Testing theories of mire development using multiple successions at Crymlyn Bog, West Glamorgan, South Wales, UK


  • P. D. M. Hughes,

    Corresponding author
    1. Department of Geography, University of Southampton, Highfield, Southampton, SO17 1BJ, and
    • Paul Hughes, Department of Geography, University of Southampton, Highfield, Southampton SO17 1BJ, UK (e-mail Paul.Hughessoton.ac.uk).

    Search for more papers by this author
  • L. Dumayne-Peaty

    1. Department of Geography, University of Southampton, Highfield, Southampton, SO17 1BJ, and
    2. School of Geography and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
    Search for more papers by this author


  • 1 Direct observations of long-term plant succession can be made using quantified plant macrofossil records from peat bogs, providing a means to re-evaluate theories of succession previously based on time-space substitution studies or field stratigraphic surveys.
  • 2 Multiple successions from fen towards raised mire recorded at Crymlyn Bog demonstrate that divergent pathways exist, even when initial conditions are similar within a single bog.
  • 3 Over time-scales relevant to the later stages of mire succession, allogenic forcing factors are significant and may be responsible for driving the direction and rate of species turnover in both forward and reversed hydroseral successions.
  • 4 Differences in the local climatic regime may be responsible for the contrasting character of the mid- and late-Holocene transitional mire communities represented in Core CRB93 at Crymlyn Bog.
  • 5 Plant macrofossil analyses show that Sphagnum is not always a dominant part of the mire community before the establishment of raised peats. Other species including Eriophorum vaginatum, may be equally important ‘ecosystem engineers’ at the fen–bog transition.


The early stages of both terrestrial and wetland successions are relatively well known because pioneer communities are often characterized by rapid species turnover, lending themselves to direct observational studies (e.g. Pearsall 1918; Godwin et al. 1974). By contrast, the founding theorists of plant succession struggled to describe long-term change because of the difficulties involved in tracking vegetation community turnover over millennial time-scales. Much of the theory relating to plant succession that has been so influential during the 20th century, was originally based on the principle that variation in space could be used as a substitute for change through time. Many studies have since shown that this methodology, termed the side-by-side approach by Zobel (1988), is not reliable (e.g. Walker 1970; Zobel 1988).

Modern palaeoecological techniques have equipped researchers with the appropriate means to re-evaluate theories relating to long-term vegetation change, using the unique and detailed archives of past biodiversity deposited in wetland systems (Barber 1993). Walker’s (1970) benchmark paper was the first major re-evaluation using field stratigraphy and palynological evidence, and demonstrated most notably that raised bog and not mixed oak forest developed from hydroseres in oceanic climates. Walker also demonstrated that the hydrosere could follow a number of different pathways and that reversals could occur. These generalized findings were based, not least, on the numerous field stratigraphic studies completed by Godwin (e.g. Godwin 1941) and Walker. Walker’s assessment of autogenic hydroseral succession, although a great step forward, was limited by the taxonomic precision afforded by pollen analysis and the uncertainties introduced by pollen transport. Plant macrofossil studies provide the most direct means of tracking long-term succession, but very few were then available. Furthermore, recent macrofossil studies of British raised mires (Leah et al. 1998; Hughes et al. 2000) suggest that it is often difficult to disentangle long-term autogenic succession from that brought about by allogenic forcing, such as climate change. Climatic regimes have varied continuously and significantly over time-scales relevant to the later stages of succession (Barber 1981; Barber et al. 1994). Such macrofossil records indicate that it may not be safe to assume that the later stages of a forward succession are principally driven by autogenic processes (sensuWalker 1970).

This paper aims to test directly whether wetland succession is deflected along new pathways as a result of allogenic disturbances, or merely set back to an earlier stage on a single pathway, by comparing the directions of multiple successions in a single mire. The null hypothesis states that: allogenic disturbance results in an initial retrogressive succession followed by renewed forward succession dominated by autogenic development. Preliminary investigations show that Crymlyn Bog contains an appropriate developmental sequence for the present study, as it records at least two major phases when the mire acidified, separated by a period of reversed succession, characterized by a return to Phragmites-dominated fen (Barber & Hughes 1995). These stratigraphic sequences effectively allow the ‘successional clock’ to be wound back and rerun. If allogenic inputs to the system, of sufficient magnitude to affect long-term succession, represent occasional isolated disturbances, which are followed by predominantly autogenic development, then the two fen–bog transitions should be very close in character. If, on the other hand, the effects of allogenic factors are both significant and continuous in directing both forward as well as reverse successions (sensuGleason 1926), then the observed mire sequences might be expected to differ substantially.


site description

Crymlyn Bog (SS6994) is a grade 1 Site of Special Scientific Interest, National Nature Reserve and RAMSAR site located in the Lower Swansea Valley, 4 km east of Swansea (Fig. 1). The modern surface of the peat complex lies at c. 7 m (OD) and could be classified as floodplain mire (cf. Giller & Wheeler 1988). The 268-hectare site (Gilman 1994) is the third largest fen in the United Kingdom and the largest in Wales. The surface vegetation was described in detail by Meade (unpublished data) and Headley et al. (1992). Much of the peatland is rich fen containing abundant Phragmites australis, Juncus spp. and Cladium mariscus. Betula and Salix carr form a significant part of the site, whilst some areas support open poor and intermediate fen vegetation. The latter community contains Eriophorum latifolium, a species of national conservation status (Meade unpublished data). Mire stratigraphy has been studied previously by Godwin (1940) and Barber & Hughes (1995).

Figure 1.

Site map of Crymlyn Bog showing positions of transects A and B and coring locations, with inset map showing location of the site in Wales.

field methods

Core CRB93 (Fig. 1) was retrieved from one of the deepest parts of the mire (956 cm) using a 30 × 10 cm Russian pattern corer (Barber 1984). Preliminary macrofossil analyses of this core suggested a complex history of mire development warranting further investigation. Consequently, two transects (Fig. 1) were established across Crymlyn Bog to investigate spatial variability in mire development, and whether the stratigraphy of CRB93 could be replicated across the bog. Eight cores were taken using a 50 × 5 cm Russian corer at approximately 250-m intervals on a transect from south-west to north-east (transect A), with eight further cores being obtained from a north to south transect (transect B). The south-western end of transect A was surveyed to base level using an electronic distance measurer, but dense tree cover over the rest of the site prevented further levelling.

laboratory methods

The cores from transects A and B were analysed using a stereozoom microscope at × 10 magnification to determine the composition of the peat sequences. The properties of the peat were logged using the classification system formulated by Tröels-Smith (1955).

Macrofossil analysis of core CRB93 was performed using the quantitative quadrat and leaf count (QLC) macrofossil technique (Barber et al. 1994) on samples 4 cm3 in volume at 16-cm intervals, except between 648 and 560 cm and 332 and 360 cm, where analyses were at 8-cm intervals. Once the percentage cover of the main peat components had been established, the abundance of other small macrofossils (e.g. seeds, bud scales, seed boxes) was recorded on a five-point scale (rare = 1, occasional = 2, frequent = 3, very frequent = 4, abundant = 5). Quantification of Sphagnum and other bryophyte leaves was undertaken by mounting a random selection of at least 100 leaves onto a microscope slide using Aquamount, and identifying them at either × 100 or × 200 magnification. Where monocotyledons dominated the peat, at least 50 fragments were mounted onto microscope slides using Aquamount and identified at either × 100 or × 400 magnification. The macrofossils were identified with reference to Grosse-Brauckmann (1968, 1972), Katz et al. (1977), Smith (1978) and Daniels & Eddy (1990).

Pollen analysis of CRB93 was undertaken on samples of peat 1 cm3 in volume at 8-cm intervals, except between 164 and 44 cm and 44 cm and 0 cm, where analyses were undertaken at 4-cm and 2-cm intervals, respectively. The pollen was prepared according to standard methods (Barber 1976), although the siliceous material in the core between 952 and 792 cm was removed by leaving the samples in cold 40% hydrofluoric acid for 24 h. At least 300 dry land pollen (total land pollen minus Cyperaceae, Ericaceae and Poaceae < 25 m) were counted at magnifications of × 400 and × 1000 and identified using type slides and Moore et al. (1991). Pollen nomenclature follows Bennett et al. (1994) and plant nomenclature follows Stace (1991). Poaceae pollen was separated into Poaceae > 25 m and Poaceae < 25 m, as the latter was thought to be mainly pollen of Phragmites australis (Faegri et al. 1989).

The macrofossil and pollen diagrams were constructed using Tilia Graph version 1.25 (Grimm 1991) and zoned with the assistance of CONISS (Figs 5 and 6). Resultant dendrograms have been omitted to maintain clarity. Macrofossil zones are prefixed CRBm and pollen zones are prefixed CRB, although the depths of the zone boundaries on the macrofossil and pollen diagrams mostly coincide. Only the mire taxa with which this paper is concerned are shown on the pollen diagram, expressed as a percentage of themselves.

Figure 5.

Detrended correspondence analysis (DCA) biplot of species scores derived from core CRB93 macrofossil data. DCA axis 1 vs. DCA axis 2.

Figure 6.

Proxy-nutrient indices (DCA axis 1 sample scores from core CRB93 macrofossil data) plotted against depth for Crymlyn Bog. Individual sample scores are assumed to indicate the relative acidity of the mire surface. The zone lines from the macrofossil diagram and the date (Cal. years BP) of each zone boundary are shown.

radiocarbon dates

In order to provide an absolute chronology for CRB93, 10 samples were pre-treated at the Radiocarbon Laboratory in East Kilbride before being sent to the University of Arizona NSF-AMS laboratory for radiocarbon dating (Table 1, code AA). Where possible, assays were performed on individual leaf fragments of Phragmites australis that were thought to be stratigraphically intact. Two bulk peat samples from a monolith obtained at a location 1 m from CRB93 (76 cm depth (SWAN-92) and 54 cm depth (SWAN-91)) were radiocarbon-dated using conventional techniques at the Radiocarbon Laboratory, University of Wales, Swansea. The monolith and core were cross-correlated using pollen analysis to ensure that representative samples were dated (Rosen 1998). The dates were calibrated using INTCAL98 (Stuiver et al. 1998; Table 1) and unless otherwise stated, dates are given as calibrated years BP and are derived from extrapolation of average accumulation rates. The 2σ calibrated age range is also quoted. The two basal dates (AA-28123 and AA-28124) may have been contaminated by ancient carbon, whilst AA-28132 appears to be ‘too young’ for its stratigraphic position and may have been contaminated by modern organic material, and these three age estimates were therefore excluded when calculating accumulation rates. The use of average accumulation rates in this way is problematic because the peat is not of uniform composition throughout the core, but the approach is the best that can be achieved using the present data.

Table 1. Radiocarbon dates From Crymlyn Bog. Radiocarbon dates and their calibrated age ranges in years BP and BC/AD from Crymlyn Bog. Calibration was undertaken using INTCAL98 (Stuiver et al. 1998)
Laboratory codeDepth (cm)Material datedRadiocarbon age (yr BP)2σ calibrated range (yr Cal. BP)Mid-point of range (yr Cal. BP)2σ calibrated range (yr Cal. AD/BC)
  1. PL = Phragmites australis leaves, W = wood, SL = Sphagnum leaves, B = bulk peat.

AA-28123956PL12035 ± 75
AA-28124804PL/W10355 ± 65
AA-28125640W 5015 ± 505896–56525775 Cal. BP3946 Cal. BC – 3702 Cal. BC
AA-28126580W 4570 ± 505448–50455245 Cal. BP3498 Cal. BC – 3095 Cal. BC
AA-28127490W 3795 ± 504400–39884195 Cal. BP2450 Cal. BC – 2038 Cal. BC
AA-28128350SL 2490 ± 452729–23672350 Cal. BP 780 Cal. BC – 418 Cal. BC
AA-28129216SL 1920 ± 501950–17211835 Cal. BP   0 Cal. BC – Cal. AD 229
AA-28130164PL 1170 ± 451174–9671070 Cal. BPCal. AD 776 – Cal. AD 983
AA-28131108PL/W  980 ± 40 955–784 870 Cal. BPCal. AD 995 – Cal. AD 1166
AA-28132 44PLModern
SWAN92 74–76B  410 ± 40 520–316 420 Cal. BPCal. AD 1430 – Cal. AD 1634
SWAN91 52–54B  210 ± 60 424–0 210 Cal. BPCal. AD 1526 – Cal. AD 1955


analysis of core crb93

The major zones in the macrofossil and pollen diagrams for CRB93 are described in Tables 2 and 3, respectively. The macrofossil diagram (Fig. 2) depicts in detail the mire vegetation changes at this location. Species have been arranged so that fen types occur on the left and bog types on the right. Whilst many of the fluctuations in the pollen diagram (Fig. 3) accord with those displayed in the macrofossil record, changes in off-site (i.e. non-mire) vegetation will also be recorded in the pollen data by virtue of the regional pollen source area of Crymlyn Bog (sensuJacobson & Bradshaw 1981). Consequently, for pollen types such as Betula, Alnus glutinosa and some of the non-arboreal taxa (e.g. Plantago lanceolata, Filipendula, Rumex acetosa type), the extent to which the changes in the pollen spectra are a result of on-site or off-site vegetation change is sometimes unclear.

Table 2. Description of macrofossil diagram for CRB93 (Fig. 2)
ZoneDepth (cm)Date (Cal. yr BP)Description of main changes (Cal. yr BP)
CRBma800–7707280–6995Clay merges upwards into blue-grey clay intercalated with black organic laminae.
CRBmb770–6406995–5770 Phragmites is dominant, accompanied by Scirpus lacustris. Associated herbs: Menta aquatica, Juncus, Lychnis flos-cuculi, Hydrocotyle.
CRBmc640–5705770–5130After early dominance by Sphagnum palustre the assemblage changes to Carex spp. and Menyanthes trifoliata peat with Betula remains.
CRBmd570–5105130–4435Fen species largely disappear and are replaced by Eriophorum vaginatum, Calluna vulgaris and low levels of Sphagnum sect. Acutifolia.
CRBme510–3804435–2745 Myrica gale and then Phragmites/Carex spp. replace Eriophorum vaginatum. The reed/sedge peat also contains remains of Betula.
CRBmf380–3282745–2260The lower zone boundary is marked by Menyanthes trifoliata, Rhynchospora alba and Calluna vulgaris. Other bog taxa increase later.
CRBmg328–2322260–1895 Sphagnum imbricatum dominates, accompanied by Rhynchospora alba, Eriophorum vaginatum and E. angustifolium.
CRBmh232–1581895–1045 Sphagnum imbricatum declines to c. 10%. It is replaced by a mixture of Sphagnum papillosum and Rhynchosporaalba.
CRBmi158–1081045–870Aggregate monocotyledons reach 95%. Sphagnum and other mosses are absent. Phragmites australis increases to 90%.
CRBmj108–0 870–0 Phragmites australis dominates and fen herbs are also present. Erica tetralix and Calluna vulgaris remain but in small quantities.
Table 3. Description of pollen diagram for CRB93 (Fig. 3)
ZoneDepth (CM)Date (Cal. yr BP)Brief description (the main features of each zone)
CRBaa952–8008710–7280 Betula reaches 12% and Poaceae < 25 µm 30%. Chenopodiaceae rise to over 30%, later declining to 5%. Sphagnum spp., peaks at 12%.
CRBab800–7707280–6995The lower boundary marks the rational limit of Alnus, which later increases to 17%. Poaceae < 25 µm is dominant throughout.
CRBb770–6506995–5870Poaceae < 25 µm remains abundant (30–50%). Cyperaceae increases significantly at 750 cm, peaking at 55%.
   Chenopodiaceae remain common. Rumex, Lotus type, Galium type and Lynchnis type are well represented. At 655 cm Sphagnum spores increase to 90% of TLP.
CRBc650–5705870–5130Monolete Pteridophyte spores and Sphagnum spores increase and Calluna vulgaris rises to a peak of 20% before declining again later. Betula increases through the early part of the zone to a peak of 45%, whilst Poaceae < 25 µm declines to less than 10% of the pollen sum.
CRBd570–5105130–4435 Calluna peaks at 30% TLP. Betula and Alnus frequencies decline, whilst Sphagnum representation falls to less than 5% of the pollen sum.
CRBe510–3804435–2745 Betula representation increases steadily from 20% to 70% of the pollen sum. The base of the zone is defined by a short-lived but substantial increase in Sphagnum sprores. The middle and upper parts of the zone are characterized by peaks in Poceae < 25 µm and then Cyperaceae.
CRBf380–3282745–2260 Calluna vulgaris rises to 15% of TLP, accompanied by an increase in Alnus. Betula declines to between 20% and 40%, whilst Sphagnum increases.
CRBg328–2322260–1895 Sphagnum spores increase to 60%. Thereafter, Sphagnum varies between 10% and 30%. Calluna vulgaris pollen fluctuates between 1% and 12%.
CRBh232–1581895–1045 Betula fluctuates, reaches 33%. Poaceae < 25 µm increases to 15%. Calluna and Sphagnum spores are well represented.
CRBi158–1081045–870The zone records significant increases in Poaceae < 25 µm and Cyperaceae, whilst Calluna pollen falls to trace levels.
CRBj108–0 870–0 Betula is an early dominant, later it is replaced by Poaceae < 25 µm. Plantago and Pteridium increase and Osmunda rises near the surface.
Figure 2.

Macrofossil diagram from core CRB93 on Crymlyn Bog. Linked histograms represent percentage cover of the microscope grid graticule at × 400 magnification. Unlinked histograms represent a 5-point scale of abundance where rare = 1, occasional = 2, frequent = 3, very frequent = 4 and abundant = 5.

Figure 3.

Percentage mire pollen diagram from core CRB93 on Crymlyn Bog, constructed from the analysis of at least 300 total dry land pollen grains (i.e. Cyperaceae, Ericaceae and Poaceae < 25 µm excluded).

Zones CRBma (macrofossil) and RBaa/CRBab(Pollen): salt marsh and pioneer reedswamp

The stratigraphy indicates that Crymlyn Bog is underlain by blue-grey clay, probably of estuarine origin. The basal radiocarbon age estimates (AA-28123 and AA-28124, Table 1) may be erroneous, possibly because of contamination by hard water, marine carbon or reworked organic material, but extrapolation of the average sediment accumulation rate places the base of the pollen diagram at c. 8710 Cal. years BP (range 8830–8585 BP). This is earlier than for the base of the macrofossil diagram, which is estimated at c. 7280 BP (range 7400–7155 BP). Few macrofossils were encountered in the clay, but the pollen spectra include types consistent with open halophytic or brackish-water herb communities. This evidence indicates the onset of salt marsh formation, possibly a consequence of marine regression. Godwin (1940) suggests that sea level declined in Swansea Bay during the Boreal period. A marine regression between 7600 and 6800 BP is also reported from parts of the neighbouring Severn Estuary (Smith & Morgan 1989; Walker et al. 1998). The clay in core CRB93 merges upward at 800 cm into blue-grey clay intercalated with thin laminae, rich in Phragmites australis remains (CRBma and CRBab). The upper part of the zone contains fewer herbs characteristic of salt marshes, suggesting the gradual removal of marine influence as relative sea level fell and succession began.

CBRmb and CRBb: freshwater reedswamp

The establishment of a diverse fen herb community in this zone represents succession from reedswamp to rich fen (beginning at 6995 BP), dominated by Phragmites australis and Scirpus lacustris, with associates such as Lychnis flos-cuculi and Osmunda regalis. The increase in pollen and spores of aquatic species late in zone CRBb indicates that the bog surface gradually became much wetter. These changes may reflect regional water table fluctuations, but the pollen evidence also shows that types associated with salt marsh declined further in the zone. Thus the wetter mire surface may have been a consequence of increased run-off rather than base-level changes.

CRBmc and CRBc: mesotrophic fen and carr

At 640 cm depth (uncalibrated age 5015 ± 50 BP (AA-2815, Table 1), 2σ range 5895–5650 years (BP), the stratigraphy changes from a rich fen assemblage to peat containing sedges, abundant Betula remains and Menyanthes. This assemblage is indicative of swamp carr woodland. Betula can establish in such wet conditions, where sufficient nutrients (mainly nitrogen and potassium) are available (Atkinson 1992; Barkman 1992). By the end of the zone, carr species decline and are replaced by macrofossils of ‘brown mosses’ and small quantities of mesotrophic Sphagna, including Sphagnum recurvum, signalling the onset of acidification. The average accumulation rates (Fig. 4) suggest that the mire was mesotrophic for c. 750 years between c. 5870 BP (range 5990–5745 BP) and c. 5130 BP (range c. 5330–4930 BP).

Figure 4.

Average sediment accumulation rates for core CRB93 plotted using the calibrated radiocarbon dates presented in Table 1.

CRBmd and CRBd: dry oligotrophic mire

In CRBmd, mesotrophic species are replaced by a relatively dry tussocky oligotrophic mire assemblage, demonstrated by substantial increases in Eriophorum vaginatum and Calluna vulgaris wood, accompanied by traces of oligotrophic Sphagna. This stage occurred between c. 5130 BP (range 5330–4930 BP) and c. 4435 BP (range 4635–4235 BP). The transition to tussocky oligotrophic mire is very familiar throughout the British Isles. It has been recorded from mires in Ireland, Wales and Cumbria (Hughes 1997), and many of the mires surveyed by the North-west Wetland Survey in the north-west of England (Leah et al. 1998). This mire type is very frequently the precursor to Sphagnum-dominated raised bog.

CRBme and CRBe: poor fen carr

The Calluna/Eriophorum mire stage is cut short and replaced by fen carr (CRBme and CRBe, c. 4435–2745 BP). First Myrica gale and then Betula, Carex and Phragmites join the macrofossil assemblages. This represents a major departure from the bog development pathways of many British raised bogs (Leah et al. 1998) and strongly suggests a substantial allogenic disturbance of the bog complex. The reversal of the acidification process, and re-establishment of fen carr and then open fen, may have been triggered by nutrient enrichment. Clearance of woodland by Bronze Age people could have accelerated surface run-off and soil erosion, leading to increased nutrient supply to the mire (Lisa Dumayne-Peaty, unpublished data) or, alternatively, an increase in relative sea level may have led to a rise in ground-water levels. Evidence for a marine transgression at c. 3500–2800 BP is reported from elsewhere in South Wales (Smith & Morgan 1989; Walker et al. 1998).

CRBmf and CRBf: acid, wet pseudo-raised mire

This zone marks a return, over 500 years, to more acidic, poor fen conditions. By contrast with the first phase of acidification, the macrofossil assemblage maintains a rather wet character right through to the establishment of true ombrotrophic conditions in zone CRBmg. This is indicated by the presence of Menyanthes trifoliata, Rhynchospora alba, Drepanocladus spp., Erica tetralix, Eriophorum angustifolium, Sphagnum sect. Cuspidata and Sphagnum imbricatum. Sphagnum sect. Cuspidatum and Menyanthes trifoliata macrofossils in particular, suggest that open pools were present. The zone began c. 2745 BP (range 3125–2765 BP) and ended c. 2260 BP (range 2375–2145 BP).

CRBmg and CRBg: ombrotrophic mire

Sphagnum imbricatum is dominant throughout much of zone CRBmg and it is associated with Rhynchospora alba, suggesting that the mire developed into an ombrotrophic bog capable of maintaining a near-surface water level between c. 2260 BP (range 2375–2145 BP) and c. 1890 BP (range 2010–1780 BP). Once Sphagnum begins to invade a peatland, acidification tends to be self-perpetuating, as the uptake of cations from solution and the release of hydrogen ions by Sphagnum leads to further acidification and lowering of pH. The ability of Sphagnum to retain water in its hyaline cells also leads to an increase in surface wetness to which many other plants are intolerant. The average accumulation rate for this zone is 3.8 years cm−1, which is much faster than that of other raised bogs in Britain, where the average rate of peat accumulation is between c. 9 and 11 years cm−1 (Barber 1981; Barber et al. 1994; Mauquoy & Barber 1999). These observations are important because ombrotrophic conditions are not found on Crymlyn Bog today.

CRBmh and CRBh: ombrotrophic mire

CRBmh and CRBh are dated to between c. 1895 BP (range c. 2010–1780 BP) and c. 1045 BP (range c. 1130–960 BP). The principal feature of the zone is the decline of Sphagnum imbricatum, which is replaced by Sphagnum papillosum before a general decline in Sphagna. The species assemblages of CRBmh and CRBh represent the transition to mire communities typical of modern British lowland raised mire vegetation (Rodwell 1991). The recent major decline in Sphagnum imbricatum, and its replacement by either S. papillosum or S. magellanicum, is documented from elsewhere in the British Isles (e.g. Stoneman 1993; Mauquoy & Barber 1999). The results from Crymlyn Bog support the conclusions of Mauquoy & Barber (1999) that the decline in S. imbricatum may be associated with high mire surface wetness. However, the replacement of S. imbricatum occurs somewhat earlier at Crymlyn Bog (Cal. AD 990–820) than at sites in northern England and southern Scotland (Cal. AD 1030–1485) (Mauquoy & Barber 1999). Although Phragmites australis macrofossils are virtually absent from the zone, the presence of Poaceae < 25 µm pollen suggests that Phragmites was growing elsewhere on and/or near the site, and that ombrotrophic conditions may not have occurred everywhere. This interpretation is supported by the field stratigraphy data presented later.

CRBmi and CRBi: poor fen

Sphagnum macrofossils that are present in the previous zone are absent in CRBmi, which is characterized by a sudden increase in Phragmites australis macrofossils. The rise and fluctuations in pollen of Poaceae < 25 µm (pollen zone CRBi) support the macrofossil evidence for a rise in Phragmites australis, and the presence of other monocotyledons is supported by fluctuations in Cyperaceae pollen. Between c. 1045 BP (range 1130–960 BP) and 870 BP (AA-28131, range 955–785 BP) the pollen and macrofossil evidence suggests a return to a drier, species-poor Phragmites australis community. The presence of Calluna vulgaris pollen and Sphagnaceae spores suggests that these taxa were growing elsewhere on the site. The decline in Sphagnaceae suggests that it was out-competed by Phragmites australis (Haslam 1969). Once Sphagnum establishes, ombrotrophic mire usually persists (Walker 1970), unless there are strong allogenic influences on mire development. Thus, the transition to poor fen communities recorded here is a diversion from the normal course of succession. These changes may have been a result of an incursion of base-rich run-off water, as a consequence of woodland clearance by people in Medieval times. This is evident in the full pollen diagram by a decline in dry-land tree pollen and increases in open-ground taxa (L. Dumayne-Peaty, unpublished data).

CRBmj and CRBj: rich fen

The uppermost zone records the re-establishment of Phragmites-rich fen with Cyperaceae species and fen herbs. Several acidic taxa remain in the assemblage, as shown by the presence of Myrica gale and Calluna vulgaris. In places, swamp or fen pools may have re-established, in which Sparganium erectum, Equisetum spp., Potamogeton, and Typha spp. could survive. Records of Betula pollen indicate recent colonization of the mire and its margins by birch woodland, in which Carex paniculata, Pteridium and Osmunda possibly grew. Increased nutrient input to ground and surface water from agriculture, and the construction of the Glan-y-Wern and Port Tennant Canals in the AD 1790s (John & Williams 1980; Boeye et al. 1995), may have led initially to eutrophication. Pollution as a consequence of the Industrial Revolution and later industrial developments (Rosen 1998) is not coincident with changes in the mire flora, but could have aided development of rich fen because mine drainage, run-off from the municipal rubbish tip and from the oil refinery, probably changed water quality locally.

multivariate statistical analysis

The interpretation of CRB93 is supported by detrended correspondence analysis (DCA) of the macrofossil data (Table 2, Fig. 2), undertaken using the computer programme canoco (Ter Braak 1988) (Fig. 5). Axis 1 accounts for three-quarters of the variability in the macrofossil data set. The positions of the taxa along axis 1 suggest that it represents a nutrient gradient, with taxa characteristic of salt marshes and nutrient-rich fens to the right and those representative of oligotrophic conditions towards the left. The eigenvalues suggest that this is the most important influence on taxa variability, which is consistent with data from modern ecological studies (e.g. Wheeler 1980a; Daniels & Eddy 1990). The overall scores on DCA axis 1, for each of the macrofossil samples, were plotted against depth (Fig. 6) to provide what might be regarded as a proxy-nutrient index. This provides a summary of variation in mire nutrient status over time.

the stratigraphic transects

The cores from transects A and B were divided into homogenous horizons according to the Tröels-Smith (1955) peat classification system (Figs 7 and 8). Although the two transects of cores reveal spatial and temporal variability in gross mire stratigraphy, they suggest that most stages of development, depicted in the pollen and macrofossil records from CRB93, can be traced and replicated across the mire. This implies that there was some degree of spatial homogeneity in mire development.

Figure 7.

Sediment stratigraphy of transect A (see Fig. 1), classified using the Tröels-Smith system: Sh = Substantia humosa (degraded peat), Tb (Spa) = Turfa bryophytica (Sphagum peat), Tl = Turfa lignosa (wood peat), Th = Turfa herbacea (Monocotyldeon peat), Th (Phr) = Phragmites peat, Ag = Argilla granosa (Silt), As = Argilla steatodes (Clay). Symxbols appear in mixtures representing the proportion of each sediment type in the horizons.

Figure 8.

Sediment stratigraphy of transect B (see Fig. 1), classified using the Tröels-Smith system (key as Fig. 7).

There are, however, three exceptions. First, the cores in the south-west of the mire show a second period of clay deposition shortly after the initial reedswamp, indicating that a marine incursion could have occurred in this area. Secondly, the transition to ombrotrophic mire only occurred in the western part of the site. A lens shaped body of ombrotrophic Sphagnum-rich peat is evident in stratigraphy cores CRBa2-CRBa5 and CRBb3 (the cross point of the transects). This suggests that ombrotrophic conditions were not achieved in the east of the mire, which remained poor fen perhaps because run-off from adjacent land or saline/brackish water had a greater influence locally. Thirdly, clay horizons occur near the surface of core CRBa1. The sediment was probably eroded from the adjacent hillside.


the development of crymlyn bog and models of hydroseral succession

The definition of vegetation succession has been debated for over 100 years (Cowles 1899; Walker 1970). Clements (1916) popularized the concept of succession as an autogenic, unidirectional and progressive turnover of species, resulting in the development of a stable climax community. Gleason (1926) argued from the opposite standpoint, stressing the role of chance disturbances to the community structure that could potentially deflect succession along any number of developmental pathways. He also suggested that the concept of a climax community might be erroneous and that succession may be continuous. Despite the efforts of Gleason and a mounting list of critics (e.g. Whittaker 1953; Walker 1970; Connel & Slatyer 1977; Finegan 1984), aspects of the simple model of autogenic succession offered by Clements (1916) have persisted in the literature, albeit in a somewhat altered form. For example, Walker (1970) examined the nature, direction and rate of succession in wetland communities, and followed the basic assumption that a set of mires could be identified that were relatively free of allogenic disturbances, so enabling a study of long-term autogenic succession. Although Walker’s study lacked the most direct form of palaeoecological data for tracking mire succession, namely plant macrofossil records, the work remains an influential and rare re-assessment of succession theory using palaeo-mire communities.

Recent studies of the fen–bog transition (Hughes 2000) in a number of the mires used by Walker (1970) suggest that the assumption of predominantly autogenic development is not necessarily a safe one, even when there are no obvious reversals in the succession. For example, Hughes (2000) suggests that the frequently observed Eriophorum/Calluna mire type (‘Pseudohochmoor’, Rybniček 1973) lying at the fen–bog transition (FBT) is a response to the peat surface becoming perched above the general water table. This could happen for several reasons, such as the autogenic development of tussocks and subsequent leaching of nutrients from their crowns by rainwater; or, alternatively, similar conditions could occur after a drop in the mire water table. A falling mire water table could be triggered by a number of factors (e.g. a change in the base level or a reduction in effective precipitation). Tregaron Bog is a good example of a bog that became ombrotrophic following a phase of water table oscillation, with the FBT coinciding with a period of low water levels (Godwin & Mitchell 1938).

The results of the macrofossil investigation at Crymlyn Bog (Figs 2 and 6) show at least two major reversals in the course of succession. As a result, the mire sequence contains two transitions towards raised bog communities, the first in zone CRBmd and the second in zone CRBmf. In between (zone CRBme), there is a return to more nutrient-rich conditions. The mire sequence can be used therefore to make a comparison of the pathways of succession towards raised mire in two contrasting periods of the Holocene at the same sample site. The first oligotrophic transitional mire community (the first FBT) developed on the mire at c. 5000 BP (just above radiocarbon date AA-28126, 5245 BP). The species present indicate that the water table of the mire at this point was either deep or fluctuating, laying down a layer of well-humified fibrous Eriophorum/Calluna peat, containing rather few Sphagna. Similar phases are found in many of the raised mire sequences of north and west Britain that became ombrotrophic in the early to mid-Holocene (Leah et al. 1998). The lowland Eriophorum/Calluna mire has no natural modern analogue in Britain and closely resembles the ‘Pseudohochmoor’ of central Europe described by Rybniček & Rybničkova (1968). A feature of this mire type is that it contains some moss species indicative of slight nutrient enrichment (e.g. Aulacomnium palustre), possibly resulting from peat mineralization during periods of desiccation. Cenococcum spp. soil fungi, abundant in aerated conditions, are also characteristic of this peat type.

The succeeding zone (CRBme) represents a period of reversal in the mire succession at Crymlyn Bog, with the reappearance of fen communities, indicating that the growing surface became reconnected to the nutrient-rich groundwater supply at c. 4435 BP. This reversal may have been a response to an allogenically driven increase in the mire water table. Disturbance to the catchment through woodland clearance, an alteration in the configuration of sand bars in the estuary, or base level change could all account for the observed mire development sequence (see Fig. 2, zone CRBme, and Fig. 6). Godwin (1941) recorded similar reversal events in mire succession at Shapwick Heath in the Somerset Levels.

The second phase of development towards ombrotrophy (the second FBT), registered at the beginning of zone CRBmf and dating to 2600 BP, contrasts markedly with that of the first phase of acidification (Fig. 6). Although Eriophorum vaginatum is still a major part of the flora, there is evidence that the tussocks were surrounded by wet mire surface conditions, as indicated by the arrival of, first Menyanthes trifoliata and then Rhynchospora alba and Erica tetralix. The stratigraphic transects in Figs 7 and 8 also show that large areas of the mire remained Phragmites swamp throughout the second acidification phase, although ombrotrophic communities can be traced for over 1 km between stratigraphic cores CRBa2 and CRBb3. The record outlined above contrasts markedly with many inland bogs, where ombrotrophic peats often permanently replace the majority of the preceding fen.

In sharp contrast to the lower FBT, the upper one appears to have developed from Eriophorum vaginatum tussocks whilst the water table remained near the surface of the mire. Leaching of nutrients from the tops of the tussocks and the cation absorption capacity of Eriophorum vaginatum may have been the dominant processes causing acidification. In these conditions, high levels of precipitation might aid the process of raised mire formation. The recorded date for the change to wet acidified mire (very poor fen) in zone CRBmf is 2729–2367 BP (2σ range), which may correspond to the widely recognized period of increased effective precipitation at the beginning of the Sub-Atlantic period (van Geel et al. 1996). There are problems associated with the calibration of radiocarbon dates in this period, caused by a plateau in the radiocarbon calibration curve. Consequently, the 2σ confidence interval for the date quoted above occupies an age range of c. 360 years. Despite the uncertainty associated with the date of the shift to wet poor fen, the most likely explanation for the major increase in mire water levels would appear to be climate change. Thus not only the character, but also the timing and the rate of development of both FBTs, could have been determined by the allogenic control of regional climatic change. This finding is important because it highlights the significance of allogenic factors in directing both forwards as well as reversed mire successions over long time-scales.

the role of eriophorum vaginatum and sphagnum in the two fbts

The traditional model of autogenic raised mire development (Walker 1970) stresses the importance of Sphagnum as an ‘ecosystem engineer’ in acidifying the mire (Bellamy 1968; Walker 1970). Although Sphagnum is present in the transitional community at the first FBT, it is not a major part of the second one until the establishment of full ombrotrophic conditions. Careful examination of the macrofossil diagram shows that Sphagnum sect. Acutifolia alone was recorded as a trace in the macrofossil assemblage of zone CRBmf. The lack of Sphagnum in the upper FBT is unlikely to be a result of differential preservation, as the few remains that were recorded were well-preserved delicate leaves. The pollen and spore data (Fig. 3) support this interpretation, as Sphagnum spores represent less than 10% of pollen and spores in zone CRBf. Many of the cores in the two transects also contained few ombrotrophic Sphagna at the upper FBT, suggesting that Sphagnum was then poorly represented across significant areas of the mire. The findings may indicate that whilst Sphagnum is important in some FBTs, the development of pioneer oligotrophic mire is not dependent upon the presence of this genus in the community. Studies of raised mires in Shropshire (Leah et al. 1998) are in agreement with this finding, indicating that the coastal site is not a special case. Further detailed macrofossil work will be required to examine the role of Sphagnum at the FBT. The present paper suggests that the arrival of Eriophorum vaginatum is an important precursor to the establishment of Sphagnum-dominated true raised mire, both in pathways developed from allogenically influenced dry oligotrophic mire (FBT1 at Crymlyn Bog) and from wet tussocky poor fen (FBT2 at Crymlyn Bog).

the demise of raised bog and development of phragmites fen at crymlyn bog

Walker’s model of the hydrosere suggests that mire sequences in Britain usually culminate in ombrotrophic communities that are rarely invaded by trees. Thus, raised bog has been proposed as the climax of the hydrosere (Walker 1970). The model is intended principally to describe autogenic processes, and assumes that allogenic controls have not been dominant during the course of succession unless major reversals are evident. Both the reversal in succession in zone CRBme and the second reversal in zone CRBmi demonstrate that on at least two occasions Crymlyn Bog was subjected to perturbations large enough to deflect the development pathway onto a completely new course. The final decline of ombrotrophic mire communities began in the Medieval period at c. 1045 BP (range 1130–960 BP). The full pollen diagram (Dumayne-Peaty, unpublished results) shows a clearance of dry land tree species before the expansion of Phragmites and Carex communities. Increased run-off from the steep slopes surrounding Crymlyn Bog may have triggered the observed changes.

The medieval date for the uppermost change from raised mire to poor fen in CRB93 refutes the view that the demise of raised bog was coincident with the Industrial Revolution and its consequent pollution (Meade unpublished data; Headley et al. 1992; Gilman 1994). However, the radiocarbon chronology pertains to one discrete location (CRB93), and the transition from ombrotrophic mire to fen could have been time-transgressive, perhaps explaining why historical and botanical accounts describe raised mire communities at Crymlyn Bog in the 18th century AD (Meade unpublished data). Nevertheless, the change began considerably earlier than the Industrial Revolution at CRB93.

Palaeoecological investigations of many other raised bogs (e.g. Barber et al. 1994; Leah et al. 1998; Mauquoy & Barber 1999) demonstrate that mire communities have changed dramatically in the last couple of centuries. For example, Molinia caerulea and Betula spp. have invaded numerous formerly Sphagnum-rich raised mires, altering the community structure significantly. Some of these changes may be a response to new allogenic triggers, such as increased nitrogen deposition. The continuous variation in allogenic factors highlighted by the analysis of core CRB93 at Crymlyn Bog, and the other palaeoecological records mentioned above, emphasize the problem with the definition of climatic climax communities sensuClements (1916). It can rarely be assumed that allogenic forces have remained unchanged over time-scales relevant to mire succession. The palaeoecological record, however, only demonstrates the mire’s response to forcing factors, and it is therefore often difficult to ascertain the exact cause and nature of external influences.


The two fen–bog transitions (FBTs) present in core CRB93 from Crymlyn Bog demonstrate contrasting pathways to ombrotrophy. The first transition was initiated from dry hummocky communities and suggests a relatively rapid acidification process under the influence of falling or unstable water tables. This transition may have been allogenically driven (e.g. reduction in effective precipitation). The second FBT was initiated from tussocky Eriophorum vaginatum communities surrounded by shallow bog pools, possibly aided by a significant increase in effective precipitation at c. 2650 BP (van Geel et al. 1996). Such a change would have caused leaching of nutrients from the tussock tops prior to the widespread establishment of Sphagnum imbricatum, the main raised peat former. The stratigraphic sequence illustrates that divergent successions may occur from similar initial communities at different times in the Holocene.

The two successions at Crymlyn Bog suggest that allogenic as well as autogenic controls are frequently important in determining both the character and timing of forward successions. Although reversals in the direction of succession are generally the result of external disturbances, here too, the rule may be broken. For example, the development of trees on schwingmoor may result in an entirely autogenic reversal of succession (Tallis 1973). Consequently, it may be unwise to use the term succession to imply any particular process or mechanism, particularly in the context of long-term (millennial-scale) change, but rather it can be used as a descriptive term simply meaning the sequential replacement of vegetation communities.


The authors thank the Countryside Council for Wales and British Petroleum, Llandarcy, for part-funding the project and allowing access to the site, the NERC Steering Committee for AMS radiocarbon dates and the Swansea Radiocarbon Laboratory for conventional radiocarbon dates. Andrew Moss prepared the pollen samples. We thank Mr Rhodri Evans, warden of Crymlyn Bog, and undergraduates and postgraduates who assisted in the field. The Cartographic Unit, School of Geography and Environmental Sciences, University of Birmingham, drew Figs 1, 4, 5, 6, 7 and 8.