The effects of flooding on survivorship in overwintering larvae of the large copper butterfly Lycaena dispar batavus (Lepidoptera: Lycaenidae), and its possible implications for restoration management

Previous work suggests that submergence of Lycaena dispar larvae during overwintering may play a significant role in this butterfly’s population dynamics. Since potential re-introduction sites in eastern England are prone to regular seasonal flooding, we further studied the species’ submergence tolerance with a view to formulating management protocols conducive to larval survi­ vorship under periodic flood conditions. Simulated flooding regimes using captive-reared larvae showed that enforced submergence has a twofold effect: firstly, a direct increase in mortality after 28 days under water and, secondly, a longer term, post-diapause increase in mortality; manifest either as an inability of larvae to resume feeding, or a failure to complete development. Additionally, there was a marked difference in the response of “early” and “late” diapause larvae; the latter generally succumbing after shorter periods under water, and suffering higher total mortalities. Behavioural investigations suggest that, if afforded the opportunity, diapausing larvae can evade submergence by climbing onto the exposed sections of partially flooded host plants. Significantly, survival on partially flooded plants was found to be comparable to that on unflooded controls. Further re-introductions of L. dispar in the U.K. will probably necessitate a direct translocation of wild Dutch stock. As the flood tolerance of this source population remains largely undetermined, and given that re-introduction site hydrology will be generally unamenable to conservation-oriented manipula­ tion, it is recommended that restoration management be directed towards creating structural diversity in the vegetation of overwin­ tering habitats, thereby providing potential “flood refugia” for hibernating larvae.


INTRODUCTION
Two specimens collected during the 19th century from the Broadland region of eastern England -one labelled "Ranworth, I860" the other "Woodbastwick, 1864"may well represent the last reliable evidence of the (now extinct) English race of the Large Copper butterfly, Lycaena dispar dispar Haworth (Irwin, 1984).The demise of L. d. dispar in Broadland seems almost certain to have been driven by the extensive loss of former fenland habitats (through successional change), hastened by the abandonment of traditional management practices (Pullin et al., 1995).In contrast, elsewhere within its his toric range large scale drainage, and the subsequent recla mation for agriculture, was probably the principal factor contributing to the butterfly's decline (Duffey, 1968).Though still extant throughout much of mainland Europe, eastwards across temperate Asia to Korea, other subspe cies of the Large Copper have undergone marked declines in the wake of wetland degradation and fragmentation (Pullin et al. 1998).In the Netherlands, for example, the endemic subspecies batavus has suffered considerable losses, to the extent that it may now be represented by just two self-sustaining populations -inhabiting the Weerribben and Wieden National Parks (Tax, 1989;Oostermeijer, 1996;G. Padding, pers. comm.).Currently listed as "endangered" in the IUCN Red List of threat ened animals (IUCN, 1990), L. dispar's imperilled status also merits its scheduling under the 1979 Convention on the Conservation of European Wildlife and Natural Habi tats (Bern Convention) and its inclusion in Annexes II and IV of the 1992 European Community Habitats Direc tive.In addition, since 1991, L. dispar has featured as one of the target invertebrates under the umbrella of English Nature's Species Recovery Programme, and this distinc tion has fostered a renewed impetus in researching the butterfly's ecology and habitat requirements (Pullin, 1997).
In its native wetland habitat, L. d. batavus is univoltine and host-specific, ovipositing on a single larval food plant, the Great Water Dock Rumex hydrolapathum Hud son.Eggs hatch in August and larvae feed for approxi mately four weeks, prior to entering winter diapause as second instars.Diapause is characterized by a cessation of feeding, followed by migration to the base of the host plant.Here, the cryptic larvae seek out senescent Rumex leaves that will serve as winter hibernacula.Subject to local seasonal climate, dormancy termination and resump tion of feeding occur between late March and early May.Larvae subsequently undergo a further two moults prior to pupation in June.Adults are on the wing from late June through to late August, peaking in abundance during the latter half of July (Pullin et al., 1998).
Throughout the present century considerable conserva tion effort has been expended in the shape of a succession of attempts to re-introduce L. dispar to the UK (Duffey, 1968;Webb & Pullin, 1996).To date, none of these have resulted in establishment of a self-sustaining population.Given the prominent role wetland drainage is traditionally presumed to have played in the butterfly's historical decline, it is perhaps somewhat ironic that several re establishment attempts have been thwarted, not so much by a deficiency of water, but by its unseasonal excess (Ellis, 1951(Ellis, , 1965;;Duffey, 1968).As an acknowledged wetland specialist, it seems not unreasonable to expect L. dispar to exhibit traits adapting it both physiologically and behaviourally to a seasonally flooded fenland habitat.Indeed, a longstanding consensus held that overwintering larvae displayed tolerably good resistance to periodic sub mergence (Anon, 1929;Purefoy, 1929;Duffey, 1968).More recent research has, however, challenged this assumption.In a comparative field ecology study, Webb and Pullin (1996) reported mortality during the overwin tering stage to be markedly higher at a flood-prone Eng lish site than at an otherwise comparable non-flooding Dutch site.This circumstantial evidence for the detri mental affects of flooding gained support from comple mentary laboratory investigations (Webb & Pullin, 1998) in which it was shown that episodes of inundation, if suf ficiently prolonged, could significantly increase the risk of larval death during hibernation.Moreover, it became evident that the incidence of mortality recorded immedi ately post-submergence is frequently not a sufficient pre dictor of the eventual flood-induced fatalities (Webb & Pullin, 1998).This latter observation suggests a twofold submergence response, which can be defined as acute and chronic.The former represents immediate mortality, ascribed to the submergence event per se, where larvae succumb whilst still under water (e.g. from anoxia).Chronic symptoms are manifest over a longer timescale in, for example, an inability to resume feeding, a failure to complete development or, conceivably, more subtle effects, such as a perturbation of the phenological syn chrony evolved between the insect and its host plant (cf.Chippendale, 1982;Tauber et al., 1986;Danks, 1987;Woiwod, 1997).Post-submergence chronic effects, analo gous to that seen in L. dispar, have also been described in other wetland invertebrates, notably the larch sawfly Pristophora erichsonii Hartig (Lejeune & Filuk, 1947;Lejeune et al., 1955), and the large heath butterfly Coenonympha tullia Muller (Joy & Pullin, 1997, 1999).
As some of the potential re-introduction sites in Broadland are prone to seasonal flooding (Giller & Wheeler, 1986;George, 1992), further investigations of the capacity of larval populations to withstand protracted inundation, allied to an evaluation of possible floodmitigation measures, are needed prior to undertaking any further re-establishment attempts.
In striving to experimentally simulate natural field con ditions, an additional layer of complexity arises from the growth habit of the food plant itself.R. hydrolapathum generally flowers during its second year of growth (Duf fey, 1971) and, following autumnal senescence, the tall (up to 2m) flower stalk (typically bearing a few withered bract leaves) often persists in situ throughout the winter months.In most circumstances a substantial proportion of the senescent stem would be expected to protrude above the level of all but the most exceptional of floods.The potential "flood-refuge" value of these inflorescent growths to overwintering larvae has been commented on in the past (e.g.Purefoy, 1929;Duffey, 1968;Duffey & Mason, 1970;Tinning, 1975), as has the considerable movement of which larvae appear capable, during what is normally considered a period of overwintering dormancy (Duffey, 1968;Tinning, 1975).Given the potential impediment of flooding to L. dispar re-establishment, the present study was conceived with the following aims: (1) To assess the impact of varying periods of enforced submergence on overwintering larvae, in the context of both short and long-term survival.
(2) To assess the impact of time of flooding, by com paring the responses of early and late diapause larvae to submergence.
(3) To investigate the behavioural responses of larvae under conditions of partial -as opposed to completefood plant submergence.(4) To consider management options that might enhance the viability of field populations under flood conditions.

METHODS
All larvae used in the investigation were derived from a captive-reared colony of L. d. batavus maintained at Wood walton Fen National Nature Reserve, Cambridgeshire, U.K. since 1970, and transferred to the University of Birmingham for the purposes of this study.The Woodwalton colony can itself be traced back to a small founding population of Dutch batavus, originally released onto the fen in 1927 (Duffey, 1968).

The impact of varying durations of submergence on larval survival
Assessment of flood tolerance under conditions of enforced submergence incorporated a parallel set of experiments in which the response of both early and late diapause larvae (see below) were examined.Prior to the imposition of treatment conditions, larvae were subjected to a preliminary pre-conditioning or diapause-induction phase devised (from previous studies) to take them through to the desired stage of development (Webb & Pullin, 1998).Larvae, raised from egg eclosion to the 2nd instar stage under short day illumination (9L : 15D) at room tempera ture (~ 20oC), were transferred to senescent leaves of R. hydro lapathum (10 larvae per leaf), which would serve as overwintering hibernacula.Leaves were placed into filter paperlined perspex entomological boxes, and retained within con trolled environment cabinets, running on a photoperiod/ temperature regime (9L : 15D, 5°C) designed to simulate over wintering conditions.Early diapause pre-conditioning was ter minated after a period of four weeks, by which time it is assumed larvae have attained full metabolic suppression (e.g.Tauber & Tauber, 1976;Tauber et al., 1986;Danks, 1987).Late diapause individuals were retained under induction conditions for a further 16 weeks.This period was calculated to take them through to a stage of development equivalent to that of immi nent dormancy termination in field populations when metabo lism will be more active.To achieve synchrony in the subsequent submergence treatments, late diapause pre conditioning was commenced 16 weeks in advance of that for early diapause larvae."Times (days) required to kill 10, 50 and 90 percent of the population, respectively.
Following pre-conditioning, leaf hibernacula harbouring dia pause larvae were pinned to short cane supports, and these were inserted amongst the withered foliage of potted senescent Rumex plants (one hibernacula per plant).Host plants were enmeshed with muslin sleeves to counter any tendency for larvae to wander, and then placed into a large plastic water tank.Muslin-filtered lake water (pH 6.5) was progressively added to the tank to a final depth that ensured all host plant material was completely immersed.Temperature and illumination were held constant (9L : 15D, 5°C) for the duration of the experiment.Larvae were assessed for their ability to withstand submergence over five treatment periods (3, 7, 28, 56, 112 days).These were selected as a range of times culminating in an interval repre senting a full winter (which has been recorded in some poten tial re-introduction sites) and additionally allowed statistical estimation of a lethal time at which 50% of the population perish (LT50).Submergence treatments and an unflooded control were run in duplicate, such that totals of 20 larvae were sub jected to each period of immersion.After flooding treatments, designated plants were carefully withdrawn from the tank, and plants allowed to drain for 24 hours.An initial assessment of larval viability was undertaken with the aid of a low power bin ocular microscope.Those larvae showing either signs of move ment, or no outward evidence of mortality (e.g.waterlogging, fungal growth) were recorded alive at this stage.Surviving lar vae, retained on their leaf hibernacula, were then either returned to normal overwintering conditions (9L : 15D, 5°C) until dia pause termination (early diapause) or, to simulate a late dia pause situation experienced by field populations, provisioned with fresh growths of Rumex, and exposed to a regime of increasing photoperiod and temperature (increments of 30 min and 2°C per day, over a period of 8 days) (late diapause).There after, larval viability was monitored at regular intervals to ascer tain any delayed effects of submergence on long-term survival.

Behaviour and survival in response to partial submergence of host plant
Over two consecutive years (1997 and 1998) a complemen tary investigation was undertaken to examine the behavioural response of larvae to submergence.Early diapause larvae (pre conditioned as described above) were assigned to one of two submergence treatments -either partial or total.For each treat ment, 10 larvae were transferred to the leaves of individual semi-senescent pot-grown Rumex plants.Plants, bearing larvae, were placed within 25 litre glass tanks to which water was sub sequently added to a level, that either completely immersed plants (total submergence treatment), or left approximately one third of the foliage protruding above the water surface (partial submergence treatment).In the case of partial submergence, five of the ten larvae (marked with a small spot of white correction fluid) were positioned so that initially they fell below the water line, whereas the remaining five were placed on leaves pro jecting above the waterline.Unflooded control plants and larvae were retained within additional "dry" tanks.Owing to the small number of larvae available in 1997, treatment duplication was only possible during the second year of the investigation.Thus, for each flooding regime, the fate of 10 and 20 larvae were fol lowed in the first and second year, respectively.The experiment was conducted in an external poly tunnel which, though affording protection from wind and rain, was otherwise subject to the ambient seasonal climate.Observations of larval position and movement, and records of maximum and minimum tem perature, were made daily for 45 days (22/10/97 to 5/12/97) and 58 days (14/10/98 to 11/12/98).Subsequently, plants were with drawn from the tanks, and an initial estimate of larval survival was undertaken.Thereafter, plants were retained within the poly tunnel for a further 16 weeks prior to assessing final overwin tering mortalities.

The impact of varying durations of submergence on larval survival
The pattern of larval mortality under conditions of enforced submergence was found to be determined by two aspects of the flooding regime: its duration and its timing in relation to the stage of larval development.Sta tistical evaluation of survivorship data, using a G-test for frequency analysis (cf.Fowler & Cohen, 1990), revealed both highly significant associations between treatment (length of submergence) and survival (early diapause G = 57.16,df= 4,p < 0.001; late diapause G = 131.81,df= 4, p < 0.001), and a highly significant difference in the responses of early and late diapause larvae to the treat ments (early vs. late diapause G = 29.58,d f = 5, p < 0.001).Comparison with controls indicates that, for early diapause larvae, periods of enforced submergence, up to at least 28 days, effect no increase in either immediate or long-term larval mortality.But, although short-term sur vival remains unaffected after 56 days under water, it is at this stage that the delayed repercussions of submergence become evident.Only a single larva exposed to the 56 day treatment successfully resumed feeding and com pleted development.After 112 days submergence, imme diate survivorship had undergone a marked decline to 30%, and of those larvae still alive at this stage, all had perished by the termination of diapause.In contrast, and consistent with a priori assumptions (see discussion), the detrimental effects of enforced submergence on late dia-early (X) and late (□) diapause L. d. batavus larvae.Shaded areas represent the time larvae spent under water.pause larvae were observed over considerably shorter durations of exposure.Recorded mortalities after 28 and 56 days were 65% and 100%, respectively.Control mor tality rates over the same time periods were 15% and 25%.Even relatively brief periods of immersion (3 and 7 days), although having a negligible effect on immediate survival, caused substantial mortality in the longer term (Fig. 1).A probit analysis (Finney, 1971) of the survivor ship data (Table 1) gave estimated LT50 values (the period of submergence required to kill 50% of the population) of 90.8 and 19.4 days, for early and late diapause larvae, respectively.Corresponding values based upon the num bers of larvae that actually recommenced feeding were 36.9 and 12.9 days.

Behaviour and survival in response to partial submergence
In circumstances where larvae were given an opportu nity to evade submergence, there was a certain degree of bi-directional migration across the water barrier.Correla tion analysis using the Spearman rank correlation test, revealed a significant relationship between larval position and mean daily temperature (Spearman rank correlation coefficient : 1997, rs = 0.694, n = 25, rScrit = 0.526, p < 0.01; 1998, rs = 0.759, n = 51, rscrit = 0.368, p < 0.01).Thus, a sufficient depression in temperature might induce larvae to seek shelter beneath the waterline, whereas at elevated temperatures there could be a tendency for indi viduals to climb onto exposed portions of the host plant.Whereas a response to fluctuating temperatures remains highly plausible, this does not of course rule out the pos sibility of partial, or indeed exclusive, reaction to addi tional unmonitored physical variables (e.g.variations in light level).Furthermore, since individual larvae could not be reliably identified, it was not always possible to ascertain which larvae were actually moving across the water barrier.Nonetheless, given that survivorship on partially submerged plants (recorded at the conclusion of the experiment) was equal to that on control 'dry' plants in 1997: 4 larvae (40% survival), and only marginally less in 1998: control = 7 larvae (35% survival), partially sub merged = 6 larvae (30% survival), whereas survival under conditions of complete submergence was zero in both years, it is evident that an opportunity for larvae to escape flooding could have important implications with respect to the vegetation management of overwintering habitats (Table 2).This contention is borne out by complementary observations made under field conditions, both in the UK and the Netherlands (Nicholls, 2000).

DISCUSSION
An evaluation of flood tolerance in Lycaena dispar should ideally take into consideration not only the dura tion of the submergence period itself, but also its timing relative to the stage of larval development.Although the precise mechanistics underlying submergence-induced larval mortality remains to be conclusively resolved -and past workers have sometimes presented conflicting accounts of larval survival under flood conditions (Anon, 1929;Purefoy, 1929;Duffey, 1968Duffey, , 1977;;Webb & (Danks, 1987).Nevertheless, towards the end of normal diapause, resource-depleted larvae would predictably be acutely vulnerable to the debilitating consequences of prolonged submergence.The data presented here, indicating a mark edly greater susceptibility in late diapause, with periods of less than 28 days causing over 50 % mortality, are con sistent with this premise.Furthermore, the laboratory results also find support in several documented field observations.In 1951, for example, two years after the successful introduction of subspecies batavus to a site in the Yare Valley, Norfolk, post-diapause larval emergence coincided with a period of extensive spring tidal flooding, and this was thought to be instrumental in wiping out the established colony.Similarly, at Woodwalton Fen in 1966, flood conditions persisted well into the month of May, delaying larval emergence, and as a consequence led to poor survival (Duffey, 1968).
Clearly, a degree of caution needs to be exercised when invoking physiological concepts like a 'flood tolerance threshold' in the context of field populations.Tenuous relationships between the field and laboratory will inevi tably be confounded by habitat features and environ mental exigencies that are largely unamenable to ex situ replication.Irregularities in fenland relief, for instance, tend to generate a mosaic of spatially-dispersed microto pographic and microclimatic situations in which R. hydrolapathum grows.This spatial matrix may in turn give rise to a marked temporal variability in the degree and the length to which food plants are subject to flood ing.Rumex plants growing along watercourses may, for example, be either partially or totally submerged for periods exceeding that of the longest treatment time (112 days) imposed during the current investigations.The putative avoidance of these "riparian" Rumex by ovipo siting females (Duffey, 1968), and the suggested prefer ence for those plants growing on the slightly raised sub stratum of mixed herbaceous fen, could arguably be inter preted as an adaptive strategy designed to evade the worst excesses of unpredictable inundation.However, recent research on host plant selection (Pullin, 1997;Webb & Pullin, 2000) has found little evidence to substantiate this supposition.
Results from the behavioural experiments suggest that larval movement is of frequent occurrence during over wintering and is at least partially temperature dependent.This conclusion finds support in Duffey's (1968) obser vation that diapausing larvae in captivity became active during the winter if the air temperature rose above 7°C.Comparable manoeuvrability has been noted in overwin tering larvae of C. tullia (Joy & Pullin, 1997, 1999), which are also able to feed when conditions are suffi ciently mild -an option which, given the annual senes cence of Rumex, is generally not available to overwin tering L. dispar.
A propensity for diapausing larvae to migrate to ele vated "flood refuges" (whether temperature-induced or not) could go some way in reconciling the disparate opin ions previously expressed on flood tolerance in L. dispar.Assessment of larval viability by earlier workers, most notably Purefoy (1929) and Duffey (1968), was based largely upon observations made under conditions of incomplete submergence.Viewed from this perspective, Duffey's ostensibly paradoxical observation, that survival on "flooded" plants exceeded that recorded on controls, is perhaps more readily explicable.Of the 45 larvae sur viving his flooding regime, 35 were found to be located on the upper, emergent extremities of the host plants.Subsequent work on submergence tolerance in L. dispar indeed led Duffey to a re-evaluation of his initial find ings (Duffey, 1977).If results obtained from the current behavioural research are taken to be broadly representa tive of behavioural patterns in the field then, implicitly, there may be little difference between partially flooded and unsubmerged host plants in terms of potential over winter larval survival.
A further important question arising from recent work on flood tolerance in L. dispar (cf.Webb & Pullin, 1998), and one that has a significant bearing on the proposed re introduction of the butterfly to the UK, concerns the genetic fitness of the larval population used.The current research population is derived from a colony reared under captive conditions at Woodwalton Fen since 1970.This population was in turn descended from individuals taken from an independent captive population established during the early 1960s from wild Woodwalton stock.The latter was inaugurated as an insurance policy against extinction of the butterfly on the fen (a fate which indeed was to befall the Woodwalton population following unprecedented summer flooding in 1968).Sustained over successive generations in novel ex-situ environments, small populations may undergo evolutionary changes that contribute to an erosion of genetic variability and indi viduals may progressively develop traits which, although adaptive under artificial confined situations, render them less well adapted to survival in the wild (Frankham, 1995).Adaptation of invertebrate populations to captive environments has been widely documented (e.g., Ayala, 1965;Briscoe et al., 1992;Frankham & Loebel, 1992;Latter & Mulley, 1995) and, given strong directional selection, shown in some instances to occur relatively rapidly (Miyatake & Yamagishi, 1999;Lewis & Thomas, 2001).Although lacking critical evidence, a comparative study of wild and captive bred L. dispar has led us to argue previously (Nicholls & Pullin, 2000) that these two populations may now have diverged genetically as a con sequence of experiencing contrasting genetic milieus.In this context, a potentially significant early observation was that made during the winter of 1927-28, shortly fol lowing the original introduction of subspecies batavus to Woodwalton Fen.Hibernating larvae were at this time subjected to 60 days continuous flooding, without appar ently a marked impact on their subsequent survival (Purefoy, 1929;Riley, 1929).This circumstantial evidence suggests a superior flood tolerance prior to the start of the captive rearing programme and, together with our more recent research, leads us to speculate that prolonged exsitu propagation may in some respects have now compro mised flood tolerance in the captive stock.At present the ability of wild populations of L. dispar to survive extended periods under water remains undetermined.Regrettably, owing to the endangered status of the but terfly in its native habitat, a potentially definitive com parative study of flood tolerance in captive and wild populations cannot at present be justified.Nonetheless, given the suspected "genetic inferiority" of the captivereared butterflies, it now seems inevitable that any future re-introductions will necessitate a direct translocation of individuals from the wild Dutch population.

Implications for habitat management
Implications for conservation management have to be viewed with some caution as we cannot be sure if wild populations will be more or less susceptible to flooding than the captive bred population used in this study.Despite this, the results of the submergence experiments do highlight the necessity to include an appraisal of local hydrology (particularly the timing, frequency and magni tude of flood episodes) as an integral element in the assessment of potential re-introduction sites.The current restoration programme for L. dispar in the UK is focusing primarily on the floodplain fens of Broadland; now probably the only region within the historical range of subspecies dispar retaining sufficient habitat to support a viable population(s) of this characteristically dispersive, low density butterfly (Pullin, 1997).Significantly, how ever, from a restoration viewpoint, many of the Broad land fens experience a considerable seasonal flux in water levels (amplitude > 50 cm) and therefore pose potential problems for remedial management (Giller & Wheeler, 1986).The broader conservation relevance of our find ings can perhaps be put into perspective by reference to a recent series of experimental releases in Broadland (Nicholls & Pullin, 2000).These showed significantly greater larval survivorship over the relatively dry winter of 1996/1997 compared with the winters of 1995/1996 and 1997/1998, during both of which prolonged flooding occurred.
A stated objective of the ongoing Fen Management Strategy for the Broadland region (Broads Authority, 1997) is the reinstatement of a natural hydrological regime to the fen drainage system.As a consequence, any water level manipulation targeted specifically for the benefit of L. dispar, although technically feasible on a limited scale, is unlikely to constitute an exclusive long term conservation option.Management of the vegetation communities, of which the butterfly's food plant R. hydrolapathum is a typical associate (notably the Peucedano -Phragmitetum herbaceous fen community (Wheeler, 1978)) would seem to offer a more practical and sustainable solution.Essential as vegetation manage ment will almost inevitably be (in the short-term at least), a detailed understanding of site hydrology, and predicting how this might change with time, will, nevertheless, remain integral to any future remedial management.Given predictions emanating from some of the current cli matic models, there seems a strong likelihood that both the frequency and magnitude of extreme weather eventscomparable to the summer flood of 1968 (Duffey & Mason, 1970), and those of spring 1998 (Nicholls & Pullin, 2000) -are set to increase over the coming dec ades.Among the forecasts, set out in a recent review highlighting the potential effects of climate change in the UK, are projected increases in precipitation of 5% by the year 2020 and 10% by 2050 (Department of the Environ ment, 1996).Taken in conjunction with the continual rise in relative sea levels (George, 1992), it would appear likely that the low-lying valley fens of Broadland are des tined to become increasingly flood-prone.Accommo dating projected climatic change will unquestionably constitute an important element in any long-term conser vation strategy drawn up for L. dispar.Any future trend towards wetter, warmer winters in the Northern Hemi sphere might not only seriously undermine continuing attempts to re-establish the butterfly in the UK, but could equally imperil the extant native population in the Nether lands.This underlines the current emphasis being attached to a "landscape scale" approach to L. dispar res toration (Pullin, 1997).The establishment of multiple, spatially-dispersed populations, in providing potential "flood refugia" for hibernating larvae, and thereby offset ting the likelihood of catastrophic localized extinction, will lie at the core of this approach.

Table 1 .
Estimates of the capacity of Lycaena dispar batavus to survive submergence during overwintering.Data presented are from probit analysis oflarval survival (95% fiducial limits are shown in parentheses).

Table 2 .
Survivorship of early diapause L. dispar larvae overwintered under conditions of total and partial submergence.Numbers refer to the number of larvae subjected to each treatment and those recorded as alive at the termination of diapause.