Arthropod distribution on an alpine elevational gradient : the relationship with preferred temperature and cold tolerance

The distribution of arthropod species on a 400 m elevational gradient (equivalent to a temperature decrease of 2.5°C) on Snowdon, North Wales, was examined and compared with the British distribution. Preferred temperature, an indication of optimal body temperature (Tb), and supercooling point (SCP), an indication of cold tolerance, of several species on the gradient were deter­ mined experimentally. The alpine beetle species Patrobus assimilis and Nebria rufescens had low preferred Tb, of 5.6 and 7.1°C respectively, whereas the more widespread upland species had higher preferred Tb, between 12.9 and 15.5°C. The SCP ofboth alpine and widespread beetles were similar, being between -6.9 and -5.8°C. The alpine species, which were smaller, were freeze intolerant, whereas the widespread species, which were larger, were freeze tolerant. On the national scale there was significant correlation between preferred Tb and species elevation, but no correlation with SCP. It is concluded that the alpine species survive on Snowdon because their optimal Tb is close to the ambient temperature at the time of day and year when they are active and because they are able to tolerate winter temperatures, by a combination of cold tolerance and shelter. Although a species’ optimal niche will tend to shift upwards as mean temperatures rise with global climatic change, complex microclimatic and biotic factors make changes in dis­ tribution difficult to predict.


INTRODUCTION
The distribution o f a species within its historical range is usually connected with certain, more or less narrowly defined, habitat conditions (Uvarov, 1931); thus, "m oun tain" carabids tend to occur at higher elevations and at northerly latitudes (Luff, 1998).Despite temperature being only one component o f the environmental factors (weather, food, other organisms and a place in which to live (Andrewartha & Birch, 1954)) which interact with the physiology and behaviour o f a species to determine its distribution, there is evidence that increasing global tem perature is resulting in species shifting northward and to higher elevations (Parmesan, 1996;Parmesan et al., 1999).Potential migration rates o f insects are considered to be sufficiently fast to track such changes (Lawton, 1995).
This study investigates the involvement o f preferred temperature, i.e. the temperature selected in the laboratory (e.g.Herter, 1926;Thiele, 1964;Thiele & Lehmann, 1967;Laudien, 1973), and cold tolerance, i.e. the ability to survive at low temperature (e.g.Block, 1990;Lee & Denlinger, 1991;Leather et al., 1993;Somme, 1999), in the elevational distribution o f selected arthropod species.The selected 400 m elevational gradient on Snowdon in North Wales is in the alpine zone, i.e. above the timber line; in Britain, because deforestation occurred c. 2000 years ago (Birks, 1988), this zone above the potential tree line is frequently termed "montane" (Ratcliffe & Thomp son, 1988).The gradient provides a theoretical mean tem perature difference o f 2.5°C between its limits, given a decline o f 0.6°C per 100 m (Mani & Giddings, 1980;Lennon & Turner, 1995).This is similar to the predicted increase in temperature due to global warming in the next 100 years (IPCC, 1996).The elevational distribution is compared with the British distribution o f each species, particularly in relation to elevation and temperature.
Although the preferred temperatures o f m any arthropod species are related to their habitats (Herter, 1953;Thiele, 1977), the reason was unclear and so the measure fell into disrepute.Its relevance is supported by Cossins & Bowler (1987).Because a species' physiological performance is optimal at or near a specific temperature, then body tem perature (Tb) is more relevant than environmental tem perature (Ta).Thus, its preferred Tb and optimal Tb [at which individuals leave m ost descendants (Begon et al., 1990)] are likely to be coincident.Consequently, in an ideal temperature gradient, such as fresh water (Christie & Regier, 1988), an individual will select a Ta that results in a specific Tb.Although a species might have optimal performance over a wide range o f body temperatures, the influence o f biotic factors, such as food availability, pre dation and competition, might result in a narrower niche than expected.Similarly, optimization o f the balance between costs and benefits may necessitate a change, by acclimatization, in preferred Tb from the optimal.
To some extent, arthropods survive winter cold by occupying insulated microhabitats, such as grass tussocks (Luff, 1965) or leaf litter (Danks, 1991); in the mountains, they occupy cavities under stones, or crevices in rocks or in the soil (Mani, 1980).Snow is a particularly important insulator (Strathdee & Bale, 1998).Extreme cold can be survived by freeze avoidance (intolerance) or freeze tolerance (Bale 1993;Sinclair, 1999;Somme, 1999).In the former, freezing, which is lethal, is avoided by supercooling, although non-freezing mortality and sub-lethal injury m ay occur above the supercooling point (SCP) (Strathdee & Bale, 1998).In freeze tolerant species, which are less common, freezing is restricted to the body fluids around the cells, which thereby remain undamaged (Block, 1990).SCP often represents the lower limit o f survival in freeze intolerant species (Somme, 1999) whereas freeze tolerant species may sur vive much lower temperatures.As with preferred Tb, the effect o f a specific temperature will vary w ith the stage in the life cycle and acclimatization.

Study sites
The elevational gradient is on a north-facing, broad shouldered ridge extending from 610 m to 1085 m on Snowdon (Yr Wyddfa) in North Wales.It is bordered by crags and glacial cirques (Campbell & Bowen, 1989) and has areas of bouldery glacial terrain.The mean annual rainfall, recorded nearby between 1945 and 1965, was 3 820 mm (Perkins, 1978).The tree-line is estimated to have been at 635 m (Birks, 1988) and so the main study sites, at 660, 860, 980 and 1055 m, are in the alpine zone.All sites are in "montane" Festuca grassland, which accounts for about 6% of the British alpine zone (Thompson & Brown, 1992).The area is grazed by sheep from about May to early October.Boulders and surface stones provide shelter for arthropods.
Preferred T Preferred Tb was measured on a thermal gradient (-4 to +35oC) along a 100 x 10 cm aluminium bar bedded in polysty rene foam, supporting a double-glazed glass cover 2 cm above the bar.The bar was heated at one end, by ten 10 ohm, 11 watt, ceramic-bodied resistors controlled via a transformer by a vari able voltage source, and cooled at the other by four thermoelec tric, 50-watt cooling modules, attached to heat sinks with the fins force-cooled.Temperatures were recorded by eleven ther mocouples set into the bar at 10 cm intervals and by four sen sors positioned 2 mm above the bar.
Invertebrates collected from the five sites (origins) on the field gradient in June and September 1992 were acclimated at about 8°C (close to mean field temperature) for at least five days.Each experimental "run" involved introducing individuals of one species on to the thermal bar and, after 30 min, recording the position (later converted to its equivalent temperature) at which each had settled.The number of individuals introduced in each run depended on the size and availability of the species; the difference in numbers was taken into account in the analysis.To prevent starting temperature affecting the final temperature selected, approximately equal numbers of "runs" were initiated at each of three positions, labelled A (about 5°C), B (15°C) and C (25°C).The order of testing species and of starting positions were, depending on availability, randomised.The bar and air temperatures were recorded at the start and end of each run.As the species were predominantly nocturnal, and to preclude response to uneven light, all runs were in the dark.
The availability and size of the species resulted in the number of individuals introduced in each "run" varying from one to 17. Adults of C. problematicus, N. rufescens, P. assimilis, N. germi nyi, B. pilula, H. riparius and M. morio, and larvae of C. prob lematicus were used.There was no evidence from preliminary experiments, or runs with differing numbers of individuals, that interactions had a significant effect on the results.Although alu minium is an unnatural substratum for the arthropods, it pro vides a consistent temperature gradient for all species and enables thorough cleaning between runs.

Cold tolerance
Specimens collected from March to September 1994, during twice weekly searches at each site, were retained in the field in small, open-top cages (180 x 180 x 100 mm, floor 50% capil lary matting, containing soil, vegetation and stones, covered in "Agro-fleece" and buried to soil level).They were transferred, at field temperature, to the laboratory in June and September 1994, and in February 1995.
Supercooling was determined by the method of Worland et al. (1992).The arthropod being tested was carefully inserted into a tapering plastic vial, secured near the apex with foam plastic and a thermocouple attached, using petroleum jelly.On inserting the tube into a cooling bath, the monitored temperature was stabilised briefly at 8°C and a controlled rate of cooling of 1°Cmin-1 applied, the preset lower limit being -40°C.An indi vidual was considered to have survived freezing if it showed voluntary mobility for a period of 24 hr following transfer to 0°C (3 hr), 4°(3 hr) and finally 11°C (permanently).

Arthropod distribution
At each site, ten pitfall traps (55 x 55 mm, containing 10 mm of ethanediol) were inserted at 1 m intervals in a row from east to west.A large stone, supported on smaller stones, above each trap prevented disturbance by animals.Weather permitting, the traps were emptied weekly from July to October 1989 and April to October 1990.The limitations of pitfall trapping are well known (e.g.Adis, 1979;Buse, 1988;Halsall & Wratten, 1988;Vandenberghe, 1992), the main advantages being continuous sampling and economy of labour.The influence of differential behaviour and activity were avoided by collecting in one habitat only and by making only intraspecific comparisons.
The British distribution of the relevant species was deter mined from the records of the Biological Records Centre (Centre for Ecology and Hydrology Monks Wood).Presence or absence of each species from the 2763 10 km squares contain ing <98% sea were recorded, together with the mean elevations, mean winter and summer temperature, and mean hours of sun shine in these squares.

Temperature measurements
Because the study spanned several years, site temperature records are intended only to assist in interpreting results, rather than to identify detailed correlations.From mid-July to mid-October 1990, the maximum and the minimum temperature for each week at each site were recorded, within a 0.5 m radius, in shaded turf, in sun-exposed turf, and under a boulder.One maximum and one minimum reading were retrieved for the winter period from November 1990 to March 1991.Throughout 1991, daily changes in temperature were monitored at 15 min intervals at the 660 and 860 m sites, using "Icespy" temperature loggers (Silvertree Engineering Ltd).

Statistical analysis
The statistical analysis in the preferred Tb experiment concen trated on the mean temperature, and its standard deviation, from each run, using a general linear model with origin as a cate gorical variable.Weighted regression, where weights were the numbers of individuals in each run, was used.For each species, the mean preferred Tb from each run was analysed to examine the differences between the origins and any A/B/C effect.The analysis of standard deviation was, obviously, restricted to those runs containing >1 individuals.A comparison of the species for both mean temperature and standard deviation was made by weighted regression, using the mean or the standard deviation from each run.The latter measure is an indication of the spread along the temperature gradient and, as such, is a measure of temperature specificity.The comparison between origins and species was made on the basis ofleast square (predicted) means.
In the supercooling experiments, one-way ANOVA was used to compare the SCPs of the species and to examine the effect of elevation and season on the results.
The significance of the differences in the counts of a species at different elevations on the elevational gradient was tested for each species individually using a Chi-squared test, with a null hypothesis that counts did not differ at each elevation.
Pearson correlation was used to test the relationship between Tb, SCP and the mean elevation, mean summer temperature and mean winter temperature of the British distribution of several species.Spearman's rank correlation was used to test the rela tionship between the temperatures recorded at the field sites.

Preferred Tb
The means of the temperatures which individual species 'select' on the thermal bar differ significantly (p < 0.001) (Table 1).M itopus morio accumulated at the lowest mean temperature (3.4°C) with P. assimilis and N. rufescens several degrees higher.The remainder of the species had higher preferred Tb, with means between 12.9 and 15.5°C.
The point of introduction on to the thermal bar had no significant effect on the final position of the species along it, except for N. germinyi and M. morio (p < 0.05) (mean final positions 8.6, 15.6, 13.5°C and 2.5, 3.6, 5.4°C respectively, when introduced at A, B, C).The site of col lection had little effect on the final position of individuals on the thermal bar, except for summit (1055 m) speci mens of M. morio in warmer temperatures (Table 2): this species also had a more variable response when collected from the two highest sites.

Cold tolerance
The mean supercooling points (SCP) differed signifi cantly between the species tested (Table 3 and 4).The elaterid H. riparius had a particularly low mean of -15.1°C, followed by B. pilula at -8.0°C, but the other

Arthropod distribution
The number of arthropods collected at the 660 m site was disproportionately low (Fig. 1), probably because of a lack of shelter.Differences in the distribution between the sites were significant for all species (p < 0.001).Pterostichus madidus (n = 74 at 660 m) was found solely in the lowest site, whereas P. assimilis and N. rufescens were almost exclusively above it.Notiophilus germinyi and B. pilula gradually increased in numbers from the lowest site, whereas H. riparius decreased from the 860 m site upwards.Carabus problematicus and M. morio were fairly evenly distributed.
The adult occurrence of N. rufescens, B. pilula and H. riparius peaked in late spring, N. germinyi in summer and P. assimilis in both (Fig. 2).The peak of C. problem a ticus was in summer, but it spread into late spring and early autumn.The harvestman M. morio was distributed evenly throughout the sampling period, but included both young and adult stages.
N. rufescens and P. assimilis have high mean elevations in Britain as a whole (Table 5).As these means are for entire km squares, they are lower than they would be for point measurements.Carabus problematicus, N. germi-

Site temperatures
There was significant correlation (p < 0.01) between sites and maximum temperature.Although the higher sites tended to have lower maximum temperatures (Fig. 3), the 860 m site was coldest.The range of maxima between sites, for a particular 14 day period, varied from 10°C in August to 2°C in early October.The minimum temperatures were more in order of elevation, but the highest site (1055 m) was significantly warmer than the second highest (980 m) (p < 0.001).The range varied from about 5 to 1°C.The coldest winter temperature (-6.3°C) was recorded in the second lowest site (860 m).
During the example of a sunny day (Fig. 4a), the maxi mum, minimum and widest range of temperature (16°C)  were in the turf in the sun.The highest night-time, lowest day-time and least range of temperatures were under the boulder: the shade temperatures were similar.The maximum difference at one time was 9°C.On the cloudy day (Fig. 4b), the range was only 3.5 °C; least fluctuation was again under the boulder.

Correlation between preferred Tb, cold tolerance and British distribution
There is a significant correlation between preferred Tb and the mean elevation of the km2 cells in Britain in which the species had been recorded (Table 6).The corre lation between preferred Tb and mean summer and winter  temperatures also tends towards significance (p = 0.08).
There is no correlation between SCP of the species and the environmental factors related to their British distribu tion.As might be expected, winter and summer tempera tures are correlated with elevation.

DISCUSSION
The relationship between the distribution of arthropod species and elevation (Fig. 1) was not clear-cut, partly because the selected elevational gradient was not a simple temperature gradient.One of the lower sites was signifi cantly colder than the higher sites (Fig. 3), probably because of greater wind chill.The Snowdon and British Fig. 3.The maximum temperature (-) and the minimum temperature (-----) recorded at four sites of increasing elevation during fortnightly periods in 1990.The maximum and minimum temperature at each site during the winter of 1990 -1991 are also shown.Recordings were in shaded turf.There was significant correlation between site and maximum and minimum temperatures (p < 0.01 and p < 0.001 respectively).(Lindroth, 1974).
The harvestman M itopus morio is ubiquitous in Britain (thus explaining the lower mean elevation in Table 5), occurring up to 3000 m in Europe (Hillyard & Sankey, 1989).It has, however, various upland varieties, such as M. morio alpinus (Herbst) and, in Britain, M. morio ericaeus Jennings (Hillyard & Sankey, 1989).The alpine N. rufescens and P. assimilis had lower pre ferred Tb (7.1 and 5.6 respectively) than wide-ranging species such as C. problem aticus (15.5°C).K rogerus' (1960) value o f 8°C for N. rufescens suggests that pre ferred Tb is broadly constant for individual species.The lowest m ean preferred Tb (3.4°C) was for M itopus morio; an earlier value o f 11.8°C (Todd, 1949) suggests that it differs between upland and lowland varieties.The alpine species, being crepuscular, were active when the mean site temperatures were 4.5°C and 4.6°C, at 980 m and 1055 m respectively (Fig. 3), close to their preferred Tb.The low mean temperatures for their British distribution (Table 5) were less clear because such records are com piled on a km 2 basis.Although the complex temperature regime precluded significant correlation between pre Table 6.The results of the Pearson correlation between pre ferred Tb, SCP, and environmental parameters of the km2 cells in Britain in which the selected species (see  2), this relationship was significant for Britain as a whole (Table 6), supporting the hypothesis that species at higher elevations had lower preferred Tb than those at lower ele vations.Correlation o f preferred Tb with winter and summer temperatures also tended towards significance.
In winter, survival depends on cold tolerance.Surpris ingly, there was little difference in mean SCP (-5.8 to -8.0°C ) between the smaller alpine and the larger upland species (Table 3).(Adult H. riparius had an exceptionally low SCP o f-1 5 °C , but do not overwinter.)In this study, SCP tended to rise from June to February (Table 4); gen erally it decreases in winter, e.g. by 6°C in the arctic Pterostichus brevicornis (Baust & Miller, 1970).Minimum temperatures (-6°C ) in tu rf in the winter o f 1990 (Fig. 3) were close to the highest SCP.Air tempera tures at the summit during the cold tolerance experiment fell to -8.0°C , the minimum for 1994-2000 being -12°C (J.Williams, pers.com.).Snow is a good insulator (e.g.S0mme, 1982S0mme, , 1999S0mme, : Danks, 2000)), for example, the tem perature at 10 cm soil depth was 1.1°C when air tempera ture was -9.2°C (Danks 1991), but, as it is infrequent on Snowdon, overwintering arthropods are subject to con tinuously changing soil temperature profile as air tem perature changes.The turf provides some protection; Rosenberg (1974) reports the equivalent o f-1 0 °C at 1 cm and -6 °C at 40 cm in bare ground being -7°C and -3°C under turf.The small size o f the freeze intolerant alpine species suggests that they survive on Snowdon because they penetrate the soil.Their slightly lower mean SCP, as normal for freeze intolerant species (S0mme, 1999), adds further protection, but fatalities may occur above this as lethal temperatures above SCP were not investigated.The freeze tolerance o f the larger carabids may contribute to their more widespread distribution.Their lowest lethal temperatures were not measured and so may be below SCP.However, as few British insects are freeze tolerant (Hart & Bale, 1997), further examination o f these large carabids would be worthwhile.
The alpine species survive on Snowdon because o f the lower preferred/optimal Tb at higher elevations, as hypothesised, and their ability to tolerate winter tempera tures.Their response, however, is not solely to tempera ture, but to the whole gamut o f environmental conditions (Cossins & Bowler, 1987), including available shelter and competing/enemy species (Davis et al., 1998).The advan tages o f proximity to the optimal Tb are illustrated by the widespread C. problematicus having a life-cycle o f one year at lower elevations, where ambient temperatures are close to its preferred Tb (c.15°C), but two years at higher elevations where they are lower (Butterfield, 1986;Sparks et al., 1995).The hypothesis that there is greater cold tolerance at higher elevations is rejected as the SCP was similar in both alpine and wide-ranging species.The alpine species were freeze-intolerant, depending on shelter to protect them from cold, whereas the wideranging species were, by the criteria used in this study, freeze-tolerant.The close relationship identified in this study between arthropod species and temperature sug- gests that the expected shift of temperature zones with global climate change will lead to movement of species to higher elevations, with the loss of sensitive species (Brown, 1992).However, it also confirms that current species' distributions, because of their underlying com plexity, are no guide to what they might be under global climate change (Davis et al., 1998).

Fig. 1 .
Fig. 1.The elevational distribution (% combined catch in 1989 and 1990) of arthropod species collected by pitfall trapping at four sites on Snowdon.The complete comparison is significant at P<0.001 in all cases.The number of individuals of each species col lected is shown in parentheses in the key.

Fig. 2 .
Fig. 2. The temporal distribution (% total catch in 1990) of arthropod species collected by pitfall trapping on Snowdon.The number of individuals of each species collected is shown in the key.

Table 1 .
A comparison of the mean temperatures for each species of the final positions on the thermal bar in the tempera ture gradient experiments.n=number of runs contributing to the mean, ( ) = mean number of individuals per run, SE = the simple average of the standard error of the least square means, R2 based on a model including species terms only.°C.Only the SCPs of H. riparius show an elevational effect (Table3), although those of C. problematicus tend towards significance, indi viduals from the middle elevation sites having lower SCPs than those from the summit.The seasonal effect on SCP (Table4) is pronounced in the smaller carabids, N.
rufescens and P. assimilis, and also in B. pilula', the values tend to be lower before than during winter.Using the criterion of voluntary mobility for > 24 hr following freezing, the larger carabids, P. adstrictus, P. aethiops, P. niger and C. problematicus, were freeze-tolerant, whereas the smaller species, H. riparius, B. pilula, N. rufescens and P. assimilis were freeze intolerant.

Table 2 .
The effect of origin (site of collection) on the final position of species on the thermal bar in the temperature gradient experiments.( ) number of runs, temperatures in °C.
nyi and M. morio have lower mean elevations.No British distributional data were available for B. pilula or H. riparius.

Table 3 .
The effect of elevation on the supercooling point (SCP) of various arthropods.( ) = n, temperatures in °C.* = species included in preferred Tb experiment

Table 4 .
The effect of season on the supercooling point (SCP) of various arthropods.( ) = n, temperatures in °C.* = species included in preferred Tb experiment.

Table 5 .
The relationship of some arthropod species with the mean elevation, temperature and sunshine of the km2 cells in Britain in which they had been recorded.All the species are included in the preferred Tb experiment.

Table 5 )
have been recorded.