Will loss of snow cover during climatic warming expose New Zealand alpine plants to increased frost damage?
- First Online:
- Cite this article as:
- Bannister, P., Maegli, T., Dickinson, K.J.M. et al. Oecologia (2005) 144: 245. doi:10.1007/s00442-005-0087-3
- 338 Views
If snow cover in alpine environments were reduced through climatic warming, plants that are normally protected by snow-lie in winter would become exposed to greater extremes of temperature and solar radiation. We examined the annual course of frost resistance of species of native alpine plants from southern New Zealand that are normally buried in snowbanks over winter (Celmisia haastii and Celmisia prorepens) or in sheltered areas that may accumulate snow (Hebe odora) and other species, typical of more exposed areas, that are relatively snow-free (Celmisia viscosa, Poa colensoi, Dracophyllum muscoides). The frost resistance of these principal species was in accord with habitat: those from snowbanks or sheltered areas showed the least frost resistance, whereas species from exposed areas had greater frost resistance throughout the year. P. colensoi had the greatest frost resistance (−32.5°C). All the principal species showed a rapid increase in frost resistance from summer to early winter (February–June) and maximum frost resistance in winter (July–August). The loss of resistance in late winter to early summer (August–December) was most rapid in P. colensoi and D. muscoides. Seasonal frost resistance of the principal species was more strongly related to daylength than to temperature, although all species except C. viscosa were significantly related to temperature when the influence of daylength was accounted for. Measurements of chlorophyll fluorescence indicated that photosynthetic efficiency of the principal species declined with increasing daylength. Levels of frost resistance of the six principal alpine plant species, and others measured during the growing season, were similar to those measured in tropical alpine areas and somewhat more resistant than those recorded in alpine areas of Europe. The potential for frost damage was greatest in spring. The current relationship of frost resistance with daylength is sufficient to prevent damage at any time of year. While warmer temperatures might lower frost resistance, they would also reduce the incidence of frosts, and the incidence of frost damage is unlikely to be altered. The relationship of frost resistance with daylength and temperature potentially provides a means of predicting the responses of alpine plants in response to global warming.
KeywordsFrost resistanceNew ZealandAlpineSnowClimate change
The islands of New Zealand are highly oceanic, but are subjected to cold weather at any time of year, particularly when blocking anticyclones to the west of New Zealand direct southerly flows of cold Antarctic air onto the eastern side of South Island (Brenstrum 1998; Sturman et al. 1999). Consequently, the mountain ranges of this region experience frost and snow at any time (Mark et al. 2000) and alpine plants would be expected to maintain a high level of frost resistance throughout the year. There is relatively little information on the frost resistance of alpine plants, both in New Zealand and in temperate latitudes elsewhere, with most investigations of frost resistance of alpine plants having been concerned with upper limits of forest and scrub. In contrast, the frost resistance of herbaceous alpine plants is relatively neglected (cf. Körner 1999), though, arguably, the frost resistance of tropical alpine plant species have attracted more attention (e.g. Beck et al. 1984; Goldstein et al. 1985; Azocar et al. 1988; Squeo et al. 1991; Lipp et al. 1994; Rundel et al. 1994; Rada et al. 2001).
Prior to the current study, the greatest frost resistance recorded in any New Zealand plants was from species growing at the treeline or in subalpine shrubland. The coniferous shrub, Halocarpus (Dacrydium) bidwillii resisted damage at temperatures below −20°C to −25°C (Wardle and Campbell 1976; Sakai and Wardle 1978; Reitsma 1994); the ericaceous shrub Dracophyllum filifolium showed 50% damage at −22°C (Reitsma 1994); other coniferous shrubs, Phyllocladus alpinus and Podocarpus nivalis (Sakai and Wardle 1978), and the fern, Blechnum penna-marina (Bannister and Fagan 1989) were undamaged at ca. −20°C in winter. Reitsma (1994) provides further records of the frost resistance of woody New Zealand species growing at or above the treeline and for the native tall tussock grass, Chionochloa rubra.
Global warming has been perceived as a threat to the continued survival of some alpine plants, as it is anticipated that they will be driven to higher altitudes as a response and also as a result of competition from more vigorous species migrating from lower altitudes (Körner 1999). Areas suitable for alpine species will be reduced and fragmented in relation to local and regional topography (Halloy and Mark 2003) and patterns of alpine biodiversity will be altered (Grabherr et al. 1994, 1995; Gottfried et al. 1998). A reduction of snow cover would expose plants that are currently protected by snow in winter and spring to greater extremes of temperature and an increased risk of damage by frost, desiccation and insolation (Neuner et al. 1999a). For example, Polwart (Unpublished PhD thesis) found that the frost resistance of the dwarf shrubs, Vaccinium myrtillus and V. vitis-idea, growing at tree-line and covered by snow in winter, had lower frost resistance than those growing on exposed snow-free sites at lower altitudes. Moreover, it is generally considered that species which are normally covered by snow in winter are less frost resistant than those which occupy snow-free environments at the same altitude (Ulmer 1937).
Our hypotheses are as follows: (1) alpine plants that are normally buried by snow in winter will have both lower frost resistance, and a narrower range of frost resistance, than those occupying adjacent areas, which accumulate relatively little snow; (2) compared to more continental environments, alpine plants from southern New Zealand will have relatively high frost resistance during the growing season as cold events can occur at any time during the year; (3) the seasonal course of frost resistance will be more strongly related to daylength than to temperature because of the unpredictability of frost events and warm spells in the alpine environments of southern New Zealand (cf. Mark et al. 2000). Consequently, we measured the seasonal course of frost resistance of selected alpine plant species, determining (a) whether those from beneath snowbanks have lower frost resistance than those from exposed alpine areas; (b) the frost resistance in a variety of other alpine plants during the growing season; and (c) how frost resistance related to environmental temperatures and daylength. We also evaluated the potential effect of increased temperatures on frost resistance of alpine plants.
2 Materials and methods
The principal study site was on the Rock and Pillar Range (45°25′05′′S, 170°05′29′′E, 1,240 m a.s.l) although some plant material was also collected from the alpine zone of the Old Man Range (45°19′35′′S, 169°12′22′′E, 1,650 m a.s.l) and at three altitudes on the Pisa Range principally at the lowest elevation (44°53′06′′S, 169°05′56′′E, 1,520 m a.s.l) with occasional collection at two higher levels (44$52′59′′S, 169°10′44′′E, 1,940 m a.s.l; 44°52′59′′S, 169°08′32′′E, 1,700 m a.s.l). The principal vascular species were three indigenous alpine daisies (Celmisia viscosa, C. haastii (C. haastii var. tomentosa on the Rock and Pillar Range) C. prorepens), a grass (Poa colensoi), and two woody species, Hebe odora (a subalpine shrub) and Dracophyllum muscoides (a prostrate, cushion-forming, ericaceous shrub). C. viscosa occurs in exposed herbfield, but may be found at the edges of early snowbanks; C. prorepens is found in shallow snowbanks and under moderate depths of snow in larger snowbanks, and C. haastii occurs at the greatest depths in snowbanks. Talbot et al. (1992) recorded average annual snow cover of 111, 163 and 199 days, respectively, for the communities dominated by these three Celmisia species on the Rock and Pillar Range. H. odora occurs here in less exposed areas, but may be covered for up to 117 days in winter, whereas D. muscoides and P. colensoi are common in exposed areas with less than 100 days of snow cover (Talbot et al. 1992).
Air temperatures were measured hourly at 1.3-m above ground level with sensors that were protected by a ventilated plastic radiation shield and recorded by data loggers (Hobo H8 Pro Series). The instruments were located at the previously listed sites on the three mountain ranges. On the Old Man and Pisa ranges, plant samples for frost resistance determinations were collected in the vicinity and at the same altitude as the temperature recorders but, on the Rock and Pillar Range, plants were collected at a lower elevation (1,240 m) than the data logger (1,332 m). Minimum temperatures for this site were adjusted using a lapse rate of 0.475°C 100 m−1 (derived from temperature records of loggers from the three sites recording air temperatures on the Pisa range).
Altitude above sea level, annual means of daily air temperature minima (°C) and annual totals of days with air frosts from the Rock and Pillar (R and P), Pisa and Old Man (OM) ranges
R and P
Altitude above sea level (m)
Mean minimum temperature
Highest minimum temperature
Lowest minimum temperature
Days with air frost
The plant material used for the determination of frost resistance was collected from the field sites at intervals between March 2003 (late summer) and June 2004 (midwinter) and transported to the laboratory in an insulated container to minimise overheating in transit. Plant material was collected at the end of the day and either subjected to the freezing programme directly on return to the laboratory or held in a domestic refrigerator at ca. 4°C overnight and subjected to the freezing programme. The sites of collection were between 100 km and 200 km distant from the laboratory (2–4 h travelling time). There was no evident damage to samples as a result of their transit. A variety of other alpine species from the three mountain ranges were collected in late summer (5–18 March 2004). Whole rosettes of Celmisia species, excised shoots of H. odora, D. muscoides and other cushion plants, tillers of grasses, and leaves and rosettes of herbaceous species were tested. Inner and outer leaves of C. viscosa, C. haastii and C. proprepens, younger leaves and buds of H. odora, leaf bases and tips of P. colensoi, and shoot tips of D. muscoides were tested. Three replicates of each species were used in each low temperature treatment. Four programmable, thermostatically controlled, freezers provided a range of temperatures. After initial cooling to a temperature close to freezing, plants were cooled to various target temperatures at a rate of ca. 5°C h−1. They were held for 8 h at the target temperature and allowed to warm to ambient temperature before further treatment (4–7°C h−1, depending on target and ambient temperature). Control samples were held in a domestic refrigerator at ca. 4°C. Plant material was wrapped in dampened paper towels in the dark at room temperature (ca. 20°C). As visual damage was not immediately obvious, damage was assessed after 3 days, using a Teaching-PAM chlorophyll fluorometer (Walz, Effeltrich, Germany), to determine the ratio of variable to maximum fluorescence of the sample (Fv/Fm) of dark-adapted photosynthetic organs (Bolhar-Nordenkampf et al. 1989). The mean of three measurements from each sample was used for further calculations. As completely dead material effectively had a Fv/Fm of zero, percentage damage was calculated as 100(1− Ff/Fmax): where Ff is the Fv/Fm of the frozen sample and Fmax is the maximum value of Fv/Fm for all samples of the tested species. The temperature at 50% damage (LT50) was determined by linear interpolation between the temperature of the highest Fv/Fm under 50% of Fmax and the lowest Fv/Fm over 50% Fmax (cf. Bannister et al. 1995). Extrapolation was used only if the Fv/Fm from the coldest treatment approached 50% of Fmax. On the few occasions when the lowest temperature of a freezing treatment caused no damage to a particular species, that temperature was taken as the best estimate of freezing tolerance. The Fv/Fm of control (unfrozen) dark-adapted samples used in the determination of frost resistance was used as an estimate of Photosystem II efficiency.
As Neuner and Buchner (1999) had found that indices of damage derived from Fv/Fm overestimated winter frost resistance, and considered that visual damage quantified by digital image analysis provided a more reliable estimate of LT50, we tested 58 paired samples of LT50 calculated from damage estimated both visually and using chlorophyll fluorescence (Fv/Fm). A paired t test showed no overall difference between the two sets of estimates. The mean LT50 for visual estimates was −8.0°C, compared with −8.4°C for measurements derived from Fv/Fm (t57=0.40, P=0.69). The correlation between the two estimates was significant r=0.72, P<0.001) and indicated that visual estimates rather than fluorescence measures tended to overestimate frost resistance at lower temperatures. Electrolyte release methods have been tried on New Zealand alpine plant species, but with limited success, as a number of the species tested showed no perceptible electrolyte release (Reitsma 1994).
2.1 Statistical analyses
Multiple regression analyses were used to relate the seasonal variation of frost resistance (LT50), and the Fv/Fm of the unfrozen samples, to the highest minimum temperature (HMT) in the preceding 4 weeks and the daylength on the day of collection (cf. Bannister and Polwart 2001). Daylength on the day of collection was calculated from the date and latitude of particular site, using the algorithm given by Jones (1983). The residuals (i.e. the deviations from the regression lines of frost resistance on daylength) were related to the highest minimum temperature using either linear regression or a non-linear regression equation from Sigma-Plot 9.0 (modified hyperbola III, which has the form y=a−(b/(1 + cx)1/d). Paired t tests of means (Sigma-Plot) were used to compare the daily temperatures found on the Rock and Pillar Range with the Old Man and Pisa ranges, differences between the frost resistance of young and old tissue, and the differences in frost resistance determined by visual and chlorophyll fluorescence estimates of damage.
3.1 Seasonal patterns of frost resistance and Fv/Fm
The relationship of frost resistance (LT50) of older and younger tissue of alpine plants collected between March 2003 (late summer) and June 2004 (midwinter) to daylength (DAY) and the highest minimum temperature (HMT) in the 4 weeks preceding collection
In our seasonal data, the correlation between daylength and HMT was weak (r2=0.06) and non-significant (P>0.05). The effects of HMT were separated from those of daylength by using the residuals from the regression of frost resistance on daylength. These residuals were linearly related to HMT in the Celmisia spp. and H. odora, but the rate of change of frost resistance with HMT was either small (<0.5°C °C−1 increase in HMT) or non-significant (in C. prorepens). In both D. muscoides and P. colensoi, the regression of the residuals against daylength produced a hyperbolic curve, which showed little change in frost resistance when the HMT exceeded 4°C, and a rapid increase in frost resistance at when the HMT was below 4°C (Fig. 6). All these regressions indicate that frost resistance decreases with increased HMT and increases with decreased HMT. The multiple regressions for Dracophyllum and Poa (Table 2) used the reciprocal of HMT (1/HMT), to account for this curvilinear relationship. The residuals (the deviations from the regression line of frost resistance on HMT) all showed positive linear relationships with daylength that were significant for all species except C. viscosa (data not presented).
The pattern of frost resistance from summer to winter was strikingly similar in both years (2003 and 2004) and indicates a strong influence of photoperiod on the development of frost resistance (Fig. 3). In the Celmisia spp. and Dracophyllum, analyses of co-variance showed no differences in either slope or elevation (P>0.05) of the relation between frost resistance and date in the 2 years. The slopes for Hebe and Poa showed significant differences in elevation indicating that frost resistance of these species in the first half of 2004 (February–June) was greater than that in 2003 (P=0.05 and P<0.02, respectively).
Maximum frost resistance of the principal species ranged between −17°C and −33°C and their minimum resistance ranged between −8°C and −13°C. The range (maximum to minimum) of frost resistance for the six principal species varied from 8.5°C in C. prorepens to 13.5°C in D. muscoides, and 24.8°C in P. colensoi (Table 4).
Measurements of frost resistance, over the period of study, showed that all species developed a level of frost resistance that enabled them to survive the coldest prevailing air frosts. The difference between minimum air temperature and the species’ frost resistance was least in spring (October–December) and autumn (March); hence, the risk of frost damage was greatest at these times (Fig. 7).
3.2 Frost resistance during the growing season
Frost resistance (°C) of alpine plant species from the Rock and Pillar (R and P) Old Man (OM) and Pisa ranges during late summer (5–18 March 2004)
R and P
3.3 Effects of snow cover
Seasonal variation in minimum, maximum and range of frost resistance (°C) of old (o) and young (y) tissue from alpine plants collected from the Rock and Pillar, Old Man and Pisa ranges
Seasonal variation in frost resistance
Species from sheltered habitats
Species from exposed habitats
Comparison of frost resistance (FR) and Fv/Fm of unfrozen shoots or leaves of Celmisia viscosa, Hebe odora, Dracophyllum muscoides, and Poa colensoi buried beneath and exposed above snow during winter (July–August) and spring 2003 (September–October)
30 Jul 2003
6 Aug 2003
16 Sep 2003
15 Oct 2003
20 Oct 2003
4.1 Seasonal course of frost resistance
The regression equations in Table 2 indicate that the seasonal course of frost resistance is determined by both daylength and temperature (HMT). Frost resistance increased with shortening days and lower temperatures. The lack of a significant correlation between seasonal temperature and daylength reflects the vagaries of an oceanic climate where cold periods often occur in summer and warm spells are not infrequent in winter. Consequently, daylength is a more reliable predictor of season than temperature and appears to be the most important factor in controlling the annual cycle of frost resistance (Table 2). The frost resistance of each species in Table 2 was significantly related to photoperiod, most strongly in younger tissue, and the frost resistance of each species also showed some significant relationship with temperature (in young material of C. viscosa, C. haastii and D. muscoides and older material of C. prorepens, H. odora and P. colensoi). Schwarz (1970) found that photoperiod, rather than temperature, was the dominant influence in determining the seasonal course of frost resistance of woody species (Pinus cembra and Rhododendron ferrugineum) at the timberline in Austria, although Neuner et al. (1999b) concluded that temperature was the more dominant influence in R. ferrugineum. Senescent and older tissues are usually less frost resistant than younger tissues and appear to lose frost resistance readily in warmer temperatures, which accelerate senescence of older leaves. This feature was particularly noticeable during the growing season (e.g., in the outer leaves of C. prorepens and older leaves of P. colensoi) and the decreased frost resistance at higher temperatures accounts for the wider range of frost resistance that occurs in older tissues (Table 4).
The pattern of frost resistance from summer to winter (Fig. 3) was remarkably similar in 2 years (2003–2004). Neuner et al. (1999b) also noted a high degree of consistency between eight annual records of the course of frost resistance in R. ferrugineum made over the last 60 years.
4.2 Comparisons with plants from other alpine areas
In any comparisons of frost resistance of species, the use of different critical levels of damage and a variety of techniques must be taken into account. We used the temperature causing 50% damage (LT50) as our measure of frost resistance. Other authors have used the lowest temperature at which plants escape damage (LT0, e.g. Sakai and Wardle 1978), or suffer initial damage (LTi, e.g. Warrington and Southward 1995), or the highest temperature causing complete damage (LT100, e.g. Wardle and Campbell 1976). Both LT0 and LT100 are difficult to determine because of the sigmoid relationship between temperature and damage, which produces extensive ranges of higher and lower temperatures that cause either no damage or complete damage. In contrast, there is only a narrow range of temperatures between LT0 and LT100, which allows a relatively precise determination of LT50. Consequently, we used the temperature causing 50% damage (LT50) as our basis for comparison.
We used chlorophyll fluorescence (Fv/Fm) of leaves from cut shoots to estimate frost resistance and found no significant difference in LT50values derived from visual estimates of damage or chlorophyll fluorescence (see Methods). Visual assessment of damage was difficult in many species. Leaves of C. viscosa are dark green, whereas leaves of P. colensoi retain their bright green colour, and leaves on winter shoots of D. muscoides are entirely brown (Calluna vulgaris shows a similar “winter browning”, Bannister 1964). A lack of visual changes makes it difficult to judge progressive damage and hence, frost resistance is overestimated, but chlorophyll fluorescence of the same samples shows progressive changes in Fv/Fm . Many of the species examined are rosette or cushion plants in which the older leaves are senescent or dead so that visual assessment of damage in older tissue is difficult. For this reason, we have concentrated on estimating frost resistance in younger tissues.
We measured frost resistance of leaves from excised shoots. Pellet et al. (1981) concluded that frost resistance measured on cut shoots overestimated that determined from observations of damage in the field, whereas Bannister (2003) found that 75% of frost resistance determinations (made on cut shoots) were consistent with observed damage in the field. In contrast, Taschler et al. (2004) and Taschler and Neuner (2004) considered that measurements of leaves from cut shoots produced lower estimates of frost resistance than leaves of alpine plants frozen in situ. However, some of their evidence was based on comparisons of their in situ measurements of frost resistance with published data using cut shoots (e.g. Table 2 of Taschler et al. 2004; and Table 3 of Taschler and Neuner 2004). The only valid comparison is a parallel, contemporaneous, investigation of the frost resistance of cut shoots and attached shoots from the same plant. For example, Neuner et al. (1997) compared in situ determinations of the frost resistance of leaves of Nothofagus menziesii with those of excised shoots and found that the frost resistance of leaves from the latter was 1°C greater than leaves in situ. Taschler et al. (2004) conducted similar investigations on excised and attached developing shoots of Picea abies and found no significant differences between LT50 values for either resting or swelling buds, but greater frost resistance in breaking buds, and lesser frost resistance in expanding leaves from excised shoots. These differences in LT50 are only slight (<1°C) and inconsistent and, consequently, we consider that our estimates of frost resistance can be validly compared with those of other workers. However, there is a major difference in ice nucleation temperatures as water freezes at a higher temperature in situ than in excised shoots (Neuner et al. 1997; Taschler et al. 2004). Small excised plant parts have a greater capacity to supercool than large samples or intact plants (Robberecht and Juntila 1992) and it has been suggested that freezing at higher temperatures is due to an increased probability of encountering active ice nuclei in large samples or intact plants (Flinn and Ashworth 1994).
Taschler and Neuner (2004) found that the frost resistance (LT50) in a range of alpine species collected from the Austrian Alps in the growing season ranged from −4°C to −15°C (median −7°C), whereas the equivalent range in New Zealand (Table 3) was −8 to −18°C (median −9°C). The biggest differences were for herbaceous plants (−5 to −12 in Austria and −8 to −18 in NZ). Körner (1999) has listed previously unpublished data on frost resistance of alpine plants during the growing season at 2,470 m on the Furka Pass (Switzerland): herbaceous dicots had frost resistance (LT50) of about −7°C, while in graminoids it was at least −10°C (compared with −14°C in New Zealand). The frost resistance of the New Zealand plants was greater than that of plants from the Austrian and Swiss sites, which were collected at higher altitude and latitude than the New Zealand sites. The greater frost resistance of the New Zealand species during the growing season is consistent with our initial hypothesis, which posited that the prevalence of cold Antarctic air masses at any time of year would select for relatively high frost resistance in the alpine species here.
The maximum winter frost resistance in leaf bases of P. colensoi (−32.5°C) is somewhat less than that of grasses from dry alpine grassland in the Austrian Alps (e.g. Stipa capillata −37°C, S. eriocaulis −36°C, Larcher et al. 1989) but greater than that of the maritime Antarctic grass, Deschampsia antarctica (−27°C, Bravo et al. 2001). Alpine shrubs and dwarf shrubs from New Zealand are less resistant than their European counterparts: the maximum frost resistance of H. odora was −19°C and −22°C for D. muscoides, whereas the maximum frost resistance for R. ferrugineum in Austria was −30°C (Neuner et al. 1999b) and −29°C for Calluna vulgaris (Ulmer 1937). Polylepis tarapacana, grows in Bolivia at the highest altitudes (4,300–4,850 m, 18°7′S, 68°57′W) for any tree in the world (Rada et al. 2001) and has a frost resistance of −18 to −23°C, similar that of New Zealand alpine shrubs in winter.
Neuner et al. (1999a) measured Fv/Fm in plants of R. ferrugineum in winter and found that photo-inhibited leaves recovered their photosynthetic efficiency at higher temperatures (≥10°C) and low light intensities. The values of Fv/Fm for the unfrozen control plants used in our determinations of frost resistance had recovered their photosynthetic efficiency from field levels, and their seasonal values tended to decrease with increasing daylength—indicating greater photoinhibition in the longer days and higher light intensities of summer. In contrast, Neuner et al. (1999a) found that the highest values of Fv/Fm occurred in the long days of summer. Some control plants retained low values of Fv/Fm (e.g. D. muscoides on 28 August at high altitude on the Pisa Range had an Fv/Fm of 0.418) and their condition indicated physical damage due to frost, freeze-thaw cycles or desiccation.
New Zealand alpine species have a wide seasonal range of frost resistance (Table 4) in contrast to native New Zealand forest species. The seasonal amplitude of frost resistance in younger tissues of alpine species (other than P. colensoi) was between 9°C and 14°C: P. colensoi had a seasonal amplitude of 25°C. The maritime Antarctic grass, Deschampsia antarctica, increased its frost resistance by 15°C after 21 days of acclimation in long days at 4°C and attained a frost resistance of −27°C (Bravo et al. 2001), which is similar to the values obtained by P. colensoi in the New Zealand winter. In contrast, seven lowland forest species (Pittosporum eugenioides, Carpodetus serratus, Pseudopanax crassifolius, Plagianthus regius, Hoheria populnea, Streblus microphyllus and Sophora microphylla) have low seasonal maximum and minimum frost resistances (minima −1°C to −6°C; maxima −6°C to −12°C) and a narrower seasonal range, from 3.5°C in P. to 6.5°C in C. serratus (Darrow et al. 2001).
The frost resistance of New Zealand alpine plants during the growing season (Table 3) is similar to that of tropical alpine plants. For example, the LT50 of leaves of the alpine plants from the Venezuelan Andes (8°37′N, 70°12′W) ranged between −8°C and −19°C (median −12°C) (Squeo et al. 1991), whereas that of alpine plants growing at 3,200 m in the northern Chilean Andes (29°45′S, 69°59′W) ranged from −4.7°C to −20°C (median −13.3°C) (Squeo et al. 1996). The Hawaiian alpine mega-herb, Agyroxiphium sandwicense (silversword), had a frost resistance of −15°C (Goldstein et al. 1996) while the Afro-alpine mega-herbs Senecio keniodendron and S. keniensis had frost resistances of −14°C and −10°C respectively, and the giant lobelias (L. telekii and L. keniensis) resisted temperatures as low as −20°C (Beck et al. 1982). Other herbaceous Afro-alpine species had frost tolerance between −13°C and −15°C (median −14°C) (Beck 1994). Troll (1968), Halloy and Mark (1996), Mark et al. (2000, 2001) and Wardle (1998) have noted affinities between the vegetation, life forms and climate of the alpine zones in New Zealand and tropical high mountains, and the similarities in frost resistance appear to be another characteristic that these alpine floras have in common.
The current study was supported by University of Otago Research Grants in 2003 and 2004. The Department of Conservation granted permits for plant collection and experimental plots in areas under their control. We thank Norman Mason, Stewart Bell, Rob Daly and Laura Harrison for their assistance, particularly during winter; J and E O’Connell for access through their property on the Rock and Pillar Range; and J and M Lee for access and use of sites on the Waiorau Snowfarm on the Pisa Range.