability is likely to become more important. The validity of these predictions depends, however, upon a number of uncertainties: (1) the robustness of the model; (2) the effectiveness of management; and (3) the severity of future climate change.

First, the model is essentially a correlative one (Pearce-Higgins et al. 2010), which potentially imposes limits on its applicability into the future, although the model is underpinned by detailed ecological understanding (Pearce-Higgins & Yalden 2004). Importantly, it incorporates process error (based on fluctuations in August temperature), which reflects the previous variability in the time-series, as well as measurement error (Pearce-Higgins et al. 2010). Incorporating these errors usefully quantifies some of the uncertainty associated with projecting the future, although not all; the evolutionary adaptability of the species (Visser 2008) and the potential for climate change to impact upon populations in multiple ways (Mustin et al. 2007) remain unknown. These issues are discussed in detail in Pearce-Higgins et al. (2010) and the fact that the model has good predictive ability over a 34-year period does add confidence in relation to its usefulness for projecting into the future, with appropriately wide margins of uncertainty (Fig. 5).

Secondly, the accuracy of the projections about the success of counteracting adaptation management depends upon the assumptions about the effects of management upon cranefly abundance. This is the least certain component of the model, as little information about the effects of land management upon craneflies has been published, and nothing that examines how management may affect the sensitivity of cranefly populations to temperature. To produce plausible estimates of the potential magnitude of management to increase cranefly populations, I have simply modelled cranefly abundance as a function of habitat, as this acts as a crude surrogate for the long-term effects of management upon soil moisture. Craneflies are therefore most abundant on wet sites but are at low densities on the wettest Sphagnum-dominated peatlands (Coulson 1962).

Extensive areas of the UK peatlands have been historically drained or suffered significant erosion as a result of inappropriate management (Bragg & Tallis 2001, Holden et al. 2004), leading to a shift from Sphagnum- and sedge-dominated bog to dwarf shrub- and grass-dominated vegetation (Coulson et al. 1990, Stewart & Lance 1991, Bragg & Tallis 2001), which our model suggests could have reduced cranefly abundance, potentially mimicking or exacerbating the likely effects of summer warming and drought as a result of climate change. Management to restore peatland hydrology and habitat condition (Holden et al. 2004) may therefore reverse these effects. However, there is considerable uncertainty regarding the length of time required for such management to be effective, ranging from 12 years (Wilcock 1979) to in excess of 25 years (Van Seters & Price 2001), which may be most achievable on sites with shallow slopes and low hydraulic conductivity (Holden et al. 2004). Because of these uncertainties, I have not attempted to incorporate a likely time-lag over which peatland restoration will increase cranefly abundance, but instead have simply compared simulated Golden Plover population trends from Snake Summit based upon cranefly densities estimated from its current vegetation with simulated trends from an equivalent site with 100% bog vegetation where cranefly abundance is predicted to be 39% greater. Whilst such a change may be argued as being different from increasing the quality of all habitats, it is currently the only plausible way of deriving an estimate of the potential magnitude of benefit associated with such adaptation management, which is projected to make the population resistant to an increase in August temperature of about 0.4 °C (Fig. 2). Pearce-Higgins et al. (2010) suggested that some aspects of peatland restoration, such as the blocking of drainage ditches, might further increase the resilience of cranefly populations to increasing temperature and alter the slope of the relationship between temperature and cranefly abundance. This has yet to be parameterized but, once done, could be easily incorporated into this model and would be likely to result in a greater apparent benefit of counteracting adaptation management than that suggested here.

The second potential component of management examined was that of compensatory adaptation, mediated through simulated predator control to increase Golden Plover nest survival rates. There is considerable evidence from observations of nest and chick survival (Parr 1992, Pearce-Higgins & Yalden 2003), the spatial association between Golden Plover abundance and grouse moor management (Tharme et al. 2001, Pearce-Higgins et al. 2009), and the results of experimental manipulation (Fletcher et al. 2010) that Golden Plover populations benefit from predator control. Although none of these studies directly relates management intensity to productivity as output by the model (daily nest or chick survival rates, or the number of fledglings per pair), I also produce an estimate of the proportion of pairs successfully raising a fledgling, equivalent to the breeding success measure of Fletcher et al. (2010), who estimated that 18 ± 8% of Golden Plover pairs successfully fledged young on non-predator control sites, but that 75 ± 8% were successful under predator control management. These estimates are reassuringly similar to the 23 and 71% of pairs modelled to successfully fledge young when August temperature is fixed to the 1971–2005 mean of 17.1 °C from the scenarios of the highest and lowest rates of nest and chick predation, respectively. The overall estimate of 69% of pairs fledging chicks under the adaptation management scenario, again with August temperature fixed at 17.1 °C, therefore realistically estimates likely productivity on an intensively managed grouse moor. This means that the implementation of this high level of predator control could result in population stability being achieved at 19.9 °C. This is a 1.9 °C increase on the estimate based on current management (Fig. 3), although as the site is already a grouse moor (Pearce-Higgins & Yalden 2003), to achieve such a rate of increase would require a particularly high intensity of nest predator control.

The third uncertainty about these projections, as with any attempt to look at the future effects of climate change on biodiversity, is over the likely magnitude of future climate change. For the purposes of illustration, I have simply extrapolated the current linear trend for increasing August temperature from 1971 to 2005 into the future, resulting in a projected 5.2 °C rise about the 1971–2005 mean by 2100, or 5.8 °C above 1960–90 levels. This means that the magnitude of climate change examined in this paper is towards the high end of the projections produced by UKCP09, with a 7% probability of being reached under a low emissions scenario, an 18% probability under a medium emissions scenario and a 33% probability under a high emissions scenario (UKCP09 2010). When modelling the potential effectiveness of adaptation management it is appropriate to use a pessimistic scenario to appraise whether management is capable of achieving real long-term benefit and, accordingly, there was a significant benefit simulated to be associated with adaptation management. It is, however, possible to examine the outputs another way, as the Golden Plover population is projected to remain largely stable until about 2035 under the maximal adaptation scenario (Fig. 5; projected population at 2035 = 80% of 2006 level, 95% CI 19–223%). This is equivalent to a 2.0 °C rise in August temperature from 1960 to 1990 levels, emphasizing the need for effective mitigation to reduce the likely severity of future climate change and increase the likely success of adaptation management.

Much modelling of future climate change impacts upon biodiversity has concerned species distributions in relation to spatial variation in climate (e.g. Thomas et al. 2004, Huntley et al. 2007), which provide only a projection of potential large-scale patterns of range change. Correspondingly, the focus of adaptation management from such an approach therefore becomes managing potential shifts in a species’ range (Opdam & Wascher 2004, Vos et al. 2008). This could be seen as management to enable the predictions from bioclimatic models to be realized, by facilitating the colonization of potential new areas of habitat as they become climatically suitable. However, as demonstrated in this paper, adaptation management may also significantly increase the persistence of a species in an increasingly unfavourable climate. Such management therefore in effect prevents the realization of predictions from bioclimatic models, by enabling species ranges to continue to exist in areas from which the models predict they should be lost. Although such an approach is regarded by some as risky and likely to be increasingly costly and challenging to maintain (Heller & Zavaleta 2009, Scott et al. 2010), the results of this modelling exercise for Golden Plover suggest that site-based adaptation management may increase the persistence of species within an increasingly unfavourable climate. Indeed, for some species with small and isolated ranges, this may be the only adaptation option facing conservationists. However, as also indicated by our results, depending upon the future severity of climate change, such management may only be effective for a limited period. Whilst it could be argued that such limited effectiveness renders the management unsuccessful in the long term, it may still provide benefit by increasing the number of potential colonizers for sites further north (Green & Pearce-Higgins 2010). Furthermore, it is difficult at present to judge with certainty which adaptation management strategies are likely to be successful and which are not.

The development of a modelling framework, such as that presented here, may be used as a quantitative basis on which to make such decisions. Given the likely limited conservation resources relative to the magnitude of the conservation problem (Scott et al. 2010), developing robust decision-making processes is likely to become increasingly important. Furthermore, in combination with future monitoring, predictive models may be used to assess when the magnitude of climate change exceeds the ability of adaptation management to infer resistance, indicating when such management is no longer sustainable (Pearce-Higgins in press). On this basis, further research should be conducted to better understand the likely limits of such adaptation, and to extend this approach to a wider range of species. Additional field studies are required to understand more fully the potential consequences of management on bird populations, and particularly whether such management can alter the relationship between climate and species demography.

I am grateful to both RSPB and BTO for supporting this work and particularly to Jeremy Wilson, Stephen Baillie and Sarah Eglington for comments to an earlier version of this manuscript. Many thanks also to Dan Chamberlain and two anonymous referees for many useful suggestions to improve the manuscript. This paper is based on work presented at the 2010 BOU conference Birds and Climate Change.