The population biology of the early spider orchid Ophrys sphegodes Mill. III. Demography over three decades


Correspondence author. E-mail:


1. Compared to animals, long-term, large-scale demographic studies based on plants are scarce. A 32-year plant-based demographic study of the rare early spider orchid Ophrys sphegodes is presented, covering periods of management by cattle grazing (1975–1979) and sheep grazing (1980–2006).

2. Annual recruitment exceeded mortality under sheep grazing, but numbers of emergent plants did not increase for many years. Eventually, following rapid population increase, numbers fluctuated strongly, with high recruitment in 1 year followed by heavy mortality the next.

3. The population’s distribution between different life states varied considerably, even between consecutive years with identical management. On average, almost 30% of plants were dormant. Most dormant periods lasted <4 years (c. 78% were ≤ 2 years), but dormancy of up to 8 years was recorded.

4. Most orchids had short lives from first to last appearance, but some lived for >20 years. Age-specific survivorship data yielded a half-life of 2.25 years.

5. Peak flowering date advanced by 0.5 day year−1 during the study. Flowering was earlier after warmer years and later after winters with more frosts. Inflorescence height and leaf number were positively correlated with rainfall during inflorescence extension, but negatively correlated with temperature and sunshine hours over up to a year before flowering. Higher temperature was associated with less of the population flowering and more being vegetative. The proportion in dormancy was unaffected by climatic variables.

6. Annual recruitment and mortality were positively correlated with temperature in the previous year, and annual recruitment was positively correlated with the number of flowering plants in each of the two previous years. In most years, <1 plant was recruited per flowering plant in each of the two previous years.

7. Despite a dramatic increase in the number of emergent plants c. 10 years after management changed to sheep grazing, and large numbers of emergent plants thereafter, mortality greatly exceeded recruitment over the last 10 years of this study.

8.Synthesis. Conservation of orchids like O. sphegodes, which have numerous ‘weedy’ life-history characteristics, is heavily reliant on appropriate management. Although previous management prescriptions for conservation remain valid, some site disturbance will be beneficial to recruitment.


Long-term demographic studies involving repeated censuses produce invaluable data for understanding how and why natural populations change in size and structure. There have been numerous studies of this type in animal ecology (e.g. Murie 1944; Sinclair 1977; Clutton-Brock, Guiness & Albon 1982; Clutton-Brock, Major & Guinness 1985; Messier 1994; Stenseth et al. 1997), and many of them are celebrated for the insight they provide, and repeatedly described in ecology textbooks, and mined for data to address new research questions. With few exceptions (e.g. Lauenroth & Adler 2008; Roach, Ridley & Dudycha 2009), plant ecology lacks comparable studies. This article describes a demographic study of the early spider orchid, Ophrys sphegodes Mill., that is similar in duration and scale to these classic studies on animals. It is based on over 30 annual censuses, often involving several hundred plants, making it virtually unique in duration and in size among plant population studies. Altogether, data were obtained from 3681 plants of O. sphegodes during the study, and full life histories were recorded for 2499 of these.

Tamm (1948, 1972; Inghe & Tamm 1988) pioneered the demographic study of terrestrial orchids. The Orchidaceae has now attracted more demographic research than any other plant family; Kull (2002) lists 42 studies on 30 orchid species in 18 different genera, and several more have been published since (Whigham & Willems 2003). However, many early studies only reported counts, often just of flowering plants, in different years (Farrell 1991; Vanhecke 1991; Silvertown et al. 1994; Øien & Moen 2002), and few studies involve censuses, from which detailed information about recruitment, mortality, life spans and other population characteristics can be obtained. Moreover, although there are exceptions (e.g. Wells 1967; Hutchings 1987b; Jacquemyn et al. 2007), many of the census-based studies gathered data from few plants in few years. Consequently, despite the effort invested in the study of orchids, and notwithstanding the contributions of recent publications (e.g. Kéry & Gregg 2003, 2004; Shefferson et al. 2003; Shefferson, Kull & Tali 2005; Gregg & Kéry 2006; Pfeifer et al. 2006; Shefferson 2006; Jacquemyn et al. 2007; Shefferson & Simms 2007; Shefferson & Tali 2007), many aspects of orchid population dynamics are still poorly understood (Kull 2002; Whigham & Willems 2003).

Like many other orchid species, O. sphegodes has declined significantly in range and local abundance in recent decades (Hutchings 1987a, 1989a; Dixon et al. 2003; Kull & Hutchings 2006), with habitat destruction, isolation of remaining populations, unsuitable management and eutrophication among the prime causes. Ophrys sphegodes is now rare in the UK, and appropriate management (Hutchings 1987a) and an understanding of the demographic effects of different management regimes are vital for its conservation. This study examines the demographic behaviour of O. sphegodes under different forms of management and its impact on conservation of the species. The investigation focussed on the largest remaining population of O. sphegodes in the UK. The study began in 1975, when the habitat was managed by cattle grazing. This was replaced by sheep grazing in 1980. From 1981 onwards, grazing animals were removed each year, from c. 1 month before O. sphegodes flowered until after seed dispersal (April to early September). Earlier analyses of the first 10 years of the study (Hutchings 1987a,b; Waite & Hutchings 1991) showed that annual recruitment fell steadily under cattle grazing, and annual mortality was high, resulting in a progressive decline in the number of plants in the population. After sheep grazing was introduced, recruitment exceeded mortality every year. Despite these beneficial effects of the change in management, counts of emergent plants gave little evidence of an improvement in the size of the population, or in its condition.

This study examines several aspects of the demography of O. sphegodes. First, population dynamics before and after the change in management are analysed in detail. Secondly, the influence of prior climate on flowering behaviour is examined, and evidence of a shift in flowering phenology is sought. Thirdly, the effects of prior climate upon the proportion of plants in different life states (vegetative, flowering, dormant), on plant performance, and on annual recruitment and mortality, are examined. Fourthly, demographic information obtained after 32 years is compared with results reported after 10 years, to determine whether the additional period of study altered understanding. Finally, previous management recommendations for conserving O. sphegodes are reviewed and reassessed.

Materials and methods

Study species

Ophrys sphegodes is a short-lived tuberous orchid of chalk and limestone grassland. It produces a rosette of leaves in September–October and flowers in the following year from late April to the end of May. Every year, some plants are dormant, and some emergent plants are vegetative. Few inflorescences exceed 15 cm in height in the population studied, and most bear four or fewer leaves and flowers. Ophrys sphegodes can self-pollinate, but pollination is also carried out by males of the rare solitary bee Andrena nigroaenea (Schiestl et al. 1997). In most years, few (6–18%) seed capsules produce seeds (Summerhayes 1951; Lang 1980). Plants that do not flower or produce seeds die back quickly, whereas those with fruits persist until after seed dispersal in mid- to late August.

Ophrys sphegodes is a species of southern and central Europe, reaching its northern range limit in southern England. Between 1930 and 1975, its UK range contracted by c. 80% towards the south-east (Hutchings 1987a). Removing or damaging plants is now prohibited by law.

Study site

The study was carried out in Castle Hill National Nature Reserve, East Sussex, UK, an area of 47 ha of ancient calcareous grassland and restored grassland on former arable land, between 100 and 190 m a.s.l. on Upper Chalk bedrock. The vegetation is primarily mesotrophic grassland of community type MG4 (Rodwell 1992), which developed during many decades of close grazing by sheep. More recently, cattle have been used for grazing. Further details of the soils and vegetation can be found in Gay, Grubb & Hudson (1982) and Hutchings (1983, 1987a).

The population of O. sphegodes is spread over several hectares of a south-west facing hill slope with an inclination of c. 16°. It consists of many thousands of plants. Because of the area it covers and the number of plants it contains, data were collected from a small proportion of the population within a permanently marked plot. Photographs and floristic records since 1980, when sheep grazing began to be used again as a management regime, show that there has been little change in the grassland, or in its composition or standing crop, during this time.

Data collection

Methods of data collection have been consistent since the study began in 1975, enabling the demography of O. sphegodes over more than 30 years to be directly compared with behaviour over the 10-year period reported in Hutchings (1987a,b). Although repeating censuses within years (Wells & Cox 1991; Sanger & Waite 1998) and the use of mark–recapture analysis (e.g. Kéry & Gregg 2003, 2004, Kéry et al. 2005) might have yielded additional information, logistics, practicality and time constraints limited data collection to one census per annum. The large number of plants censused in most years and the duration of the study, combined with the short life span of O. sphegodes, are presumed to compensate for any limitations in this approach.

In 1975, a 20 × 20 m plot was permanently marked out by driving brass rods into the bedrock until flush with the soil surface. From 1975 until 2006, the plot was systematically and meticulously searched at the peak of the flowering season for all flowering, grazed and vegetative plants of O. sphegodes. Thirty-one censuses were carried out in 32 years. Access to the countryside was banned by the UK Government in 2001 because of an outbreak of foot-and-mouth disease, preventing data collection in that year.

The position of every emergent plant was recorded each year by triangulation from two of the corners of the plot, to an accuracy of 0.5 cm, and rectangular coordinates calculated. The life-history stage of each plant (flowering or vegetative), its condition (intact, inflorescence grazed) and number of leaves were recorded. Flower number and height of inflorescences were also recorded. The life histories of all recorded orchids were then reconstructed by examining the maps and status of plants between years.

Some orchid species pass through a subterranean phase between germination and first appearance above-ground that can last for several years (Wells 1981). This phase appears to be short (two years or less) in at least some species of Ophrys (Ziegenspeck 1936). Because of this, recruits can only be registered from their first appearance in the emergent population, which may not correspond with the year of germination. Therefore, ages and life spans were calculated from first appearance. Previously recorded plants can also become dormant for one or more years, after which they may re-emerge, flower and set seed again. Several studies have assumed that orchids that fail to appear above-ground for three consecutive years are dead (Hutchings 1987a; Kull 2002; Pfeifer et al. 2006). However, longer dormancy can occur in some species, including O. sphegodes, although it may be unusual (Shetterson 2009). In this study, as in Hutchings (1987a,b), orchids that did not appear for three consecutive years were assumed dead unless later records contradicted this assumption. It is therefore not possible to be certain whether orchids that were absent for 1 or 2 years at the end of the study were dead or dormant. Similarly, the first plants to be treated as new recruits were those recorded for the first time in the fourth year of the study. Retrospectively, for each year of the study, plants that were found to have been dormant were assigned to one of three categories (dormant 1, dormant 2, dormant >2) depending on how many consecutive years they had been dormant.

Population size was determined, together with the numbers and proportions of plants in the emergent and dormant fractions of the population and in the different states (flowering, grazed, vegetative, dormant 1, dormant 2 and dormant >2). Annual and accumulated recruitment and mortality, and the accumulated balance between the two were estimated. The frequency distribution of life spans and lengths of episodes of dormancy were analysed, and the half-lives of each annual cohort of newly recruited plants, and the proportions of each cohort that flowered in the first year above-ground, were also calculated.

The data were examined to determine whether the proportions of plants in the vegetative, flowering and year 1 dormancy states in a given year were correlated with the proportions in each of these states in the next year. In addition, relationships were sought between the proportions of vegetative, flowering and dormant plants in each year and mean number of leaves, inflorescence height and number of flowers per inflorescence in the same year. Bonferroni adjustments for the level of significance were made in each of these three sets of analyses, with the critical value for P raised to 0.005 (there were nine tests in each case).

Relationships between population performance and selected climate variables were examined using data from the nearest Meteorological Office weather station, at Eastbourne, East Sussex UK, 21 km east of the study site and 7 m a.s.l. The climate data used were the number of air frosts in the winter preceding flowering, and summed monthly rainfall totals, mean monthly temperature and summed sunshine hours for 12 (May–April), 8 (September–April), 4 (January–April) and 2 (March–April) month periods prior to flowering each year. These periods were selected to represent, respectively, climatic conditions (i) for the whole year between two flowering periods, (ii) from leaf emergence until flowering, (iii) during the period of temperature increase before spring flowering and (iv) during inflorescence extension. Correlations were sought between each climate variable and the percentage of the whole population that was vegetative, in flower, or dormant, the mean number of leaves borne by emergent plants, and mean inflorescence height. As O. sphegodes produces a new tuber every year, it was predicted that, unlike the situation in some other orchid species (Leeson, Haynes & Wells 1991), there would be no significant effects of climate further in advance of flowering on the performance of established plants. To test this prediction, analyses were also carried out using the same climate metrics for 12–24, 12–19, 12–15 and 12–13 month periods prior to flowering. In addition, correlations were sought between climate variables during the 12 months from May to April of years t – 1 and t – 2 and the number of new recruits observed in year t, to allow for the possibilities of recruitment to the emergent population both within 12 months of the climatic cues, and after a delay of 1 year. The effects of climate in year t – 1 on mortality in year t were also examined. As climate metrics for periods of different duration were correlated, relationships between climate and performance were assessed separately for the 12-, 8-, 4- and 2-month periods prior to flowering. For each climate variable and for each time period considered, a total of five tests were carried out. Bonferroni adjustments for significance were made in each set of analyses, with the critical value for P raised to 0.01. Finally, the effects of the selected climate variables on peak time of flowering, as reflected by the date on which each census began, were also examined. All statistical analyses were carried out with systat 10.2 (Systat Software Inc., Richmond, CA, USA).


Time of recording

The peak of the flowering season, as reflected by the date on which recording began each year, became significantly earlier (= −0.69, < 0.001) as the study progressed (Fig. 1), advancing by an average of 0.5 day year−1. Flowering was later after winters with more frosts (r = 0.55, < 0.01) and earlier after years with higher mean monthly temperatures during the previous 12 months (r = −0.77, < 0.001). Rainfall prior to flowering did not influence recording date.

Figure 1.

 The relationship between the date of peak flowering in the population of Ophrys sphegodes (measured from 1 May each year) and year of the study. The regression line is significant (= −0.69, < 0.001).

Effects of climate on population performance

Although flowering was later after winters with more frosts, number of frosts had no significant effect on other aspects of population performance. Mean number of flowers per inflorescence was not significantly correlated with any climate variable and, as predicted, climate more than 12 months before flowering had no significant influence on the performance of established plants.

Population performance was only significantly influenced by rainfall during inflorescence extension (March–April). Mean inflorescence height and mean number of leaves per plant were both positively correlated with rainfall during these months (Table 1). In contrast, performance was affected by mean temperature and sunshine hours across wider periods of time. Mean inflorescence height and number of leaves per plant were negatively correlated with both mean monthly temperature and sunshine hours, but the relationships with sunshine hours were stronger and more consistently significant (Table 1). Mean monthly temperature had a stronger influence on the proportions of vegetative and flowering plants in the population. Higher temperature was associated with a smaller proportion of the whole population flowering and a greater proportion being vegetative. The proportion of dormant plants was not significantly affected by any of the variables considered.

Table 1.   Correlations between climate variables and various measures of population performance
PerformanceClimate variable
∑ rainfall, Mar–AprMean monthly temp, May–AprMean monthly temp, Sept–AprMean monthly temp, Jan–AprSunshine hours, May–AprSunshine hours, Sept–AprSunshine hours, Jan–AprSunshine hours, Mar–Apr
  1. Climate was measured in terms of summed rainfall, mean monthly temperature and sunshine hours, across four different periods (May–April, September–April, January–April and March–April) prior to flowering. Results are shown only for periods when there was at least one significant correlation with population performance. Significant correlations, after Bonferroni adjustments for multiple tests, are marked in bold. Positive and negative relationships are denoted by the use of + or − symbols, respectively.

% Non-flowering plants, year t−0.01+0.45+0.48+0.53+0.33+0.36+0.09+0.10
% Flowering plants, year t+0.13−0.370.470.48−0.310.55−0.20−0.12
Mean number of leaves+0.540.44−0.34−0.380.640.680.620.54
Mean inflorescence height+0.51−0.16−0.22−0.25−0.39−0.43−0.25−0.38

Population size and structure

The mean number of emergent plants within the recorded plot was 96.3 ± 7.3 (mean ± SE) from 1975 until 1989, almost a decade after the change in management from cattle to sheep grazing. However, the number of emergent plants increased exponentially after 1989, reaching a maximum of 757 by 1993 (Fig. 2a). There was a decline in numbers of emergent plants between 1998 and 2003, followed by recovery to 755 plants by the end of the study.

Figure 2.

 (a) The number of emergent plants and dormant plants, and the total number of plants recorded in the population each year from 1975 to 2006, (b) the number of flowering and vegetative plants in the emergent population, (c) the number of plants in the first, second and >second years of dormancy, (d) the proportion of the whole population in each life state (black = flowering, pale grey = vegetative, mid-grey = dormant 1, off white = dormant 2, dark grey = dormant >2).

Throughout the study, flowering plants accounted for 55.6 ± 4.3% of the emergent population, and 41.4 ± 4.0% of the whole population. However, only 20–30% of the emergent plants flowered during the period of population increase in the early 1990s. In contrast, towards the end of the study, when the emergent population increased for a second time, the proportion of flowering plants also increased, reaching 71.4% in the final year (Fig. 2b).

An increasing proportion of the population was dormant between 1997 and 2003, and the number of plants in the emergent population declined sharply during these years (Fig. 2c,d); whereas a mean of 28.7 ± 2.7% of the population was dormant throughout the study, 67.7% of the population was dormant in 2003. After accounting for dormant plants, the decrease in population size between 1997 and 2003 was much less than suggested from counts of emergent plants alone (Fig. 2a). In 2006, the final year of recording, the total population, including dormant plants, reached a record size of 1123 plants. The distribution of dormant plants between dormancy classes showed considerable variation. In particular, there was a sharp increase in the proportion in the dormant 1 and 2 categories after the early 1990s (Fig. 2d). Overall, the proportions of plants in different life state categories (flowering, vegetative, dormant 1, dormant 2, dormant >2), and in the emergent and dormant fractions of the population, were highly variable, even between consecutive years (Fig. 2d).

The proportions of the total population appearing as vegetative plants and as flowering plants in consecutive years were positively correlated (Table 2). The proportion of vegetative rosettes in any year was negatively correlated with the proportion of flowering plants in the next, and the proportion of flowering plants in any year was negatively correlated with the proportion of vegetative plants in the next. There were negative correlations between the proportion of vegetative plants in any year and the mean number of flowers on inflorescences in that year, and between the proportion of plants in year 1 dormancy in any year and mean inflorescence height. In contrast, there was a positive relationship between the proportion of the population that flowered in a given year and the mean number of flowers per inflorescence in that year (Table 2).

Table 2.   Correlations between (i) the distribution of plants between the vegetative, flowering and year 1 dormancy states in a given year, and the proportion of plants found in each of these states in the subsequent year, and (ii) between the proportions of vegetative, flowering and dormant plants in each year and the mean plant performance (number of leaves, inflorescence height and number of flowers per inflorescence) in the same year
RelationshipCorrelation coefficient
  1. All tabulated correlations are significant after Bonferroni adjustments for multiple tests.

Proportion of non-flowering rosettes in year t vs. proportion of non-flowering rosettes in year + 1+0.67
Proportion of flowering plants in year t vs. proportion of flowering plants in year + 1+0.62
Proportion of non-flowering rosettes in year t vs. proportion of flowering plants in year + 1−0.66
Proportion of flowering plants in year t vs. proportion of non-flowering rosettes in year + 1−0.58
Proportion of non-flowering rosettes in year t vs. mean number of flowers per flowering plant in year t−0.63
Proportion of flowering plants in year t vs. mean number of flowers per flowering plant in year t+0.76
Proportion of plants in first year of dormancy in year t vs. mean height of inflorescences in year t−0.49

Dormancy, life spans and survivorship

Of all the recorded periods of dormancy, 52.7% were 1 year long, and 77.8% of dormant episodes lasted 2 years or less (Fig. 3a). Observation frequency declined exponentially as the duration of dormancy increased. Dormancy of up to 8 years was recorded, but only 3.8% of dormancy episodes exceeded 4 years in length.

Figure 3.

 Frequency histogram of (a) observed episodes of dormancy of different duration, and (b) life spans of plants, reckoned from the first year of observation until final sighting. In each case, the black portion of each bar represents periods of dormancy, or life spans, of which both the beginning and end have been recorded. The grey portions represent periods of dormancy, or life spans, of which the beginning or end may not have been recorded because plants could have been either dead or in a state of dormancy. See text for further details.

Most plants were short-lived from first appearance. Of those for which both recruitment and death were recorded, 51.7% appeared above-ground only once (Fig. 3b). Several of the small annual cohorts of new recruits recorded prior to 1990 had completely expired within a few years of recruitment (Fig. 4a). In contrast, some individuals from all of the larger cohorts establishing after 1990 were still alive at the end of the study. Thus, a small proportion of plants lived for a long time (Fig. 3b). The mean half-lives of cohorts recruited during years when the site was grazed by cattle were significantly shorter than those of cohorts recruited under sheep grazing (1.64 ± 0.26 vs. 2.75 ± 0.45 years, = 6.15, P << 0.001, d.f. = 23). In common with many other herbaceous species, survival during the year after first emergence was relatively poor, but analysis of log survival vs. age, using data from all cohorts, indicated that mortality risk was largely independent of age (r = −0.99, P << 0.001, Fig. 4b). The overall half-life, calculated from age-specific survivorship data for all plants of which recruitment and death were recorded, regardless of year of establishment, was 2.25 years.

Figure 4.

 (a) Depletion curves (in the first years of the study) and survivorship curves for cohorts of Ophrys sphegodes that were recruited in each year of the study. (b) Age-specific survivorship curve for all plants of which the first year of emergence and last year spent above-ground were recorded with certainty (y = 3.38 − 0.13x, r = −0.99, P << 0.001).

Analysis of population flux

Annual recruitment, annual mortality and net annual change in number of plants in the population were all low during the first half of the study (Fig. 5a,b). During the years of cattle grazing, the population accumulated a numerical deficit, but an increasingly positive net balance was observed after the change to sheep grazing. Population flux became much more dynamic from 1989 onwards, with recruitment considerably exceeding mortality in several consecutive years (Fig. 5a). Following this, there was a series of years in which high recruitment in one year was succeeded by high mortality the next (Fig. 5a,b), producing alternating years of net increase and net decrease in population size. Over the whole study, there was a strong positive relationship between recruitment in year t and mortality in year + 1 (= 0.86, < 0.001); overall, 52.0 ± 13.1% of newly recruited plants died each year within 12 months of their first appearance. Despite this, recruitment throughout the study exceeded mortality by 612 plants. However, most of this excess was achieved between 1991 and 1996, and there was a net loss of 226 plants during the last 10 years of the study.

Figure 5.

 (a) Annual recruitment and annual mortality, and (b) net annual change in number of plants in the population.

The number of recruits in any year t of the study was positively correlated with mean temperature from May to April both in – 1 and year – 2 (= 0.55, < 0.01, and = 0.38, < 0.05, respectively), but negatively correlated with the number of frosts in the preceding winter (r = −0.41, < 0.05). Annual mortality was positively correlated with temperature in the preceding year (= 0.46, < 0.01).

Flowering and fitness

Despite a trend of increased numbers of flowering plants throughout the study, there was also a gradual decline in the proportion of the population initiating flowering (Fig. 6a). Until 1998, an increasing proportion of these plants were grazed, and in 1993 and 1998, more than half of the inflorescences were grazed. Throughout the study, there was a significant positive relationship between the number of inflorescences initiated and the proportion grazed (Fig. 6b). Abortion of flowering was also observed, but this usually accounted for <10% of flower initials.

Figure 6.

 (a) Percentages of emergent plants with flower initials (•), actually flowering (bsl00072) and in a vegetative condition (○), in each year of the study; (b) the relationship between the number of flower spikes initiated in each year and the proportion of the flower spikes that were grazed (= 3.60 + 0.06x, = 0.51, < 0.01); (c) the proportion of emergent plants that flowered in each year of the study; and (d) the relationship between the proportion of newly recruited plants that flowered in the first year of observation and their number (= 70.4 − 0.16x, = −0.62, < 0.001).

Until the mid-1980s, most plants flowered on first appearance (Fig. 6c). However, when the population began to increase rapidly in size, the proportion doing this fell sharply before increasing again at the end of the study. There was a significant negative correlation between the proportion of new recruits that flowered in their first year and their number (Fig. 6d).

Although the time elapsing between seed production and emergence of new recruits is unknown, there were significant positive correlations between the number of new recruits in year t and the number of flowering plants in years − 1 and − 2 (r = 0.43 and 0.39, respectively, both < 0.05). In most years, <1 new plant was recruited per flowering plant in years − 1 and − 2, although in a few years, notably 1988–1993, recruitment per flowering plant in earlier years was much higher. Calculated over the whole study, the mean number of recruits per flowering plant, assuming 1- and 2-year delays in appearance, was 1.11 ± 0.18 and 1.31 ± 0.25, respectively.


Effects of climate on flowering date and population performance

This study provided evidence of a significant advance in flowering date in O. sphegodes. Several previous studies have inferred phenological responses to a warming climate on the basis of changes in first flowering dates (e.g. Fitter et al. 1995; Fitter & Fitter 2002; Root et al. 2003; Parmesan 2007). Miller-Rushing, Inouye & Primack (2008) have cautioned that first flowering dates can be unreliable for inferring phenological shifts, especially if populations are small or observations infrequent. The change in flowering time reported here was identified from peak flowering dates in a large population, which is a strong basis for inferring changes in phenology (Miller-Rushing, Inouye & Primack 2008). Despite variation, peak flowering advanced by c. 0.5 day year−1. Although O. sphegodes can self-pollinate and is cross-pollinated by the rare solitary bee A. nigroaenea (Schiestl et al. 1997), fertilization rates are low in the UK (Summerhayes 1951; Lang 1980). Unless the phenologies of O. sphegodes and A. nigroaenea respond similarly to climate, there is a risk of even lower fertilization in future, posing an additional threat to O. sphegodes. The onset of flowering was later after winters with more frosts, and earlier after warmer years. The strongest relationship was with mean temperature over the whole year prior to flowering, rather than just when parts of the plant were above-ground (September onwards) or during inflorescence extension (March–April). This suggests that, even when the plant is below ground, climate still exerts either a direct influence on the plant itself or an indirect influence via its mycorrhizal associates (Wells et al. 1998).

New tubers are produced annually in the genus Ophrys. Inflorescences with embryonic flowers are visible in new tubers in November prior to flowering, and may be present as early as the preceding summer (Wells & Cox 1989). Unless the plant’s state (flowering, vegetative, dormant) in the previous year, or its capacity to produce a tuber, is affected by conditions before that, behaviour is only likely to be influenced by climate after tuber formation. The prediction that climate >12 months before flowering would not influence behaviour was supported. In contrast, the proportions of the total population (i.e. emergent plus dormant plants) in different states were significantly influenced by mean temperature over periods of up to a year before flowering (Table 1). Higher temperatures were associated with a lower proportion of the population flowering, and a higher proportion being vegetative. Similar effects have been reported in Herminium monorchis (Wells et al. 1998) and Gymnadenia conopsea (Øien & Moen 2002). These species produce leaves in spring and flower in summer, whereas O. sphegodes produces leaves in autumn, and flowers in spring. The proportion of plants flowering in H. monorchis and G. conopsea populations was also positively correlated with rainfall prior to the flowering season, but no such correlation was found for O. sphegodes. It has been suggested that lower rainfall and higher temperature in spring limit growth and hasten leaf senescence in H. monorchis and G. conopsea, resulting in less carbohydrate to support summer flowering (Wells et al. 1998; Øien & Moen 2002). In O. sphegodes, inflorescence height and mean leaf number in the flowering season were positively correlated with rainfall, and negatively correlated with sunshine hours, although the latter correlations were not all significant. The genus Ophrys has its centre of distribution in the Mediterranean, suggesting that it should be adapted to wet winters and hot, dry summers. In light of this, the results presented here show some unexpected effects of climate on O. sphegodes, with no effect of winter rainfall on the proportion of the population that flowered, but higher rainfall at the time of flowering being associated with the production of larger inflorescences, and higher temperatures and more sunshine, respectively, tending to have negative effects on the proportion of plants flowering and on inflorescence height. At the same time, years with high proportions of flowering plants were also years when plants were largest, as judged from mean flower numbers and inflorescence heights (Table 2).

The proportion of O. sphegodes plants in dormancy each year was not correlated with the climate variables examined. This contrasts with results for orchid species that invest heavily in vegetative propagation. For example, the proportion of dormant ramets in Cypripedium calceolus ssp. parviflorum was positively correlated with both the number of spring frosts and precipitation, and negatively correlated with spring temperature (Shefferson et al. 2001), and more ramets of Cleistes bifaria were dormant after dry years (Kéry et al. 2005). Outside the Orchidaceae, wet summers and dry autumns increased dormancy in Silene spaldingii (Caryophyllaceae) (Lesica & Crone 2007). Clearly, it is not yet possible to draw general conclusions about the control of vegetative dormancy by climate in herbaceous species.

Population behaviour

Hutchings (1987a) reported that most O. sphegodes plants had very short lives from first appearance until their last observation; newly recruited cohorts of O. sphegodes had half-lives ranging from 1.5 to 2.3 years from first appearance. This agrees closely with the values of 1.6 to 2.8 years calculated over the longer time frame of the current study. Age-specific survival data for all cohorts yielded a half-life of 2.25 years, making O. sphegodes one of the shortest-lived orchid species studied to date (Hutchings 1989b; Kull 2002) although, as also reported by Hutchings (1987a), a few plants survive for well over a decade from first appearance.

The distribution of plants between emergent and dormant states, flowering and vegetative states, and different dormancy categories was highly variable, even between years with the same management. Part of the variation was associated with the rapid increase in population size during the late 1980s and early 1990s, which was both preceded and accompanied by a decline in flowering in the emergent population. There was also a sharp decrease in the proportion of new recruits flowering, commencing shortly after the start of management by sheep grazing. During the first 10 years of the study, means of 84% of the emergent population flowered (Hutchings 1987a) and 71% of newly recruited plants flowered on first appearance (Hutchings 1987b). Mean values calculated across the whole study showed that only 56% of emergent plants flowered and 52% of new recruits flowered on first appearance. After the increase in population size, a higher proportion of the population was in the dormant 1 category. Towards the end of the study, there was an increase in the proportion of emergent plants that flowered and a higher proportion of dormant plants were in later years of dormancy. Some very long periods of dormancy (4–8 years) were observed, as in some other orchid species (Kull 2002). During the first 10 years of the study, c. 50% of the population was dormant (Hutchings 1987a), but by the end of the study this figure had fallen to <30%. The dramatic temporal variation in population state structure suggests a need for caution in predicting future population behaviour from small-scale, short-term demographic studies.

Little is known about the time terrestrial orchid species spend below ground between germination and first emergence (Wells 1981; Kull 2002). It has been suggested that Ophrys apifera can produce leaves after only 1 or 2 years (Ziegenspeck 1936), and in O. sphegodes there were significant correlations between the number of new recruits in 1 year and the numbers of flowering plants recorded 1 and 2 years earlier, suggesting that the subterranean phase can be similarly brief. In most years, flowering plants of O. sphegodes left an average of <1 recruit, with occasional years of higher recruitment. Throughout the study, assuming 1- and 2-year delays between seed production and emergence of recruits, there were, respectively, means of only 1.1 and 1.3 recruits per flowering plant. Many of these recruits did not flower in their first year. An inverse relationship between the proportion of new recruits that flowered and their number (Fig. 6d), and a positive relationship between grazing of inflorescences (probably by rabbits) and the number of inflorescences produced (Fig. 6b), suggests that successful flowering is inversely density-dependent, even though the mean densities of new recruits and of plants initiating flowering never exceeded 1 and 1.5 m−2, respectively. Hutchings (1987a) reported that c. 65% of O. sphegodes plants flowered only once before dying, and a substantial proportion die without ever flowering. Thus, this population appears to be almost constantly at risk of failing to replace itself with descendants. The inevitable conclusion is that its survival is still precarious despite the increase in its size during the late 1980s and early 1990s. This conclusion is borne out by the analysis of population flux, which shows mortality considerably exceeding recruitment from 1995 to 2006. There were striking changes in population flux after the change in management from cattle grazing to sheep grazing (Hutchings 1987a,b). Annual mortality was high under cattle grazing, annual recruitment fell progressively as the years passed, and the population accumulated a growing numerical deficit. When sheep grazing was introduced, recruitment exceeded mortality each year, resulting in a net increase in plant numbers from 1975 to 1984. However, the emergent population did not become larger until many years after the change in management (Hutchings 1987a), when a rapid and spectacular increase in numbers was observed. Detailed flux analysis (e.g. Fig. 5c) reveals that there was both a growing excess of recruitment over mortality, and a slow increase in the number of dormant plants, prior to the increase in the emergent population (Fig. 2a,e).

Lauenroth & Adler (2008) have commented that plant ecologists still have very few data on the three demographic traits of survival, life expectancy and life span. This study reports details of all three (see also Hutchings 1987a,b). However, these data come from a single population at the edge of O. sphegodes’ current geographical range. Ideally, comparative values from other populations would have been obtained, because demographic properties can differ between populations and environments (e.g. Davison et al. 2010). Some results from the first 10 years of the study, such as the estimated half-life of O. sphegodes, remain robust after the collection of additional data, whereas others, such as how long plants can be dormant and still reappear, the proportion of the population in dormancy, the proportion of emergent plants that flowered and the proportion of new recruits that flowered on first appearance, changed considerably.

Implications for demographic studies and for management and conservation of O. sphegodes

The combination of (i) a subterranean phase, during which mortality may be high and reproduction is impossible, (ii) a short life span, with few flowering episodes before death, (iii) a low level of fertilization and (iv) limited vegetative propagation inevitably means that periods of abundant recruitment from seed, such as that observed between 1989 and 1995, are vital to the persistence of O. sphegodes. Hutchings (1989b) has previously described O. sphegodes as a ‘weedy’ species. Given its life-history and demographic traits, management for the conservation of this, and other orchid species with similar characteristics, should aim at favouring seed production and seedling recruitment. Previous management recommendations (Hutchings 1987a,b) remain broadly valid; grazing prevents more aggressive and taller species from outcompeting O. sphegodes, and sheep grazing is more beneficial than cattle grazing. Cattle inflict greater mechanical damage on the habitat and vegetation than sheep, causing high mortality and low recruitment in the orchid population. The grazing animals should be withdrawn from the habitat during flowering and seed set. However, recruitment during the late 1980s and early 1990s may have been promoted by the presence of gaps created during the cattle grazing regime. Some artificial gap creation might therefore be beneficial. It should be emphasized that the population will rapidly decrease in size under inappropriate management, simply because the species is short-lived and dependent on regular recruitment. Flowering, seed production and seedling establishment are vital, annually, to ensure population persistence.

Some aspects of the demography of O. sphegodes, including recruitment and mortality, were also strongly influenced by conditions that cannot be altered by management. Annual recruitment was positively correlated with temperature in each of the previous 2 years, and mortality was positively correlated with temperature in the previous year. Thus, the same climate variable influenced recruitment and death in a similar way. Winters with fewer frosts were associated with more recruitment in the following year, suggesting that climate warming may promote establishment and perhaps range extension in the UK.

This study also shows that population behaviour is very variable between years. Such variability may be more characteristic of short-lived species like O. sphegodes than of longer-lived species, but there are few data for comparison. Whether this is true or not, the insight gained from this long-term study suggests that predicting future population performance from short-term studies involving small numbers of plants is likely to be subject to considerable error (Bierzychudek 1999). Such predictions should therefore be treated with extreme caution.

As one of the longest and most extensive investigations conducted to date on any plant population, this study joins a catalogue of long-term demographic studies hitherto almost exclusively limited to animal species. Future papers will report detailed analyses of further aspects of population behaviour, and predictive modelling of population responses to different management and climate regimes.


I am grateful to Fiona Coates, Ana Mendoza, Roy Turkington and Steve Waite for comments on an earlier draft of this manuscript, and to two anonymous referees and the Handling Editor for their suggestions for improving the submitted version. I am especially indebted to many people who helped in the collection of data during the course of the study, including Dave Aplin, Cat Back, Richard Bickers, Karen Booth, Chris Budd, Kieron Day, Dave Graham, Chris Hance, Ben Haines, George Holdsworth, Ana Mendoza, Vicky Newington, Jo Nightingale, Liz Price, Andy Slade, Jun-Ichirou Suzuki, Steve Waite, Penny Wallis, Kevin Warwick, Belinda Wheeler and, especially, Martyn Stenning. I am also greatly indebted to Kevin Warwick for help with processing of field records during earlier years of the study.