Introduction
Variation in human activities has greatly impacted the processes and intensities of erosion, sediment transport and storage throughout the Late Holocene, and many lowland rivers around the world have responded to these variations (Notebaert & Verstraeten, Reference Notebaert and Verstraeten2010; Brown et al., Reference Brown, Lespez, Sear, Macaire, Houben, Klimek, Brazier, Van Oost and Pears2018; Macklin & Lewin, Reference Macklin and Lewin2019). The process–response relationship on a long timescale has been well established: The general concept of fluvial response following human disturbances in NW Europe is that of continuously increasing human impact since Neolithisation which caused a continuous aggradation in the floodplain, although delayed (e.g. Houben, Reference Houben2007; De Moor et al., Reference De Moor, Kasse, van Balen, Vandenberghe and Wallinga2008; Dotterweich, Reference Dotterweich2008; Lespez et al., Reference Lespez, Clet-Pellerin, Limondin-Lozouet, Pastre, Fontugne and Marcigny2008; Notebaert & Verstraeten, Reference Notebaert and Verstraeten2010; Fuchs et al., Reference Fuchs, Will, Kunert, Kreutzer, Fischer and Reverman2011; Verstraeten et al., Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017). Changes in climate and land use in the river catchment could also cause changes in river channel patterns (e.g. De Moor et al., Reference De Moor, Kasse, van Balen, Vandenberghe and Wallinga2008; Słowik, Reference Słowik2015; Candel et al., Reference Candel, Kleinhans, Makaske, Hoek, Quik and Wallinga2018). Other studies point to the importance of infrastructural works in the floodplain (e.g. watermills and bank protection measures) for changes in floodplain morphology and channel pattern (e.g. Walter & Merritts, Reference Walter and Merritts2008; Słowik, Reference Słowik2013; Hobo et al., Reference Hobo, Makaske, Wallinga and Middelkoop2014; Maaß and Schüttrumpf, Reference Maaß and Schüttrumpf2019). For the Dijle catchment, located in the Belgian loess belt, Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b) have studied the floodplain response to human disturbance and have shown how the Dijle floodplain changed from a natural to a human-dominated environment. By evaluating detailed data on floodplain changes with detailed data on human impact and regional vegetation changes, they showed that natural floodplains of the Dijle valley consisted of strongly vegetated marshy environments dominated by an alder carr forest. This changed when human impact in the catchments crossed a threshold, such that sediments eroded on cultivated hillslopes entered the floodplains. As a result, floodplain geomorphology and ecology (hereafter called ‘geo-ecology’) changed from a marshy, forested environment towards a more open floodplain dominated by clastic overbank deposits (Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b).
Superimposed on these long-term trends of increasing human impact and the fluvial response to it, short-term fluctuations (c.200 years) are also present. The effects of such shorter-term changes in climate and human impact on the fluvial system during the Holocene period are harder to study, due to preservation and resolution issues. Insight into these effects is needed, however, to fully understand the process–response relationship on shorter timescales, and to understand the possible effect of future changes in human impact or climate on river systems. For instance, it can be questioned whether short-term climate oscillations in the Late Holocene are intensive enough to trigger changes in fluvial systems or whether these impacts are buffered in the catchment. The first millennium AD in NW Europe is characterized by such short-term changes in human impact and climate. Particularly between 250 and 700 AD, cultural change, population decline and widespread forest regeneration are observed in NW Europe (e.g. Zolitschka et al., Reference Zolitschka and Schneider2003; Cheyette, Reference Cheyette2008; Wickham, Reference Wickham2009; Forster, Reference Forster2010) and therefore this period is often called the ‘Dark Ages’. In Belgium, forest regeneration and recovery started between 250 and 700 AD (Tack et al., Reference Tack, Van Den Bemt and Hermy1993). In France, the Netherlands and Belgium, overall population decline is estimated at c.40% between 500 and 650 AD compared to the Roman period (Russell, Reference Russell and Cipolla1972). In the Rhine–Meuse delta (the Netherlands), Groenewoudt and Van Lanen (Reference Groenewoudt and van Lanen2018) and Van Lanen et al. (Reference Van Lanen, De Kleijn, Gouw-Bouman and Pierik2018) estimated the population decline between 270 and 725 AD at c.80%. Moreover, the first millennium AD in NW Europe was characterised by climate variability with consecutively a warmer period (between c.1 and 250 AD), a colder period (between c.250 and 700 AD) and another warmer period (between c.700 and 1000 AD) (Riechelmann and Gouw-Bouman, Reference Riechelmann and Gouw-Bouman2019), although there is no consensus about their precise temporal and spatial extent (see e.g. Helama et al., Reference Helama, Jones and Briffa2017; Neukom et al., Reference Neukom, Steiger, Gómez-Navarro, Wang and Werner2019; Riechelmann and Gouw-Bouman, Reference Riechelmann and Gouw-Bouman2019).
Only a few previous studies have been done on the effect of these changes in human impact and climate, during the first millennium AD, on changes in river systems in NW Europe (e.g. Lang and Nolte, Reference Lang and Nolte1999; Houben et al., Reference Houben, Schmidt, Mauz, Stobbe and Lang2013; Pierik et al., Reference Pierik, Cohen and Stouthamer2016; van Dinter et al., Reference van Dinter, Cohen, Hoek, Stouthamer, Jansma and Middelkoop2017). In the Belgian loess belt, indications of decreasing human impact during the first millennium AD come from pollen records in Dijle (Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c) and Mombeek catchment (Heyvaert, Reference Heyvaert1983), and are briefly linked with changes in the river system (Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b). A detailed understanding of these links is not yet available. Therefore, this research seeks to study the changes in floodplain geo-ecology in the Belgian loess belt during the first millennium AD and the role of changing human impact in it. For this purpose, an integrated approach is needed, and detailed data on floodplain changes need to be evaluated against detailed data on human impact (see e.g. Foulds and Macklin, Reference Foulds and Macklin2006; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b; Verstraeten et al., Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017). Several studies (Diriken, Reference Diriken1981; Heyvaert, Reference Heyvaert1983; Rommens et al., Reference Rommens, Verstraeten, Bogman, Peeters, Poesen, Govers, Van Rompaey and Lang2006; Notebaert et al., Reference Notebaert, Houbrechts, Verstraeten, Broothaerts, Haeckx, Reynders, Govers, Petit and Poesen2011a, b; Broothaerts et al., Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013, Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b) have previously presented data on floodplain changes and vegetation changes in the Belgian loess belt. In the present study we combine these data, together with newly collected data, to provide an understanding of the significance of the changes in human impact during the first millennium AD on the river valleys in the Belgian loess belt.
Study sites
This study focuses on three river catchments (Fig. 1), the Dijle catchment south of Leuven (c.750 km2), the Gete catchment (c.800 km2) and the Mombeek catchment (c.100 km2), all part of the Scheldt catchment (c.20,000 km2). The catchments are located in the Belgian loess belt, characterised by an undulating plateau in which several rivers are incised. Soils are mainly Luvisols, developed in Pleistocene loess deposits. The elevation of the Dijle catchment ranges from 25 m a.s.l. at the outlet to 165 m a.s.l. in the south of the catchment. The Dijle River has at the outlet a base discharge of 4 m3 s−1 and a peak discharge of 25 m³ s−1. The elevation of the Gete catchment ranges between 160 and 21 m a.s.l.; base discharge at the outlet of the Gete catchment is c.2 m³ s−1 and peak discharge c.22 m³ s−1. In the Mombeek catchment the elevation ranges between 112 and 34 m a.s.l.; base discharge at the outlet of the Mombeek catchment is c.0.1 m³ s−1 and peak discharge c.3 m³ s−1. Floodplain slope gradient in all three catchments ranges between 0.1% in the main valleys and 1.5% in the headwaters. Floodplain width in the headwaters does not exceed 150 m; in the downstream part of the catchments, floodplains can reach 1000 to 1500 m.
Location of the studied catchments, with indication of the locations of the coring transects. From west to east: Dijle, Gete and Mombeek catchments.
The current land use is dominated by cropland (c.40%), built-up area (c.25%) and grassland (c.20%). Floodplains are mainly dominated by grassland and plantation forests. Previous palynological studies in the Dijle and Mombeek catchments show that the catchments were mainly forested during the first half of the Holocene (Heyvaert, Reference Heyvaert1983; Broothaerts et al., Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013, Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). The Neolithic Linearbandkeramic arrived in the Belgian loess belt around 5200 BC (Vanmontfort, Reference Vanmontfort2007, Reference Vanmontfort2008). During the Neolithic period (5200 BC until 1900 BC; Table 1) human impact in the catchment was probably limited to local disturbances and small-scale forest clearance (Heyvaert, Reference Heyvaert1983; Broothaerts et al., Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013, Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). From the Bronze Age (1900 BC until 700 BC; Table 1), vegetation gradually changed under the influence of increasing human impact in the catchment. Human impact first peaked during the Roman period (50 BC until 250 AD; Table 1). During the Medieval period (from 1000 AD; Table 1) human impact increased further to reach its highest values in the Modern period (Heyvaert, Reference Heyvaert1983; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). Archaeological and historical data for the Belgian loess belt are limited and fragmented. For the Flemish part, an extensive database is available (Van Daele et al., Reference Van Daele, Meylemans and de Meyer2004; CAI, 2019); but for the larger Walloon part, such a detailed database is missing.
Archaeological periods in the Belgian loess belt. Based on CAI (2019)
Materials and methods
Previous studies have provided data on past changes in floodplain morphology, local floodplain vegetation, regional vegetation and human impact in the Dijle, Mombeek and Gete catchments. In this study, we chronologically combine these data on local floodplain vegetation with the available reconstructions of floodplain morphology to reconstruct the floodplain geo-ecology in these catchments. In a next step, these data on floodplain geo-ecology are chronologically evaluated with detailed data on human impact and regional vegetation changes, to provide insight into the temporal relation between human impact and changing floodplain geo-ecology. The temporal focus in this study is on the first millennium AD. As such, for the first time, an integrated approach is provided in which detailed data on floodplain changes are evaluated against detailed data on human impact for the first millennium AD. The temporal details of the available data allow this evaluation to be made on a 200-year resolution.
An overview of how the floodplain morphology, local floodplain vegetation, regional vegetation and human impact were reconstructed in the Dijle, Mombeek and Gete catchments is given below. For a more detailed description of these methods and datasets, we refer to the original papers.
Reconstructing floodplain morphology
Floodplain transects are available for the Dijle catchment from Rommens et al. (Reference Rommens, Verstraeten, Bogman, Peeters, Poesen, Govers, Van Rompaey and Lang2006), Notebaert et al (Reference Notebaert, Houbrechts, Verstraeten, Broothaerts, Haeckx, Reynders, Govers, Petit and Poesen2011a,b) and Broothaerts et al. (Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013, Reference Broothaerts, Notebaert, Verstraeten, Kasse, Bohncke and Vandenberghe2014a); for the Mombeek catchment from Diriken (Reference Diriken1981) and Vervoort (Reference Vervoort2018); and for the Gete catchment from Quintens (Reference Quintens2019). Altogether, 16 transects are available for the Dijle catchment, 11 transects for the Mombeek catchment and 3 transects for the Gete catchment (Fig. 1). Coring density is around one coring every 20 m for most of the transects. The alluvial architecture was reconstructed based on this data set of, in total, c.400 corings. For each coring, a lithological field description is available with a vertical resolution of 5 cm, containing texture and sorting determined by palpation, colour description, description of soil horizons and identification of inclusions, following United Nations Food and Agriculture Organization (FAO) guidelines (Jahn et al., Reference Jahn, Blume, Asio, Spaargaren and Schad2006). Based on the detailed field descriptions, textural data and organic matter content, the sediments were grouped in lithostratigraphical units, using floodplain architecture concepts (Houben, Reference Houben2007; De Moor and Verstraeten, Reference De Moor and Verstraeten2008) and following Notebaert et al. (Reference Notebaert, Houbrechts, Verstraeten, Broothaerts, Haeckx, Reynders, Govers, Petit and Poesen2011a) and Broothaerts et al. (Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013) (Table 2).
Lithostratigraphical units identified in the Dijle, Mombeek and Gete catchments. Based on Notebaert et al. (Reference Notebaert, Houbrechts, Verstraeten, Broothaerts, Haeckx, Reynders, Govers, Petit and Poesen2011a) and Broothaerts et al. (Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013)
Chronology of floodplain changes
A detailed chronology of the floodplain changes in the Dijle catchment is provided by Broothaerts et al. (Reference Broothaerts, Notebaert, Verstraeten, Kasse, Bohncke and Vandenberghe2014a) and Verstraeten et al. (Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017). Radiocarbon dates are available for the Mombeek valley from Diriken (Reference Diriken1981), Heyvaert (Reference Heyvaert1983) and Vervoort (Reference Vervoort2018). For the Gete catchment, radiocarbon dates are available from Quintens (Reference Quintens2019). In total, 160 radiocarbon dates are available for the studied catchments for the entire Holocene period, and provide a chronological framework for the floodplain changes and lithostratigraphical units. All ages were calibrated using the IntCal13 calibration curve (Reimer et al., Reference Reimer, Bard, Bayliss, Beck, Blackwell, Ramsey, Grootes, Guilderson, Haflidason and Hajdas2013) and Oxcal 4.3 software (Ramsey, Reference Ramsey2009). The calibrated radiocarbon dates were used to obtain a chronology for the floodplain changes during the Holocene period and more specifically during the first millennium AD. Timescales for the available pollen sequences and floodplain accumulation rates were made based on four to seven radiocarbon dates for each pollen sequence, dating each important change in lithology and pollen signal. Finally, the organic-rich overbank deposits observed in the studied floodplains were dated using 14 available radiocarbon dates from this layer (Table 3).
Radiocarbon dates for the organic-rich overbank deposits, in Dijle, Mombeek and Gete catchments (total: 14)
Reconstructing vegetation changes and human impact
Holocene vegetation changes in the Dijle catchment were reconstructed based on pollen data of six alluvial sites by Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). Reconstruction of Holocene vegetation changes in the Mombeek catchment comes from pollen data of one alluvial site analysed by Heyvaert et al. (Reference Heyvaert1983). Pollen data for the Gete catchment are currently not available.
In this study, data on human impact in the Dijle and Mombeek catchments are extracted from the regional pollen signal based on non-metric multidimensional scaling (NMDS), a statistical ordination technique. NMDS has successfully been applied to pollen data in previous studies (Ghilardi and O′Connell, Reference Ghilardi and O’Connell2013; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c, Reference Broothaerts, Robles-López, Abel-Schaad, Pérez-Díaz, Alba-Sánchez, Luelmo-Lautenschlaeger, Glais and López-Sáez2018; Woodbridge et al., Reference Woodbridge, Roberts, Palmisano, Bevan, Shennan, Fyfe, Eastwood, Izdebski, Çakırlar, Woldring, Broothaerts, Kaniewski, Finné and Labuhn2019). Full explanation of NMDS can be found in Legendre & Legendre (Reference Legendre and Legendre1983) and McCune & Grace (Reference McCune and Grace2002). In this study, NMDS was applied to the pollen data of the Dijle catchment (six sites; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c) and Mombeek catchment (one site; Heyvaert, Reference Heyvaert1983) respectively, using R-package vegan (Oksanen et al., Reference Oksanen, Blanchet, Kindt, Legendre, Minchin, O′Hara, Simpson, Solymos, Stevens and Wagner2012). Bray–Curtis dissimilarities were used to calculate the distance matrix for ordination. All regional pollen taxa that occurred in more than 5% of the samples were included. Wisconsin double standardisation and a square-root transformation were performed in vegan, if the pollen values were larger than common abundance class scales. A two-dimensional solution was used, since the measure of fit of the data versus the number of dimensions shows that two axes provide a greater reduction in stress than a higher number of axes. Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c) showed that scores on NMDS axis 1 can be used as a proxy for human impact in the catchment: low negative scores correspond to tree pollen types such as Ulmus, Tilia, Corylus and Quercus and indicate forested landscapes, while high positive scores correspond to cultural indicators such as Centaurea cyanus, Plantago lanceolata and cereal types and indicate deforested landscapes and human activities. The scores on NMDS axis 1 were plotted as a function of time, using the timescale constructed for the pollen data by Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c) and Heyvaert (Reference Heyvaert1983) for the Dijle and Mombeek catchments respectively. Average NMDS scores were calculated with a time step of 200 years, and trends are discussed on a 200-year resolution. Plotting the scores on NMDS axis 1 in function of time can be used as an indication of the evolution of human impact through time, as shown by Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c).
In addition, the available pollen data for Dijle and Mombeek were also used to reconstruct changes in local floodplain vegetation throughout the Holocene and the first millennium AD.
Results
Regional vegetation changes and human impact
The pollen records from the Dijle and Mombeek catchments indicate that these catchments were forested during the Mid-Holocene, dominated by deciduous trees, and that human impact indicators were absent or low (Fig. 2). This deciduous forest is considered as the natural vegetation of the catchments in the Belgian loess belt (see Heyvaert, Reference Heyvaert1983; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). A clear decrease in forest cover is observed from the Bronze Age onwards, with an increase in grasses and anthropogenic indicators (Fig. 2). The trend of increasing human impact and deforestation is interrupted after the Roman period (Figs 2 and 3). More specifically the scores on NMDS axis 1, an indicator for human impact, show a decrease in human impact between 200 and 400 AD in the Dijle catchment and between 200 and 800 AD in the Mombeek catchment (Fig. 3). During that time, the pollen data and regional vegetation reconstructions show an increase in tree taxa and a decrease in human impact indicators (Fig. 2). Within the Dijle catchment there are some local variations, with a more pronounced signal in the downstream part of the catchment, as for example in Archennes (Fig. 2A), and a less pronounced signal in the upstream part (Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). The period of decrease in human impact and renewed forest growth is, however, rather short, only 200 to 600 years. By c.400 AD, at most study sites in the Dijle catchment and by 800 AD in the Mombeek catchment, forest cover is again decreasing (Fig. 2) and human impact is increasing (Fig. 3). Human impact finally reaches its highest values during the Medieval to Modern period (Fig. 3).
(A) Typical pollen diagram in Dijle catchment, near Archennes (based on Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). Black rectangles indicate location of the available radiocarbon dates; numbers refer to Table 3. (B) Typical pollen diagram in Mombeek catchment, near Vliermaal (based on Heyvaert, Reference Heyvaert1983). Black rectangles indicate location of the available radiocarbon dates; numbers refer to Table 3.
Human impact scores for Dijle catchment (updated from Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c) and Mombeek (Vliermaal, based on pollen data of Heyvaert, Reference Heyvaert1983).
Floodplain geo-ecology
Reconstructions of the floodplain morphology (Fig. 4; Table 3) show a complex of organic-rich deposits and peat at the base of the Holocene deposits. These organic-rich deposits are overlain by clastic overbank deposits, which are linked to increased erosion on the hillslopes due to increasing human impact (Broothaerts et al., Reference Broothaerts, Verstraeten, Notebaert, Assendelft, Kasse, Bohncke and Vandenberghe2013). In turn, these clastic overbank deposits contain organic-rich overbank deposits locally grading into peaty units (Fig. 4). These peat and peaty units are in situ, as determined in the field and described in the original publications (Diriken, Reference Diriken1981; Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b). These organic-rich overbank deposits indicate a phase with increased accumulation of organic material or even regrowth of peat in the distal parts of the floodplains. Radiocarbon dates from these organic-rich deposits range between c.1 and 1300 AD, with a peak between 600 and 800 AD (Fig. 5). During that time, we also see a decrease in sedimentation rates at the well-dated sites in the Mombeek catchment (Fig, 4B). Such a decrease is not observed in the data of the Dijle catchment (Fig. 4A), probably due to the rather limited time resolution of the sediment stratigraphy and averaging effects (see e.g. Notebaert et al., Reference Notebaert, Houbrechts, Verstraeten, Broothaerts, Haeckx, Reynders, Govers, Petit and Poesen2011a). These peaty deposits are mainly found in the distal parts of the floodplain and backswamps, and do not cover the entire floodplain width. Near the river channel, clastic overbank deposits can still be found (Fig. 4).
Typical lithostratigraphical transects of the studied floodplains. (A) Dijle (based on Broothaerts, Reference Broothaerts2014); (B) Gete (based on Quintens, Reference Quintens2019); (C) Mombeek (based on Diriken, Reference Diriken1981). Black rectangles indicate location of the available radiocarbon dates of the organic-rich layer; numbers refer to Table 3.
Probability density function (PDF) of all radiocarbon ages of Unit 5 (organic-rich overbank deposits) in Dijle, Gete and Mombeek catchments (n = 14).
Also in the first millennium AD, the local pollen data clearly show a regrowth of the alder carr forest at several study sites (Fig. 2). In Archennes, Dijle catchment, this regrowth is dated between 300 and 900 AD, and in Mombeek catchment, between 400 and 750 AD (Fig. 2). Similar observations are made at the other sites in the Dijle catchment (Broothaerts, Reference Broothaerts2014).
Discussion
Changes in floodplain geo-ecology during the first millennium AD
The general pattern of increasing human impact and consequent changes in floodplain geo-ecology is interrupted during the first millennium AD. Whereas from the Bronze Age onwards the strongly vegetated marshy floodplain environment dominated by an alder carr forest changed under the influence of increasing human impact towards a more open floodplain dominated by clastic overbank deposits, this trend is reversed during the first millennium AD. During the first millennium AD, the floodplain geo-ecology changed again (Fig. 6) with a regrowth of the alder carr forest (Fig. 2), organic-rich deposits and locally a regrowth of peat (Fig. 5), and a decrease in sedimentation rates (Fig. 2). We attribute these changes in floodplain geo-ecology to changes in human impact in the catchment during the first millennium AD. The observed decrease in human impact between 200 and 400 AD in the Dijle catchment and between 200 and 800 AD in the Mombeek catchment (Fig. 3) is suggested to cause a decrease in soil erosion. Similar changes in human pressure on the landscape during the first millennium AD were observed in the Netherlands (van Dinter et al., Reference van Dinter, Cohen, Hoek, Stouthamer, Jansma and Middelkoop2017; Pierik et al., Reference Pierik, van Lanen, Gouw-Bouman, Groenewoudt, Wallinga and Hoek2018; Van Lanen et al., Reference Van Lanen, De Kleijn, Gouw-Bouman and Pierik2018) and SW Germany (Lang & Nolte, Reference Lang and Nolte1999; Houben et al., Reference Houben, Schmidt, Mauz, Stobbe and Lang2013). De Brue & Verstraeten (Reference De Brue and Verstraeten2014) showed for the Dijle catchment the non-linear relation between human impact and sediment delivery to the fluvial system. It is likely that even a minor decrease in agricultural practices and local soil erosion resulted in an important decrease in runoff and sediment delivery to river channels. This can be related to the regeneration of valley-side vegetation reducing hillslope–floodplain connectivity again, as is also demonstrated for the Wetterau catchment in Germany (Houben et al., Reference Houben, Schmidt, Mauz, Stobbe and Lang2013). As a result, flood frequency and sediment input in the floodplain decreased, leading to more stable floodplain environments in which the organic matter content could increase again and alder carr forests could re-establish.
Conceptual overview of the main regional changes (vegetation, tree cover and human impact) and local floodplain changes (probability density function (PDF) of radiocarbon ages of Unit 5, lithology, and floodplain geo-ecology) for river catchments in the Belgian loess belt for the last 6000 years. The conceptual model of floodplain response is partly based on Broothaerts et al. (Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b).
However, the current dataset is too limited to identify the exact mechanism behind these local geo-ecological changes. Indeed, next to changes in sediment input, reduced intensities of human impact may also have resulted in subsequent changes in groundwater fluxes leading to a rewetting of the floodplain and, as such, promotion of organic matter sequestration. Hydrological and sedimentological modelling studies at the catchment scale (e.g. Notebaert et al., Reference Notebaert, Verstraeten, Ward, Renssen and Van Rompaey2011c; De Brue and Verstraeten, Reference De Brue and Verstraeten2014; Swinnen, Reference Swinnen2020) may be better suited to identify causal factors and to study the sensitivity of floodplain geo-ecology to the different changing factors.
Organic-rich deposits from the first millennium AD are mainly found in the distal parts of the floodplain and are not continuously observed in the floodplain transects (Fig. 4). As such there is not a complete return to the original, natural state of the floodplains, i.e. marshy environments without a clear river channel. There is rather an increased accumulation of organic material in the distal parts of the floodplain. Moreover, there is also a high variability within the studied catchments: the described changes in floodplain geo-ecology are observed in 21 out of the 30 studied floodplain transects (70%) and do not show a continuous pattern (Diriken, Reference Diriken1981; Broothaerts, Reference Broothaerts2014). As a result, extrapolation of local trends to the whole catchment is not always justified. A catchment-wide approach, with results from multiple corings within one cross-section and cross-sections from different study sites in the catchment, is needed to gain full insight into the effect of short-term changes in human impact on the fluvial system.
Time lag in floodplain response during first millennium AD
For the Dijle catchment, human impact decreased between 200 and 400 AD whereas the renewed expansion of the alder carr forest is seen in the period 300 to 900 AD (Figs 3 and 6). In the Mombeek catchment, human impact decreases between 200 and 800 AD, whilst alder carr forests return between 400 and 750 AD (Figs 2 and 6). Thus, a time lag of 100 to 200 years can be observed between regional catchment changes (decrease in human impact, regrowth of the forest) and local floodplain changes (renewed peat formation, regrowth of the local alder carr forest). Such a temporal offset between changes of human impact and local floodplain changes is observed in several catchments in West and Central Europe (Lang & Nolte, Reference Lang and Nolte1999; Trimble, Reference Trimble1999; De Moor & Verstraeten, Reference De Moor and Verstraeten2008; Macklin et al., Reference Macklin, Jones and Lewin2010; Houben et al., Reference Houben, Schmidt, Mauz, Stobbe and Lang2013) and has been demonstrated for the Dijle catchment by De Brue & Verstraeten (Reference De Brue and Verstraeten2014), Verstraeten et al. (Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017) and Notebaert et al. (Reference Notebaert, Broothaerts and Verstraeten2018). The observed time lag can indicate that floodplain changes are only triggered when changes in human impact in the catchment are large enough and thus reach a threshold. Once the threshold is reached, even a minor decrease in agricultural practices can result in an important decrease in runoff and sediment delivery to the fluvial system (see Verstraeten et al., Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017; Notebaert et al. 2018). For this, also the regeneration of valley-side vegetation and the abandonment or less intensive use of the network of dirt roads and sunken lanes are important, as they reduce hillslope–floodplain connectivity (see e.g. Houben et al., Reference Houben, Schmidt, Mauz, Stobbe and Lang2013). However, the observed time lag can also indicate that changes in floodplain geo-ecology during the first millennium AD are not only caused by changes in human impact but also driven by other factors such as climate change. Indeed, during the first millennium AD, fluctuations in temperature and precipitation are observed in NW Europe (see e.g. Riechelmann & Gouw-Bouman, Reference Riechelmann and Gouw-Bouman2019) which can cause changes in both catchment and floodplain hydrology. However, there is currently no consensus about the precise temporal and spatial extent of the climate variability (see e.g. Helama et al., Reference Helama, Jones and Briffa2017; Neukom et al., Reference Neukom, Steiger, Gómez-Navarro, Wang and Werner2019; Riechelmann & Gouw-Bouman, Reference Riechelmann and Gouw-Bouman2019). To identify the exact role of climate change, a more detailed temporal framework (<200 years) and more detailed and independent databases on climate, land cover and fluvial changes are needed, as well as hydrological and sedimentological modelling studies at the catchment scale (see previous subsection).
Also, with the increase in human impact at c.400 AD (Dijle) and c.800 AD (Mombeek) and the associated floodplain changes towards a more open floodplain dominated by clastic overbank deposits, there is a delay in the floodplain response (Fig. 6). Again, this can be attributed to the threshold that needs to be crossed before floodplain changes are triggered (see also Verstraeten et al., Reference Verstraeten, Broothaerts, Van Loo, Notebaert, D′Haen, Dusar and De Brue2017; Notebaert et al. 2018;). Moreover, in the wide floodplains in the downstream parts of the studied catchment, the peaty layer from the first millennium AD can be found, especially in the distal parts of the floodplains. A small increase in sediment input will trigger floodplain sediment deposition close to the river channel, while peat growth continued for the more distal parts of the floodplain. Only when human impact in the catchment becomes overwhelming the massive floodplain sedimentation will be triggered and the entire floodplain will become dominated by clastic overbank deposition (Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014b).
Significance of the Dark Ages in the Belgian loess belt
This study demonstrates clear changes in the landscape during the first millennium AD in the Belgian loess belt (Fig. 6). The regional vegetation reconstructions show an increase in tree cover and decrease in human impact between 200 AD and 400 AD, respectively, and 800 AD, in the Dijle and Mombeek catchments. These regional changes in vegetation and human impact are suggested to cause a decrease in soil erosion, a decrease in hillslope–floodplain connectivity and a resulting decrease in sediment delivery to the fluvial system. These regional changes are therefore suggested to be linked to changes in local floodplain geo-ecology, i.e. a regrowth of the alder carr forest and an increase in the organic matter content of the alluvial deposits with a local reactivation of peat growth (Fig. 6). These changes can still be recognised in the current floodplain transects (Fig. 4), although discontinuously. The observed changes in floodplain geo-ecology during the first millennium AD are, however, not observed in the downstream part of the Scheldt catchment (Meylemans et al., Reference Meylemans, Bogemans, Storme, Perdaen, Verdurmen and Deforce2013; Storme et al., Reference Storme, Louwye, Crombé and Deforce2017), probably due to the larger catchment area and the accumulated human-induced sediment supply in the downstream part of the catchment (Broothaerts et al., Reference Broothaerts, Notebaert, Verstraeten, Kasse, Bohncke and Vandenberghe2014a).
Quantitative estimates of forest regeneration or population decline during the first millennium AD cannot be made based on the data presented in this study. The presented reconstruction of human impact is semi-quantitative and cannot be considered as an absolute quantification of human population or percentage of deforestation (see also Broothaerts et al., Reference Broothaerts, Verstraeten, Kasse, Bohncke, Notebaert and Vandenberghe2014c). Such quantitative reconstructions of population densities and deforestation are, however, needed in order to fully understand human–environment interactions in this period, and to discuss in quantitative terms how ‘dark’ this period really was for the Belgian loess belt. To come up with quantitative reconstructions of population densities, the current fragmented archaeologic and historic datasets in the Belgian loess belt should be integrated and analysed in depth (see e.g. Shennan, Reference Shennan2017; Van Lanen et al., Reference Van Lanen, De Kleijn, Gouw-Bouman and Pierik2018). Moreover, more archaeological and historical data are needed to understand the nature of the changing human impact signal in terms of subsistence economy and land use strategies.
The signal of decreasing human impact and increasing forest cover in the Belgian loess belt during the first millennium AD is rather short and took only 200 to 600 years. From 400 AD (Dijle catchment) and 800 AD (Mombeek catchment), human impact increases again and, as a result, forest cover decreases again, with lowest forest cover and highest human impact during the Medieval to Modern period (Fig. 6). This relatively short signal of 200 to 600 years was, however, enough to trigger changes in the floodplains. Once the threshold in the landscape system was crossed, changes in floodplain geo-ecology were triggered rather quickly and floodplains responded to limited changes in the catchment. Defining such threshold values of human impact or vegetation cover in quantitative terms remains problematic. To do so, there is a need for a more detailed temporal framework, more information on the processes involved, such as the hillslope–floodplain connectivity and the sedimentological and hydrological responses to changes in human impact, as well as a more detailed archaeological and historical dataset.
The observation that short-term fluctuations in human impact can have an important impact on the fluvial system and can trigger changes in floodplain geo-ecology has important implications for future floodplain management. Our results show local floodplain managers and restoration projects that rewetting of the floodplain or an increase in carbon storage can be attained even with a limited decline in human impact in the catchment. Exactly how large this decline should be and how future changes in the catchments will trigger changes in floodplain geo-ecology, however, should be quantified with hydrological and sedimentological modelling studies – for which the data presented in this study can be very useful.
Acknowledgements
This research is part of two projects funded by Fonds Wetenschappelijk Onderzoek (Research Foundation Flanders – application G0A6317N and S003017N). The authors thank Neil Quintens, Laurens Vervoort and Katrien Wouters for their assistance during the field campaigns.
Open access
