Our study population of Platanthera chlorantha occupies a strip of calcareous downland and gradually encroaching scrub, situated at the foot of the north-facing slope of Wolstonbury Hill. Managed by the National Trust, the Hill forms a salient along the South Downs chalk escarpment of East Sussex, 10 km North of Brighton, England (N 50°54’, W 0°10’; ca 103 m asl). The Wolstonbury population was first utilised scientifically in 2007, when it contributed to a citizen science project designed to explore variation across Europe in spur length in both P. chlorantha and P. bifolia (Bateman and Sexton, 2008; Bateman and Sexton, 2009; Bateman et al., 2012). In 2008, the initial short-term project was expanded to include a further three morphological variables plus annual counts of the number of flowering plants ( Table 1 ).

By the end of data collection in 2022, the Wolstonbury population of P. chlorantha was estimated by us to consist of at least 500 mature individuals. However, the population extends over an area exceeding one hectare; this fact, combined with habitat disturbance resulting from constant public access, soon dissuaded us from pursuing more than superficial attempts to monitor the long-term performance of individual plants. We were therefore obliged to focus on annual fluctuations in the overall performance of the population, an approach that at least offered the advantage of reliably delivering a statistically robust sample size (e.g. Lindley, 1997).

Our study was entirely non-destructive. Morphometric data were collected in situ in mid-June of 16 successive years between 2007 and 2022. The chosen date of measurement varied from June 1st in unusually early seasons to June 23rd in unusually late seasons, the mean date being June 11th. Following an initial count of numbers of flowering plants, only plants approximating full flower were included in the otherwise randomly selected annual sample of 60 plants (though only 36, 25 and 45 plants were measured in 2007, 2008 and 2009, respectively – a period when encroaching brambles were gradually cleared from the site by conservation volunteers). In total, 886 data-sets were acquired, collectively yielding 3472 morphological data points.

The four traits selected by us reflect a desire for relative ease of measurement combined with a desire to examine as many properties of the plants as possible using as few traits as possible. Many prior morphometric studies performed by RB and colleagues on populations of a wide range of European orchid species, typically scoring approximately 50 characters per plant (e.g. Bateman and Rudall, 2011; Bateman et al., 2013), had already identified consistent patterns of variation among contrasting traits in several European orchid genera, including Platanthera (Bateman et al., 2012; Bateman et al., 2013).

The strongly nocturnally scented white flowers of P. chlorantha are typical of orchids pollinated predominantly by moths ( Figures 1A–C ). The precise length of the extensive nectar-rich spur is widely regarded as influencing the spectrum and efficiency of pollinating insects (e.g. Claessens et al., 2008; Claessens and Kleynen, 2011; Esposito et al., 2017). It is therefore important that it should be measured with precision ( Figure 1D ); details of the optimal procedure were given by Bateman and Sexton (2008). Because orchid flowers reliably diminish in size from the base to the apex of the inflorescence (Bateman and Rudall, 2006), for each plant the spur was measured on a flower located midway along the inflorescence. The number of flowers (including any unopened buds) on the inflorescence was recorded, being the trait that largely determines the reproductive output of the individual. From 2008 onward, we also measured stem height from ground level to the tip of the inflorescence ( Figure 1A ). This character not only determines inflorescence density but also the height at which the flowers are presented to prospective pollinators and, subsequently, the distance over which the resulting dust-seeds are likely to disperse from the dehisced seed-pods.

The final trait to be measured was the maximum width of each expanded leaf (as seen in the lower part of Figure 1A , almost all mature plants produce a pair of expanded basal leaves, though occasionally a comparatively weak plant will produce only one). Measuring this trait was simpler than attempting to estimate in the field the aggregate surface area of the leaves, which would have constituted the best proxy for photosynthetic potential. In order to gain a more precise yardstick for leaf area, we therefore measured the maximum width and length of 22 leaves sampled in 2020, before using pixel numbers in scanned outlines to determine their precise areal extent. We were thus able to determine that multiplying maximum width by length yielded on average a figure that exceeded the true areal extent of the leaf by 30%, and that the squared value of the maximum width yielded a figure that was 60.9 ± 8.5% of the true areal extent. This figure formed the basis of our post hoc conversion of measured actual width to yield an estimate of the areal extent of each leaf, subsequently summating the values of the (usually) paired leaves to reflect the likely relative photosynthetic potential of the plant in question.

The monthly figures summarized in Table 1 for rainfall, sunshine and maximum daytime temperature during the main growth period of March–June throughout the 16-year period of study were downloaded from internet repositories for the viable meteorological stations located closest to Wolstonbury. Data for rainfall were obtained from Plumpton Agricultural College, located 6 km east of Wolstonbury Hill and representing a broadly similar aspect and altitude. Unfortunately, the nearest site providing reliable sunshine and temperature data, along with total frost days per winter, was the meteorological station at Eastbourne, 36 km ESE of Wolstonbury, where the lower altitude and coastal location are likely to have yielded sunshine figures somewhat higher, and frost figures marginally lower, than the number of hours experienced by the Wolstonbury plants. Comparison of March–June rainfall data from the two stations gathered since 2005 showed rainfall to be on average 29% lower at Eastbourne, though intriguingly, the respective annual means converged significantly during the final four years of the present study.

Microsoft Excel was used to calculate mean, sample standard deviation and coefficient of variation for each of the nine variables included in Table 1 , for each of the 16 years, together with mean values for the entire time-series. Linear correlations were performed in Excel (v15.40 and v16.70) before being double-checked and presented in Red Rock’s Deltagraph (v7.1.3) package. All six possible pairwise combinations of the four morphological traits were plotted for all plants measured in each year of study.

This exercise was then repeated for annual mean values for each trait, together with the number of plants that flowered annually in the population, thus requiring ten pairwise combinations. The comparisons were conducted not only on a same-year basis (year t) but also with the data staggered to allow comparisons with the previous year (t–1) and the year before that (t–2), thereby increasing the number of valid comparisons to 65.

Although this project was conceived partly to simply explore the properties of the four contrasting morphological traits, it also aimed to address more ambitious questions, such as whether climatic conditions and/or photosynthetic potential during the critical period of one annual growth cycle may impact on plant performance during the following annual growth cycle. Therefore, annual mean values for the four morphological traits across the 16 years of study, together with number of flowering plants within the population, were plotted against aggregate March–June rainfall (at Plumpton) and aggregate March–June sunshine, mean maximum monthly temperature, and number of frost days (at Eastbourne).

Tabulated r values for all correlations were subsequently subjected to two-tailed p tests (with 58 degrees of freedom for samples of 60 plants in the case of within-year comparisons, and 14 degrees of freedom for among-year comparisons of the 16 successive years of mean values).

Seeking a more sophisticated check on conclusions drawn from these simple linear regressions, we then repeated our time-displaced comparison of each trait with each of the four categories of meteorological data through autocorrelation – specifically, applying temporal cross-correlation methods to covariance values in order to generate sliding dot products of values scaled to a range between –1 and +1 relative to the average for the recording period (e.g. Derrick and Thomas, 2004; Tahmasebi et al., 2012).

Given that our study is confined to a single population of P. chlorantha, it is important that we demonstrate that it is typical of the species as a whole if our conclusions are to have more general validity. Certainly, the habitat – calcareous scrub and grassland – falls within the normal range. Admittedly, the absence of data for Wolstonbury plants on DNA-based properties or common-garden experiments means that we are unable to directly address issues such as plasticity, thus confining our comparisons to morphological traits.

Fortunately, we have access to directly comparable morphometric data gathered by Bateman et al. (2012) from 79 plants collectively representing 13 English populations of P. chlorantha. Comparing mean values for those other populations with the overall mean values for Wolstonbury summarised in Table 1 shows acceptably close similarity: respective values are 32.5 mm (reduced to ca 30 mm if only open-habitat populations are considered) versus 29.0 mm for spur length, 12.5 versus 13.5 flowers per inflorescence, 36.7 mm versus 36.8 mm for stem height, and estimated leaf areas of approximately 60 cm² versus 52 cm². A further slight difference apparent in the data was that 3.8% of the plants measured by Bateman et al. (2012) bore three leaves rather than two, whereas such plants represented only 0.1% of the Wolstonbury population; there, 1.6% of plants bore only one leaf ( Table 1 ).

It is also important to rule out the possibility that the Wolstonbury population could have been rendered atypical of the species through past hybridization. Although Platanthera L.C. Rich. is a genus of approximately 60–80 species (cf. Hapeman and Inoue, 1997; Bateman et al., 2009; Efimov, 2016), only two of those species are confirmed as occurring naturally in the British Isles (e.g. Bateman, 2022). Admittedly, they frequently hybridize when co-occurring (e.g. Mötlep et al., 2021), but the closest population of the second British species, P. bifolia, throughout the last quarter-century has lain 25 km to the west of Wolstonbury Hill (BSBI DDb database, consulted February 2023) – a distance sufficient to give confidence that the Wolstonbury population is genetically cohesive.

The morphology of an organism is dictated by its genetics as interpreted and modified through epigenetic and environmental factors (ecophenotypy) and through the degree of maturity and vigour already achieved by the organism (ontogeny).

In the present study, early stages in the life history from mycotrophic seed germination to mature tuber (e.g. Rasmussen, 1995) must be set aside, as our data relate only to mature individuals able to develop an inflorescence. Once having reached that stage of maturity, the Butterfly-orchid’s life history ( Figure 7 ) is one of annual somatic replacement (e.g. Tatarenko and Kondo, 2003). Having over-wintered, the fusiform root-stem tuber produces a shoot that breaks ground in early spring, simultaneously generating a pair of basal leaves, and a radiating horizontal set of stem-roots, throughout the spring. At the same time, the over-wintered tuber begins to generate one or more new tubers. In England, plants that have “decided” to flower typically begin to elongate their stem above ground in April or early May; the shoot initially bears a very few small bract-like leaves before reaching anthesis in June. Successfully fertilized ovaries swell during June and July, at which point resourcing of the new root-stem tuber and its associated terminal root is also completed below ground. The ovaries dehisce their dust-seed in August and September, as that year’s vegetative growth – both above and below ground – withers and then gradually decays ( Figure 7 ).

Obviously, the timing of annual measurement is important in ensuring consistency when gathering data on properties that are likely to differ only subtly from year to year. By the time we conducted our annual measurements in early–mid June (e.g. Figure 1A ), leaf growth had ceased, and the number of flowers per inflorescence had long since been determined (see below). However, gradual stem growth continues during anthesis, as does spur length at a lower, more subtle level of variance (Bateman and Sexton, 2008; Bateman et al., 2012). These features therefore constitute a (slowly) moving target and require carefully timed measurement. A recent multifaceted study by Cullen et al. (2023) shows that early stages of spur elongations reflect enhanced rates of cell division but that the later stages that are of interest in the present study largely reflect cell elongation, supporting a conclusion made earlier by Bateman et al. (2012) using Platanthera as their preferred model.

Coefficients of variation for the four traits measured here echo both relative and absolute values for these characters in previous morphometric studies of tuberous temperate orchids (e.g. Bateman and Rudall, 2011; Bateman et al., 2013).

The substantially lower coefficient of variation for spur length relative to the remaining three morphological traits reflects contributions from several different factors. Dimensions of the components of a flower are relatively highly constrained, as the flower is a disposable unit that reflects determinate growth and terminates a developmental cascade within the inflorescence. In addition, the relatively short period of the year allocated to flower development in general, and spur expansion in particular, limits opportunities for environmental modification of its mature morphology. Moreover, conventional wisdom states that the size of the overall floral display and the size and shape of those floral components that influence pollination success, notably the gynostemium and nectar-rich labellar spur, will be at least periodically subject to stabilizing selection, driven by the necessity of assuring frequent reproduction through seed dispersal (cf. Boberg and Ågren, 2009; Boberg et al., 2014; Ambroise et al., 2020; Brzosko et al., 2022). This innate conservatism inevitably limits variance in floral morphology, and therefore in our study it weakened all correlations that involved spur length – a phenomenon that is readily apparent in Tables 2 – 5 .

Nonetheless, despite its comparatively low variance, spur length retains some residual interest. Previous studies of spur length in both P. chlorantha and P. bifolia (sister species to P. chlorantha: e.g. Bateman et al., 2009) showed that populations occupying shaded habitats have longer spurs than those in the open, and also revealed a latitudinal gradient of 2.2% increase in spur length per 100 km, as estimated along an approximate North–South transect from Scandinavia to North Africa (Bateman and Sexton, 2008; Bateman and Sexton, 2009; Bateman et al., 2012). Using data for 61 populations of P. chlorantha accumulated by Bateman et al. (2012), we can now demonstrate that labellar spurs of 21 populations growing in shaded habitats were on average approximately 12% longer than spurs of 40 populations occupying more open habitats – a figure that is reduced to 10% when adjusted for latitude. Indeed, this figure approximates the coefficient of variation calculated annually for the Wolstonbury population ( Table 1 ), suggesting that preferred habitat has as great an influence on spur length as do the aggregate effects of innate properties such as tuber size and genetic composition.

We are unable to explain three statistically significant (albeit modest in scale) correlations involving spur length: the same-year positive correlation between spur length and stem height observed during six of the 15 relevant study years ( Table 2 ; Figure 2B ), and the apparent modest increases in stem height and flower number two years after a short-spur year ( Tables 2 , 3 ; Figures 3G, H ). The equally modest but statistically significant negative correlation observed between spur length and spring insolation ( Tables 4 , 5 ; Figure 4A ) may relate to details of the process of spur growth, which is believed to reflect not only cell division but also cell elongation (Bateman and Sexton, 2008). The longer spurs of plants growing in shade suggest a minor, small-scale effect analogous with the etiolation observed in entire plants growing in shade. If so, this effect could represent the antithesis of the slightly reduced spur lengths evident in data for those years when the plants were subjected to unusually sunny springs.

Within particular years, we expected a particularly strong positive correlation between stem height and flower number, as previous studies have suggested that in terrestrial orchids the proportion of the stem allocated to flower production is usually fairly consistent, measured at between one quarter and one third (e.g. Willems, 1978; Bateman et al., 2012; Bateman et al., 2021). Bateman et al. (2012) showed that the inflorescence of P. chlorantha averages a relatively stable 24 ± 6% of total stem length. Past studies of several other European orchid genera also show that spacing between flowers within inflorescences is fairly consistent among plants, thereby achieving an equally consistent floral display (e.g. Bateman et al., 2013; Bateman et al., 2021). We were therefore surprised to find that the inflorescence length–flower number relationship proved to be weaker on average than relationships between stem length and flower number or especially between stem length and leaf area ( Tables 2 , 3 ; Figures 2 , 3 ).

Given that the leaves emerge before the stem expands to form an inflorescence, we also predicted a comparatively strong positive same-year relationship between leaf area and stem height, assuming that larger leaves would help to resource a taller stem and perhaps a larger number of leaves. This positive correlation was statistically strongly significant, both for individuals within years ( Figure 2F ) and population means analysed among years ( Figure 3A ). Wells and Cox (1989) were able to predict the likelihood of successful flowering of a plant of Ophrys apifera based on the number of rosette leaves that it produced (unlike P. chlorantha, its leaves – four or more in number in flowering individuals – are produced in the previous autumn). However, given that 98% of mature P. chlorantha plants bear two leaves, we were not afforded an opportunity to test this approach to predicting subsequent performance. Janeckova and Kindlmann (2002) found leaf area to be a strong predictor of whether individual plants of the terrestrial orchid Dactylorhiza fuchsii would flower during the following summer.

Overall, the behaviour of leaf area appears more complex than we initially anticipated. It was no surprise to find a strong positive same-year correlation between leaf area and stem height ( Figure 3A ), but the negative same-year correlation between leaf area and frequency of flowering ( Figure 3E ) demands a more complex explanation. We suspect that in years of few flowering plants, only the most vigorous plants that have produced the largest leaves are able to successfully generate flower spikes, whereas in fecund years, plants that are somewhat less well-resourced are also by some means encouraged to flower. In order to explore these intriguing issues further it would be necessary to measure leaf area of all the non-flowering plants present in the orchid population, in order to determine whether non-flowering plants that are relatively large-leaved in year t increase their tuber size and then proceed to flower preferentially in year t+1.

The low level of variation evident throughout the study period in mean spur length and mean flower number per inflorescence ( Table 1 ) shows that there is little year-on-year difference in the amount of resources expended on reproduction by a typical flowering individual. It might be argued that contrasts among years in the availability of potential pollinators could nonetheless impact strongly on reproductive output. However, although this might be true of pollinator-limited orchids that rely on food deception or sexual deception, P. chlorantha offers a substantial nectar reward to any insect competent to access the apical half of the long labellar spur. Claessens and Kleynen (2011) listed as observed pollinators of P. chlorantha an impressive 29 species of 15 insect genera. They also reported fruit-set figures for 26 populations that we calculate collectively average 78 ± 13%. This figure is typical of food-rewarding (and also of autogamous) temperate-terrestrial orchid species, and is much higher than fruit-set frequencies calculated for orchid species that indulge in food deception or sexual deception (Neiland and Wilcock, 1998; Bateman, 2022). Substantial annual fluctuations in frequency of fruit-set are therefore unlikely.

Hence, we suspect that contrasts in annual overall reproductive effort are dictated almost entirely by the proportion of mature plants in the population that “choose” to flower in any particular year – a figure that proved highly variable at Wolstonbury ( Table 1 ). This conclusion renders particularly intriguing our evidence that both relatively low spring rainfall and high spring sunshine strongly encourage plants to flower, but only when these environmental cues have been experienced by the plants during the previous spring ( Figures 4B, D ) when they were utilising the resources of the previous tuber ( Figure 7 ). Destructive studies of numerous plants would be required in order to determine whether these apparently favourable climatic conditions in year t–1 lead to preferential resourcing of that year’s crop of replacement tubers, thereby increasing the total population biomass for deployment in year t.

The t–1 results highlighted in Figure 4 do strongly indicate that the “decision” to flower is taken by a plant during formation of the new tuber, almost a year prior to its actual flowering. In this context, the embryonic inflorescence in P. bifolia, sister-species to P. chlorantha, was described by Tatarenko and Kondo (2003) as forming its inflorescence within the tuber during the autumn of the year that immediately preceded flowering. We have therefore initiated a programme of micro-dissection of early-stage above-ground shoots in the genus Dactylorhiza, which is closely related to, and has a very similar life history to, that of Platanthera (e.g. Bateman et al., 2018) but constitutes a more amenable experimental organism. It is possible that the number of flowers to be borne by the embryonic inflorescence is also determined during the previous summer.

If the commitment to flower is indeed made in the previous year, we suspected that a secondary “failsafe” mechanism might exist that offers the plant the ability to rescind that decision in the face of disadvantageous conditions encountered during the following spring. Such a mechanism would allow the plant to generate and retain leaves to aid immediately subsequent tuber formation but would permit early-stage abortion of the embryonic flowering stem under unsatisfactory environmental conditions. However, were this hypothesis correct, we would have expected to find at least occasional statistically significant same-year correlations between climatic variables versus stem height, flower number and/or leaf area, but we did not ( Table 4 ).

An alternative mechanism of de facto abortion was reported by Sanger and Waite (1998) after monitoring a population of the tuberous European orchid Ophrys sphegodes. They observed frequent abortion of the entire above-ground parts of plants of prior to flowering in years when the population experienced suboptimal growing conditions during that same spring. Stroh (2019) suspected that similar casualties, perhaps caused primarily by frosts, occurred in his long-term study population of the orchid Anacamptis morio. Our approach to data collection was not designed to track premature leaf loss, which may similarly affect Platanthera, though the fact that Ophrys species produce leaf rosettes that differ from the paired leaves of Platanthera in being winter-green may increase their vulnerability to harsh winter climates.

Such thoughts encouraged us to take one final look at our demographic data. Specifically, we find particularly remarkable the relationship through time between spring rainfall and flowering success that is evident in Figure 5 . These graphs give the impression of a background state of comparatively low numbers of flowering plants broadly tracking spring rainfall. But two sudden radical rises in numbers of flowering plants are also evident, occurring ten years apart. Both are initiated in a relatively dry spring but then persist at a high level for two further years irrespective of amount of spring rainfall (compare the low rainfall of 2011 and exceptionally high rainfall of 2012, both accompanied by over 300 flowering plants). Do 2010 and 2020 – the years of exceptionally high recruitment to the cadre of plants successfully flowering – also represent years when there was a large-scale increase in the number of plants that possessed tubers sufficiently large to permit flowering? Is it even possible that the periodicity between the two phases of mass flowerings actually reflects successful establishment from seeds of large numbers of juvenile plants about ten years previously, stimulated by environmental conditions (more likely in the autumn than the spring) that proved unusually conducive to successful pairing of seeds with appropriate mycorrhizae and/or root-bacteria?

Unfortunately, our approach to data collection precluded estimation of the half-life of Platanthera chlorantha. Until a long-term demographic study is conducted that monitors specific plants from year to year, we cannot estimate the half-life of a P. chlorantha plant, which would have helped us to interpret the kind of complex patterns summarized in Figures 5 , 6 . Data on broadly comparable species yielded surprisingly short half-life estimates of approximately two years for Dactylorhiza viridis (Willems and Melser, 1998), Neotinea ustulata (Tali, 2002) and Ophrys sphegodes (Hutchings, 2010). However, the majority of tuberous terrestrial orchid species – including Orchis anthropophora, Spiranthes spiralis (Wells, 1981) and Ophrys apifera (Wells and Cox, 1989) – approximate six to seven years. Our anecdotal observations on the Wolstonbury population would support a similar figure for P. chlorantha.

Our demographic investigation of Platanthera chlorantha makes an interesting comparison with that performed during the period 1966–1995 by Wells et al. (1998) on two populations of Herminium monorchis. Our results suggest that flowering of Platanthera is encouraged by low spring rainfall, high spring sunshine and perhaps also frequent frosts experienced in year t–1, whereas Wells et al. (1998) found that Herminium responded best to high rainfall, not only in year t–1 but especially in the subsequent year t (also responding positively to lower summer temperatures). Herminium favours habitats even more prone to desiccation than those occupied by Platanthera, and its flowering typically peaks one month later. Consequently, its two or three leaves often wither prior to flowering. And as already noted, tuberous orchid genera such as Ophrys and Spiranthes produce prostrate rosettes of several leaves during the autumn of year t–1 that are then obliged to withstand the succeeding winter conditions. When combined with different habitats preferences, such demographic contrasts are likely to strongly influence how particular species react to changing climate, suggesting that even vaguely accurate predictions of future behaviours will need to be species-specific.

We concluded data collection for this study intrigued to discover whether a 16-year run of trait data would reveal any directional trends through time that would leave the population in a condition materially different from that in which it was first examined. But the only vaguely directional trend that we were able to identify appears trivial – specifically, the absence of years incurring unusually high coefficients of variation in flower number and leaf area during the last six years of the study. We suspect that a far longer period than 16 years is needed to modify a population through directional selection, and that the spotlight of selection rarely shines sufficiently long on any of the traits we have studied to influence their time-averaged mean values. The primary threats to the survival of plant populations such as Wolstonbury are more likely to be changes in their immediate environment. Some such changes would appear catastrophic, such as an intensive wildfire or a radical change in the dominant grazing regime. More gradual environmental changes such as increasing atmospheric CO₂ levels would need to persist on a timescale of at least centuries in order to induce meaningful directional evolution.

In this context, we note with interest the strong, albeit largely indirect, influence of precipitation and insolation on reproductive success in our study population, and the similar inferences drawn from previous studies of genetic and epigenetic variation in other European orchids. As noted for an orchid genus closely related to Platanthera by Paun et al. (2010), “water availability in combination with temperature appears to be a key factor causing environmental allopatry in Dactylorhiza, being identified both at the epigenetic level and by transcriptome profiling” (as shown subsequently by Paun et al., 2011). Similarly, a recent study of the European rhizomatous orchid genus Epipactis by Evans et al. (2023) concluded that although a combination of temperature and altitude have strongly influenced genotypes close to the Mediterranean, further north it is precipitation that has been the primary influence. In summary, the current emphasis on research related to climate change appears justified.