Mid- to late-Holocene vegetation history of Greater Exmoor, UK: estimating the spatial extent of human-induced vegetation change
- First Online:
- Cite this article as:
- Fyfe, R.M., Brown, A.G. & Rippon, S.J. Veget Hist Archaeobot (2003) 12: 215. doi:10.1007/s00334-003-0018-3
- 479 Views
This paper presents the results from three pollen profiles from a group of small spring mire sites on the southern edge of Exmoor in south west England. The size and topography of these sites allow detailed local landscape histories around each site to be reconstructed which broadly cover the mid- to late-Holocene. Comparison of the individual local landscape histories demonstrates the scale of spatial variation in vegetation around the upland edge, and facilitates understanding of human-landscape interactions from the early Neolithic onward. In the early Neolithic significant short-term woodland disturbance is recorded around the upland fringe, including clearance of oak-hazel-elm woodland, suggesting that the shift from Mesolithic to Neolithic is not marked by a gradual environmental transition. Following this, there is clear evidence of Neolithic management of upland heath using fire, presumably for the management of upland grazing. Woodland clearances are recorded throughout the later Prehistoric period; however, the use of multiple profiling suggests that woodland clearance is spatially discrete, even within an area of 4 km2. Pastoral land use is dominant around the uplands until around 900–1,000 a.d., and there is no discernible Roman or post-Roman period impact in the vegetation, suggesting cultural stability from the late Iron Age to the early Medieval period. By 1,100 a.d., there is a shift to mixed arable-pastoral farming which appears to continue well into the post-Medieval period.
One of the more problematic issues in studying human-environment relations around upland areas is the use of palynological sites which record vegetation history at an appropriate spatial scale, and which are located within suitable areas of the landscape (Edwards 1999). Following the pioneering work of Pennington (1970) and Turner (1970) pollen analytical studies in upland areas have tended to concentrate on data from either blanket mire sequences (see reviews in Simmons and Innes 1987; Tipping 1994; Caseldine 1999) or lakes (Pennington 1965; Pennington et al. 1972). Although these study sites have provided detailed information about the nature and timing of human impact it is difficult to make inferences about the spatial extent of human impact on the landscape (Edwards 1979; Brown 1999). This is due to the scale of their pollen source areas, which is likely to be regional in nature (Jacobsen and Bradshaw 1981; Prentice 1985). Most work has also concentrated on the prehistoric period, although such sites are located in areas likely to be have been marginal in terms of prehistoric domestic settlement. As a result, our understanding of landscape change and human impact on vegetation tends to be at a regional level, and there are few data to demonstrate the importance of local changes within landscapes, and crucially, the spatial dimensions of human-induced vegetation change. Dumayne-Peaty and Barber (1998) have demonstrated the importance of using multiple pollen profiles from single sites to demonstrate vegetation differences on a small spatial scale. However, this work was undertaken on raised mire sites, which reflect regional vegetation history, although multiple site profiling has proved useful in understanding woodland disturbance at a small spatial scale (e.g. Smith and Cloutman 1988; Simmons and Innes 1996). An attempt to use small pollen sites in conjunction with surrounding archaeological data for an upland fringe has been published by Skinner and Brown (1999).
This paper uses multiple pollen profiles to highlight the importance of local landscape histories in developing detailed vegetation reconstructions. Palaeoecological analyses on three small spring mire sites have been used to identify local differences in vegetation history. The sites used in this study are small soligenous mires, between 20 and 40-m wide, but up to several hundred metres long. Owing to their restricted width, the pollen source distance for the sites should be small. The question of the size of the pollen source area has been discussed by Jacobsen and Bradshaw (1981), and more recent attempts have been made to quantify aspects of pollen source area (Prentice 1985; Sugita 1993, 1994, 1998). For sites within the size range used in this study, the local and extra-local relevant source area should be in the order of hundreds of metres (Jacobsen and Bradshaw 1981). The sites are located beyond the present-day margins of the enclosed landscape on the southern upland fringes of Greater Exmoor (around 300 to 340-m OD). As a result they have the potential to record landscape continuity and/or change in the area of the landscape sensitive to fluctuations in land use pressure or settlement over the last 5,000 years. This period is particularly significant as palaeoecological studies in south west Britain have traditionally focussed on the upland area of Dartmoor (e.g. Simmons 1964; Smith et al. 1981; Simmons et al. 1983; Caseldine and Hatton 1993), and there are few securely dated sequences spanning the latter half of the Holocene (Caseldine 1999).
Materials and Methods
Site stratigraphy was investigated using a 5-cm-wide gouge corer, and sites were sampled using a 5-cm-wide Russian-type corer (Jowsey 1966). The choice of sample core was made based upon the deepest peat sequence at each site. Sub-sampling for pollen analysis was undertaken in the laboratory using 0.5-cm peat slices, taken at 4-cm intervals from each core; pollen preparation followed standard procedures (Faegri et al. 1989). Pollen and spores were identified using the keys in Moore et al. (1991), Andrew (1984) and the University of Exeter reference collection. A minimum sum of 500 land pollen grains was counted. Results are shown as percentage total land pollen for land pollen types, and percentage total land pollen and spores for spores. Pollen nomenclature follows Bennett (1994); differentiation of Poaceae and cereal types follows Andersen (1978). Charcoal was also counted within two size classes: 10–50 μm and >50 μm, and is expressed as concentration data. Slices of sediment 5-cm thick were used for dating; samples were submitted to the Radiocarbon facility at Waikato, New Zealand. It had been the intention to use conventional dating on these samples; however, insufficient carbon necessitated their analysis using AMS radiocarbon dating. Five dates were submitted from Long Breach, four dates from Gourte Mires and six from Ansteys Combe. Dates are expressed as uncalibrated b.p., calibrated age ranges expressed as years b.c./a.d. (to 2 sigma) are also provided in the discussion. Calibration was performed with the Calib 4.0 program (Stuiver and Reimer 1993).
Radiocarbon results and time-depth chronology
Results of AMS radiocarbon dating
Age 14C b.p.
Calibrated age range (2σ)
50 b.c.–a.d. 240
The results from the six-radiocarbon samples from Ansteys Combe reveal a problem in the construction of a time-depth series by linear interpolation for the site. The dates show two reversals, and the middle two dates produce similar results (Fig. 3). The lower two dates show a reversal (1,920±60 b.p. at 145–150 cm; 3,830±60 b.p. at 134–139 cm). The older of the two dates can be questioned through examination of the loss-on-ignition data for this section of the core (Fig. 3). The date 3,830±60 b.p., at 134–139 cm is associated with a fall in loss-on-ignition values, suggesting that this sample could contain old carbon from inwashed soils. The date of 1,920±60 b.p. is associated with higher loss-on-ignition values, and as such is more likely to represent the true age of the sediment. The upper two dates also show a reversal (420±60 b.p. at 35–40 cm; 1,120±60 b.p. at 25–30 cm). The date at 25–30 cm can be rejected and the date at 35–40 cm accepted by plotting the curve for the early 18th century a.d. Pinus sylvestris rise, which suggests the sample at 25–30 cm to be around 800 years too old. The middle two dates from Ansteys Combe are similar. Correlation of the pollen sequence from Ansteys Combe and Gourte Mires suggests the date of 1,160±70 b.p. at 80–85 cm is likely to represent an accurate estimate of the age of the sample. As a consequence, the date of 1,110±70 b.p. at 60–65 cm depth is unlikely to represent an accurate estimate of the age of the sample, and probably reflects contamination through inwash of older soil material.
LBMC-1 (184–150 cm), ca. 5,400–4,650 b.p., Quercus-Corylus-Succisa.
LBMC-2 (150–130 cm), ca. 4,650–4,240 b.p., Poaceae-Quercus-Corylus.
LBMC-3 (130–58 cm), ca. 4,240–2,400 b.p., Poaceae-Corylus-Quercus.
LBMC-4 (58–26 cm), ca. 2,400–600 b.p., Poaceae-Cyperaceae.
LBMC-5 (26–0 cm), ca. 600–0 b.p., Poaceae-Calluna
The start of the zone is characterised by declines in Quercus and Corylus to low levels (both below 5%). Towards the top of the diagram Pinus rises, reflecting the development of pine plantations in the area. Heath vegetation increases, and there are declines in grassland herbs, including Centaurea nigra and Lactuceae. Charcoal concentrations drop to very low levels.
Four stratigraphic cross sections were examined across Gourte Mires. The width of the mire varies from around 80 m at the top of the site to around 30 m at the lowest transect. The site is around 150-m long, draining to the south into the headwaters of the Yeo valley. The stratigraphy is predominantly comprised of monocotyledonous peat (Fig. 4). Towards the upper end of the site the monocotyledonous peat overlies stiff silty sand, which includes abundant weathered bedrock clasts.
GM1–1 (200–150 cm), ca. 4,100–3,100 b.p., Poaceae-Quercus-Corylus.
GM1–2 (150–70 cm), ca. 3,100–1,950 b.p., Poaceae-Corylus-Quercus.
GM1–3 (70–30 cm), ca. 1,950–1,000 b.p., Poaceae-Calluna.
GM1–4 (30–10 cm), ca. 1,000–200(?) b.p., Poaceae-Calluna-Cyperaceeae
The start is marked by a decline in grassland herbaceous species (including Centaurea nigra and Succisa), and the appearance of cereal types (Hordeum-type, Avena/Triticum-type, Secale cereale) and weeds of arable cultivation (Anthemis-type). The zone is characterised by low woodland representation, and elevated levels of Cyperaceae. Charcoal concentrations fall to low levels.
Ansteys Combe is a small, narrow valley head draining south into the Yeo valley. A single coring transect runs down the site. The site is approximately 10 m in size at its widest point, and extends for 150 m, with a slope of around 0.11 m m−1. The site is comprised of well-humified monocotyledonous peat, with frequent inwash bands made up of sand and small grit lenses.
AC1–1 (160–138 cm), ca. 2,100–1,800 b.p., Calluna-Quercus-Poaceae.
AC1 −2 (138–82 cm), ca. 1,800–1,150 b.p., Quercus-Calluna-Poaceae.
AC1–3 (82–62 cm), ca. 1,150–800 b.p., Poaceae-Calluna.
AC1–4 (62–30 cm), ca. 800–200 b.p., Calluna.
AC1–5 (30–0 cm), ca. 200–0 b.p., Calluna-Poaceae.
The start of the zone is characterised by a decline in heath pollen, although Calluna remains the dominant taxon. Increases occur in both grassland (including Poaceae, Ranunculus acris-type and Lactuceae), and woodland taxa (including Quercus and Corylus). There is an increase in Pinus at the start of the zone.
The data sets from Gourte Mires and Long Breach were subjected to a PCA, executed in CANOCO (ter Braak 1987). The pollen data were run separately to investigate within-site variability between samples, and as a combined dataset to investigate between-site variability. Combined datasets must be used to compare between-site variability as this method ensures both datasets are plotted on the same axis, and are responding to the same environmental variables. The data from Ansteys Combe were not included: visual inspection of the data shows significant difference from the other pollen sites, in particular the presence of significant amounts of Quercus until around 1,000 b.p. and the subsequent dominance of Calluna in the record. Dominance of individual taxa in an analysis has been shown to lead to meaningless results in statistical analysis (ter Braak 1987). Results presented here are therefore from the analysis of combined data from Long Breach and Gourte Mires.
PCA species-scores for first two axis, combined analysis, Long Breach and Gourte Mires
Axis 1 (eigenvalue=0.515)
Axis 2 (eigenvalue=0.218)
Interpretation and discussion
Vegetation character during the later Mesolithic on southern Exmoor
The oldest deposits in the study are from Long Breach, where peat accumulation started during the later Mesolithic, around 5,400 b.p. (4,750–4,350 b.c.). At this time the landscape was predominantly wooded (pollen zone LBMC-1), with oak and hazel the dominant trees and elm present at low levels, probably at the edge of its distribution. The open part of the landscape around Long Breach was dominated by open damp heath, including heather and sedges. Archaeological evidence of any Mesolithic presence in the surrounding landscape is absent, and the pollen evidence does not suggest any human impact on the vegetation in the local or extra-local area.
At Exebridge, to the south of Exmoor, Fyfe et al. (2003) recorded later Mesolithic woodland disturbance by fire starting at 7,710±60 b.p. (6,650–6,440 b.c.). In a re-examination of the development of peat and heath vegetation on Dartmoor, Caseldine (1999) argues that later Mesolithic communities used fire to manipulate the vegetation of the uplands, maintaining openness at the ecotone between upland woodland and the open summit areas. At Black Ridge Brook, Bellever and Pinswell, charcoal records suggest this was most prevalent between ca. 7,700 and 6,100 b.p. (5,750–4,150 b.c.). However, at Long Breach there is no evidence that fire played a role in the management or creation of open heathland at the ecotone edge on Exmoor during the end of the later Mesolithic period.
The character of the upland vegetation during the Neolithic
At Long Breach there is an abrupt change (occurring over at most 85–100 years) in the character of the vegetation around Molland Common shortly before 4,700±60 b.p. (3,640–3,360 b.c.). This is marked by several palynological features, starting with a short-lived decline in the dominant oak-hazel woodland, along with a permanent decline in elm at the end of the clearance event. This is taken to mark the change from the Mesolithic to the Neolithic. The clearance episode may have lasted between 50 and 200 years; woodland regeneration occurs at 4,700±60 b.p. (3,640–3,360 b.c.). During the phase of reduced woodland, representation of heathland within the pollen record increases. It is possible that the increase in heathland represents a change in pollen recruitment to the site: woodland clearance may have opened the site environs resulting in an increase in pollen from the uplands. The evidence from The Chains suggests that the central uplands of Exmoor were dominated by wet heath communities before 4,170±75 b.p. (2,910–2,500 b.c., Moore et al. 1984). Alternatively, the damp heath communities around Long Breach may have expanded onto formerly wooded areas in the local surroundings. At the end of the clearance episode, the oak-hazel woodland returns to former levels, suggesting only a temporary clearance (or clearances) around the site at the start of the Neolithic. Elm does not recover to former levels when oak-hazel regenerates, and does not appear at significant levels in the vegetation history of the area at any later date.
The timing of the elm decline fits the broad pattern seen for southern Britain (Beckett and Hibbert 1979; Scaife 1988; Peglar 1993). However the nature of the decline, in association with a decline in oak-hazel woodland, is unusual for British sequences. Relative percentage values of non-elm deciduous woodland species usually remain steady throughout the elm decline, although there is evidence of small pre-elm decline clearances, for example at Hockham Mere, Norfolk (Sims 1978) and Gatcombe Withy Bed, Isle of Wight (Scaife 1988). The strong element of woodland clearance around Long Breach broadly supports models of human impact, possibly combined with pathogenic attack (Peglar and Birks 1993; Peglar 1993; Parker et al. 2002). It has been argued that small openings in the canopy would have assisted the vectors of pathogenic attack. The scale of clearance around Long Breach would support this possibility although the elm decline is more gradual than the decline in oak and hazel, and the final elm decline occurs after oak and hazel return to higher values in the pollen sequence. Regeneration of elm might have been prevented by continued human activity, particularly woodland grazing, through exploitation of the gaps left by dead elms (Edwards and MacDonald 1991), grazing preventing re-growth of shoots, or continued low levels of disease. The surviving elm may have left a trace lasting several hundred years in the pollen record.
Post-dating the re-expansion of oak-hazel woodland around Long Breach, there is a phase of further vegetation change and disturbance. The character of the upland heath vegetation around Molland Common changes at 4,700±60 b.p. (3,640–3,360 b.c.). The shift is characterised by a change from heather-dominated heath to grass-dominated heath, and is closely associated with the start of significant levels of burning in the palaeoecological record (zone LBMC-2). A direct link between the levels of microscopic charcoal and the shift to grass-dominated moorland vegetation cannot be established although the timing strongly suggests that there is a causal link between the increased burning and a shift away from heather-dominated vegetation. There is a strong association of Calluna and fire in the palaeoecological record (e.g. Odgaard 1992; Simmons 1996). However this tends to relate to spread of blanket mire vegetation, in particular the removal of trees and the creation of open land. Modern studies of burning on heather moorland show that heavy burning may lead to a shift to grass-dominated vegetation: Rodwell (1992) states that the National Vegetation Community type U3 (Agrostis curtisii grass heath), shows a strong relationship with burning regimes. Todd (1996) suggests that Molinia caerulea will out-compete Calluna in stands that are heavily burned, with heavy grazing regimes also playing a role in control of heather, and there is evidence to suggest that uncontrolled or erratic fire may hamper regeneration of Calluna. Palaeoecological studies have demonstrated a possible strong causal link between charcoal and Molinia from macrofossil work (e.g. Chambers et al. 1999), and the deciduous nature of Molinia may lead to positive feedback in the frequency of fire on grass-heath. The upland vegetation of Dartmoor shows a strong relationship between charcoal (taken to be a proxy for fire frequency) and grass-heath (Caseldine and Hatton 1993).
It has been argued that fire history on Dartmoor, albeit in the later Mesolithic rather than the Neolithic, is related to hunting strategies to make animal movement and location more predictable in the landscape (Caseldine and Hatton 1993; Caseldine 1999). This is supported by earlier work by Mellars (1976), who argues that fire played a role in improving hunting strategies by maintenance and improvement of suitable grazing habitats, making the location of grazing herds more predictable. As well as fire, trampling and grazing by Red deer (and sheep) have been shown to be important in shifting Calluna heath to grass-heath (most commonly dominated by Agrostis capillaris and A. canina) (Hester and Baillie 1998; Palmer and Hester 2000). Increased faecal material may also promote a shift towards more species-rich, grass-heath dominated vegetation. It is likely, therefore, that if the increase in charcoal reflects deliberate burning over parts of Molland Common during the Neolithic, it was undertaken to manage and maintain the upland for grazing. The presence of stone rows and circles across Exmoor (Riley and Wilson-North 2001) suggest that the upland may also have been part of a ritual or ceremonial landscape during the Neolithic period.
Spatial aspects of woodland clearance during the early Bronze Age
There is little evidence from Long Breach for significant vegetation changes around Molland Common between the later Neolithic and the middle Iron Age, with the exception of very gradual reduction in woodland cover during this period (pollen zones LBMC-3). However, the evidence from Gourte Mires strongly suggests reduced woodland cover at 3,560±60 b.p. (2,120–1,730 b.c.), during the start of the Bronze Age. Representation of oak-hazel-alder woodland falls, whilst there is an increase in the representation of heather-dominated heath, together with a period of reduced microscopic charcoal. A stepped increase in heather is also recorded at Long Breach (dated by interpolation at around 3,500 b.p. i.e. ca. 2,000 b.c.).
The discrepancy between the data from Long Breach and Gourte Mires suggests that the focus of valley woodland clearance at the start of the Bronze Age was towards the surrounding lower areas to the south of Molland Common. Long Breach does not reflect woodland changes; it is likely that the site reflects woodland from around the easterly draining Dane’s Brook catchment, rather than the slopes and valley draining to the south into the Yeo valley. The difference between the sites allows some estimation of the patch size and scale of vegetation disturbance between these sites. Assuming the relevant source area (sensu Sugita 1994) of the sites to be similar, this must be at least half the distance between the sites, as they record different vegetation patterns. Thus the scale of disturbance, and relevant source area for Gourte Mires must be less than 500 m (half the distance between the two sites).
A number of undated field monuments that are ascribed to the later Neolithic and early Bronze Age are recorded from Molland Common. These include a stone setting comprising two upright and one recumbent stone, and ten round barrows which extend from East Ansteys Common to the east across the higher part of the study area (Riley and Wilson-North 2001). These monuments must have been sited within open landscapes on the flat interfluves, which supports the model of an open heath with wooded valleys. There is little archaeological evidence to suggest why woodland clearance occurs as recorded at Gourte Mires and not Long Breach; there are no archaeological field data to suggest the establishment of field systems on this part of the moor at this time. The palynological data may suggest that different areas of this part of the moor were utilised in different ways, hence the patchy woodland clearance. By contrast there is considerable evidence to show that comparable upland areas on Dartmoor were part of an agricultural landscape in the second millennium b.c. (Fleming 1979, 1988, 1994). The relict field systems (reaves) on Dartmoor are most likely to have been established between 3,250–3,050 b.p. (1,550–1,350 b.c., Fleming 1994), although these may have represented a short period of landscape exploitation. The evidence for species-rich grassland around Gourte Mires and Long Breach is strong at the start of the Bronze Age, and it is possible that field-systems were established in the lower parts of the landscape, possibly extending onto the higher moorland, although there is no trace in the archaeological record of this. The results from earlier work on Exmoor also suggest Bronze Age landscape clearance. On The Chains woodland is reduced at 3,505±120 b.p. (2,190–1,520 b.c., Moore et al. 1984), while at Hoar Moor woodland declines and grassland expands at ca. 3,500 b.p. (2,000–1,500 b.c., Francis and Slater 1990). As these sequences are from blanket mire their pollen source area is more likely to be regional, suggesting a general phase of landscape clearance in the uplands during the early Bronze Age. Evidence from the lowlands to the south of Exmoor also shows an open landscape by the middle Bronze Age (Fyfe et al. 2003). The timing of clearance in the blanket mire sequences, and the Gourte Mire sequence, is also broadly coincident with a phase of increased alluviation in the Barle valley to the north of Molland Common. This suggests increased sediment delivery to the river caused by accelerated woodland clearance (Fyfe et al. 2003). Woodland appears to regenerate around Gourte Mires by ca. 3,000 b.p. (1,350–1,100 b.c., dated by interpolation), that is towards the end of the Middle Bronze Age, and there is little significant change in the vegetation of the upland area until the mid- to late-Iron Age.
Woodland clearance during the later Iron Age
Both Long Breach and Gourte Mires show reduced woodland cover towards the end of the late Prehistoric period. At Long Breach there is a distinct drop in the levels of woodland shortly after 2,380±60 b.p. (770–370 b.c.), which is characterised by a decline in oak, hazel and alder. At Gourte Mires woodland is similarly reduced from 2,230±70 b.p. (410–90 b.c.). The decline in oak is rapid, while hazel and alder decline slowly. This suggests either selective removal of oak from mixed woodland with hazel and alder preferentially retained, or clearance of stands of oak. There is no overlap at one standard deviation between the dates from Long Breach and Gourte Mires for this clearance. However, after calibration there is some slight overlap between the age ranges. At Long Breach the date precedes the woodland decline while at Gourte Mires the position of the date is during the woodland decline. This strongly suggests that the event occurred at both sites at around the same time. Later Iron Age woodland clearance on Exmoor is also recorded at Moles Chamber and is characterised by a reduction in oak-hazel woodland at around 400-m OD (Fyfe et al. submitted). The evidence from Exmoor for woodland clearance during this period supports a wider regional pattern of later Iron Age woodland clearance. This clearance was common across northern and western Britain, being recorded at South Cumbria (Wimble et al. 2000), Lancashire (Mackey and Tallis 1994), North Cumbria (Dumayne and Barber 1994), the Solway Firth (Tipping 1995), Wales (Turner 1964), Shropshire (Leah et al. 1988) and southern Scotland (Tipping 1994).
Although the sequence at Ansteys Combe does not start until around the end of the Iron Age, it is clear that significant oak-hazel woodland is recorded in the sequence at the end of the Iron Age and the start of the Romano-British period. At first glance this contradicts the evidence from Gourte Mires and Long Breach. However, the major pollen source area for Ansteys Combe will have been the valley below the site, owing to the very enclosed nature of the steep-sided valley head. The evidence therefore suggests that this area and other steep valley sides are likely to have retained woodland through the later Iron Age and into the Romano-British period. Woodland would have been an important component of the Iron Age economy as a source of timber used in construction, of charcoal for industrial production such as iron working, and of fuel for domestic cooking and heating (Ralston 1999).
The evidence from Long Breach and Gourte Mires suggests a shift to wetter heath vegetation towards the later Iron Age. This shift may reflect the general decline in climate observed across much of North-West Europe during the later Prehistoric period (Lamb 1982; van Geel et al. 1996). Despite this shift to wetter heath, grass-heath was still dominant around the area, and the record of herbaceous taxa suggests that species-rich grass-dominated vegetation persisted. The continuous record of microscopic charcoal strongly suggests burning continued to be an important tool in managing heath vegetation. It is possible that the apparent shift to damper heath vegetation, with a possible re-expansion of heather in the vegetation, did not reflect a climatic shift, but partitioning of the upland fringes and the establishment of “enclosed” and “unenclosed” compartments in the landscape. This may account for the continued presence of species-rich grassland in the vegetation history (within an enclosed landscape) whereas in an unenclosed landscape heather and wet-heath vegetation will have regenerated or developed. It is likely that enclosure began in the lowlands and expanded into the uplands up river valleys, whilst the higher parts of the landscape may have remained unenclosed.
Later Iron Age to early Medieval landscape continuity
From the later Iron Age to the early Medieval period there is very little change in the pollen record from Molland Common (pollen zones LBMC-4, GM1–3, AC1–1 and −2). This strongly suggests continuity and stability in the land use of the southern fringes of Exmoor for a period of up to 1,000 years, and there is no reflection of the Roman period within the landscape. This supports the general concept of stability in rural south west England from the later Iron Age to the early Medieval reflected in the cultural record: there is no non-military domestic evidence for Roman presence throughout North Devon and Exmoor. The primary agricultural system during this period is pastoral, and the evidence from Ansteys Combe strongly suggests some woodland survival in the uplands draining the southern fringe of Exmoor. This woodland is most likely to have been part of a managed landscape, as a source of fuel, timber or for industrial processes. There is increasing evidence of iron production on Exmoor during the Roman period, requiring significant volumes of charcoal, although there is no archaeological evidence from the study area to suggest this activity was important there.
This open pastoral landscape with the use of fire in land management continues into the early Medieval period, supporting Francis and Slater’s (1990, 1992) evidence from both Hoar Moor and Codsend Moor on central Exmoor. In contrast, Moore et al. (1984) identify a period of woodland regeneration at The Chains dated to 1,500±60 b.p. (420–660 a.d.), which they suggested provided evidence for “large scale migration of people to Brittany because of plague and famine” (Moore et al. 1984: Table 2). The evidence from both this study and the work of Francis and Slater (1990, 1992) clearly refutes the idea that de-population occurred across Exmoor in the early post-Roman period.
Medieval expansion of cultivation
By ca. 1,000 b.p. (1,100 a.d.) the evidence from all three sites indicates a change in the pattern of land use on the upland around Molland Common. The data from Gourte Mires exemplifies the shift (zone GM1 lpaz-4) with a rise in cereal pollen types (Secale cereale and Avena/Triticum types, sensu Andersen, 1978) and associated weeds of cultivation, dated to 1,020±60 b.p. (890–1,170 a.d.). Species associated with pastoral land use decline and there is a reduction in microscopic charcoal, suggesting a cessation of burning on the upland. At Ansteys Combe the rise in cereal cultivation (mostly Secale cereale and the Hordeum group, sensu Andersen, 1978) is dated to 1,160±70 b.p. (680–1,020 a.d.). The Hordeum group of Andersen (1978) provides difficulties in interpretation as it includes wet flush species, including Glyceria, and therefore may not reflect cultivation. However, the strong presence of Secale cereale and associated weeds of cultivation (e.g. Anthemis-type) from the start of zone AC1–3 must indicate cereal cultivation around Ansteys Combe. The onset of cultivation at Ansteys Combe also marks a dramatic decline in oak-dominated woodland around the site, with a strong shift to a heather-dominated heath. Cereals are also recorded from Long Breach during this period.
The evidence of cereal cultivation from around the tenth century a.d., on Molland Common (at 340-m OD) is not restricted to sporadic grains of cereal-type pollen, rather all three sites record significant levels of cereal pollen, in particular rye (Secale cereale). Pollen of cereals is notoriously poorly dispersed beyond the immediate location of cultivation (Vourela 1970). Therefore it is highly unlikely that the evidence from the Molland Common sites represents cereal grains blown up the combes to the sites, especially given the position of the valleys in relation to the prevailing winds, but rather reflects the extension of cereal cultivation onto the upland. The archaeology of the surrounding upland strengthens this interpretation and documentary evidence indicates that the nearby farms at Lyshwell, Cloggs and Shircombe (1.5 km downstream from the site at Long Breach), are the remnant of a larger group of farmsteads (up to nine). However the historical data are insufficient to allow reconstruction of the number of settlements active during the later Medieval period (Riley and Wilson-North 2001). Extensive relict medieval field systems extend across Molland Common and it is likely these relate to this large group of holdings. The decline in woodland around Ansteys Combe further suggests increased pressure on either woodland resources during the later Medieval period, or a requirement for increased open land for pastoral activities.
The pollen data from Long Breach show lower levels of cereal pollen than Gourte Mires or Ansteys Combe. This local difference may reflect different uses of parts of the upland fringes, for example pastoral activities around Long Breach, with arable (or possibly some form of rotational pastoral and arable) focussed around Gourte Mires and Ansteys Combe. Hatcher (1988) and Fox (1991) argue that arable cultivation in the South West would have formed part of a convertible husbandry scheme. Under this system fields were subject to short cultivation periods of 1 to 3 years, followed by extensive ley periods of around 5 and 7 years. It is possible that the pollen analyses from Ansteys Combe and Goure Mires reveal a 5–10 year rotation of fields on the upland, with permanent pasturing around Long Breach, at least until 650±60 b.p. (1,270–1,420 a.d.), when arable agriculture may have expanded up towards Long Breach.
The evidence for the timing of cessation of arable agriculture around Molland Common is less clear than the evidence for the start of cultivation. At Ansteys Combe and Long Breach the decline in Secale cereale, which is taken to represent the end of rye cultivation on the uplands, is broadly synchronous with the rise of pine. This is estimated to have been at around 1,750–1,800 a.d., representing the establishment of plantations around the fringes of the upland. The timing of the expansion of arable cultivation into the upland fringe, together with the apparent continuous presence of arable agriculture throughout the medieval and later period, suggests that a model of high medieval expansion and contraction of arable into the upland margins must be re-examined (Henderson and Weddell 1994). In fact this evidence may support a model of continuity from the early Medieval well into the post-Medieval period.
The use of multiple pollen profiles has clearly demonstrated the potential for the reconstruction of a detailed local vegetation history for a small geographical area and allowed an estimate of the scale of human clearance. Where sites record different vegetation patterns, a maximum estimate of patch size, or the scale of human disturbance, may be defined as half the distance between each site. Between Gourte Mires and Ansteys Combe this is around 150 m. Local variation in vegetation is clear through comparison of the sequences from small mires, which reflect predominantly local, and extra-local vegetation history. For example woodland is cleared by the late prehistoric period on the higher areas of the upland; however, the sequence at Ansteys Combe demonstrates survival or preservation of woodland on the steep valley sides immediately adjacent (within 0.5 km) to the uplands until the later Medieval period. The earliest pollen data are from Long Breach, beginning during the later Mesolithic, around 5,400 b.p. (4,750–4,350 b.c.). The landscape around southern Exmoor was partially open; the woodland was predominantly oak-hazel, and low levels of elm are recorded, probably near the limits of its distribution. The open landscape was characterised by heather-dominated heath. The change from the Mesolithic to the Neolithic is clearly marked in the study area by a significant decline in all woodland species, suggesting the creation of clearings or significant open land around the upland shortly before 4,700±60 b.p. (3,640–3,360 b.c.). This occurred around the classic elm-decline date, although unlike other elm-decline events recorded from southern England there was a decline in oak-hazel woodland as well. This change does not fit with the transition model for the Mesolithic/Neolithic boundary, but suggests a sudden adoption of pastoralism or the possible immigration of pastoralists. The evidence from Long Breach shows that oak-hazel woodland regenerated shortly after the decline. This is corroborated by the Gourte Mires data, which record significant oak-hazel woodland when the peat accumulation began at 3,960±60 b.p. (2,630–2,280 b.c.). Shortly after the recovery of woodland there were significant changes in the nature of the open heathland vegetation. This was characterised by a shift to grass-dominated heath, very probably representing deliberate management of the upland vegetation through burning to facilitate upland grazing from the early Neolithic onwards. The extent and character of woodland during this shift in heath vegetation was unchanged, suggesting that the woodland was located on valley slopes and combes and around the upland fringes.
The use of multiple sites has allowed a better understanding of the spatial dimensions of woodland clearance and survival during the early Bronze Age to the late Iron Age. The evidence from Long Breach suggests that woodland persisted in and around the valley draining to the east of Molland Common. However, the data from Gourte Mires show woodland clearance around the southern edge of Molland Common at 3,560±60 b.p. (2,120–1,730 b.c.). The data from the higher blanket mire sequences on Exmoor similarly indicate a regional increase in woodland clearance during this period. The vegetation then shows remarkable continuity through time until the later Iron Age, when intensive woodland clearance is recorded at 2,230±70 b.p. (410–90 b.c.). This clearance fits an increasing body of evidence for regional woodland clearance across western and northern Britain during the mid-late Iron Age. In contrast to the evidence from Gourte Mires and Long Breach, the data from Ansteys Combe suggest that woodland persisted within the steeper-sided valley contexts draining to the south of Molland Common. It is likely that these landscape contexts would have been suitable for management of woodland as a resource, either for timber, domestic use or charcoal production for industrial uses.
From the later Iron Age through to the early Medieval period, the agricultural system around Molland Common appears to be pastoral, and there are no phases of scrub or woodland regeneration, clearly indicating that there was no cessation of activity around the upland fringe of Exmoor. This continued until around 1,000 a.d., when a significant shift to a mixed arable/pastoral regime is indicated from the Molland sites, although the data from Long Breach suggest that pastoral activities probably were more dominant towards the higher, central part of the study area. This pattern continued until well into the post-Medieval period, and most likely represents some form of convertible husbandry on the upland area. The data from the sites have demonstrated that the vegetation at the upland fringe was and is not homogenous, but most likely has been a mosaic of different vegetation communities since the earlier Neolithic.
This research was supported by the Leverhulme Trust (Grant F/00144/D). We wish to thank Rob Wilson-North (Exmoor National Park) for discussion of the archaeology, and Mark Dinnin, Zoë Ruiz and Jackie Hatton for help with field and laboratory work. Comments by Keith Barber and an anonymous referee were also most helpful.