Veget Hist Archaeobot DOI 10.1007/s00334-016-0591-x

ORIGINAL ARTICLE

Late Holocene landscape development around a Roman Iron Age mass grave, Alken Enge, Denmark Niels Emil Søe1 • Bent Vad Odgaard1 • Anne Birgitte Nielsen2 • Jesper Olsen3 Søren M. Kristiansen1

•

Received: 2 May 2016 / Accepted: 5 October 2016 Ó Springer-Verlag Berlin Heidelberg 2016

Abstract Sediments from the small lake Ilsø situated in the Illerup/Alken Enge Valley were studied in order to investigate past landscape development at the time of a probably ritual human mass burial following battle during the Roman Iron Age (AD 1–400). A pollen record from Ilsø and a number of other records from Jutland were combined using the Landscape Reconstruction Algorithm to reconstruct local vegetation changes through the last 2,800 years. These methods were supplemented by studies of catchment-related geochemistry of the Ilsø lake sediments. The results show a marked reforestation event associated with a strong decrease in erosion levels at the very beginning of the first century AD, contemporaneous with the finds of human remains at Alken Enge. Comparison with a pollen record 10 km away and with those from other sites, reveals that this reforestation occurs unusually early and rapidly, and is an unparalleled development in a Danish context. We suggest that the major landscape changes at the beginning of the Roman Iron Age and forest cover for the next few centuries comprise a possible example of ritual control of local land-use.

Communicated by M.-J. Gaillard. & Bent Vad Odgaard [email protected] 1

Department of Geoscience, Aarhus University, HøeghGuldbergs Gade 2, 8000 Aarhus C, Denmark

2

Department of Physical Geography and Ecosystem Science, Lund University, So¨lvegatan 12, 223 62 Lund, Sweden

3

Department of Physics and Astronomy, Aarhus University, Ny Munkegade 120, 8000 Aarhus C, Denmark

Keywords Iron Age  Landscape Reconstruction Algorithm  Palaeolimnology  Erosion rates

Introduction The valley of Illerup in eastern Jutland (56°030 1900 N, 9°550 3000 E) is well known for its rich Roman Iron Age weapon finds associated with warfare (Ilkjær 1991). These multiple finds of artefacts have been interpreted as clearly ritual in character suggesting a special significance for the valley landscape. This significance has been further supported by the discovery of a large concentration of human bones in a wetland area further down the valley (Alken Enge), just before it widens into the large lake Mossø (Andersen 1959). Recent osteological investigations of newly excavated human bones representing an estimated 300 individuals in Alken Enge have shown that many skeletal parts have unhealed injuries from sharp weapons, indicating a mass burial of warriors killed in battle. Numerous radiocarbon dates indicate that the event took place in the early first century AD which seems to mark the start of ritual burial in the valley system. The present study investigates whether the apparent ritual status of the Illerup/Alken Enge valley during the first half of the first millennium AD was reflected in local land-use and land cover. We use a series of palaeoecological techniques to study the sediments of a small lake in the lower part of the valley system aimed at testing the null hypothesis that the area had a vegetation succession during the first millennium AD comparable to areas of similar landscape settings in Denmark (Odgaard and Rasmussen 2000; Nielsen and Odgaard 2010; Nielsen et al. 2012). The development of the Landscape Reconstruction Algorithm (LRA) has the potential to discriminate past

123

Veget Hist Archaeobot Table 1

14

C-dates and their calibration from sediments of Ilsø and Taastrup

The 14C ages are given in conventional radiocarbon years BP (before present = 1950) with a measurement uncertainty of one standard deviation, calibrated ages in calendar years (Reimer et al. 2013) and depth as cm below sediment surface

123

Veget Hist Archaeobot Fig. 1 Digital elevation map (Danish Vertical Reference 90). Lakes are shown by black and light blue fill

vegetation dynamics at local as opposed to regional scales (Sugita 2007a, b). This approach works in two steps: first by reconstructing the regional vegetation from multiple pollen records (or that from only one large site); and second by using this in combination with a pollen record from a small site to isolate the local vegetation signal (Sugita et al. 2010). The method simultaneously corrects for differences in pollen production and pollen dispersal between the dominant pollen types (Brostro¨m et al. 2008). In this study we use LRA as a tool to unravel local landscape dynamics close to the spectacular Iron Age post-battle deposition site of Alken Enge. Palaeolimnology is another approach to studying past landscape dynamics. In a number of studies, lake status and hence sedimentological records have proved to be strongly dependent on catchment processes (e.g., Meyers and Teranes 2001). This provides an independent line of evidence for past landscape change. Landscape transformations reconstructed through changes in lake status have in some studies been shown to be more robust than those based on catchment evidence alone (e.g., the identification of the position of the arctic tree line in Canada through lake water dissolved organic carbon concentration, Pienitz and Smol 1993). In this study we use the reconstruction of lake status as evidence of changes in land-use regimes independent of the conclusions about local vegetational change derived from the LRA approach.

The overall aim of this study is to reconstruct past local landscape variation through the last three millennia, with a special focus on the period of the Alken Enge ritual deposition of the earliest Roman Iron Age. This is part of a larger study of the palaeoenvironment and relationship to ritual deposition. We compare pollen records and LRA reconstructions from other sites in Denmark, in particular that of Taastrup Sø, 10 km distant, with Lake Ilsø to test the null hypothesis of no divergent local landscape development.

Materials and methods Study site characteristics The landscape of the eastern part of the Jutland peninsula, Denmark, is characterised by elevated glacial moraine plateaus intersected by E-W oriented tunnel valleys. The tunnel valley bottoms typically have numerous depressions which today are occupied by either wetlands or lakes. Ilsø is located in a narrow tunnel valley that is a part of a wider tunnel valley system, in which the Alken Enge and Illerup sites are found (Fig. 1). The current lake has a surface area of 0.6 ha and although the bathymetry has not been mapped in detail, a maximum depth of 2.5 m has been recorded. The lake has

123

Veget Hist Archaeobot

no surface inlet, has a small topographical catchment of ca. 32 ha and is thought to be primarily replenished by groundwater. This assumption is supported by a maximum groundwater level approximately 5 m above the presentday lake surface in nearby wells. The present outlet at Ilsø is a ditch dug in the late 19th century that maintains a relatively constant lake level at 31.3 m a.s.l. However, a digital elevation model (DEM) suggests elevated palaeoshorelines up to 34.5 m a.s.l., reflecting a past maximum lake surface area of ca. 7.0 ha. This level corresponds to the level of the natural pre-drainage outlet observed on the DEM. Lake Taastrup, located 10 km NE of Ilsø in a wide tunnel valley (Fig. 1), was included as a reference site. Today Lake Taastrup is eutrophic, has an area of 14 ha and a maximum depth of 2 m. The topographical catchment is 11 km2. As with Ilsø, a bathymetric map is not available for Lake Taastrup, but the lake bottom of this shallow lake is very even. Sampling A 7.2 m long sediment sequence from Ilsø (denoted ‘‘Ilsø core’’) was obtained at a water depth of 248 cm from the central part of the lake using a 1 m chamber and 80 mm diameter Russian corer, with approximately 20 cm overlap between sections of two parallel cores. Deeper sediments were finely laminated (probably annual varves), but unfortunately not in the portion studied here. The core sections were accurately correlated using X-ray fluorescence (XRF) datasets and images. A HON-Kajak corer (Renberg and Hansson 2008) was used to obtain the upper 30 cm of sediment. Following careful trimming and XRF scanning, the Ilsø core was separated into 171 consecutive 5 cm segments that were sampled for analysis. Coring in Lake Taastrup was done in the centre of the lake at a water depth of 124 cm with an Usinger 8 cm diameter piston corer (Mingram et al. 2007). Upper sediments were taken using a Russian corer and the top 30 cm using a HONKajak corer. Chemical analyses Cleaned surfaces of Ilsø sediment core sections were XRF scanned on an Itrax corescanner with a Mo tube (30 kV, 30 mA). The step size was 200 lm and exposure time 3 s. XRF data were stacked to obtain average counts of elements for each 5 cm core segment (i.e., the average of 250 measurements). Digital images were obtained using an integrated optical camera. The intensities of raw XRF data are strongly affected by changes in organic matter, water content and grain size, and it is recommended that interpretations are based on log-ratios of intensities, which are

123

linearly correlated to the log-ratios of concentrations (Weltje and Tjallingii 2008). The normalizing element is advantageously chosen to be a conservative element of the lithogenic fraction, so that changes in the lithogenic composition and the lake environmental processes can be assessed (Lowemark et al. 2011). However, the very low counts of Ti, K and Rb in parts of the Ilsø core prevented a uniform normalization. The analysis of the qualitative variations of an element is therefore based on the raw counts of the element and comparisons with the variation of total organic carbon (TOC) and the counts of the other elements. All core segments were analysed for total carbon (TC), total nitrogen (TNnon-treated) and total sulphur (TS) on an Elemtar Vario Max CNS at 1,100 °C. The sediment was dried at 50 °C for 24 h and crushed before analysis. Seventeen samples had replicate measurements (n = 2) and the maximum standard deviation was 2.9, 0.25 and 2.9 % for TC, TN and TS, respectively. Carbon mass accumulation rate (C MAR) was calculated based on the TC weight percentages, the mass of volume specific samples dried at 50 °C and the deposition rates of sediment core segments based on the age-depth model: C MAR (mgC cm-2 year-1) = TC (mgC g-1)  dry mass (g cm-3)  deposition rate (cm year-1) Isotopes A total of 89 bulk sediment samples distributed along the entire Ilsø core were washed with 1 M HCl at 60 °C to remove carbonate, rinsed with deionised water and freezedried prior to analysis of total organic carbon (TOC), total nitrogen (TNacid washed), and 13C/12C (d13C) and 15N/14N (d15N) isotopic ratios on a EuroVector elemental analyser coupled to an IsoPrime stable isotope ratio mass spectrometer at the AMS 14C Dating Centre, Aarhus University. C/N ratios were derived from the TOC and TNacid washed measurements as molar ratios. Cellulose was extracted from 52 samples of 5–10 cm3 between 540 and 720 cm depth in the Ilsø core using the CUAM procedure (Wissel et al. 2008). Preliminary tests showed that insufficient amounts of cellulose could be extracted above this depth interval. Analysis of d18O (denoted d18Ocell) and d13C (d13Ccell) on the cellulose was successful on 50 and 46 samples, respectively. Replicate analyses (n = 2–7) were performed on samples with sufficient cellulose (29 for O and 32 for C), and showed maximum standard deviations of 0.34 % for d18Ocell and 0.18 % for d13Ccell. Assuming the origin of cellulose to be purely aquatic, the d18Ocell values were transformed into lake water O isotope values (d18Olw) using an isotopic fractionation constant of 1.030, as cellulose extracted using the CUAM protocol has been reported with fractionation

Veget Hist Archaeobot

constants in the range 1.0299–1.0302 (Heyng et al. 2014; Mayr et al. 2013). However, input of terrestrial organic matter may obscure the d18Olw signal, which must therefore be interpreted in relation to other proxies. As reference for the d18Olw record in the sediment, 13 modern lake water samples were collected during winter, spring and summer 2013–2014, and analysed for d18O (denoted d18Omodern). All isotope results are expressed as d-values representing deviations per ml for d13C values from the VPDB, d15N from the AIR, and d18O from the VSMOW standard. Pollen A total of 32 samples distributed along the entire Ilsø core were analysed for pollen. Mean sample interval was 88 years according to the age/depth model (see below). For the Taastrup core 42 samples were analysed covering the period 4300 BC–AD 1993 (22 samples younger than 800 BC). Volumetric samples for pollen analysis were secured from the center of each core segment using a plastic syringe with a cut-off tip. Samples were freeze dried and 0.05 g taken for preparation. Tablets with Lycopodium spores (n = 10,679) were added to allow calculation of pollen concentrations and pollen accumulation rates (PAR). Chemical preparation included initial 10 % HCl, 10 % KOH, decanting, acetolysis and transfer to silicone oil via tertiary butanol. Pollen and other microfossils were counted at a magnification of 4009, using a magnification of 1,0009 for difficult identifications. Minimum sums of terrestrial pollen were 505 for Ilsø and 955 for Taastrup. Pollen percentages were calculated relative to a sum of terrestrial pollen excluding pollen and spore types from lacustrine and wet ground taxa. Cyperaceae were considered as deriving mostly from wet ground taxa and were accordingly excluded. Humulus/Cannabis type was left out from the terrestrial pollen sum because both lakes seem to have been used for retting of Cannabis, so the majority of grains from this pollen type may derive from this process and hence bias the record of the mainly airborne types if included in the pollen sum. The identification of cereal pollen types followed Andersen (1979). Land cover of the main vegetation constituents was estimated based on terrestrial pollen data using the LRA (Sugita 2007a, b). The LRA consists of two steps, the first of which is estimating regional vegetation cover for the region surrounding the site, using the REVEALS model (Sugita 2007a) and pollen data from multiple small or large sites in the region. Here, REVEALS was applied to nine lakes in Jutland (Odgaard 1994, 1999a, b, 2000; Andersen 1995). For the LRA modelling of Taastrup, Ilsø was included in the regional estimate, and vice versa for the modelling of Ilsø. To allow for the different sampling

Fig. 2 Ilsø age-depth model and core-image scan with congruent depth-scale. Pollen assemblage zones are separated by red lines

Fig. 3 Taastrup age-depth model. Red outlier date. Green dates not included in the model, but used for validation

123

Veget Hist Archaeobot

resolution at different sites, the regional vegetation was estimated for 17 time slices: AD 2010–1900, 1900–1750, 1750–1550, 1550–1350, 1350–1150 and 1150–950, and then in 500 year time slices back to 1500 BC. For each of the regional sites, pollen counts within each time slice were added together, to obtain one mean REVEALS cover estimate per time window. REVEALS modelling was carried out using the software REVEALS v. 4.5 (Sugita unpublished). The REVEALS model provides estimates of the regional vegetation composition, with its error estimates, for each time slice, based on the pollen data from the nine regional sites. Once these are available, the local vegetation within the relevant source area of pollen (RSAP) around Taastrup and Ilsø could be estimated in the second step of the LRA, using the LOVE model (Sugita 2007b). This is implemented using the software LOVE v. 4.6.1 (Sugita unpublished). For each pollen sample from Taastrup and Ilsø, the regional vegetation estimate of the time slice to which the sample belonged was applied. The LOVE model requires an initial estimate of RSAP radius, which was here set to 300 m. This was re-evaluated at 20 m intervals, using a threshold of ±1 SE to evaluate whether all vegetation estimates were between 0.0 and 1.0 (Sugita 2007b). For both REVEALS and LOVE, the standard pollen productivity estimates and pollen fall speeds used in the Landclim project was applied (Mazier et al. 2012, standard 2). These are the means of available pollen productivity estimates from Northwest Europe, after removal of outlying values (Mazier et al. 2012). For Artemisia, fall speed and pollen productivity estimated in the Czech Republic (Abraham and Koza´kova´ 2012) were used. Pollen dispersal was assumed to follow the Sugita (1993) model for lakes. Wind speed was set to 3 m/s and default atmospheric parameter settings, representing neutral atmospheric conditions, were used in all model runs. The maximal extent of the regional vegetation was set to 100 km. Chronology Terrestrial plant macrofossils for radiocarbon dating were secured by screening residues from wet sieving (mesh size

500 lm) of about 100 ml fresh sediment using a stereo microscope. Identified terrestrial macrofossils were isolated, dried and weighed before being radiocarbon dated by Accelerator Mass Spectrometry at the AMS 14C Dating Centre, Aarhus University, Denmark. Based on radiocarbon dates and for Taastrup also on correlated levels (see results), age depth modelling was performed using OxCal version 4.2.4 (Bronk Ramsey 2013) and a P sequence with 0.3 events per cm. At Ilsø the onset of the increase in pollen of the introduced spruce (Picea) was set to AD 1900 ± 20 based on comparisons with published maps (cf. Rasmussen 2005) and included as an additional date in the model. Furthermore, a boundary was inserted at the depth where water content of the sediment started an upward increase to handle the effect of compaction. At Taastrup the first spheroidal carbonaceous particle (SCP, derived from high-temperature burning of coal/oil, Odgaard 1993) was likewise used to set a date of AD 1900 (Odgaard 1993). The age assigned to a core segment in the datasets is the average modelled age at the centre of the segment. Numerical analysis Zonation of the pollen records was performed on terrestrial pollen percentages using optimal splitting by information content as implemented in PSIMPOLL version 4.27 (Bennett 2015). The selection of number of zones was guided by a broken stick model (Bennett 1996) also implemented in PSIMPOLL. Since the focus of the present study was on catchment processes, the zonation based on pollen data was transferred to the XRF and isotope data as well. Ordinations were performed to summarize the multivariate data and to investigate the relations between data sets using Canoco 5.0 (Ter Braak and Sˇmilauer 2012). Linear methods (PCA, RDA) were used, since either linear responses could be expected, or since gradients (for vegetation) were shorter than 2 standard deviations. Centered log ratio transformation was used for percentage data to circumvent the closure problem of compositional data (Pawlowsky-Glahn and Egozcue 2006). Significance tests were performed using 999 Monte Carlo permutations.

Table 2 Sediment description of the Ilsø core (depth below lake level) with corresponding age according to age-depth model Depth (cm)

Corresponding age

335–250

AD

1950–2000

Very dark olive-green gyttja

375–335

AD

1850–1950

Slightly silty and clayey gyttja; upper boundary gradual

570–375

AD

950–1850

1,000–570

800

Alternating light and dark olive-grey laminated gyttja; weakly calcareous; scattered layers of dark grey clay; thick clay layer at 480 cm Very dark reddish brown gyttja; weakly calcareous below 700 cm, few gastropod shells

123

BC–AD

950

Description

Veget Hist Archaeobot Table 3 The variation of sediment composition and isotopes in pollen zones of the Ilsø core Zone

Age

4

AD

1650–2000 3

AD

1000–1650 2

AD

50–1000

1

800 BC–AD 50

Characteristic High and decreasing K, Ti, Rb. High and increasing Fe/Mn log-ratio. TOC, C MAR and C/N increase. Decreasing d15NOM and d13COM High and increasing K, Ti, Rb with peaks during AD 1000–AD 1350. Variation in Rb deviates from Ti and K. Peak in Ca and Sr at AD 1000. High Fe/Mn log-ratio. High d15NOM. Low TOC and decreasing C MAR. Low and fluctuating C/N and d13COM Low K, Ti, Rb peaks at 850 BC. High TOC, C MAR and C/N. Decreasing and low d15NOM. Peak in Fe and Mn at AD 200 and AD 300. Decreasing Fe/Mn log-ratio. Elevated counts of Ca and Sr at AD 200–400; High d18Olw and a short period of low values at AD 350 Moderate K, Ti, Rb and Sr counts that peaks at 700 BC and elevated from 150–50 BC. High TOC and C MAR. Rapid decrease at 750 BC in C/N, d13COM, Fe and Mn. d15NOM increase rapidly at 750 BC and decreases until 400 BC whereupon it stays low. Peaks in Mn after 700 BC and in Mn and Fe at 50 BC. Increasing Fe/Mn log-ratio. Decreasing Ca. d18Olw increases after 100 BC

Fig. 4 Temporal changes in X-ray fluorescence (XRF) raw counts of significant elements of the Ilsø sediments and Fe/Mn as a log ratio. Red horizontal lines separate the pollen zones

Results Age-depth models The age-depth model for the Ilsø sediments shows an approximately linear deposition rate until the 20th century (Table 1; Fig. 2). The 5 radiocarbon dates control the model between 800 BC and AD 1500, whereas the younger part of the model is based on the increase of Picea and the recent lake bottom. The age-depth model from Taastrup spans from 4000 BC to present (Fig. 3). The deepest radiocarbon date was excluded from the model as an outlier since it was younger than the two radiocarbon dates above, a decision that was corroborated by the correlated level of the decline in Ulmus. The youngest radiocarbon date is c.

1400 cal BC and above this date the model is based on correlated levels of increases in pollen percentages of Poaceae and Plantago lanceolata together with a decrease in Alnus, the first SCP and the lake bottom. Even though the upper part of the core has no radiocarbon dates, it still shows a similar linear deposition rate to the lower part and fits with the first Secale pollen grain, which in Denmark is a useful time marker for 250–50 BC (Odgaard 1994). Sediment composition The sediment description of the Ilsø core is given in Table 2 and is characterised by a visually obvious change at 570 cm corresponding to the transition between zone 2 and zone 3 (Fig. 2).

123

Veget Hist Archaeobot Fig. 5 Temporal changes in the organic fraction and stable isotopes of the Ilsø sediments. Dashed line shows variation in TOC. Red horizontal lines separate the pollen zones

Table 4 Results of RDA on Ilsø data using LRA-estimated vegetation data as response variables and sedimentary geochemical data as explanatory variables Variable

% variance explained

P level

C/N ratio

25.4

0.001

K

21.1

0.001

Ti

20.9

0.001

Rb

20.8

0.001

TOC

20.7

0.001

TC

19.1

0.01

TON

17.8

0.001

Mn d15N

12.3 12.0

0.01 0.01

9.4

0.05

Sr

Only significant results are shown

The elements in the XRF-dataset used in this paper include Ti, K, Rb, Ca, Sr, Fe and Mn. These elements have been shown to be significant proxies for detrital input, inlake productivity and redox conditions (Kylander et al. 2011). The variation of each element is summarised in Table 3 and shown in Fig. 4. The dry weight percentage of TOC and TC are almost identical in zone 3 and 4, whereas TOC is slightly higher in zones 1 and 2 indicating the presence of carbonates (Fig. 5). TNacid washed is likewise generally higher than TNnon-treated in zones 1 and 2 due to the removal of

123

Fig. 6 RDA triplot of Ilsø results with LRA-based vegetation estimates (blue arrows), significant geochemical sedimentary constituents (red arrows) and samples labelled by their estimated ages (?), negative ages are BC, positive AD. Key to taxa abbreviations: PlanLanc = Plantago lanceolata, PlanMajo = Plantago major-type, RumxAcet = Rumex acetosella-type

carbonates. A strong linear correlation between TOC and TNacid washed (q = 0.98) intercepting at 0.1 wt% of TN indicates absence of significant concentrations of inorganic

Veget Hist Archaeobot

Fig. 7 Pollen percentage (lines) and pollen accumulation rate (PAR, bars) diagram of trees and shrubs from Ilsø. PAR values have been scaled down by 500 for legibility

N. In the following we denote TNacid washed as total organic nitrogen (TON). Cellulose inferred d18Olw ranged from -4.6 to -3.5 % with an average value of -4.2 %, and was significantly higher (P \ 0.001) than the d18Omodern of the modern lake water samples, which ranged from -8.7 to -6.7 % with an average of -7.5 %. The variation of TC, C MAR, C/N, d13COM, d15NOM and d18Olw is summarised in Table 3 and Fig. 5. An redundancy analysis (RDA) of geochemical data using LRA-estimated vegetation as response variables showed significant correlations to geochemical sedimentary characteristics (Table 4). Although it is the response in geochemical sedimentary characteristics that can possibly be explained by LRA-estimated vegetation and not the reverse, we still chose vegetation data as the response variable and geochemical sedimentary characteristics as explanatory variables in the RDA because (1) the focus is on catchment vegetation, (2) LRA-estimated vegetation is compositional data whereas the geochemical sedimentary characteristics are non-compositional data, (3) the RDA still reveals the correlations between geochemical sedimentary characteristics and LRA-estimated vegetation, and (4) this reverse analysis allows the individual elements to be tested against the LRA vegetation estimates. The analysis showed that some geochemical characteristics were positively correlated to periods of dominantly open landscapes (Fig. 6: K, Ti, Rb, Sr, d15NOM). These can be considered either proxies of catchment erosion or of higher in-lake productivity with associated lake bottom anoxia caused by increased nutrient inputs deriving from such erosion. Others (Fig. 6: Mn,

TON, TOC, TC, C/N) characterize the periods of more stable, forested catchments. The obvious change of sedimentation in the lake at AD 950 (Table 2; Fig. 2) encouraged investigating correlations that only exist in parts of the core. Therefore, a series of RDAs were performed on running subsets of 20 samples throughout the core. In all subsets C/N, K and Ti showed significant correlations to LRA vegetation. The 10 elements in Table 4 all showed significant correlations to LRA vegetation in the upper subsets of the core (AD 91–AD 2006), whereas TS, C/N, K and Ti were the only geochemical sedimentary characteristics that showed significant correlation (p \ 0.05) in the lowest subset (820 BC–AD 1049). Furthermore, TS showed increasingly significant correlation towards the lower subsets and positive correlation to periods of predominantly closed landscape (see variation of TS wt% in Fig. 5).

Vegetation estimates based on pollen Pollen percentages and pollen accumulation rates Pollen percentages and pollen accumulation rates (PAR) broadly reflect identical stratigraphies with dominance of herb and grass pollen in zone IL-1, of tree pollen in IL-2, and again of herb and grass pollen in zones 3 and 4 (Figs. 7, 8). A reduction in tree pollen in the beginning of IL-1 is more strongly reflected by PAR values than by percentages. This tree pollen reduction affects Alnus, Corylus, Fraxinus, Ulmus and Tilia. In contrast, an increase is seen in grasses and a series of herbs such as Triticum-type, Artemisia,

123

Veget Hist Archaeobot b Fig. 8 Pollen percentage (lines) and pollen accumulation rate (PAR,

bars) diagram of selected non-arboreal terrestrial pollen types from Ilsø. PAR values have been scaled down by 500 for legibility

Plantago lanceolata, Filipendula, Chenopodiaceae and Rumex acetosella-type. By the transition to zone IL-2 the light-demanding elements decrease, while Quercus and Corylus in particular increase again. While AP-percentages are marginally higher in the second half of the zone (*88 %) than in the first half (*80 %), AP-PAR values are twice as high in the later half compared to the earlier. Fraxinus, Alnus, Corylus, Quercus, Betula, Salix and Fagus have peak PARs here. Poaceae, Rumex acetosella-type, Calluna and most herb types show roughly constant PAR values throughout IL 2, but percentage-wise they drop through the period. At the start of IL-3, terrestrial PAR drops to about half of the values in the late part of IL-2, especially for tree pollen types such as Quercus, Corylus, Betula, Alnus, Fraxinus and Fagus. This drop is also seen in percentage values of these taxa, although it is less pronounced. In contrast, the sum of herb and dwarf shrubs PAR increases mainly caused by an increase in Poaceae. Percentage-wise grasses increase as well and the general drop in PAR causes other pollen types to increase also, although their PARs hardly change. Examples are Hordeum-type, Triticum-type and Polypodiaceae. Humulus/Cannabis-type and Centaurea cyanus are newcomers in this zone. In the middle of zone IL-3, PAR-increases in Fagus, Salix, Betula and Alnus cause a temporary rise in tree PAR and tree pollen percentages. Zone IL-4 is characterized by high percentages and PARs of grasses (Poaceae), Rumex acetosella-type, increasing Pinus and Picea and decreasing Fagus, in percentages as well as PARs. Regional land cover reconstruction The REVEALS analysis showed that changes in vegetation cover at the regional scale were quite slow and gradual over the 3,000-year period (not shown). Regional tree cover declines from over 35 % at 1000–1500 BC to less than 20 % in the last 200 years, but throughout the period the regional vegetation seems to have been quite open, with a large proportion of Calluna heathland. This heathland signal derives mostly from the western lakes in the dataset situated in catchments with poor sandy soils. Cerealia-type undiff. (Hordeum and Avena/Triticum types) are noticeable in the reconstruction from 2500 BC onwards, with Secale cover expanding after AD 1. The combined cereal cover estimate for the last 100 years is just 6 %, indicating that

123

Veget Hist Archaeobot

Fig. 9 LOVE estimates of local vegetation cover of major plants at Ilsø (solid blue) and Taastrup (dashed green) during the period 800 BC–AD 2000

cereals may be somewhat underrepresented in the REVEALS estimates. Within the tree cover, taxa like Quercus, Betula and Corylus decline through the period. Fagus reaches its highest regional cover at AD 1400–1800, while Picea and Pinus expand in the last time slice reflecting modern coniferous plantation. The two REVEALS reconstructions (not shown) are very similar, indicating that estimates are robust to the inclusion or exclusion of single sites. Local land cover reconstruction: Ilsø The zonation of the LOVE vegetation record of Ilsø resulted in four distinct episodes for the interval 800 BC until the 20th century AD (Fig. 9). The LOVE analysis indicates that while at the regional scale, vegetation development was relatively slow, there have been rapid and large variations in the vegetation composition at a local scale around the lakes. For Ilsø, the LOVE reconstruction (Fig. 9) shows larger and more abrupt variations in vegetation composition than the uncorrected pollen diagram (Figs. 7, 8), with strong contrasts between open and forest covered periods. The estimated RSAP for this site ranges from 300 to 450 m. From the onset of the record (800 BC), tree cover was declining, and there is an open period from ca. 500 BC to AD 50, with tree cover in the later part of the period below 25 %, and the landscape dominated by Poaceae together with high amounts of Plantago lanceolata, Rumex acetosella-type and other herbs. There was a rather high cover (around 5 %) of Cerealia undiff. Secale arrived and had low cover from around 100 BC.

From AD 50 to 1000 (zone IL-2) tree cover was high around Ilsø. Quercus, Corylus and Betula in particular expanded. Fagus was present, but its cover increased slowly until ca. AD 800, after which it increased quickly to cover more than 40 % of the local landscape, replacing Quercus, Ulmus, Salix and partly Betula and Alnus. Shortly after AD 1000, landscape openness increased quickly to more than 60 %. The cover of Fagus fell by a third to around 25 %, while most other trees almost disappeared from the local area. Tree cover increased again after AD 1200. Fagus increased most, but Betula and Alnus, and later Quercus and Fraxinus also expanded. Landscape openness was reduced, but this mostly affected the cover of grasses. Cereal (Cerealia undiff.) cover even seemed to expand during this period, together with the cover of Rumex acetosella-type. Grass cover again increased after AD 1600, and remained high for the rest of the record. Cereal cover continued to be high. Fagus dominated the remaining tree cover, which was reduced to ca. 20 %. However, it declined in the topmost sample. Local land cover reconstruction: Taastrup The estimated RSAP for Taastrup ranges from a radius of 300 to 620 m. It does not seem to be correlated with landscape openness, but the highest values are obtained at transitions between forested and open landscapes (or vice versa). Tree cover in the early part of the record (800 BC) was much higher than in the region in general, with values [80 % (Fig. 9). This period was apparently locally

123

Veget Hist Archaeobot

dominated by Corylus with high amounts of Alnus and Fraxinus. A period of openness occurred after 500 BC when Poaceae dominated the land cover, while Plantago lanceolata expanded, and a significant cover of Rumex acetosella-type is reconstructed locally for the first time. After 200 BC, the cover of Cerealia-type undiff. expanded slightly and Secale appeared for the first time. All woody species were reduced in cover in this period except for Betula, Salix, Pinus (not shown) and Fagus, the latter with a low, but increasing presence. Tree cover, initially Quercus and Corylus, again increased after AD 200. Fagus increased slowly at first, but after AD 700 its cover rose quickly. The maximum of ca. 40 % Fagus was reached around AD 1200. After this, both Fagus and other trees declined, as a new period of open, grass-dominated landscape began, lasting until the present. The composition was similar to the previous open period, except that cover of cereals was higher and tree cover was even lower than before. Furthermore, much of the remaining tree cover was composed of Fagus, with a little Juniperus (not shown), Ulmus and Salix. The cover of Pinus and later Picea increased towards the present (not shown). To summarize, at both sites the local vegetation cover showed distinct variations between open periods, dominated by grasses (with varying degrees of cereal cover), and tree dominated periods. However the timing of these changes varied between the two sites, except for when both sites shared a large degree of openness in the early part of the Iron Age, ca. 500–1 BC. Furthermore, the reforestation during the period AD 1–1000 began much earlier and progressed more rapidly at Ilsø. Finally, at Ilsø the woodland cover was mostly higher during this period than at Taastrup.

Discussion Catchment/lake interactions The strong correlation of C/N ratio with forest cover suggests that this ratio reflects the relative contributions of N-rich aquatic algae and N-poor terrestrial plant material to the organic fraction of lacustrine sediments (Meyers and Teranes 2001). This conclusion is to be expected because a forest-bordered lake is likely to receive more terrestrial material through leaf fall than a lake bordered by open vegetation types. The generally low d13COM of the entire core (mean = -31.7 %), however, could indicate that the sedimentary organic matter is of mainly aquatic origin throughout (Meyers and Teranes 2001). Measured d13COM is the result of complex interactions, one of which is the ratio of carbon assimilated as HCO3- to carbon assimilated

123

as CO2. It is likely that this ratio was biased more to CO2 uptake during zone IL-2 than during zone IL-1 because of an expected drop in pH induced by lower primary production in the lake in zone IL-2. Such a process would then counterbalance the expected increase in d13COM values derived from higher input of terrestrial organic matter. During zone IL-1 a locally high linear positive correlation between C/N and d13COM (q = 0.59) supports the conclusion that lake productivity was controlled mainly by increased input of nutrients during open landscape phases. Furthermore d15NOM values suggest increased nitrogen limitation in the lake as a likely consequence of the increased productivity. Nitrogen limitation is suggested after 400 BC by low d15N values which may have been caused by dominance of nitrogen-fixing bacteria assimilating atmospheric nitrogen low in d15N (Talbot 2001). We interpret the high C/N ratios in zone IL-2 as a reflection of a relatively high terrestrial fraction of the bulk organic matter. This is in accordance with the pattern of TS variation. We attribute the stronger response in TS to estimated vegetation in the lower part of the core to terrestrial material here being the dominant contributor of sulphur to the sediment. TS levels are higher in zone IL-2 than in zone IL-1 supporting the C/N evidence of higher relative input of terrestrial material during zone IL-2. d18O of cellulose from lake sediment has been shown to record d18O of lake water through the process of algae assimilating oxygen from water (Wolfe and Edwards 1997; Wolfe et al. 2000; Heyng et al. 2014). Even cellulose of lake sediments with elevated C/N has shown correspondence to lake water (Wolfe et al. 2001). However, input of terrestrial organic matter is expected to obscure the signal, especially since vascular plant material is more celluloserich than that of aquatic algae. Relative to the modern water of Ilsø (d18Omodern), cellulose-inferred lake water (d18Olw) was significantly 18O-enriched, which may indicate that part of the sedimentary cellulose originated from terrestrial plants enriched in 18O due to evaporation-transpiration (Barbour et al. 2004). Alternatively, this apparent enrichment may be a result of bias introduced by comparison with a disturbed modern hydrology. We attribute the rise in d18Olw at the transition between zone 1 and 2 to an increased cellulose fraction of terrestrial origin, which is in accordance with the interpretation of both C/N and TS. Thus no remarkable hydrological shifts could be concluded. We evaluate detrital sediment yield to the lake and thereby the erosion levels in the catchment from the counts of K and Ti. The strong overall and local (running 20 sample) correlations between K and Ti indicate that the two elements are only recording detrital input and the RDA results showed a clear positive relationship of high erosion to periods of open landscapes (Table 4; Fig. 6). Again, this

Veget Hist Archaeobot

relationship seems persistent for the entire period, as reflected by significant correlations in all subset RDA analyses. Rubidium and strontium are also part of the detrital input and showed significant correlation in the RDA, but these elements are also affected by other processes. Sr is part of the carbonate geochemistry and Rb is influenced by chemical weathering by being selectively adsorbed on weathering products, i.e. clay minerals, and preferentially released to the soil solution during microcline weathering relative to potassium (Nesbitt et al. 1980). The peaks of K and Ti at 700 BC probably represent a pulse of erosion that is expected at upland landscape transitions when sediments are remobilized following destabilization connected to deforestation (Church and Slaymaker 1989). Hereafter the counts of K and Ti indicate a moderate erosion level, which again increases for a period in the last part of zone 1, when the low tree cover suggests maximum pressure on the landscape. By contrast, K and Ti counts are below the detection limit in zone 2 and indicate very low erosion levels when woodlands were expanding. Here TOC is similar to the level in zone 1, thus a dilution effect due to increased content of organic matter cannot explain the very low counts of K and Ti. A short erosional event during zone 2, with high tree cover is reflected by K and Ti peaks at AD 850. It corresponds to a single low measurement of d18Olw, and we speculate that the detrital input is a response to a hydrological event, e.g. lake shore erosion following a lake level increase. Increased erosion levels in zone 3 arise as a response to the deforestation starting at AD 1050. Deviating variation in Rb counts from those of Ti and K suggest intensified chemical weathering of the mineral material, and deposition of a clay layer (Table 2) is previously described as a response to forest clearance (Snowball et al. 1999). The erosion level remained elevated during the forest recovery after AD 1300, while erosion levels peaked again at AD 1700 as a consequence of further deforestation. The following decrease in K and Ti is likely caused by increasing contents of undecomposed organic material and decreasing bulk density, and therefore does not reflect a decrease in erosion level. Landscape dynamics The local landscape development at Ilsø can be interpreted as having had four main phases during the last 2,800 years based on the records of pollen percentages, PARs and the LOVE reconstruction (Figs. 7, 8, 9). The geochemical results document shifting surface erosion to the lake and this could also mean changing amount of pollen grains brought to the lake by surface runoff. Terrestrial PARs are clearly higher during woodland phases than during open phases (Figs. 7, 8), contrary to what

would be expected if surface runoff was a major mechanism of pollen transport to the lake sediments. We conclude that airborne pollen is the major source of pollen for Ilsø sediments. We have not been able to quantify any pollen contribution coming from runoff but this must be kept in mind as a possible mechanism during open phases exactly like in any other study of lake sedimentary pollen from open landscapes. The special case of pollen entering through retting is discussed under phase IL-3. IL-1 800

BC–AD

50

This period can be subdivided into two phases, 800–600 BC and 600 BC–AD 50. The decrease in woodland cover during 800–600 BC was accompanied by rising frequencies of ribwort plantain (Plantago lanceolata) suggesting that the woodland decline was mainly brought about by an increase in grazing pressure. Estimated ribwort plantain cover frequencies of up to almost 20 % suggest that grazing pressure was kept high during the following 650 years resulting in a very open pastoral landscape. Some crop cultivation was also practiced during this period. Hazel (Corylus avellana) was the most prominent woody taxon suggesting that the remaining and continuously declining woodland was mainly composed of copses, probably heavily affected by domestic grazing. Increasing pollen percentage and PAR values of Calluna vulgaris through the period may suggest progressive leaching of the soil, although this Callunasignal is filtered out in the LOVE reconstruction because of the inclusion of pure heathland sites used in the REVEALS data set. IL-2

AD

50–1000

This was a long period with dominant woodland cover initiated by a rapid rise up to about AD 100 and followed by a slow general increase until AD 1000. Elevated PAR values for the tree pollen sum support the interpretation of the woodland cover as denser after AD 600. The initial increase of woodland primarily comprised hazel, birch, willow and oak, which may all act as pioneer species invading abandoned pastures. Beech subsequently took over being dependent on forested environments for establishment. These patterns of vegetation succession probably reflect a combination of processes on both well-drained and poorly drained soils. The increase in birch and to a lesser extent, alder, suggest that poorly drained soils were also abandoned to allow encroachment of woodland. On welldrained soils, the initial oak-dominated forest was replaced by beech in the later part of this period.

123

Veget Hist Archaeobot

IL-3

AD

1000–1650

The rapid deforestation in the 11th century reflected in percentages, PARs and LOVE estimates of primarily birch, hazel and beech is quite early compared to the other vegetation records from medieval Denmark that generally record this event around AD 1100–1200 (Aaby 1986a, b; Andersen 1989; Rasmussen 2005; Nielsen and Odgaard 2010). Furthermore, the following open period is shortlived, being interrupted by a reforestation event by around 1300. Within the dating uncertainty, this event could perhaps be linked to the economic crisis accompanying the plague of the 14th century (Benedictow 2004). After AD 1300 Fagus became a very dominant tree. Alongside this reforestation there was also an expansion of arable farming in the second half of the period, as indicated by increases in Secale, Fagopyrum and Centaurea cyanus. However, high percentages of Humulus/Cannabis-type indicate that the lake was used for retting of Cannabis sativa during this period and it is possible that some of the signal of increased arable farming may derive from pollen contributed through the retting rather than from airborne sources. The period ends with a strong deforestation event in the 17th century. IL-4

AD

1650–2000

The modern, very open landscape is consolidated during this phase. The steep slope bordering the lake to the south continues to support a small grove of mostly Fagus. Small plantations of introduced conifers are established in close vicinity. Gradually arable farming becomes more and more important at the expense of animal husbandry. Ilsø and Taastrup are both situated at the eastern fringe of the hilly landscape of Central Jutland. When combining the vegetation records of Ilsø and Taastrup, the late Holocene dynamics of the two sites are typical for such regions in Denmark that are situated at the transition between flat, fertile landscapes and more hilly ones (Odgaard and Rasmussen 2000; Nielsen and Odgaard 2010). Nielsen and Odgaard (2010) identified this landscape setting as the buffer zone that during some periods in prehistory absorbed the expanding need for pasture for livestock in particular. Similarly, during periods of unrest or other crisis this buffer zone was the first to experience a decrease in land use intensity. However there is a difference in timing and rate of change of the Iron Age reforestation events at the two sites. Areas in Denmark showing reforestation during the first millennium AD can be divided into an early group showing initial reforestation at about AD 200 (e.g. Aaby 1986a: Fuglsø Mose; Aaby 1986b: Abkær Mose) and a late group beginning at AD 500–600 (e.g., Aaby 1986a: Holmegaard Mose; Rasmussen 2005: Dallund Sø; Andersen 1989: Næsbyholm Storskov; Nielsen and

123

Odgaard 2010: Dallerup Sø). The lack of radiocarbon dates in the upper part of the series of Taastrup potentially increases dating uncertainties for the period younger than 1000 BC, but the inferred age of the first Secale pollen grain corroborates the general precision of the Iron Age chronology. Accordingly, Taastrup seems to fit into the group of early reforestation starting by AD 200. Although LRA reconstruction of past forest cover has only been performed at a few of the sites mentioned above, it is clear from the percentage pollen records alone that in both groups the reforestation was relatively gradual and nowhere as rapid as at Ilsø. The pollen evidence of strong and early reforestation at Ilsø is supported by the independent lake sediment geochemistry results, which reflect abandonment of agricultural activities by the very late last century BC. Such abrupt landscape changes at the transition to the Roman Iron Age are unparalleled in Danish vegetation history and these results point to possibly unique conditions at Ilsø. The LRA reconstruction of the local vegetation succession at Ilsø reflects the relevant source area of pollen (RSAP) determined here to be within a radius of 300–450 m. However, it is difficult to imagine that such strong landscape changes affected just the RSAP of Ilsø and not a wider area as well, including the Illerup/Alken valley. The radiocarbon dates of carefully selected terrestrial macrofossils from the Ilsø sediment are consistent, show no outliers and reflect an almost linear sedimentation rate except in the uppermost unconsolidated layers. This results in an age/depth model with very tight 95 % credibility intervals. We conclude that, within the uncertainties of calibrated radiocarbon ages and age-depth modelling, the onset of agricultural abandonment and of subsequent reforestation at Ilsø is contemporary with the high-precision dated burial of human bones in Alken Enge. We hypothesize that the importance of the valley wetlands for ritual burial, starting with the Alken Enge event at the beginning of the first century AD and continuing with the Illerup massive weapon burial by the third and fifth century AD (Ilkjær 1991), posed restrictions on land-use that were active throughout the entire Roman Iron Age and possibly longer.

Conclusions Reconstruction of local vegetation combining a pollen record with the recently developed LRA shows a surprisingly early and rapid reforestation to have taken place in the Ilsø catchment during the first century AD. This result is corroborated by analysis of sedimentary geochemistry showing a strong relaxation of surface erosion and an increase of supply of terrestrial organic material interpreted to have woodland origin. To test the hypothesis that such a

Veget Hist Archaeobot

development was characteristic of the landscape setting in a wider area at the fringe of the Central Jutland uplands we compared this result with the catchment development of Taastrup, 10 km away. Taastrup showed a contrasting slow reforestation during the Iron Age starting at around AD 200 and followed the pattern documented at many other Danish sites. We conclude that the landscape changes at Ilsø during the Early Roman Iron Age are so far unparalleled in a Danish context and we attribute this finding to the setting of Ilsø in the Illerup/Alken Enge valley, which was repeatedly used for large-scale, post-warfare burials of first bodies and later weapons during the Roman Iron age. The new results support the conclusion that the special status of the valley system persisted at least throughout the Roman Iron Age and they provide a possible example of prehistoric control of land-use for ritual purposes. Acknowledgments This research was financially supported by The Carlsberg Foundation (2012_01_0495) and carried out during a Ph.D. co-funded by the Graduate School of Science and Technology, Aarhus University. We thank Andreas Lu¨cke for advice and analysis of stable isotopes on cellulose samples at the Ju¨lich Research Center, Germany. Richard Bradshaw is acknowledged for improving the English.

References Aaby B (1986a) Trees as anthropogenic indicators in regional pollen diagrams from eastern Denmark. In: Behre K-E (ed) Anthropogenic Indicators in Pollen Diagrams. Balkema, Amsterdam, pp 73–93 Aaby B (1986b) Mennesket og naturen pa˚ Abkæregnen gennem 6000 a˚r. Sønderjysk Ma˚nedsskrift 1986:277–290 Abraham V, Koza´kova´ R (2012) Relative pollen productivity estimates in the modern agricultural landscape of Central Bohemia (Czech Republic). Rev Palaeobot Palynol 179:1–12 Andersen H (1959) Voldsom død. Skalk 1959:8–9 Andersen ST (1979) Identification of wild grass and cereal pollen. ˚ rbog 1978:69–92 Danmarks Geologiske Undersøgelse, A Andersen ST (1989) Natural and cultural landscapes since the Ice Age shown by pollen analysis from small hollows in a forested area in Denmark. J Dan Archaeol 8:188–199 Andersen ST (1995) Pollenanalyser fra Ovesø. Geol Surv Den Kunderapport 1995–12:36–55 Barbour MM, Roden JS, Farquhar GD, Ehleringer JR (2004) Expressing leaf water and cellulose oxygen isotope ratios as enrichment above source water reveals evidence of a Pe´clet effect. Oecologia 138:426–435 Benedictow OJ (2004) The black death 1346–1353: the complete history. Boydell Press, Woodbridge Bennett KD (1996) Determination of the number of zones in a biostratigraphical sequence. New Phytol 132:155–170 Bennett KD (2015) Manual for psimpoll and pscomb. http://www. chrono.qub.ac.uk/psimpoll/psimpoll_manual/4.27/psimpoll.htm. Accessed 29 July 2015 Bronk Ramsey C (2013) OxCal version 4.2.4. https://c14.arch.ox.ac. uk. Accessed 25 May 2015 Brostro¨m A, Nielsen AB, Gaillard MJ, Hjelle K, Mazier F, Binney H, Bunting J, Fyfe R, Meltsov V, Poska A, Ra¨sa¨nen S, Soepboer W, Von Stedingk H, Suutari H, Sugita S (2008) Pollen productivity estimates of key European plant taxa for quantitative

reconstruction of past vegetation: a review. Veget Hist Archaeobot 17:461–478 Church M, Slaymaker O (1989) Disequilibrium of Holocene sediment yield in glaciated British Columbia. Nature 337:452–454 Heyng AM, Mayr C, Lu¨cke A, Wissel H, Striewski B (2014) Late Holocene hydrologic changes in northern New Zealand inferred from stable isotope values of aquatic cellulose in sediments from Lake Pupuke. J Paleolimnol 51:485–497 ˚ dal 1–2: die Lanzen und Speere. Jysk Arkæol Ilkjær J (1991) Illerup A Selsk Skr 25:1–404 Kylander ME, Ampel L, Wohlfarth B, Veres D (2011) Highresolution X-ray fluorescence core scanning analysis of Les Echets (France) sedimentary sequence: new insights from chemical proxies. J Quat Sci 26:109–117 Lowemark L, Chen HF, Yang TN, Kylander M, Yu EF, Hsu YW, Lee TQ, Song SR, Jarvis S (2011) Normalizing XRF-scanner data: a cautionary note on the interpretation of high-resolution records from organic-rich lakes. J Asian Earth Sci 40:1250–1256 Mayr C, Lu¨cke A, Wagner S, Wissel H, Ohlendorf C, Haberzettl T, Oehlerich M, Scha¨bitz F, Wille M, Zhu J, Zolitschka B (2013) Intensified southern hemisphere westerlies regulated atmospheric CO2- during the last glaciation. Geology 41:831–834 Mazier F, Gaillard MJ, Kunesˇ P, Sugita S, Trondman AK, Brostro¨m A (2012) Testing the effect of site selection and parameter setting on REVEALS-model estimates of plant abundance using the Czech Quaternary Palynological Database. Rev Palaeobot Palynol 187:38–49 Meyers PA, Teranes JL (2001) Sediment Organic Matter. In: Last WM, Smol JP (eds) Tracking environmental change using lake sediments: physical and geochemical methods, vol 2. Kluwer Academic, Dordrecht, pp 239–269 Mingram J, Negendank JF, Brauer A, Berger D, Hendrich A, Ko¨hler M, Usinger H (2007) Long cores from small lakes—recovering up to 100 m-long lake sediment sequences with a high-precision rod-operated piston corer (Usinger-corer). J Paleolimnol 37:517–528 Nesbitt HW, Markovics G, Price RC (1980) Chemical processes affecting alkalis and alkaline earths during continental weathering. Geochim Cosmochim Acta 44:1659–1666 Nielsen AB, Odgaard BV (2010) Quantitative landscape dynamics in Denmark through the last three millennia based on the Landscape Reconstruction Algorithm approach. Veget Hist Archaeobot 19:375–387 Nielsen AB, Giesecke T, Teuerkauf M, Feeser I, Behre KE, Beug HJ, Chen SH, Christiansen J, Do¨rfler W, Endtmann E, Jahns S, De Klerk P, Ku¨hl N, Latałowa M, Odgaard BV, Rasmussen P, Stockholm JR, Voigt R, Wiethold J, Wolters S (2012) Quantitative reconstructions of changes in regional openness in northcentral Europe reveal new insights into old questions. Quat Sci Rev 47:131–149 Odgaard BV (1993) The sedimentary record of spheroidal carbonaceous fly-ash particles in shallow Danish lakes. J Paleolimnol 8:171–187 Odgaard BV (1994) The holocene vegetation history of northern west Jutland, Denmark. Opera Bot 123:1–171 Odgaard B (1999a) Pollenanalytiske undersøgelser 1998. Pollenanalyser fra Dallerup Sø, Østjylland. Danmarks og Grønlands Geologiske Undersøgelse Rapport 1999/14:1–10 Odgaard BV (1999b) Fossil pollen as a record of past biodiversity. J Biogeogr 26:7–17 Odgaard BV (2000) Fra skov til hede: Vegetationens historie i Ulfborg herred. In: Dalsgaard K (ed) Mellem hav og hede. Aarhus University Press, Aarhus, pp 28–35 Odgaard BV, Rasmussen P (2000) Origin and temporal development of macro-scale vegetation patterns in the cultural landscape of Denmark. J Ecol 88:733–748

123

Veget Hist Archaeobot Pawlowsky-Glahn V, Egozcue JJ (2006) Compositional data and their analysis: an introduction. In: Buccianti A, Meteu-Figueras G, Pawlowsky-Glahn V (eds) Compositional data analysis in the geosciences: from theory to practice, vol 264. Geological Society Special Publication. Geological Society, London, pp 1–10 Pienitz R, Smol JP (1993) Diatom assemblages and their relationship to environmental variables in lakes from the boreal forest-tundra ecotone near Yellowknife, Northwest Territories, Canada. Hydrobiologia 269–270:391–404 Rasmussen P (2005) Mid- to late-Holocene land-use change and lake development at Dallund Sø, Denmark: vegetation and landuse history inferred from pollen data. Holocene 15:1116–1129 Reimer PJ, Bard E, Bayliss A et al (2013) IntCal13 and Marine13 radiocarbon age calibration curves 0–50,000 years cal BP. Radiocarbon 55:1869–1887 Renberg I, Hansson H (2008) The HTH sediment corer. J Paleolimnol 40:655–659 Snowball I, Sandgren P, Petterson G (1999) The mineral magnetic properties of an annually laminated Holocene lake-sediment sequence in northern Sweden. Holocene 9:353–362 Sugita S (1993) A model of pollen source area for an entire lake surface. Quat Res 39:239–244 Sugita S (2007a) Theory of quantitative reconstruction of vegetation. I: pollen from large sites REVEALS regional vegetation composition. Holocene 17:229–241 Sugita S (2007b) Theory of quantitative reconstruction of vegetation. II: all you need is LOVE. Holocene 17:243–257 Sugita S, Parshall T, Calcote R, Walker K (2010) Testing the Landscape Reconstruction Algorithm for spatially explicit

123

reconstruction of vegetation in northern Michigan and Wisconsin. Quat Res 74:289–300 Talbot MR (2001) Nitrogen isotopes in palaeolimnology. In: Last WM, Smol JP (eds) Tracking environmental change using lake sediments: physical and geochemical methods, vol 2. Kluwer Academic, Dordrecht, pp 401–439 Ter Braak CJF, Sˇmilauer P (2012) Canoco reference manual and user’s guide: software for ordination, version 5.0. Microcomputer Power, Ithaca, USA Weltje GJ, Tjallingii R (2008) Calibration of XRF core scanners for quantitative geochemical logging of sediment cores: theory and application. Earth Planet Sci Lett 274:423–438 Wissel H, Mayr C, Lu¨cke A (2008) A new approach for the isolation of cellulose from aquatic plant tissue and freshwater sediments for stable isotope analysis. Org Geochem 39:1545–1561 Wolfe BB, Edwards TW (1997) Hydrologic control on the oxygenisotope relation between sediment cellulose and lake water, western Taimyr Peninsula, Russia: implications for the use of surface-sediment calibrations in paleolimnology. J Paleolimnol 18:283–291 Wolfe BB, Edwards TW, Aravena R, Forman SL, Warner BG, Velichko AA, MacDonald GM (2000) Holocene paleohydrology and paleoclimate at treeline, north-central Russia, inferred from oxygen isotope records in lake sediment cellulose. Quat Res 53:319–329 Wolfe BB, Edwards TW, Elgood RJ, Beuning KR (2001) Carbon and oxygen isotope analysis of lake sediment cellulose: methods and applications. In: Last WM, Smol JP (eds) Tracking environmental change using lake sediments: physical and geochemical methods, vol 2. Kluwer Academic, Dordrecht, pp 373–400