Introduction

In northwest Europe, late Holocene vegetation composition and changes are considered to have been mainly controlled by human activities (Marquer et al. 2014). In general, this is shown by a decrease in woodland cover and an increase in cultural landscapes (Nielsen et al. 2012; Marquer et al. 2014; Githumbi et al. 2022a, b). In large parts of northwest Europe this decrease in woodland was temporarily reversed during the Late Roman period and Early Middle Ages (Teunissen 1990; Andersen and Berglund 1994; Bunnik 1999; Berglund 2003; Wick et al. 2003; Dreßler et al. 2006; Kalis et al. 2008; Rösch and Lechterbeck 2016). After this short phase of regeneration, there was a further rapid reduction of woodland as a result of medieval and modern clearances (Janssen 1974; Bunnik 1999; Litt et al. 2009; Whitehouse and Smith 2010; Nielsen et al. 2012; Fyfe et al. 2013; Marquer et al. 2014). In this paper, the phase of woodland regeneration or regrowth from the Late Roman period and Early Middle Ages is referred to as the Dark Ages (DA) woodland regeneration, in accordance with the overlapping Dark Ages Cold Period (DACP) of ad 250–700 (Helama et al. 2017; Riechelmann and Gouw-Bouman 2019) and the historical Dark Ages (James 1988). The Dark Ages are also referred to as the Migration Period in other parts of Europe, although this can cover a longer historical span (Halsall 2007). The DA woodland regeneration has been recognised in pollen studies from countries including Norway (Hjelle et al. 2022), Germany (Zolitschka et al. 2003; Dreßler et al. 2006; Rösch 2007; Litt et al. 2009; Rösch and Lechterbeck 2016; Gerlach et al. 2022; van der Knaap et al. 2020), Belgium (Broothaerts et al. 2014a, b; Deforce et al. 2020), France (Etienne et al. 2013), Switzerland (Tinner et al. 2003; Wick et al. 2003), Poland (Gałka et al. 2014; Brown et al. 2019; Pędziszewska et al. 2020), Britain (Dark 2000; Forster 2010) and the Netherlands (Teunissen 1990; Bunnik 1999). In the Netherlands, this Dark Ages phase of expanding woodland is dated around ad 300–700, covering the Late Roman period (ad 270–450) and the first part of the Middle Ages (ad 450–1050) (Teunissen 1990; Bunnik 1996; Kalis et al. 2008). In spite of its widespread occurrence, little is known about the regional pattern and extent of this phase of woodland regeneration, or the role of climate, environmental and human changes as its primary causes (Fig. 1).

Fig. 1
figure 1

Climate, population, vegetation and landscape changes in the Netherlands for the period 500 bcad 1000 (2,500–1,000 cal bp). On left, IA, Iron Age; ROM, Roman period; EMA, Early Middle Ages. Climate warm in red and cold in blue; RWP Roman Warm Period; DACP Dark Ages Cold Period; MCA Medieval Climate Anomaly, after Gouw-Bouman et al. (2019). People changes in population density in the Netherlands, data from Groenewoudt and van Lanen (2018). Woodlands changes in AP percentages from amalgamation of data from this study. Landscape changes blue, the largest environmental changes in the coastal area; green the river area, avulsions are the formation of new river channels and the abandonment of old ones; yellow, the sand area, data from Pierik (2017). On right, the main pollen zones; Va, Subboreal; Vb1, Early Subatlantic; Vb1r, DA woodland regeneration; Vb2, Late Subatlantic (de Jong 1982; van Beek et al. 2015)

The DA woodland regeneration directly followed a period of great change to human societies such as the collapse of the Roman Empire, which resulted in a population decline (van Es and Verwers 2010; van Beek and Groenewoudt 2011; Pierik and van Lanen 2019; Groenewoudt and van Lanen 2018; Pierik et al. 2018). Therefore, the cause of this regrowth phase is often considered to be the decreased human pressure on the landscape during this period. However, climate records also indicate a colder and wetter climate (Ljungqvist 2009; Büntgen et al. 2011, 2016; Helama et al. 2017; Gouw-Bouman et al. 2019; Riechelmann and Gouw-Bouman 2019), while geomorphological records point to a changing landscape (Vos 2015; Pierik et al. 2017a, b). How and to what extent these climatic and environmental factors contributed to the changes in the vegetation or even the decline of the Roman Empire is largely unknown. Since significant cultural, landscape and climate changes occurred simultaneously during the Late Roman period and Early Middle Ages, this period can be used to study the intricate relationship between vegetation and its influencing factors. The varied landscape and cultures of the Netherlands during the Late Roman period offer a good area to explore the influences of landscape setting and population dynamics on the DA woodland regeneration. Existing data from the Netherlands suggest that the degree of woodland regrowth and its timing varied spatially (Teunissen 1990; Bunnik 1999), which might be related either to variations in the abiotic (non living) landscape or human impact there.

The aim of this study was to map the spatial pattern and differences in the degree of DA regeneration of woodland and to determine the main factors explaining these, using many Dutch pollen records from various landscape and environmental settings. Based on these available pollen records, an overview of the changes in the main vegetation structure from the Iron Age/Roman period (IA/ROM) to the Dark Ages (DA) is provided. In this study the increase in woodland in the DA and its regional patterns were investigated, using the changes in AP (Arboreal Pollen) percentages as an indication. The reconstructed changes in woodland and overall vegetation composition were then compared to population estimates and landscape characteristics to identify the main controlling factors for this phase of late Holocene vegetation development.

Research area

Landscape development in the Netherlands during the first millennium ad

In general, the landscape of the Netherlands during the first millennium ad can be subdivided into four zones (Berendsen 2008; Pierik 2017), the coastal, river, sand and peat areas (Fig. 2). During this time a large part of the landscape was covered by peat. The peatlands in the northern part of the Netherlands and at some distance from rivers consisted of oligotrophic peat bogs. Closer to the rivers and coastal creeks, peat formed under more eutrophic conditions as a result of flooding with associated nutrient input. The peat area was generally flat and conditions were continuously wet, so that extensive natural woodlands could not develop and there was mostly swamp and bog vegetation instead. In the Roman period, parts of the peat area near the coast were flooded as a result of sea ingressions and became part of the coastal area (Pierik et al. 2017a).

Fig. 2
figure 2

Palaeogeographical map of the Netherlands for ad 100 (adapted from Vos et al. 2011) showing the main landscape elements and pollen sites (ESM Table S1). In red, sites shown in Fig. 6. N northern sand area; M middle sand area; E eastern sand area; S southern sand area. Red dashed line, position of the Limes, northern border of the Roman Empire

Behind a sandy coastal barrier there were salt marshes, shoals, small tidal creeks and larger tidal inlets. The subsurface was mainly composed of nutrient- and clay-rich sediments, and conditions were relatively wet since there were minimal differences in surface height, and flooding with salt water occurred regularly, daily to annually. As a result of the wet and saline conditions, extensive woods were not present in the coastal area and the vegetation there mainly consisted of open salt marshes with species such as Aster tripolium (sea aster) and Crithmum maritimum (samphire). From ad 600 onwards flooding increased and new coastal inlets emerged (Fig. 1; Vos and Heeringen 1997; Vos 2015; Pierik et al. 2017a).

The river area was characterised by active river channels bordered by sandy natural levees that were embedded in the clay-rich flood basins. The most important river systems in the area were the Rhine and Maas (Meuse). In the subsurface, sandy and clay-rich sediments were present and nutrient availability in the entire area was high. Flooding occurred with a seasonal to decadal frequency. There were height differences of 0.5–1 m between the natural levees and the lower flood basins resulting in relatively dry conditions on the higher ridges of active and palaeo-channel belts, with relatively wet conditions in the lower-lying flood basins (van Dinter 2013; Pierik et al. 2017b). In the river area the natural vegetation during the late Holocene can be characterised as a mixed deciduous woodland with Quercus (oak), Fraxinus (ash) and Corylus (hazel) or wet woodlands (carrs) with Alnus (alder); open areas were either covered by marsh, grassland or cultivated land. In the Rhine-Maas (Meuse) delta, the sediment supply increased from the Roman period onwards (Erkens et al. 2006), coinciding with increased flooding after ad 300 (Toonen et al. 2017) and there were a large number of severe floods in the period ad 185–282 (Fig. 1; Jansma 2020). Additionally, new river channels such as the Waal, Lek and IJssel developed around ad 300, reorganising the delta landscape (Pons 1957; Tornqvist 1993; Berendsen and Stouthamer 2001; Pierik and van Lanen 2019).

The Dutch sand area is part of the western section of the northwest European sand belt (Koster 2005) and it consists of Pleistocene ice-pushed ridges and plains. Differences in height vary between 1 and 4 m, but can reach over 100 m for ice-pushed ridges. The subsurface mainly consists of sand and conditions are relatively dry and often nutrient poor. The sand area of the Netherlands can be further subdivided into four zones, the southern; middle, eastern and northern sand areas (Pierik 2017) (Fig. 2). Glacial features are present in all except the southern sand area. In the northern sand area the substrate is loamy and wet due to the presence of an impermeable till of Saalian age (Rappol 1987; Bosch 1990). In the middle and eastern sand areas large ice-pushed ridges are present, which contain coarse and gravelly sediments (Busschers et al. 2008). The substrate in the southern sand area is relatively loamy, resulting in relatively wetter conditions (Bisschops et al. 1985; Schokker et al. 2007). In this study the southern loess area, located in the far south of the Netherlands, is included in the southern sand area since both have a topsoil which is well drained and nutrient rich, and were within the Roman Empire (Schokker et al. 2007; Stouthamer et al. 2015). The southern loess area is hilly, reaching more than 100 m, and the substrate mainly consists of nutrient rich loess (Schokker et al. 2007; Stouthamer et al. 2015). Conditions in the first millennium ad in the sand area remained relatively stable compared to the coastal and river zones. During the investigated period, mixed deciduous woodlands with Quercus (oak), Tilia (lime) and Corylus (hazel) or woods of Pinus (pine) and Betula (birch) grew in this region, while the open landscape was covered by heathlands, grassland or cultivated land.

Late Holocene vegetation history in the Netherlands

In pollen diagrams from the Netherlands, human impact on the vegetation becomes visible from the Atlantic period (middle Holocene; ca. 9,000 cal bp) onwards (Polak 1959; Janssen 1974; de Jong 1982; Kalis et al. 2003). These earliest impacts are often shown by a decrease in arboreal pollen (AP) and the occurrence of indicators of disturbance such as Rumex acetosella (sheep’s sorrel, Pteridium aquilinum (bracken) and Plantago lanceolata (ribwort plantain) (Janssen 1974; Behre 1981; Bos and Janssen 1996; Bos and Urz 2003; Bos et al. 2006). The subsequent increases in human impact and associated decreasing woodland cover are well recognisable in pollen data in reduced percentages of AP and an increase in cultural indicators (de Jong 1982; Berendsen and Zagwijn 1984; Behre 1988). From ca. 7,000 cal bp onwards the change to a more agricultural and settled society accelerated the loss of woodland (Polak 1959; Zolitschka et al. 2003). A first minimum in AP values in pollen diagrams from the Netherlands is reached in the Iron Age or Roman period (Fig. 1; Teunissen 1990). Its exact timing and degree varies from place to place as a result of spatial differences in the amount of human impact. In the Late Roman period this decrease in woodland reversed, with increasing AP and decreasing crop and disturbance indicators. Overall, AP percentages for this phase of woodland expansion are like those for the Bronze Age, when human impact was still minimal in the Netherlands (Fig. 1; Teunissen 1990; Kalis et al. 2008). It therefore seems that the vegetation during the Dark Ages was able to return to a state that was similar to the situation when human interference was minimal. This 300–500 years long woodland regeneration lasted until the Early Middle Ages, after which evidence of human presence increases rapidly. Pollen records from then on show the resulting reduction in woodland in rapidly decreasing AP values and increasing amounts of crop pollen, especially cereals, and disturbance indicators. The large increase in cereal pollen was partly caused by the start of Secale cereale (rye) cultivation from this period onwards (van Zeist 1976) and not solely as a result of an increase in arable land. Rye is a wind-pollinating species and thus distributes large quantities of pollen, in contrast other cereals such as Triticum (wheat) or Hordeum (barley) which release most of their pollen only during threshing (Joosten and van den Brink 1992). In subsequent centuries there were some short phases of woodland regeneration shown by increasing AP percentages, such as around ad 1400 during the Little Ice Age, at ad 1663–1664 in the plague epidemic and an increase of Pinus pollen from ad 1700 onward as a result of the establishment of pine plantations (de Jong 1982). However, the Holocene vegetation cover in northwest Europe has never fully recovered from these medieval and later clearances of woodland. Based on vegetation reconstructions in Twente, eastern Netherlands, the period around ad 1900 can be identified as the most treeless period in the vegetation history of the Netherlands since the climax wildwood of the Atlantic period in the middle Holocene (van Beek et al. 2015).

Cultural developments in the Netherlands during the first millennium ad

In the Netherlands, the Roman period started in 12 bc and ended in ad 450 (RCE 2016). The Limes, the northern frontier of the western Roman Empire, was established along the river Rhine during the first century ad (Bosman and de Weerd 2004; Polak 2009; van Dinter 2017). During the first and second centuries ad Roman society thrived, cities were established as at Ulpia Noviomagus Batavorum (Nijmegen) and Forum Hadriani (Voorburg), and there was a centralised administration resulting in a planned landscape with roads built and surplus production of food and consumables (van Es 1981; Willems 1981; van Lanen 2017). Population numbers increased during this period as a consequence of the influx of the Roman military and a continuation of the population growth since the Bronze Age (Fig. 3). In the sand areas, population density increased from 2.4 to 2.5 people/km2 during the Late Iron Age to 4.9 around ad 200 (Pierik et al. 2018). In the eastern river area, population estimates increase from 14.5 people/km2 in the Early Roman period (12 bc to ad 70) to 25.5 in the Middle Roman period (ad 70–270) (Groenewoudt and van Lanen 2018). The western part of the river area lacks data, but the population increase there is presumed to have been smaller than in the eastern part. From ad 270 onward, Roman presence in the Netherlands strongly decreased, causing a drop in settlement numbers (Louwe Kooijmans 1995; van Lanen 2017). This decline occurred across the Netherlands, however, its timing and magnitude varied (Fig. 3; Willems 1981; Groenewoudt and van Lanen 2018). In the southern and middle sand areas the estimated population density was 0.5 people/km2 around ad 500, whereas the densities in the northern and eastern areas at this time returned to values of 2.4–2.5, which are similar to the estimates for the Late Iron Age (Pierik et al. 2018). Although population density around ad 500 in the eastern river area, with an estimated 3.6 people/km2, was still higher than in the sand areas, the decrease in population density there was the strongest (Groenewoudt and van Lanen 2018). Following the collapse of the western Roman Empire, several tribes such as the Franks migrated into the Netherlands. However, population numbers remained low during the 4th and 5th centuries ad. From the 6th century onwards settlement numbers increased and the Netherlands became a focal point of medieval trade along the river Rhine with the emergence of Oegstgeest near the coast in the 6th century (Theuws et al. 2021) and Dorestad (Wijk bij Duurstede) in the central part of the delta in the 8th century (van Es and Verwers 2010).

Fig. 3
figure 3

Population density estimates for the eastern river and all sand areas (Groenewoudt and van Lanen 2018; Pierik et al. 2018; Pierik and van Lanen 2019)

Climate in northwest Europe during the first millennium ad

Several climate fluctuations occurred in northwest Europe during the first millennium ad which can be subdivided into the Roman Warm Period (RWP; 300 bc to ad 250), the Dark Ages Cold Period (DACP; ad 250–700) and the Medieval Climate Anomaly (MCA; ad 700–1000) (Büntgen et al. 2011, 2016; Riechelmann and Gouw-Bouman 2019). A lake temperature record inferred from chironomids in Uddelermeer, central Netherlands, indicates that the RWP started in 290 bc and lasted until ad 190 and the DACP lasted from ad 190 to ad 670 (Gouw-Bouman et al. 2019). The reconstructed summer temperature values indicate a drop of 1.5 °C from the RWP to the DACP. Corresponding precipitation records for the Netherlands are currently not available, and those from elsewhere in northwest Europe do not indicate a spatially consistent trend over time (Riechelmann and Gouw-Bouman 2019). The increase in flooding frequency and severity of the Rhine in the Late Roman period nevertheless suggests wetter conditions due to increased effective precipitation (Toonen et al. 2013; Jansma 2020).

Methods

Data collection

Pollen data were collected from the Dutch Pollen Database (Donders et al. 2010) and literature, and in addition five new sites were analysed for this study. Pollen records from large lakes or bogs, which are considered to record regional vegetation (Sugita 2007), do not exist for large parts of the Netherlands. Nevertheless, an indication of the regional vegetation can also be achieved by compiling pollen data from various different environmental settings (Fyfe et al. 2015). Pollen diagrams were selected that at least covered the Iron Age and Roman period up to the Early Middle Ages. Furthermore, only profiles where the period of woodland regrowth has been preserved in organic sediment such as peat have been used. Not all the datasets are supported by independent dating but undated sequences were only included when they could be relatively dated by using the pollen composition. To establish a minimum Roman age, the following criteria were used: (1) the presence of Carpinus (hornbeam) and/or Juglans (walnut) and (2) the presence of low values of Secale (rye). Hornbeam is present in pollen records from the Netherlands from the start of the Subatlantic pollen zone (ca. 2,000 cal bp) onwards (de Jong 1982). Rye and walnut are both Roman time introductions (van Zeist 1976; Zeven 1997). To exclude a medieval or later date for the pollen sequence, the following criteria were used: (1) the absence of Centaurea cyanus (cornflower) and/or Fagopyrum (buckwheat) and (2) the absence of high values of Secale (rye). Cornflower is common in pollen data in the Netherlands from the 10th to 11th century onwards and buckwheat pollen is common from the 12th to 13th century and later (van der Linden 2008; Brinkkemper 2013). Rye cultivation and consequently its pollen deposition in the Netherlands increased massively from ad 900 onward (de Jong 1982).

In total, pollen data from 38 sites were selected; 26 of these sites were located in the sand area, five in the peat area and seven in the river area (Fig. 2; ESM Table S1). It is possible that the DA expansion of woodland was covered by more pollen sequences, but that this relatively short time period was not sampled or the records are insufficiently detailed for it to be clearly detectable. The coastal area was not included in this study since there are so few of these pollen records, due to later large-scale flooding and erosion resulting in a lack of organic deposits (Pierik et al. 2017a). In addition, this area probably did not have extensive woods. Due to its open character and consequently larger source area for pollen, pollen studies from the peat area mostly reflect the vegetation of the surrounding landscape and were therefore assigned to the nearest landscape unit. Although peat profiles provide perfect pollen archives and large parts of the Netherlands were covered by extensive peat bogs, only a few records are available due to large scale peat digging and oxidation as a result of artificial drainage since medieval times. Using the former Roman Limes as a boundary between north and south, which follows the river Rhine, 15 sites are located in the north and 22 within the former Roman Empire. In order to give an overview of the vegetation history of each landscape type, six sites were selected, which are approximately comparable in size, resolution and time period covered. Pollen diagrams were drawn for these sites, covering the entire research period, the Bronze Age up to the Middle Ages.

Pollen sum

Pollen percentages from all records were recalculated using uniform pollen sums. Since the pollen data were derived from various studies, authors, sites, landscape types and environmental settings, the original pollen sums of the data differ greatly. If possible, original count data have been used, but if not, percentages were derived directly from the pollen diagrams or digitised from the available pollen diagrams using Surfer (Golden Software 1996) or Getdata (Fedorov 2002–2013). These digitised pollen percentages were then converted to pollen counts using the original pollen sums and then recalculated using the two pollen sums described below.

In this study, two pollen sums were selected to investigate the regional regrowth of woodland: (1) a terrestrial pollen sum (Pterra) and (2) a total pollen sum (Ptotal). The terrestrial pollen sum comprised of two groups: (1) Arboreal pollen (APterra) with trees and shrubs (excluding Alnus and Salix) and (2) a non-arboreal group (CUL) with crops and terrestrial (dry land) herbs. The total pollen sum comprises four groups (1) all trees and shrubs (APtotal), (2) herbs and crops (CUL), (3) Ericaceae (heathers) (HEA) and (4) Poaceae (grasses) (GRA). Due to limitations in data availability it was not possible to calculate and derive pollen percentage values for all sites using the total pollen sum. The CUL group includes crops and the pollen types of herbs such as Plantago lanceolata and Asteraceae which can be associated with cultural landscapes such as agricultural land, ruderal areas and grazed grassland (Behre 1981). Therefore, the CUL group can be used as an indicator of human impact.

Vegetation change

From the pollen records, the pollen percentages of AP, Alnus, NAP including herbs and crops, heathers and grasses were derived. These values were determined for two phases of vegetation development; the first minimum in AP in the IA/ROM and then the maximum of AP during the DA woodland regeneration. Instead of dated intervals, the minima and maxima in tree and shrub pollen percentages have been chosen to reduce dating errors, since some records do not have independent dating. Because the DA regrowth was a short-lasting phase, pollen values from a single depth were used instead of amalgamated data from a larger depth interval. However, the original samples were often collected in 1 cm slices and so they already contained data from a larger, decadal time interval. Not all the datasets have a very high resolution, so it is therefore possible that they do not capture the absolute highs and lows of the DA regrowth, leading to an underestimate of the growth and landscape openness. All landscape regions provide both low and high resolution records. Due to local disturbances and the effect of the pollen sum on the vegetation reconstruction it was not always straightforward to determine the minimum AP during the IA/ROM. To obtain a uniform dataset the following criteria have been used: (1) the depth at which the minimum occurred is based on the terrestrial pollen sum reconstruction (see pollen sum); (2) it should occur in the Iron Age or Roman period and (3) it should occur directly before the DA regeneration phase.

Although pollen percentages cannot be used to estimate absolute vegetation or tree cover, the AP/NAP ratio can serve as an indicator of the openness of the landscape (Groenman-van Waateringe 1986; Frenzel et al. 1992; Frenzel 1994; Sugita et al. 1999; Doorenbosch 2013) and has been used to map the effects of the DA regrowth in Poland (Pędziszewska et al. 2020). To map and compare its extent between sites in this study, the increase in AP percentages was expressed as a growth percentage (GP). In addition, changes in pollen percentages for all vegetation groups were also expressed as a GP. The following formula was used:

$$GP=\frac{Pb-Pa}{Pa}*100\text{\%}$$

with Pa as the pollen percentage for the IA/ROM and Pb as the pollen percentage for the DA.

Vegetation reconstruction using REVEALS

Differences in pollen productivity and dispersal of individual plant taxa can affect the changes in AP and NAP percentages (Andersen 1970). To evaluate the results based on pollen percentages, the data of 13 sites, for which pollen counts were digitally available, were recalculated using REVEALS to convert the data into estimates of vegetation abundance and to correct for biases in pollen productivity and dispersal. For this we used the REVEALSinR function in the DISQOVER package (Theuerkauf et al. 2016). The REVEALSinR model was run using the Lagrangian stochastic dispersal model which simulates turbulence in the atmosphere (Kuparinen et al. 2007). Pollen Productivity Estimate (PPE) values were taken from Fyfe et al. (2013) (Table 1). The model was run using 19 taxa which were arranged into the vegetation groups used in the total pollen sum (Table 1). Apart from Uddelermeer (12) which is a lake, all sites used in the model were peatlands. The diameter of each site as included in the model is given in ESM Table S1. Error estimates were obtained from 1,000 model runs after which the 10th and 90th percentiles were taken as boundaries for estimation of the standard deviation error (Theuerkauf et al. 2016). From the REVEALS dataset, pollen percentages were selected from the same levels (IA/ROM and DA) as in the original data, and growth percentages were also calculated in the same manner.

Table 1 Pollen taxa included in REVEALS by vegetation group with fall speed (m/s), pollen productivity estimates (PPEs) and their error margins (Fyfe et al. 2013)

Results

Vegetation development during the first millennium ad

The sites Engbertsdijksveen (9), Uddelermeer (12), Kootwijkerveen (14), Giesbeeksche broek (17), Kleefsche beek (23) and Moerkuilen (27), located from north to south across the Netherlands, were selected to describe and compare the vegetation development in different landscape types and environmental settings during the first millennium ad (ESM Table S1; Fig. 4). Uddelermeer is a lake and Kootwijkerveen a bog, both located on the Veluwe in the middle sand area. Engbertsdijksveen is a large raised bog located within the eastern sand area and Moerkuilen is an organic valley fill in a stream valley in the southern sand area. The site Kleefsche Beek is an organic channel fill in an abandoned channel of the Maas and Giesbeeksche broek is a peat sequence in the flood basin of the Rhine. All records are supported by radiocarbon dating for chronology. The main vegetation components for these sites are shown in Fig. 4 for the period Bronze Age up to the Middle Ages using the two pollen sums, Pterra and Ptotal, described in ‘Pollen sum’, as well as the REVEALSinR (PREVEALS) modelled data when available.

Fig. 4
figure 4

Summary pollen diagrams showing regional vegetation history in the Netherlands, from north to south. A Engbertsdijksveen (9); B Uddelermeer (12); C Kootwijkerveen (14); D Giesbeeksche broek (17); E Kleefsche beek (23); F Moerkuilen (27), the diagrams from the terrestrial land (Pterra) and total (Ptotal) pollen sums and on the data modelled by REVEALS (PREVEALS). In green, AP, arboreal pollen; yellow, CUL, herbs and crops; purple, HEA, Ericaceae; yellow-green, GRA, Poaceae. Alnus is included in the AP in the Ptotal and PREVEALS percentages, and is shown as a line graph (LC). Locations of sites shown in Fig. 2. No digital pollen percentage data was available for the sites Engbertsdijksveen (9) and Moerkuilen (27), so it was not possible to model the data from these sites with REVEALSinR

Engbertsdijksveen, eastern sand area

The pollen profile from Engbertsdijksveen was taken from a depression in a large peat bog and was studied by van der Molen and Hoekstra (1988). From this record vegetation history from around 2,900 to 1,000 cal bp is shown, covering a large part of the late Holocene, the Iron Age to the Late Middle Ages (ESM Table S1; Fig. 4). In comparison to the other sites, AP values (using Ptotal and Pterra) at Engbertsdijksveen are high, indicating minimal local human impact, which is not surprising since this site is within a large peat bog. Large bogs, apart from being rather inaccessible, also capture mostly wind-transported pollen of which the majority comes from trees and shrubs. From the bottom of the profile upwards, AP values start relatively high at 80–90% and slowly decrease to 75–85%. The spectrum is dominated by Quercus (oak), Betula (birch) and Corylus (hazel), while Fagus (beech), Fraxinus (ash), Pinus (pine) and Ulmus (elm) are present in lower percentages (van der Molen and Hoekstra 1988). A minimum woodland cover is represented at around 80–65 cm depth in the profile (~ 2,000 cal bp, Roman period) due to increases in pollen from herbs from dry localities and crops as well as an increase in Poaceae (grasses) (Ptotal). Ericaceae (heather) values (Ptotal) show little variation throughout this period. These results show that the landscape became more open with ruderal, agricultural and grassland areas during the Roman period. Above 55 cm, AP values, first mainly of Quercus but later also of Betula, Fagus and Carpinus (hornbeam) gradually increase, reaching values of 90–95% at 40–30 cm (~ 1,400 cal bp, Early Middle Ages). This increase, although small, is interpreted as the Dark Ages woodland regeneration. Local trees and shrubs, mainly Alnus (alder) but also Corylus, did not contribute significantly to these increases in AP values. The abundance of ruderal, agricultural and grassland pollen types decrease in this section, indicating that woodland spread over previously open areas. After this phase of regrowth, AP values gradually decline. The pollen percentage curves in the Ptotal representation are less smooth in comparison to the terrestrial one, which is probably the result of some local presence of Poaceae (grasses) and Ericaceae (heather).

Uddelermeer, middle sand area

In the record from Uddelermeer, vegetation development for the period 2,500–700 cal bp is shown, representing the Iron Age to late Middle Ages. Uddelermeer, originally a pingo, a mound with an ice core, has been a lake since the Late Glacial and currently has a diameter of 300 m with a maximum water depth of 1.3 m. The pollen results are very similar to those of Engbertsdijksveen bog, but with lower AP values throughout, which indicate a more open landscape and consequently more (local) human influence there (ESM Table S1; Fig. 4). AP values for the Roman period from Uddelermeer are 80% (Pterra), 60% (Ptotal) and 30% (PREVEALS) with mainly Quercus and Corylus and smaller values of Betula and Fagus (Gouw-Bouman et al. 2019). The landscape around Uddelermeer during this period consisted of deciduous woodland, arable fields and heathlands. The DA return of woodland is visible between 425 and 340 cm (~ 1,800– ~ 1,700 cal bp) with an increase in AP values to 90% (Pterra); 75% (Ptotal) and 45% (PREVEALS). Even though AP values are lower in the modelled PREVEALS data, the increase in woodland is still visible, and the overall pollen trend of the different vegetation groups including Poaceae and Ericaceae is comparable to the unmodelled data as shown in the Ptotal representation.

The increase in AP can mainly be attributed to Quercus and Fagus suggesting the expansion of the deciduous woods which replaced arable land and grasslands, as reflected in the decreasing values of herbs and grasses (Gouw-Bouman et al. 2019). Some local over-representation of Poaceae is visible at 385 cm; however, the overall trends are not affected. Grasslands and heaths were equally present in the area during the late Roman period, and while grasslands decreased during the DA woodland regeneration there, the extent of heaths remained stable. Alder only contributed minimally to the woodland regeneration at this site.

Kootwijkerveen, middle sand area

In this record from the middle sand area, vegetation history from ~ 1,800 to ~ 800 cal bp is covered. The Kootwijkerveen bog is 150 m wide and 600 m long and located 4.5 km to the southeast of Uddelermeer. Both the pollen trends and the range of pollen taxa are similar to those at Uddelermeer, but the high AP values are comparable to Engbertsdijksveen (ESM Table S1; Fig. 4). In the diagrams calculated with the total pollen sum and the REVEALS dataset, the large role of Ericaceae in the vegetation around the Kootwijkerveen bog becomes apparent. It is likely that heather grew locally on the bog, but the similarly high percentages in the Uddelermeer record suggest that there were significant regional heathlands in this area during the first millennium ad. In contrast, grasses seem to have played a minor role in the vegetation at Kootwijkerveen and the low values of herbs and crops indicate that there was little human influence on the direct surroundings of the bog.

The start of woodland expansion shows up at 130 cm (~ 1,350 cal bp) and it reaches its maximum at 90 cm (~ 1,125 cal bp). In the diagram calculated with the terrestrial pollen sum the DA regrowth is only marginally visible, whereas in the total and REVEALS representation there is a clear increase of AP percentages. Just as in the Uddelermeer record, the regeneration is mostly seen as an increase in Quercus and Fagus, with decreasing values of Ericaceae. Alnus increases in the last part of the period of regrowth, but its contribution to it is minimal.

Giesbeeksche broek, river area

The Giesbeeksche broek record shows vegetation development from ~ 2,900 to ~ 1,500 cal bp (Teunissen 1990). The sampling site was in a peaty flood basin in the river area and was presumably continuously wet as a result of flooding. Input of clastic (mineral material) into the sediment is minimal in the sequence at 145–132 cm and the influence of water-transported pollen grains on the vegetation reconstruction is therefore negligible in this part of the record. From 132 cm upwards there is a slight input of clay and associated nutrients and potentially some fluvial transport of pollen (by river water) in this part of the sequence. Above 116 cm the sediment changes to clay, and since this occurs after the woodland regeneration phase, this site is included in this study. This increase in fluvial activity and the deposition of clay as a result of flooding is seen across the entire river area in this period (Pierik et al. 2017a, b).

AP values at the start of this record are low 65–75% (Pterra); 75–90% (Ptotal) and 45–65% (PREVEALS), with Quercus, Corylus, Betula and to a lesser extent Fagus as the most important trees and shrubs, while Alnus is also present in large percentages (Teunissen 1990). In the diagram from the terrestrial pollen sum (Pterra), AP values for the IA/ROM period are relatively low (65%), whereas in the diagram using the total pollen sum, they are significantly higher as a result of a large contribution of (local) alder (ESM Table S1; Fig. 4). The low AP percentages indicate a relatively open landscape and substantial human impact in this area during the Roman period and Late Middle Ages. Poaceae values are high before and after the DA return of woodland and even more so in the REVEALS data. Grassland thus formed a large part of the vegetation, especially during the periods with less woodland. The landscape around this site consisted of deciduous woods and arable land in the drier places, with grassland and alder carr in the wetter parts. The DA woodland regeneration is clearly visible, starting at 130 cm (~ 2,000 cal bp) and reaching its maximum values at 120 cm (~ 1,700 cal bp). The increase in AP is mainly of Quercus and Corylus, and to a lesser extent Betula, as well as a large increase in alder during this period. The increase of alder and decrease of grass during the regrowth phase suggests that the grasslands were replaced by alder carr.

Kleefsche beek, river area

The Kleefsche beek diagram shows the vegetation changes from ~ 3,000 to ~ 900 cal bp, from the Bronze Age until the Late Middle Ages in the river area (Teunissen 1990; Kalis et al. 2008). The site was an organic infill of an abandoned channel of the Maas. The section of the channel infill representing the first millennium shows no signs of clay input or of fluvially transported pollen. The record is similar to that from Giesbeeksche broek both in pollen taxa and vegetation changes (ESM Table S1; Fig. 4). Low AP percentages for the Roman period indicate a relatively open landscape where grasslands formed an important part of the vegetation in the river area. The return to woodland is clearly noticeable, starting at 208 cm (~ 1,900 cal bp) and reaching its highest AP values at 193 cm (~ 1,400 cal bp). In contrast to the sites in the sand area, the reconstruction using the total pollen sum (Ptotal) shows a large contribution of Alnus which coincides with a decrease in Poaceae, showing that grasslands were replaced by alder carr during the woodland regeneration.

Moerkuilen, southern sand area

The diagram from Moerkuilen shows the vegetation from ~ 2,000 to ~ 600 cal bp (Janssen 1972). The Moerkuilen site was in the former peaty flood basin of a small stream or brook valley which was often flooded and most probably covered by wet grassland or marsh vegetation with Phragmites (reed).

Based on the AP percentages (ESM Table S1; Fig. 4), the vegetation around Moerkuilen was relatively open during the Late Iron Age and Roman period. Important trees and shrubs in this region were Quercus, Corylus and Fagus (Janssen 1972). There are several peaks in the Poaceae curve during the entire period, which represent local grassland vegetation. Ericaceae pollen values vary between 5 and 10% and are continuously present, showing that there were also heaths in the surroundings with a relatively stable presence during the investigated time period. The increase in woodland is visible at 40–25 cm (~ 1,700 to ~ 800 cal bp). In the terrestrial representation, a gradual rise in AP can be seen during this period, mostly of Fagus and Corylus (Janssen 1972). In the diagram from the total pollen sum, the return of woodland is less evident due to the local presence of grasses, which is so strong that to some extent it even reverses the AP/NAP ratio of the terrestrial reconstruction. The Dark Ages change is nevertheless still visible in the reduced presence of herbs and crops, indicating less human influence and suggesting the spread of woodlands onto former arable land. In contrast to the other sites from the river area, the grasslands here were not replaced by alder carr during the DA regrowth.

Regional differences in the degree of woodland regeneration

The pollen percentages of the various vegetation groups and growth percentages are shown as pollen diagrams in Fig. 5 (terrestrial, 38 sites), Fig. 6 (total, 26 sites) and Fig. 7 (REVEALS data, 14 sites), showing the vegetation composition during the Iron Age/Roman period (Figs. 5a, 6a and 7a) and during the Dark Ages woodland regeneration phase (Figs. 5b, 6b and 7b). The pollen percentage values and growth percentages are also shown in Tables 2, 3, 4 and the AP percentage increases are given in Figs. 5b, 6b and 7b. Slight variations in the AP/NAP ratio may not indicate significant changes in the vegetation cover, since fluctuations in pollen percentages occur even when the vegetation cover remains stable (Favre et al. 2008). Therefore, a percentage increase below a threshold value of 5% is not considered to represent a change to the woodland cover. When calculated using the terrestrial pollen sum, nine out of 38 sites, all located in either the peat or sand area, have a growth of less than 5% and only one, Balooerveld (1), is also below 5% when calculated from the total pollen sum. Of the 14 sites in the REVEALSinR model run, only one shows a growth less than 5%, Poppendamme (28). This is likely to be an effect of the absence of the majority of herbs from dry localities (CUL) in the REVEALSinR run, as they could not be included in the model since pollen productivity estimate (PPE) values were not available. Using the total pollen sum, the CUL group is present at 22%, whereas in REVEALS it is only 1%. Overall, we conclude that the majority of sites in the Netherlands show woodland regeneration in this period.

Fig. 5
figure 5

Pie diagrams showing the regional vegetation composition at the various sites based on the terrestrial pollen sum; a during the IA/ROM, Iron Age and Roman period; b the DA phase of woodland regeneration, including percentage increases. The terrestrial pollen sum includes, in dark green APterra, trees and shrubs excluding Alnus and in yellow CUL, herbs and crops. Other details as in Fig. 2

Fig. 6
figure 6

Pie diagrams showing the regional vegetation composition based on the total pollen sum; a during the IA/ROM, Iron Age and Roman period; b DA woodland regeneration period, including percentage increases. The total pollen sum includes, in dark green, APtotal (all trees); yellow, CUL, herbs and crops; purple, HEA Ericaceae; light green, GRA, Poaceae. Other details as in Fig. 2

Fig. 7
figure 7

Pie chart overview showing the regional vegetation composition based on the total pollen sum and modelled with REVEALSinR, including percentage increases; a during the IA/ROM period; b DA regeneration period. In dark green, all trees, APtotal; yellow, CUL, herbs and crops; purple, HEA, Ericaceae; light green, GRA, Poaceae. Other details as in Fig. 2

Table 2 Pollen and woodland regeneration percentages calculated from the terrestrial pollen sum
Table 3 Pollen and woodland regeneration percentages calculated from the total pollen sum
Table 4 Pollen and regeneration percentages calculated from the total pollen sum with data modelled in REVEALSinR

The DA reforestation is clearly marked by increasing AP values across the entire Netherlands. Overall, the growth obtained using the terrestrial pollen sum varies from 1 to 82%, averaging at 19%. IA/ROM AP values vary between 20 and 97%, average 79%, Dark Ages ones between 66 and 100%, average 92%. The AP values for the IA/ROM period are much lower in the southern part of the Netherlands than in other regions and consequently, the increases are higher there. The various sand areas have average IA/ROM AP values of 86% in the north, 83% in the middle and 80% in the south, while the Dark Ages figures are 92% (N), 96% (M), 93% (S), with average growth increases of 8% (N), 16% (M) and 17% (S). The river area has an average IA/ROM AP of 64%, Dark Ages 90% and an average growth of 41%. All sites with a growth less than 10%, according to the terrestrial pollen sum, were located in the sand area, but all sites in the river area have a growth of more than 24%. All sites with low increases appear to have high AP values for the IA/ROM period, indicating that there was minimal human influence at these sites before the DA.

The pollen percentages obtained from the total pollen sum reveal a similar trend (Fig. 6a, b). AP values vary from 17 to 81% in the IA/ROM to 35% to 97% during the DA woodland regeneration phase and growth percentages vary from −1 to 131%. The difference between the total pollen AP percentages of the various landscape units is smaller when they are calculated with the terrestrial pollen sum. The sand areas have average IA/ROM AP values of 54% (N), 61% (M) and 64% (S), for the Dark Ages 59% (N), 81% (M) and 81% (S), and average growth of 9% (N), 37% (M) and 34% (S). The river area has average IA/ROM AP values of 58%, for the Dark Ages 94% and an average increase of 64%.

REVEALS modelling produced AP values varying from 8 to 67%, with an average of 35% and growth of 2% to 355%, average 102%. Overall, the model reduces the woodland component of the vegetation, so that the AP increase showing the DA regeneration then becomes more apparent. In addition, REVEALS produced higher percentages for the heath and grassland groups, whereas the percentages for the CUL group remained relatively stable.

From the pollen percentage data, it becomes apparent that the largest regrowth of woodland and thus extent of DA regeneration was in the river area along the Rhine and Maas (Meuse). This is reflected in the greater percentage increases there, which are all more than 49% (Ptotal). During this woodland regeneration phase, a decrease of grassland and consequently an increase in AP values is indicated at all investigated sites. Overall, it appears that the vegetation remained more open in the northern and middle sand areas during the regrowth phase, with areas of woodland, grassland and substantial heathlands. Heather was an important part of the vegetation particularly in the northern and central Netherlands. This is also evident in the data from Kootwijkerveen and Uddelermeer. The records from Berkenheuvel (4), Kraloo (7), Sleenerzand (5) and Emmerdennen (6) investigated by Castel (1991) also show high percentages of Ericaceae (heathers) then. Unfortunately, these records could not be included for the total pollen sum data since accurate pollen percentage data could not be collected from them. Interestingly, Ericaceae percentages show little change during the return of woodland to the southern Netherlands, and heathers do not seem to have been such an important part of the vegetation then. In contrast, grasslands seem to have been important in the southern sand area as well as the river area.

Discussion

From the overview presented in this paper it is clear that the regeneration of woodland in the Dark Ages is a characteristic phase in the vegetation history of the 1st millennium ad in the whole of the Netherlands. The dataset does not show extreme outliers and the trends within the separate landscape regions are consistent.

When we compare the modelled and unmodelled pollen percentage data, it appears that the values modelled by REVEALS generally have lower AP values; percentages of the modelled HEA and GRA groups were larger, but the CUL group had similar values. These differences are to be expected since trees and shrubs are usually over-represented in pollen percentage data (Andersen 1970; Broström et al. 2004). The CUL group in the REVEALS model includes only five taxa, although there are many more pollen types present in this group in the pollen percentage data; in contrast, nearly all the trees and shrubs in the pollen percentage data are also included in the REVEALS modelled data. Therefore the CUL group as modelled by REVEALS is not always a good representation of it in the actual vegetation. In general, the increased growth percentages showing the expansion of the regenerated woodland are higher when calculated in REVEALS. But overall, the modelled data mostly agree with those from pollen percentages. There were several regional differences in vegetation composition and development in the first millennium ad. In the records from the northern and middle sand areas, Ericaceae pollen percentages are higher than in the other areas during the IA/ROM period and in the river area DA woodland regeneration was most pronounced. The reconstructed growth percentages of AP serve as proxy evidence for the amount of woodland regeneration during the DA.

The varying degrees of this regrowth during the DA in different parts of the Netherlands might have various causes. The vegetation changes of the various regions is discussed below in the context of their landscape settings. In addition, the population density estimates are compared with the vegetation changes, and the influence of climate is also discussed.

Vegetation composition and DA woodland regeneration in the sand areas

Pollen percentages of Ericaceae, mainly Calluna (heather), which is the dominant plant in Dutch heathlands (Weeda et al. 1988) and thus the dominant pollen type in the late Holocene pollen data, are noticeably higher in the records for the first millennium ad from the northern and middle than from the southern sand area, suggesting that each had a different landscape structure. The true extent of heathland is probably underestimated in pollen percentages, as the values modelled by REVEALS show on average four times more Calluna. Considering its relatively low pollen production and dispersal, these higher percentages are to be expected (Bunting and Farrell 2022).

The high Calluna values can be explained by two options, either there were large areas of heathlands on dry, nutrient poor and acidic soils in the regional vegetation, or Calluna grew locally on the sampled bog or fen. Both regional and local source areas can be represented in a single record (Bunting and Farrell 2022). Calluna can grow on bogs and fens during drier stages, or on raised and therefore drier hummocks (van der Molen and Hoekstra 1988; Weeda et al. 1988). Records from the northern and middle sand areas are mainly from lakes and bogs, and those from the southern sand area come from bogs and stream valleys. Increased amounts of Ericaceae however are not restricted to bogs, so it is very likely that the high percentages in the northern and middle sand areas indicate large areas of heath there. In other studies reconstructing regional vegetation development, Ericaceae pollen is also mostly interpreted as originating from regional vegetation (Mazier et al. 2012; Fyfe et al. 2013; Lagerås et al. 2016). In this study, the unmodelled AP values from the sand area varied between 40 and 80% and Ericaceae between 2 and 50%. Characteristic unmodelled present-day AP values of sites within heathlands mostly vary between 10 and 60%, and Ericaceae pollen percentages are typically twice as high as AP percentages (Mulder and Janssen 1999; Doorenbosch 2013). The range of AP percentages in this study combined with the high Ericaceae percentages suggest the presence of large areas of heath in the immediate surroundings of most of the investigated sites in the northern and middle sand areas. REVEALS-based reconstructions of vegetation in northern and central Europe over the period 50 bc to ad 450 also demonstrate the dominance of Ericaceae on sites on sandy soils near the coast and in Denmark, the contribution of Ericaceae to the vegetation composition in the late Holocene is estimated at 40–60% (Nielsen et al. 2012), which is similar to the REVEALS estimates in this study.

It is remarkable that such large percentages of Ericaceae are not obtained from the southern Netherlands. A study by Doorenbosch (2013) showed that during the Bronze Age, heathlands already formed a substantial part of the vegetation in both the southern and central Netherlands. The greater extent of heathlands in the middle and northern sand areas found in this study might be related to the presence of glacial till in the northern sand area and ice-pushed older deposits in the middle sand area, resulting in more acidic soils in both, favouring heaths (Hoek 2000). Conversely, the cover sands in the south have a higher loam content providing more nutrient rich and wetter soils which are less prone to nutrient depletion. Calluna is usually outcompeted by other plants on carbonate-rich soils and therefore its occurrence on these soil types is limited (Weeda et al. 1988), which probably restricted the development of heathlands in those areas. This distribution pattern of Ericaceae was also found in the Netherlands during the late Glacial (Hoek 2000).

The occurrence of extensive heaths is often linked to strong human impact resulting in soil degradation as a result of agricultural activities (Weeda et al. 1988; Spek et al. 2003). When comparing pollen percentage values of the CUL group as an indicator of human impact between the northern and southern Netherlands, the data show that human impact was generally slightly lower in the north. The population estimates for the IA/ROM period for all sand areas are similar up to ad 200. Based on these indicators, it is unlikely that the more widespread heathlands in the northern part are the result of greater human impact there. Therefore, we may conclude that heaths developed on a larger scale in the northern and middle sand areas as a result of the sediment and soil characteristics there, while grasslands were more important in the south. In all sand areas during the DA regeneration, deciduous woods rich in oak and beech spread, replacing open grasslands and arable land, but the area of heathland remained rather unchanged. It is likely that the nutrient poor and acidic soils of the heathlands in the north were unfavourable for redevelopment of woodland. The amount of new woodland growth was similar in the southern and middle sand areas, with average increases of 34% (S) and 37% (M) Ptotal. This corresponds to reconstructed population changes in these areas, which show a drop from 4.9 to 0.5 people/km2 at the onset of the DA (Pierik et al. 2018). The lesser extent of woodland regeneration in the northern sand area, with an average growth of 9% Ptotal, corresponds well to the smaller decrease of population from 4.9 to 2.4 people/km2 in this period. This implies that the main cause of the DA regeneration of woodland in the sand areas was a reduction in human impact there.

Vegetation composition and DA woodland regeneration in the river area

Apart from the largest woodland regeneration in the river area during the DA for the Netherlands, there was also a change from grassland to alder carr on the floodplains. Pollen-based reconstructions of vegetation cover in north-central Europe by Nielsen et al. (2012) suggest that grasses contributed only minimally to the regional vegetation, with estimates of less than 10% cover in the period 50 bc to ad 450. Nevertheless, in the river area there seem to have been substantial amounts of grassland in the regional vegetation during the Roman period, which may be because of the swampy flood basins there. The high AP percentages during the regeneration phase suggest that a densely wooded landscape developed in the river area. Previous studies, however, indicate that in river areas, active floodplains were more open as a result of flooding and river changes and remained less wooded than drier areas even at the time of maximum woodland cover and without extensive human impact in the middle Holocene (Atlantic period) (Svenning 2002; Whitehouse and Smith 2010). It is therefore likely that such more open areas remained in the vegetation, and that the high AP percentages are partly from dense alder carrs growing at the sampling sites.

The change from grasslands to alder carr during the regrowth phase could be a result of the decrease in population density in the eastern part of the river area, from 25.5 people/km2 down to 5.4, and therefore with fewer activities associated with humans in these areas, such as keeping livestock. Population data from the western part of the river area are not available; however, all but one of the sites, Vinkeveen, in a peat area near a river (13), are located within the eastern river area where population data is available. The drop in human impact indicated by the population data is reflected in the pollen data by a similar drop in cultural indicators (CUL) with average (negative) growth of −71% (Pterra) and −78% (Ptotal).

The strong human influence on the vegetation in the river area introduces several additional influences to the representation of the vegetation in the pollen record. The large area of grasslands present during the IA/ROM probably indicates grazing activities. Heavy grazing, which occurred from the Neolithic onward, prevents grasses from flowering and thus from producing pollen (Groenman-van Waateringe 1993). Consequently, when intensive grazing is a factor, the lowered percentages of Poaceae in the pollen record could under-estimate the area of grasslands that was actually present and so influence the reconstruction (Groenman-van Waateringe 1993). Since the Poaceae pollen percentages are very high and together with the high values of cultural indicators and population estimates they suggest that it is unlikely that heavy grazing influenced the pollen signal substantially. There was probably some grazing, but not so much that it prevented or minimized grass pollen production. The pollen data are thus expected to reflect the actual area of grassland present. Coppicing of trees is also known to suppress their production of pollen and may have resulted in lower AP percentages for the IA/ROM period which in turn may thus have influenced the reconstructed woodland growth. Visser (2010) found evidence for woodland management including coppicing and selection for use in the Roman Empire. Alder, oak, birch, ash, willow, lime, elm and hazel are all suitable for coppicing. It is possible that the reduction of such practices at the end of the Roman period allowed pollen production by trees and shrubs to increase thus resulting in higher AP percentages in the record. One would then assume that the ending of woodland management and an associated increase in pollen production would have been largely local. However, the uniformity of the regrowth of woodland in the river area indicates a regional change, although a local increase in pollen production might have further raised AP percentages in some places.

Additionally, flooding and tree ring records have demonstrated that in the late Roman period, the Rhine flooded more (Toonen et al. 2013, 2017). The increased occurrence of floods, especially in contrast to the previously drier conditions in the Early and Middle Roman periods, may have promoted the growth of alder carrs due to both wetter conditions and increased nutrient input, as well as limiting the use of lower-lying areas such as meadows which would have been possible under the drier conditions. However, in areas further from direct river influence, such as the sites Berkenhof (21), Duckenburgh (20), Escharen (22) and Zeist (15), both a decrease in grassland and an increase in alder woods are indicated, refuting a direct link between increased flooding and expansion of alder carr. Therefore, while it is likely that the increased flooding frequency supported the DA expansion of woodlands in the river area, the change from grassland to alder woods was mainly the result of a decline in population.

The possible influence of climate change

The varying extent of woodland regeneration during the Dark Ages in the different parts of the Netherlands and its strong correlation with the changes in population and differences in types of landscape suggest that population and/or landscape factors were more likely causes for these changes than climate. Temperature changes during this period occurred across northwest Europe, but are not expected to differ substantially within the Netherlands, considering the amount of change there and in surrounding areas (Gouw-Bouman et al. 2019; Riechelmann and Gouw-Bouman 2019). With a reconstructed July temperature reduction of only 1.5 °C, and summer temperature staying within the optimum range of most of the vegetation, its effect on the vegetation is expected to have been minimal. However the influence of this temperature change, either as more frequent crop failures or lower pollen productivity, cannot be excluded. Clear evidence of the effects of the ad 536 dust veil (Gräslund and Price 2012; Helama et al. 2018) were not observed, but they would have occurred after the start of the DA woodland regeneration phase. Apart from temperature, the increases in flooding frequency and degree suggest a greater precipitation, which seems to have substantially affected woodland regeneration in the river area.

The Dark Ages return of woodland in northwest Europe

Studies making Holocene vegetation reconstructions based on pollen often use amalgamated pollen data in large (350–1,000 year) time segments (for example, Nielsen et al. 2012; Giesecke et al. 2014; Marquer et al. 2014; Githumbi et al. 2022a, b), so they do not detect shorter phases of vegetation change, such as the DA regeneration of woodland. Studies that are sufficiently detailed to reconstruct regional vegetation during the first millennium ad which show the DA regrowth of woodland are therefore scarce (Kaplan et al. 2009; Kleinen et al. 2011; Pędziszewska et al. 2020).

Pędziszewska et al. (2020) mapped vegetation change in Poland using a similar approach to this study and reconstructed a clear regrowth phase during the Migration Period. As with this study, changes in the degree of woodland regrowth were mostly attributed to differences in human presence.

Kleinen et al. (2011) modelled woodland cover in Eurasia during the last 8,000 years using climate as the driving force of vegetation development, and compared the modelled results to pollen data from Holzmaar in Germany. The AP values from Holzmaar and woodland cover modelled from the climate data diverge in the late Holocene as a result of the change from climate to human activities as the dominant factor determining woodland development as reflected in the pollen data. Remarkably, their results show a convergence of the pollen data and the climate-driven vegetation reconstruction during the DA woodland regeneration as a result of decreased human impact, which supports the idea that during the DA woodland regeneration the vegetation returned to a state resulting from less human interference. This was also reflected in the AP percentages from the Netherlands, which return to the levels that were recorded during the Bronze Age when human impact was much smaller.

Kaplan et al. (2009) reconstructed European woodland cover using a population-driven vegetation model, resulting in a continuous record for average woodland cover in central and western Europe as well as presenting regional estimates for the Netherlands for ad 1 and ad 500, among other places. Their model reconstructs regeneration of woodland in central and western Europe starting at ad 400 and reaching its maximum at ad 500. The modelled woodland cover of the Netherlands at ad 1 and ad 500 does not show regeneration, while it does for Belgium/Luxembourg, France and England. This is the result of an assumed population increase in the model for the Netherlands in this time period. Nevertheless, the model is able to detect this short phase of vegetation development and it could also be expected to reconstruct regrowth of woodland in the Netherlands with improved population estimates for this region. Even though our study covers a small area and short time period, it shows its value as an independent test for data driven vegetation models.

Conclusions

The vegetation has been reconstructed for the regeneration of woodland in the Dark Ages at a detailed regional scale. A set of 38 pollen records from different landscapes demonstrate that this regrowth took place in the whole of the Netherlands. There were regional differences in its extent, which are consistent in the studied records, with the strongest return of woodland in the river area and the least woodland recovery in the sand areas. The varying degree of woodland return in different parts of the Netherlands suggest that climate was not its only controlling factor. Both the differences in extent and vegetation composition during this period could have been caused by differences in landscape type and amount of human impact. In addition, it is demonstrated that the vegetation history before the Dark Ages and environmental changes such as flooding were also important in determining what vegetation grew in various places.

The Dark Ages return to woodland was strongest in the river area, which had previously lost more of its woodland during the IA/ROM period, as can be expected from the higher population density then. The significant decline in population density, a direct effect of the abandonment of the Roman Limes, resulted in a strong regrowth of woodlands during the Dark Ages. In addition, the change from surplus production to supply the Roman army to a self-sustaining subsistence economy reduced the pressure on the landscape. It is possible that an ending of woodland management such as coppicing could also have increased local evidence of woodland expansion by increasing pollen production. The spread of woodland was also intensified by a change in flooding regime and consequent development of wetland woods mainly of alder which replaced grasslands in the flood basins. The presence of these dense alder carrs at most sampling sites probably increased their representation.

The higher Pleistocene sand areas were not as intensively cleared of woodlands during the IA/ROM as elsewhere, which is in agreement with the estimated lower population there at this time, and followed by a smaller decrease in population during the DA. Consequently the DA regeneration was less intense in these areas. Mixed deciduous woods with Quercus and Fagus spread over former grasslands and arable land. In the northern and middle sand areas, heather was an important element of the vegetation during the first millennium ad owing to the presence of often dry, nutrient-poor and acidic soils. These areas remained open during the regrowth phase, most probably as a result of continued human impact and the presence of depleted soils which would have hindered woodland development.

Such a high resolution study using many pollen records from different environmental settings improves the understanding of the drivers which affected late Holocene vegetation events like the DA regrowth of woodland. Its occurrence and degree is similar to results from other parts of northwestern Europe. This study once again shows the importance of the incorporation of landscape characteristics and human pressure into studies of vegetation change.