Climate and elevational range of a South African dragonfly assemblage Contribution to EU / ALARM Project

Elevation and climate are interrelated variables which have a profound aff ect on biota. Flying insects such as dragonfl ies can rapidly disperse and select optimal habitat conditions at appropriate elevations. Such behaviour is likely to be especially important in geographical areas which are subject to major climatic events such as El Niño. Accordingly, we studied dragonfl ies and environmental variables in a series of reservoirs over an elevational range of 100–1350 m a.s.l. at the same latitude on the eastern seaboard of South Africa. Th e aim was to determine how elevation and climate (as regional processes), as well as local factors, infl uence species assemblage variability, habitat preference and phenology. Certain environmental variables strongly explained the main variation in species assemblage. Th ese included local factors such as pH, marginal grasses, percentage shade, exposed rock, marginal forest and to a lesser extent, marshes and fl ow. Diff erent species showed various tolerance levels to these variables. Elevation and climate as regional processes had very little infl uence on dragonfl y assemblages in comparison with these environmental factors. Th ese odonate species are essentially sub-tropical, and are similar to their tropical counterparts in that they have long fl ight periods with overlapping generations. Yet they also have temperate characteristics such as over-wintering mostly as larvae. Th ese results indicate evolutionary adaptations from both temperate and tropical regions. Furthermore, most were also widespread and opportunistic habitat generalists. Th e national endemics Pseudagrion citricola and Africallagma sapphirinum only occurred at high elevations. However, the endemic Agriocnemis falcifera was throughout all elevations, suggesting regional endemism does not necessarily equate to elevational intolerance. Overall, the results suggest that many millennia of great climatic variation have led to a highly vagile and elevation-tolerant dragonfl y assemblage which readily occupies new water bodies. Such an assemblage is likely to be highly tolerant of global climate change, so long as there is suffi cient water to keep the reservoirs at a constant level. BioRisk 5: 85–107 (2010) doi: 10.3897/biorisk.5.844 http://biorisk-journal.com/ Copyright M.J. Samways, A.S. Niba. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. RESEARCH ARTICLE BioRisk A peer-reviewed open-access journal


Introduction
Insect phenology usually varies with topography and associated environmental factors (Wolda 1987).Interactions between temperature-dependent development and microclimate are important features of insect life-history, leading to the maintenance of considerable genetic variation in populations (Bradshaw and Holzapfel 1990;Roff 1990).Studies in insects and other arthropods suggest that microclimatic gradients sometimes can have larger eff ects on emergence phenology than do annual fl uctuations in weather conditions (e.g.Kingsolver 1979;Weiss et al. 1993).Moreover, fi eld evidence (Th omas et al. 2001) supports theoretical predictions (Th omas et al. 1999) that certain types of thermophilous insects have expanded to occupy broader niches, and hence larger patch sizes near their northern range margins in the northern hemisphere during some warm summers in recent years (Ott, this volume).
Small reservoirs are a characteristic feature of the South African agricultural landscape, acting as important reserves for dragonfl ies (Samways 1989a).Th ese reservoirs have been shown to be important in promoting the conservation of insect diversity, but mostly of generalist species (Samways and Steytler 1996).Such reservoirs increase the area of occupancy of the local species.Th us, they are present in the local area in natural water bodies and simply move across to the reservoirs.
Th e topography of KwaZulu-Natal, South Africa ranges in elevation from 0-3000 m a.s.l along a 200 km E-W transect at one latitude.Th is area is strongly modifi ed by montane climate at higher elevations, and has a sub-tropical/tropical climate at sea level.Th e study area within KwaZulu-Natal is situated at the edge of a major escarpment comprising a highly heterogeneous landscape structure with a wide variety of aquatic habitats.Th is elevational transect supports a high diversity of dragonfl y species making up close to three-quarters of the South African odonate fauna.Th is provides a basis for measuring how species phenologies and distribution respond to the seasonal (temporal) and elevational changes.Th is information can be useful for subsequent conservation action, and for providing baseline data for future studies on the impacts of global climate change.
Using a series of fi ve moderately-sized artifi cial, but well-established reservoirs (which only reach about 1400 m a.s.l.), the aim here was to determine the extent to which elevation (as a regional process), alongside local factors, infl uence habitat preferences and species distribution.Furthermore, as there is no information on the eff ects of seasonal changes on southern African odonate species, the aim was also to determine how phenology might vary with elevation.

Study area
Th e study area was in KwaZulu-Natal between the coast and the Drakensberg escarpment (<3000 m a.s.l.).Elevation exerts a major infl uence on climatic features at all spatial scales, being a barrier to rain-bearing air masses, and by altering temperature through lapse rates and aspect (Tyson 1986;Schulze 1997).
Reservoirs, all of which were over 30 years in age, were selected within this elevational gradient (Fig. 1) to be at the same latitude (with 26 minutes latitude) and to be relatively comparable (Table 1).Th e maximum elevation that could be entertained for these comparative studies was 1400 m a.s.l, even though the mountain peaks reached 3000 m a.s.l.

Methods
Each reservoir was about 1 ha, and stratifi ed into six sub-sites, each measuring 20 m length (along a line transect on the reservoir edge) by 2 m width (1 m on land and 1 m into water).Data were collected on 42 sampling occasions, and covered various stages of dragonfl y development (adults, tenerals and young adults (together here simply called 'tenerals'), larvae and exuviae).Mating or oviposition were also recorded (mostly tandem fl ights and occasional dipping of ovipositors).Environmental variables were recorded twice a month from January 2001 to December 2002, except for the winter months of June, July and August when data were collected once each month.
Adult males were recorded using close-focus binoculars, and walking along the 20 m sub-sites and counting within 6 min all individuals perching or fl ying.Counts of Anisoptera at sub-sites can be virtually 100% accurate and that of Zygoptera exceeds 80% (Moore 1991).Counts were between 10h00 and 14h00 during sunny, high activity periods of the day.
Exuviae and tenerals were recorded as an indication of successful breeding.In this study, population changes were indicated by comparing the maximum numbers of individuals (adults, tenerals and larvae) observed each month for the whole sampling period.Unidentifi ed tenerals from the fi eld were collected and reared in the laboratory until their body colour (with genitalia morphology) could be used for subsequent identifi cation.
Larvae were sampled with a dip-net (41 cm diameter × 1 mm mesh sieve).Two dips per sub-site (12 dips/site) were done within 20 min.Each dip was followed by vigorously shoving the net back and forth in water once among water weeds, along rushes and besides banks.We fully accept that no single, quantitative collection method is equally effi cient for all species of larvae, and even all ages.However, the comparative effi ciency of the collection method, being standardized, will be the same at diff erent sites.Individual larvae in the net were identifi ed using a 9× hand lens, counted and released back into water except where individuals could not be identifi ed in the fi eld, in which case, they were picked out with soft, fl exible forceps and placed in aerated plastic cages containing reservoir water.Usually only last-instar larvae were collected for subsequent rearing and identifi cation in the laboratory.
Marginal vegetation (both structural and compositional) was estimated using percentages of sub-sites they covered.At all sub-sites, aquatic plants were recorded as: marginal forest stands (Mfor), marginal grasses (Mgra), fl oating and submerged vegetation (Fsv), marginal herbs, sedges and reeds (Mhsr).
Meteorological data e.g.rainfall, ambient and water temperatures (At/Wt) collected at Goodhope Estate (GH), Cedara (CE) and the Botanical Gardens (BG) were compared with that collated by the weather bureau at Cedara Agricultural College.Also, rainfall and temperature data for Krantzkloof (KL) and Stainbank (SB) Nature Reserves were compared with that collated by the Durban Airport weather station.
Data were analysed with univariate methods for species richness and abundance relationships using diversity indices, distributional models and graphical methods.Species spatial and temporal variability was analysed using Analysis of Variance (ANOVA).Spearman's rank correlation coeffi cients were used to measure the association between variables and species abundance and richness.Th ese correlations were calculated using the software SPSS version 6.1.MINITAB and SPSS software were used to run ANOVA, relating species to sites and site variables.In addition to ANO-VA, Similarity coeffi cients calculated between every pair of samples helped facilitate a classifi cation or clustering of samples into groups which are mutually similar or an ordination plot in which the samples are 'mapped' into multidimensional space in such a way that the distances between pairs of samples refl ect their relative dissimilarity of species composition.
Hierarchical agglomerative clustering, using the program 'Cluster' in the computer software PRIMER (Clark and Warwick 1994) was used to compare sites.Th e species by sub-site (SS) data matrix was transformed using 4 th root-transformation to balance rarer and commoner species.Th e Bray-Curtis similarity index was then used to produce a similarity matrix and then fused successively through hierarchical clustering using group-average linking, to produce a dendrogram with the x-axis defi ning a similarity level at which two samples or groups are considered to have fused, and the y-axis representing the full set of samples.
Correspondence analysis (CA), operates on a site and species data matrix and represents it on a two-dimensional plane (ter Braak and Smilauer 1998).It uses a site-byspecies scores data matrix and summarises it such that increasing distance between the sites on the ordination plane means decreasing similarity in the species assemblages at the respective sites.Conversely, from a species-by-site matrix, CA ordinates the data such that the closer two species are to one another on the same ordination plane, the greater the likelihood that they will occur at the same or similar sites and vice versa.Canonical Correspondence Analysis (CCA) was used to relate species and site scores to underlying environmental variables.Th e length of an arrow representing an environmental variable is equal to the rate of change in the weighted average as inferred from the bi/triplot, and is therefore a measure of how much the species distribution diff ers along that environmental gradient.Important environmental gradients therefore tend to be represented by longer arrows than less important ones (ter Braak and Looman 1995).Th e software CANOCO version 4 and CANODRAW version 3.1 (ter Braak and Smilauer 1998) were used.

Species phenology
A total of 47 species was recorded throughout the study (Table 2).Adults of only three species (Ceriagrion glabrum, Lestes plagiatus and Crocothemis erythraea ) were recorded during winter, and then only at Mid (BG) elevation.Accumulation curves reached asymptotes for tenerals with 10-14 species, and for adults with 21-25 species, and varied with elevation (Fig. 2).

Relative proportions of adults, tenerals and larvae
Larvae stayed at about the same level all year round (Fig. 3).Tenerals and adults showed the same trends as in Fig. 2 i.e. none in July and August.However, there was trend for maximum numbers to be reached later at higher elevations, from October to December for tenerals and November to February for adults.Larval abundance varied from 20 individuals in January at Mid-low (KL) elevation to 138 in April at Mid-high (CE) elevation.Teneral counts also varied from two individuals in June at High (GH) elevation to 175 individuals in November at Mid-high (CE) elevation.Th ereafter, larval abundance at all elevations was high in November for both years.No teneral individuals were recorded at any elevations during winter (July to August).Adult abundance was greatest in November for both years and at all sites except at High (GH) and Midlow (KL) where it was in December.

Peak occurrence periods
Th ere was continual emergence over the summer months, and there was continuous presence of two or three developmental stages between September and June.Lestes plagiatus and L. tridens probably over-wintered in the egg stage.Table 3 summarises the months for peaks in adult, teneral, larval stages, and mating/oviposition in Anisoptera and Zygoptera species.Anisoptera adults from High (GH), Mid-high (CE) and Low (SB) elevations had peak occurrences mostly from December to March, although most species peaked in November during the fi rst sampling year at the High (GH) elevation.Species peaks in Mid (BG) and Mid-low (KL) elevation were also similar, occurring in November in both sampling years.Fifteen species occurred at all fi ve elevations, while 17 species were restricted to only one elevation: eight in Low (SB); four each at High (GH) and Mid (BG) and one only occurred at Mid-low (KL).15 species occurred over at least two elevations at all fi ve elevations (Table 4).
Th e dominant species at the Low (SB) elevation site was L. tridens (22%), while T. stictica dominated in Mid-high (CE) elevation.Both elevations had relatively high percentage levels of species dominance patterns compared to the other elevations.Mid-low (KL) elevation and High (GH) elevation showed some similarity in patterns of species dominance, with T. arteriosa (17%) and T. stictica (18%) being the dominant species.

Spatial variations in adults, tenerals and larvae with elevation
Larval species richness and abundance was highest at Mid-high (CE) elevation.Patterns of teneral species richness across elevations ranged between 14 and 16 species per elevation during the study, with Mid-low (KL) recording lowest individual counts.Overall number of adults species varied slightly across elevations, with Low (SB) eleva- tion supporting the most species.Adult abundance was highly variable across elevations, with Mid-low elevation (KL) recording lowest abundance.Larval species richness was signifi cantly positively correlated with elevation (F = 19.25;P = 0.002), as was abundance (F= 7.69; P=0.024).Teneral species richness was negatively correlated with elevation but not statistically signifi cantly.Th ere was weak, non-signifi cant positive correlation for teneral individuals with elevation (F = 4.73; P= 0.056).Regressions of adult dragonfl y species richness (P=0.27) and abundance (P=0.32) on elevation were not statistically signifi cant even though there was a generally decreasing trend in species as elevation increased.

Relationship between species and environmental conditions
Species associations with elevation were strongest on ordination plots when all Odonata were separated into their component sub-orders (Anisoptera and Zygoptera).CA results for Anisoptera (Fig. 4a) showed most open water species clumped at the centre of the ordination.Zygoptera species showed various trends as species were more dispersed from the centre of the ordination (Fig. 4b).Th ey were more tolerant of diverse conditions of shade as well as of open water.Separate CCA ordinations were also run for species belonging to Anisoptera and Zygoptera again for better interpretation of the eff ects of measured variables and elevation on patterns of dragonfl y assemblage composition and distribution.Species-site-variable triplots for Anisoptera (Fig. 5a) and Zygoptera (Fig. 5b) showed that most assemblages were related to a number of environmental variables, and indicated how species responded or not to gradients of these variables in space.
Accordingly, elevation, marginal grasses, pH, reservoir circumference, atmospheric temperature and percentage shade appeared on the fi rst (horizontal ordination axis) as the most important variables, while water depth, fl oating /submerged vegetation and marginal forest occurred on the second axis (vertical) and were less important in determining Anisoptera species assemblage distribution patterns.Marginal forest, percentage shade, water depth and fl oating/submerged vegetation were the most important  variables, while marginal grasses, elevation and pH were important for Zygoptera.Th e following Anisoptera species were also associated with marginal grasses of reservoirs at High (GH) and Mid-high (CE) elevations: T. stictica, Palpopleura jucunda, Acisoma panorpoides, Orthetrum caff rum.N. jonesi was associated with highly shaded conditions of sub-site three at Mid-low (KL) elevation.Low (SB) elevation species (when the elevation gradient is projected backwards on the ordination triplot) had the typical species Hemistigma albipunctum, Chalcostephia fl avifrons, Tetrathemis polleni, Diplacodes lefebvrii, Rhyothemis semihyalina and Tramea basilaris, even though the last three species were also present at higher elevations.Open reservoirs at all elevations had the following species in common, located mostly at the centre of the ordination for Anisoptera: O. julia, C. erythraea, T. arteriosa, P. lucia, A. speratus, A. imperator, T. dorsalis, N. farinosa and P. fl avescens.
High (GH) elevation zygopterans like Pseudagrion citricola and Africallagma saphirinum were strongly associated with sunny conditions, high pH and marginal grasses.Low (SB) elevation species were L. tridens and A. nigridorsum while P. hageni was associated with Middle (BG) to Low (SB) elevation shade conditions.A. elongatum, P. kersteni and P. salisburyense were associated with minimal fl ow, exposed rock and marshy conditions.Intra-set correlations of environmental gradients with axes (Table 6) showed that elevation, pH, percentage shade and marginal grasses were highly correlated with axis one for both odonate sub-orders, with marginal forest being an additional correlate to this axis for Zygoptera.Reservoir circumference for Anisoptera and exposed rock for Zygoptera were the only important correlates with axis two in both ordinations.Axes three and four were not important.A summary of weightings attributed to the fi rst two axes of ordinations for Anisoptera and Zygoptera showed that species-environment correlations using CANOCO were strong.Th e respective eigenvalues, cumulative species variances and Monte-Carlo tests for CCA are given in Table 7.With a cumulative percentage variance for species data and for species-environment relations of 89%, it meant that measured site variables were probably responsible for the main variation in species patterns for Anisoptera.A Monte Carlo permutation test of probability further strengthened this inference as the fi rst axis (Ax1: F= 5.98; P< 0.005) and all four axes (global: F = 3.140; P< 0.005) were highly signifi cant.A cumulative species variance for species data and for species-environment relations of 39.9% for Zygoptera suggests that measured site variables accounted for little variation in species assemblage distribution patterns for this taxon.Although a Monte Carlo permutation test of probability showed that the fi rst axis (Ax1: F = 1.99;P< 0.01) was signifi cant, the overall test using all four ordination axes (global: F = 1.75;P< 0.4) was not signifi cant.Table 6.Intra-set correlations between each of the site variables and Canonical Correspondence Analysis axes 1 and 2 for adult Anisoptera and Zygoptera species and site variables sampled over two years across all fi ve elevations.

Phenology
Seasonal rhythms with dormant (over-wintering) periods during winter are an integral part of the life history of temperate dragonfl ies (Corbet 1999).A similar trend was observed in this study, larvae generally being the only developmental stage sampled in winter (June and July).Th ere were no adult and/or teneral species at any elevation except at Mid (BG), where adults of three species (Ceriagrion glabrum, Lestes plagiatus and Crocothemis erythraea) overwintered.Larvae of the dragonfl y species sampled throughout this study, occurred (at various stadia) throughout the year at all elevations, but varied in diversity, richness and abundance.Th is was also the case for temperate regions where the larval stage is the most common over-wintering stage in Odonata (Norling 1984a;Corbet 1999).Some species e.g.I. senegalensis, L. plagiatus, C. erythraea, and T. stictica appeared to have several distinct generations per year.Th is may be the case when the larval population is provided for by the synchronised return of adult residents, and oviposition occurring early enough to allow more than one generation in a year (Corbet 1999).Other species appeared to have a general overlap of larval cohorts.Nevertheless, there were still noticeable peaks in adult emergence for some species, with three distinct seasonal categories of species peaks appearing at all fi ve elevations: 1) Spring peak (September-November), 2) Summer peak (early: December-March), and 3) Autumn peak (April-May).
Species with adult occurrence peaking in spring and/or summer probably overwintered between June and August as fi nal-instar larvae or intermediate stadia, resuming growth to subsequent higher-instar larvae as favourable climatic conditions and food became available from September.Autumn species perhaps over-wintered as eggs e.g.members of the family Lestidae (Corbet 1999;Norling 1984a), or as earlyinstar larvae.Th e subtropical Anisoptera species studied here were generally elevationtolerant, univoltine, yet had prolonged emergence.In contrast, most Zygoptera were multivoltine, although also highly elevation tolerant.
Since climatic changes associated with seasons act locally and its eff ects are most apparent on the level of populations and metapopulations (McCarty 2001), many factors may have accounted for species temporal variations e.g. 1) mean annual precipitation as it aff ects the long-term quality and quantity of water available (Dent et al. 1989;Pinhey 1978) with rain in this study falling in summer, 2) as there are temperature irregularities usually attributed to topographical variation (Schultze 1997) in this study area, this may have resulted in warm coastal climate with high precipitation levels versus the cooler climates at higher elevations, or, 3) simple chance migrations could also have caused variation.

Aspects of dragonfl y species adaptations in the sub-tropics
Th e centre of biogeographical distribution of a dragonfl y species is very important in determining the number of generations the species can go through in a year (Corbet 1999).Most dragonfl ies colonising the temperate zone for example, have evolved a life cycle where winter is spent in the larval stage.
Usually a large number of stadia is a means of resisting cold (e.g.Paulson and Jenner 1972;Norling 1984b).It is possible that the fi rst step in the colonisation of the temperate zone has been to evolve a mechanism where the larval stage coincides with the adverse season.According to Corbet (1957a,b;1964) and Norling (1984a), two important ecological demands are imposed upon aquatic insects like dragonfl ies in temperate climates.Th ese include the need for all members of a population to pass the winter in a stage resistant to cold, and the need for the adult, reproductive stage to be restricted to the warm season.Also, there is the subsidiary need for the adult stage to be restricted to a certain period in the warm season so that competition with sympatric species may be reduced.All these demands involve conspecifi c synchronisation and the reduction of temporal variation at certain stages of development.Larval photoperiodic responses, interacting with temperature, also provide the framework for seasonal regulation (Norling 1984b;Suri Babu and Srivastava 1990).
Although this study was carried out in a sub-tropical region, relatively close to the tropical centre of species distribution, species temporal trends refl ected some aspects of synchronisation, as with their temperate counterparts.Both the temperate and sub-tropical regions are characterised by four seasons with cold or cool winters.In contrast, the larval lifespan is very short in the tropics, where growth is usually rapid and the adult life often fairly long, bridging the dry season (Happold 1968;Gambles 1960;Corbet 1999;Hassan 1981;Van Huyssteen and Samways 2009).Th is is perhaps because of reduced fl uctuations in environmental conditions (especially temperature) leading to unsynchronised odonate emergence, and the fact that long-lived dispersal stages are probably a prerequisite for species which inhabit temporary pools in the tropics.
Most odonate species sampled here were on the wing for about nine months of the year, from September to May/June, and showed marked monthly variations in richness and abundance during this fl ight period.Th us in these sub-tropical species, the overlapping generations show similarity to their tropical counterparts by long adult fl ight periods (Parr 1984), yet like the temperate species in overwintering as larvae.
Furthermore, species that regularly move between habitats may need to adjust to climate changes that are occurring at diff erent rates in diff erent areas, such as between high, medium and low elevations (Inouye et al. 2000).Overall, the subtropical species studied here are characterised by wide elevational tolerance, as well as long fl ight period with overlapping generations.However, this does not mean that these species are tolerant of the full 3000 m elevational range, with the Alpine zone being very species poor (Samways 1989a(Samways , 1992)).

Biogeographical implication of elevational tolerance
Overall, odonate species richness ranged from 24 to 27 species between 301 and 1350 m a.s.l.However, below this (<300 m (SB)), richness increased to 31 species.Factors that may account for the high numbers of species at low elevations include high primary productivity (Connell and Orias 1964), increasingly benign, less variable and predictable environments (MacArthur 1975;Th iery 1982) and increased resource diversity (Gilbert 1984).Other processes (competition, predation and evolutionary time) may have also infl uenced species richness.Also, besides the advantages of a warm climate promoting larval development, the Low (SB) elevation site also, predictably, had a wide range of habitat types.Additionally, there are no mountain chains that might otherwise prevent either temporary or permanent movement south from the species-rich northern areas, thus maximising regional recruitment.
Th e infl uence of elevation on distribution patterns can also be highly dependent on latitude (Corbet 1999;MacDonald 2003;Koleff et al. 2003).Th is is illustrated by species which are found at progressively narrower elevations farther south.For instance, the intolerance of low temperatures by tropical species (e.g.Tetrathemis polleni in southern Africa) causes them to contract their southern range into a narrow lowland strip, extending down the eastern seaboard of southern Africa which is warmed by the south-moving Agulhas current.
Most species sampled were widespread and common African species.However, three (6.4%) of species sampled were national endemics, accounting for just 13.6% of the total South African odonate endemics.Pseudagrion citricola and Africallagma sapphirinum occurred only at the High (GH) elevation, while Agriocnemis falcifera was across all elevations, suggesting that regional endemism does not necessarily equate to elevational intolerance (Fig. 7).Interestingly, like the non-endemics, two of these are relatively common, with only A. sapphirinum being rare.
Although climate is important to odonate development, assemblage variation and geographical distribution, local factors (e.g.vegetation structure and composition) are also signifi cant in this geographical area as well as elsewhere (Steytler and Samways 1995;Samways and Steytler 1996;Osborn and Samways 1996;Niba and Samways 2001).Furthermore, water depth is also important for larval stages (Samways et al. 1996).Most adult species here responded to sub-sites refl ecting pH, open sunny versus shady and waterfall (fl ow) versus still water conditions.Infl uence of regional and local conditions were seen for example in Notiothemis jonesi which occurs only at the shady lower elevation gradient of Mid-low (KL).
Zygoptera species were more strongly elevation dependent than Anisoptera species.A. sapphirinum, A. elongatum, P. citricola, L. tridens and Azuragrion nigridosum were highly elevation-sensitive species.Elevation-tolerant species were L. plagiatus, I. senegalensis, C. glabrum, A. falcifera, Pseudagrion massaicum, P. salisburyense and P. kersteni.As well as this regional response, there was also a local response.Zygoptera species mostly showed a higher degree of habitat specifi city than the Anisopteran species.Allocnemis leucosticta, a South African endemic, for example, was restricted only to SS4 and 5 at the Botanical Gardens.One reason for this appears to be that Zygoptera are generally less vagile than Anisoptera.

Implications of results for dragonfl y response to global climate change
South African dragonfl ies are extremely sensitive to fl uctuations in water levels, with great fl uctuations being impoverishing to the odonate assemblage (Osborn and Samways 1996).Furthermore, the geographical area where this study was undertaken is subject to great variations in rainfall from one year to the next.Floods can be severe, yet the odonate assemblage can recover within a year (Samways 1989c), indicating its great resilience in this El Niño-prone area.Th is means that the eff ects of global climate change will possibly be two-fold.Firstly, changes in temperature per se would appear, from these preliminary fi ndings, not likely to have a great aff ect upon the assemblage.Th is is because the species involved, even the endemics, are vagile and opportunistic, and will simply colonize the habitats at the appropriate elevations.Secondly, but in contrast, the colonization process will depend greatly on the constancy of the water levels in the water bodies.While increased rainfall and fl ooding are likely not to be detrimental, any prolonged dry period is likely to be harmful.However, unless there is a prolonged and extreme drought, coupled loss of all local water bodies, there will almost certainly be remnant pools.Such pools would act as source habitats from which these resourceful species will disperse to new pools once the rains have returned (see Samways, this volume).

Figure 1 .
Figure 1.Th e mid-elevation site (790 m a.s.l.) with reedy margins typical of all the sites.

Figure 2 .
Figure 2. Accumulative dragonfl y tenerals (T) and adults (A) species recorded at a Low (SB), b Mid-low (KL), c Mid (BG), d Mid-high (CE) and e High (GH) elevations during the fi rst (01) and second (02) year of the study.

Figure 3 .
Figure 3. Dragonfl y species recorded at a Low (SB), b Mid-low (KL), c Mid (BG), d Mid-high (CE) and e High (GH) in terms of adults (A), tenerals (T) and larvae (L), and during the two-year sampling period.

Table 1 .
Th e fi ve elevational sites used in this study

Table 2 .
Odonata species sampled during this study with species code names * Record of adult, teneral and/or larval stage of the corresponding species. 1 Common African species whose range extends south just over the border into South Africa, but are local or rare in the country, 2 African species that are widespread and/or locally common in South Africa, 3 African species that are regularly seen in the right habitats, some of these are very common throughout South Africa, 4 Species endemic to South Africa (i.e.South of the Limpopo River).teraspecies with two peaks per year were Ischnura senegalensis and C. glabrum, each occurring at various elevations (C.glabrum was absent at High (GH) elevation).Also, Africallagma glaucum and Pseudagrion massaicum had two peak appearances in Mid-high (CE) elevation.L. plagiatus had two peaks per year at Mid (BG) and Mid-low (KL) elevation, while L. tridens and P. massaicum had two peak abundances per year in Low (SB) elevation during both years.L. tridens from Low (SB) elevation had four peaks at diff erent times during the two sampling years: April and December 2001, March and November 2002, indicating more than one generation per year.Th e number of species per family was very similar from one elevation to the next.

Table 3 .
Summary of species phenologies recorded during this study.

Table 7 .
Cumulative species variance of species data; ³ Cumulative species variance of species-environment relationship.(ns= statistically non-signifi cant at the 5% level).Summary of weightings of the fi rst two axes of CA and CCA for both Anisoptera and Zygoptera adults sampled during the study in terms of variances accounted for by the two axes.Monte Carlo probability tests of signifi cance are given for the fi rst canonical axis (AX1) and all four axes.