1 - Introduction

1The construction of a new TGV railway track between Baudrecourt (Moselle) and Vendenheim (Bas-Rhin) in the frame of the LGV Est network is going on since 2010 (fig. 1). Prior to this infrastructure project, which has destructive consequences on the sedimentary and cultural relics, extensive archaeological excavations were carried out on the future railway track by using mechanical means (fig. 1). In addition to the traditional archaeological survey, geoarchaeological and palaeoenvironmental surveys were carried out. Here we focused on the eastern part of the railway track, the Kochersberg area, covered by thick loess deposits. The Pleistocene loess stratigraphy has been extensively studied on the reference site of Achenheim, Bas-Rhin (Wernert, 1938; Heim et al., 1982). However, the morphogenesis of the Holocene and its resulting stratigraphy was never explored. Palynological data are numerous in the Vosges Mountains (Dubois & Hatt, 1930; Kalis et al., 2006), scarce in the upper Rhine Graben (Thévenin & Heim, 1983; Lechner, 2005) and no data exists in the Kochersberg region. Plant macrofossil remains and charcoal, deriving from archaeological features of different excavation sites on the railway track were investigated during post-excavation studies (Durand et al., 2011; Nocus et al., 2011; Wiethold, 2012), but palaeoecological analysis based on natural sediments and peat were still lacking. Therefore, in addition to systematic excavations, four sediment cores were sampled from four different catchments. The most interesting sediment record was found in the small valley Ungerbruchgraben (Mittelhausen, Bas-Rhin; fig. 2 & 3), where a 2 m core was sampled. Pollen and other microfossils, waterlogged plant remains, and sediment analysis were compared to the rich archaeological heritage in and around the catchment. Sites of the Michelsberg culture (Middle Neolithic, ca. 4,200-3,500 BC), the Hallstatt and La Tene period (Iron Age) and from the Middle Ages were very clearly identified in the area. Evidence of a Bronze Age occupation is almost non-existent and signs of the Roman occupation are relatively scarce. Over 6,000 years of human impact at local scale were recorded.

2 - Material and methods

2.1 - The ungerbruchgraben catchment

2The Ungerbruchgraben is a small catchment situated in the eastern part of the Kochersberg region (Bas-Rhin, France). The Kochersberg is a hilly region enclosed by two major faults situated between the Vosges Mountains (west) and the Upper Rhine valley (east). The creek flows from west to east. The catchment of the Ungerbruchgraben depression is 1,000 m long and 700 m wide, the slopes are very smooth (> 5 %) except for the western part of the catchment. The substrate, composed of sedimentary rocks from the Mesozoic, is entirely covered by a thick Quaternary loess deposit. Except from some riparian forest relicts, the entire surface is presently cultivated (ADEUS, 2011). Therefore, and despite smooth slops, soils are poorly developed because of recent intensive erosion (Auzet et al., 1995; Armand, 2009). At the place of the coring (48°41’59’’N, 07°37’49’’E), the water flow is permanent and the creek about one meter large. The floodplain is about 10 m large.

2.2 - Sedimentology

3Twenty seven stratigraphical layers were described (fig. 4). particle-size distribution was measured for each unit by laser granulometry (beckman-coulter, ls230). preliminary, the samples had been treated by h₂o₂ to destroy organic matter. the samples had been also washed successively by kcl, distillated water and sodium hexametaphosphate to deflocculate the aggregates without destructing the carbonate. on one hand, this method (1) provides a particle size distribution close to the original sediment distribution before its transportation and (2) allows a good interpretation of involved geomorphological processes. on the other hand, the interpretation of the mineral fraction distribution of organic rich sediment is more difficult. thus it should be kept in mind that for the very organic rich samples the distribution that appears often as sandy represents only a small part of the bulk sediment. sampling for organic matter (om) content measurement was done continuously with 1.8 cm wide plastic boxes (77 samples). om was measured by loss on ignition at 375 c° during 16 h.

2.3 - Palynology

4Samples were systematically taken every 4 cm and even every 2 cm in the neighbouring of stratigraphical limits. Sediment samples were processed for pollen and non-pollen palynomorph (NPP) analysis using standard techniques (Björck et al., 1978; Moore et al., 1991). They were treated with HCl (10 %), NaOH (10 %), HF (40 %), ZnCl₂ and acetolysis. The conservation of pollen and micro-fossils varies mostly between bad and acceptable. As a consequence, some samples, especially from the bottom part of the core, cannot be considered being suitable for pollen analysis. Pollen grains were identified with the aid of a reference collection of modern pollen types and photographs (Reille, 1992; Beug, 2004). Alnus, Cyperaceae, hygrophilous and aquatic taxa, spores and non-pollen palynomorphs were excluded from the pollen sum. The TILIA-programme was used to draw the pollen diagram (Grimm, 1991). The implemented programme CONISS allowed a stratigraphically-constrained cluster analysis to obtain a zonation of the Pollen diagram in local pollen assemblage zones (Grimm, 1987).

2.4 - Plant macrofossil analysis

5Sampling was done continuously with samples of 8 cm length and with a volume of about 200 ml per sample. A low vertical resolution was necessary in order to collect a sufficient amount of plant remains. The upper sample was only 6 cm high. At important stratigraphical limits, sample limits were adapted (for example: 176-183 cm, 183-188 cm, 248-252 cm). Carbonates were eliminated with HCl (5 %). Samples with high organic content were heated and treated with 5 % potassium hydroxide in order to disaggregate the material and destroy the humic acids. Then, all the samples were watersieved using standard laboratory sieves with 0.315 mm and 1.0 mm mesh sizes. Plant remains were sorted and identified under a microscope (Olympus SZ40) with 7 to 10 X magnification or 40 X for precise observations. General principle for plant remains determinations are given by Jacomet & Kreuz (1999), Marinval (1999) and Birks (1980). For the determination of waterlogged plant remains, the reference collection of modern and archaeological seeds of the archaeobotanical laboratory of Inrap in Metz (Lorraine, France) and the determination books of Beijerinck (1976), Brouwer & Stählin (1955) and Cappers et al. (2006) were used. The scientific taxinomy follows the last edition of the Flora for Belgium, Luxembourg and eastern France (Lambinon & Verloove, 2012), the English names are given according to www.tela-botanica.org.

2.5 - Radiocarbon dating

6After sorting of all plant remains carbonized seeds and wood of terrestrial species were selected for radiocarbon dating. Because only few remains were suitable for dating and only few stratigraphic units could be dated we selected additionally two extra layers on the top and in the middle of the core in order to obtain radiocarbon dates from bulk organic matter (tab. 1). The obtained dates from the bulk samples seems not to be seriously affected by hardwater effect as they fit perfectly in the general chronological framework established by the other three ¹⁴C-datings obtained from seeds, twigs and wood, and the general vegetation history evidenced by pollen analysis. Radiocarbon dates were measured by accelerator mass spectrometry (AMS) in the Poznan Radiocarbon Laboratory in Poland. Calibration was performed using the calib Rev. 6.1 calibration program and the Intcal09 calibration dataset (Reimer et al., 2009). The sediment from the core was composed of thin and relatively contrasted stratigraphic units, thus it is impossible to consider the sedimentation rate as constant. Consequently we did not calculate a time-depth age model based on radiocarbon dates but we are using the results (tab. 1) to give an approximation of the age of the dated stratigraphic units and a relative age of the other units.

2.6 - Archaeological excavations

7Before the construction of the new TGV railway line, archaeological diagnostics were conducted according to the French heritage protection law. Linear trenches, 30 m long and 2.5 m wide, were opened and examined by archaeologists. Altogether, 10 % of the impacted area was archaeologically evaluated. In these test trenches archaeological features and the related ceramics were studied in order to determine the extension of sites and their chronology. Then, based on the results of the test trenches, some sites were selected to be excavated completely. All archaeological evidence including former discoveries are reported to the archaeological map under the authority of the Service Régional d’Archéologie (SRA).

3 - Results and interpretations

3.1 - Sedimentology

8The stratigraphy is strongly dominated by silt and is a bit more clayey in the bottom part (330-238 cm) and sandy in the upper part of the core (202-133 cm). The OM content is generally high in the core but it strongly varies with a minimum at 230 cm (2.3 %) and a maximum at 180 cm (70.6 %). The 27 observed layers can be synthesized into six main stratigraphical units (S1-S6):

9- S1 (330-293 cm): at the bottom (C27, fig. 5) the particles’ size distribution is bimodal. This could indicate both some alluvial and colluvial processes. The sediment becomes progressively more silty (C26, C25), very close to pedogenetized loess material. The OM content increases as well very slightly from the bottom to the top but stays always under 10 %. In this first sequence, slope processes are active but not intense. The slopes are probably partially cleared and might be cultivated.

10- S2 (293-251 cm): the second sequence is still dominated by silt but the amount of sand increases (C22). The sediment is much more organic rich. The OM content presents two peaks at 287 cm (15 %) and 256 cm (39 %). An important part of the OM is not humified and leaves and branches can be recognized. The slope process does not occur during this sequence. The major part of deposit comes alternatively from gentle alluvial processes and organic accumulation in the floodplain. The observed sandy fraction is either composed of quartz grains which are present in scarce proportion in the loess (but roughly sorted and concentrated by alluvial processes) or of secondary carbonates that precipitated in the soils developed on loess.

11- S3 (251-214 cm): the sediment is clay loam, well sorted, typical for slope deposit derived from loess (C21, C19). Clay content decreases from the bottom to the top and OM content (< 7 %) is relatively low on the whole sequence. This sequence is mostly characterized by colluvial processes. The sharp transition with the previous sequence can be interpreted either as a fast environmental change or as a hiatus. Anyway, during this sequence the landscape is much more impacted by humans with an important part of the catchment that is unforested.

12- S4 (214-174 cm): this sequence is the most dominated by OM with a content varying between 20 % and 50 %. The mineral fraction is sandier and roughly sorted (cf. S2). The more organic rich they are, the more they are sandy (C14, C16). At 208 cm (C17), the sequence is interrupted by a thin colluvial deposit. The slopes are very stable during this sequence. Due to the high sensitivity of loess material, we can consider that the catchment is not impacted or very slightly impacted by humans and well covered by vegetation.

13- S5 (174-130 cm): this sequence is less contrasted. The silt is still slightly sandy but the organic content decreases fewer than 20 %. The mineral fraction is not sorted at all. There is no evident dominant process. Both colluvial and alluvial processes are suggested during this phase. Humans seem to be back in the catchment but only low impact is to be recognized. The slope processes are not strong and organic deposit in the floodplain is still visible.

14- S6 (130-0 cm): above the core, the sediment is dominated by fine and well sorted silt derived from loess (N86, N88). It is a massive deposit that originates from intense slope processes that are still currently active. Intensification of soil tillage, increase in field size and other agricultural practices are responsible for the intensification of these processes. Lower rates of organic matter in today’s soils also reduce the cohesion of loess material and make it more sensitive to erosion. Mechanization is most probably the first explanation of this deep sequence. But we cannot exclude that the intensification and thus the sequence started already in the Middle Ages.

3.2 - Palynology

15- Un1 (330-286 cm): the deepest part of the investigated core is characterized by a high proportion of blank samples, in which no pollen was preserved, especially between 322 cm and 286 cm (fig. 6). The three deepest samples with pollen preservation are dominated by Corylus, Ulmus and Tilia. Despite the absence of Quercus, this bottom part of the sequence may be attributed to the middle Holocene, most probably the younger Atlantic chronozone, which is confirmed by the radiocarbon dating (Poz-31237: 5,060 ± 30, calibrated with 2-sigma probability 3,958-3,779 yr cal. BC). At the end of the sequence (Un1b), Alnus, located most probably in the small floodplain of the creek, becomes more and more present. This subphase must be placed at the beginning of the Subboreal.

16In this not very well documented zone Un1 the landscape was completely forested with mixed oak forest dominating, being mainly composed of Quercus (oak), Tilia (lime) and Ulmus (elm). No evidence of human impact is detectable in the pollen diagram.

17- Un2 (286-210 cm): in this second sequence, the conservation of pollen was slightly better but still bad. No pollen was found in this upper part between 226 and 214 cm. A hiatus is clearly detectable between the subzones Un2c and Un2d of the pollen diagram.

18The pollen assemblage of Un2 is clearly different to zone Un1: Alnus is dominant at the beginning, but then progressively decreasing and the forest becomes dominated by Quercus, Pinus, Fagus and Abies. Carpinus and Juglans appear as well in this sequence. The indicators of human impact are well represented, with cereals and other associated plants (Plantago lanceolata, Chenopodiaceae, Polygonum aviculare, Rumex…). Moreover, we note the occurrence of coprophilous fungi spores related to the presence of grazing or ruminants (fig. 6).

19The landscape was still covered by forest in the subzone Un2a. Alder stands (Alnus glutinosa) are developing in the wetland zone of the valley bottom. From zone Un2 onwards the landscape becomes more open and wetland vegetation dominated by sedges (Cyperaceae) is replacing the alder stands. The presence of green algae points to the regular presence of open water. The rising values of monolete spores (spores of undeterminend ferns) are showing a strong coincidence with the occurrence of Glomus-spores, a parasite of tree roots, which is an indicator of erosion phases (van Geel, 2001). Only few cereal pollen grains are recorded, like one can observe frequently for the pre-medieval sequences in pollen diagrams of low altitudes (Laine et al., 2010). Increasing spores of coprophilous fungi at the end of the subzone Un2c are indicating the impact of grazing (van Geel & Aptroot, 2006; Cugny et al., 2010; Gauthier et al., 2010). The parasitic fungi Gaeumannomyces sp. (Hyphopodia) is normally infesting greater tussock-sedge (Carex paniculata) and cyperus sedge (Carex pseudocyperus), two wetland plants which were recorded by archaeobotanical analysis deriving from more or less the same levels of the core (van Geel, 2001). The transition of subzones Un2a to Un2b is dated by a ¹⁴C date obtained using uncarbonized wood (Poz-31236: 2,570 ± 30 BP, calibrated with 2-sigma probability 808-562 yr BC). Zone 2 is mainly covering the Bronze and the Iron Age time period.

20- Un3 (202-174 cm): this zone is characterized by a significant return of trees, mainly Ulmus (elm), Betula (birch), Corylus (hazel) and Alnus (alder). Consequently we can also observe a decline in herbaceous taxa associated with human activities. The return of forest at local scale is also confirmed at the end of the sequence by the occurrence of fungi (mainly Kretzschmaria deusta) that parasitizes broadleaved trees (van Geel 2001; van Geel & Aptroot, 2006). The reforestation is most probably resulting from slightly more dry conditions impacting the wetland at the valley bottom. This is indicated by the presence of HdV-200, a microfossil typical for drying phases of wetlands. A strong decrease of agricultural activities within the catchment of the site and a reforestation of formerly open spaces is very well visible in subzone Un3b, indicated by the disappearing of spores of coprophilous fungi, the decline of Cyperaceae and the already mentioned increase of arboreal pollen. Locally, the catchment seems to be very slightly impacted by human presence. Pastures are still present at a regional scale but cereal pollen is limited to few recording in subzone Un3b. A small peak of Cannabaceae at 194 cm is most probably indicating medieval hemp cultivation or the retting of hemp in the nearby wetlands. Fagus (beech) becomes a major component of the mixed forests at the end of the zone. The reforestation and the strong decline of agricultural indicators observed in zone Un3 are normally related to the 5^th century AD, a period characterized by a general decline of settlement activity and human impact. This interpretation is confirmed by a radiocarbon date obtained from bulk organic matter at 206 cm, just before the beginning of zone Un3. Thus, the hiatus in the pollen diagram between zones Un2 and Un3 is attributed to the Gallo-Roman period (Poz-37676: 1,745 ± 30 BP, calibrated with 2-sigma probability: 228-388 yr AD).

21- Un4 (174-130 cm): from the sample 170 cm onwards, pollen of trees and shrubs decline distinctly while Cerealia-type and other the human impact indicators are strongly increasing. The landscape is thus more open, occupied by fields, meadows and pastures in the catchment of the Ungerbruchgraben. Grazing activities are signalized by increasing values of spores from coprophilous fungi in the first subzones Un4a and Un4b. The important values of monolete spores accompanied by Glomus are indicating increasing erosion effects (van Geel et al., 1989). With the beginning of subzone Un4c, the direct environment of the coring site seems to have been surrounded by cereal fields. The forested areas seem to have been very small and probably distant to the coring site. The variations of vegetation and human impact evidenced by the pollen diagram are reflecting local farming practices, with more or less cultivated land or grassland, according to different periods. Samples at 138 cm and 134 cm appear to be the most marked by a local environment of grasslands and grazed meadows.

3.3 - Plant macrofossil analysis

22The macrofossil diagram presents, in stratigraphical order (deepest sample on the right), the results of the analysis of uncarbonized waterlogged plant remains from the obtained core (tab. 2). The counted plant remains are indicated as minimum numbers of individuals (MNI). Altogether we identified 1128 waterlogged plant remains and 15 valves of Cladocera, deriving from 24 samples, representing 40 taxonomical units. The presentation of the results follows the stratigraphical zones established by sedimentology, additionally the chronostratigraphy of the pollen diagram is indicated. The results of the analysis of plant macrofossils are reflecting exclusively the local vegetation development in the direct vicinity of the coring site. Nevertheless, some remain may have also been transported by flooding or by colluvional processes.

23- S1 (330-293 cm): this sequence has delivered a lot of calcareous concretions, but nearly no plant remains, except from few hard-coated seeds like elder (Sambucus ebulus and Sambucus nigra). Elderberry seeds could come from local environment or could be transported by zoochory or hydrochory. We note as well in these four samples, the occurrence of undetermined freshwater malacofauna species, rootlets, and rhizomes. Due to a global poor conservation of plant remains, it is difficult to conclude on the local vegetal environment for this sequence.

24- S2 (293-252 cm): the beginning of this sequence contains as well only few plant remains. Calcareous concretions in the shape of rootlets and malacofaunal remains are common and mosses are present between 284 cm and 293 cm. We regularly recorded seeds from dwarf elder Sambucus ebulus which is a heliophilous, ruderal species, mainly occurring on calcareous soils. Above 284 cm, the nutlets from Ranunculus repens are recorded and, more discretely, pips from blackberry (Rubus fruticosus agg.), nutlets from Carex flava and Carex cf. vesicaria. The few remains indicate a partially open environment which was regularly flooded.

25- S3 (252-204 cm): the plant remains increase in quantity and taxa diversity. We notice, in addition to blackberry and sedges Carex sp., pips from raspberry (Rubus idaeus), and hygrophilous herbs like Ranunculus sceleratus and Mentha aquatica/arvensis. One Chara-oogonium indicates the occurrence of open and most probably stagnant water. Nutlets from Ranunculus repens and Ranunculus acris and fruits from Carex flacca become very common at the end of the sequence. They witness wetlands and wet meadows, influenced by pastoral activities. No remains from trees were found. This is in accordance with pollen analysis indicating an increase of Poaceae and other human impact indicators.

26- S4a (204-188 cm): the tendency of sequence S3 is continued. The presence of Carex fruits (Carex flacca, Carex flava-group) suggests wet meadows and swampy vegetation accompanying the creek at the valley bottom. Fruits from Betula and Acer campestris are recognized for the first time and mosses are abundant. The floodplain is still open.

27- S4b (188-168 cm): the bottom of the valley is colonized by wet alder forest as attested by abundant wood, twigs, buds and leaves remnants as well as male and female catkins of Alnus glutinosa. These results are in accordance with pollen analysis. This vegetation type favored the accumulation of organic matter in the floodplain (cf. § 3.1). Other remains are fruits, rootlets and rhizomes from different Carex-species (Carex paniculata), which are typical for the phytosociological association of Magnocaricion, found in and nearby partially open alder stands. Woody nightshade Solanum dulcamara and gypsywort Lycopus europaeus are eutrophic plants, growing at swampy margins of the Ungerbruchgraben creek.

28- S5 (130-168 cm): remains from alder are disappearing completely, except from the upper sample (160-168 cm). Carex-fruits, seeds of Juncus articulatus and fruits of Scirpus sylvaticus become abundant, indicating open swampy areas and wet pastures. This is probably the consequence of forest clearing by Man.

3.4 - Human settlements in the catchment and in the surroundings

29At first, we examined the upstream watershed of Ungerbruchgraben, which can be regarded as the immediate environment of the core sampling site E86. Indeed, the accumulated sediments in the valley floor are originating exclusively from this sector. Four distinct settlements are listed.

30- A site attributed to the Neolithic Michelsberg culture (4,200-3,500 BC), giving evidence of storage features (silos). This site was already identified during archaeological prospection by test trenches (Jodry et al., 2008b). The site is situated about 500 m west-northwest of the coring site (fig. 3: 10.1). It expands on an almost flat area on the left bank of the Ungerbruchgraben.

31- On both sides of the valley floor (fig. 3: 10.2 & 10.3), within 100 m of the core drilling site, silos dating to the Iron Age were identified during archaeological evaluation by test trenches (Jodry et al., 2008b). There are only few indications concerning the dating of the site on the left of the valley but the silo pits on the right bank are attributed to the Hallstatt period (800-480 BC). A forest parcel, separating the two sites, has not been evaluated by archaeological test trenches. Therefore, it is not possible to judge the continuity of these two sides.

32- On the upper part of the catchment (fig. 3: 9.5), a settlement of the early La Tène period (5^th century BC) were found (storage pits “silos” and one human deposit in storage pit). This site has revealed also some evidence of a Bronze Age occupation (2^nd-3^rd millennia AD => end of 2^nd millennia BC) and of the Gallo-Roman period (Thomas & Chenal, 2008; Thomas et al., 2011).

33- Merovingian burials (400-700 AD) are mentioned by the archaeological information system at a place called Ueberjohn (Stieber, 1961), which is situated at the interfluve between Ungerbruchgraben and Vierbruckgraben (fig. 3: 2).

34If one considers a larger radius of several kilometers, the vegetation development and the human impact observed in the valley are confirmed. The Michelsberg civilization is represented by a major site of burials situated in silo pits at Gougenheim. This civilization is generally very largely represented in the eastern part of the Kochersberg region (Lefranc, 2007). The Bandkeramic period (Linear Pottery Culture, early Neolithic, 5,200-5,000 BC) is also present in the area, but remains are less abundant. The Bronze Age occupation was never well established at the Kochersberg area. The settlement indicators are often secondary or much degraded by erosion (Thomas & Chenal, 2008; Thomas et al., 2008, 2011; Michler et al., 2011). From the early Iron Age, we are witnessing a remarkable increase of the number of sites (Thomas et al., 2010). Besides the indicators present in the watershed, many indicators have been found on the track of the LGV (over ten distinct archaeological sites), indicating a high density of settlements in this micro-region. The Gallo-Roman period is also represented by several settlements (Schneikert et al., 2008). However, they appear in a wider area. The nearest indicator is a building of modest size, located south of Olwisheim (Bas-Rhin) three kilometers east of the coring (Jodry et al., 2008a). Finally, concerning the Middle Ages, it is especially the Merovingian period which is well represented at or near three neighboring villages (Berstett, Bilwisheim and Gimbrett, Bas-Rhin).

35In summary, we have evidence of three major settlement phases in the vicinity of the site: the first attributed to the middle Neolithic (Michelsberg Culture), the second to the older Iron Age (Hallstatt and early La Tene period) and the last to the early Middle Ages. The Gallo-Roman period is well represented, but not within the direct site catchment of the Ungerbruchgraben.

36

37

4 - Discussion

38The different palaeoenvironmental proxies are showing no major contradictions (fig. 7). The three palaeoenvironmental methods are complementary since they produce data at various spatial scales. Waterlogged plant remains reflect the local environment of the floodplain, sediments are correlated with the watershed land use and pollen analysis tells us about the vegetation in the watershed and generally in the Kochersberg region.

39- S1: this first sequence dated from Late Neolithic (3,960-3,770 yr cal. BC) delivered very few information about vegetation because the colluvial silts were not suitable to allow the preservation of pollen and macrofossils. The moderate slope deposits suggest a partially open landscape under a moderate human impact. Farmers of the Michelsberg culture (4,200-3,500 BC), could be responsible for this colluvial signal.

40- S2: according to the radiocarbon date obtained at 260-268 cm (808-562 yr cal. BC), the second sequence represents the younger Bronze Age (Bronze final). It reflects a very low human impact, whatever proxy we consider. This is consistent with the lack of archaeological sites and remains of the younger Bronze Age. However, from the palynological point of view, a major hiatus separates this sequence from the previous one. Therefore S2 could represent only a part of the younger Bronze Age.

41- S3: the beginning of the sequence is dated to the early Iron Age (808-562 yr cal. BC). Several remains from this period were found inside the watershed. In the Kochersberg region, we witness an explosion of archaeological remains during the latest phase of the early Iron Age and the beginning of the La Tène period (Thomas et al., 2010). Pollen and waterlogged plant remains confirm the development of agriculture. The forest cover was considerably reduced to gain agricultural surfaces. The sediments are strongly dominated by colluvial silts that resulting from agricultural erosion. Erosion processes might be amplified by climatic deterioration (van Geel & Berglund, 2000; van Geel & Magny, 2002). The second half of the sequence shows a decline of human impact. The watershed is still partially open but the slope erosion processes decrease and the human impact indicators are scarcer. It matches with the Roman period when settlements were present in the Kochersberg area, but not in the direct surrounding of the Ungerbruchgraben catchment.

42- S4: the beginning of the sequence is attributed by radiocarbon dating to the 3^rd and 4^th century AD, the Roman Empire (228-388 yr cal. AD). In the Kochersberg area, a reforestation can be observed while agriculture collapses. The Ungerbruchgraben catchment illustrates perfectly this tendency. Its floodplain becomes a swampy alder carr and slope erosion stops.

43- S5: in the sixth and in the first half of the seventh century (537-644 yr cal. AD) starts at the Ungerbruchgraben a new cycle of human impact. At first, slope erosion is still scarce. In the literature from other regions, the return of human impact is to be observed in the Carolingian period of the early Middle Age, around the eighth century AD (e.g. in the Jura moutains: Gauthier, 2004, 2006). Other geographical areas witness this return of human influence a little later (e.g. in the plains of Burgundy: Laine et al., 2010).

44- S6: this thick colluvial sequence (130-0 cm) is usually explained as result of intensive agriculture in the loess area from the high Middle Ages onwards to modern times. From this core (no macrofossils data in this upper part because uncarbonized remains were not preserved), it is difficult to evaluate the contribution of intense farming to sedimentary processes during the high and late Middles Ages. The last radiocarbon dates we obtained is from a level at the end of S5 and points to the late 7^th, the 8^th and the first part of the 9^th century (664-856 yr cal. AD). Therefore the stratigraphical unit S6, the upper part of the core, must represent the slope erosion and colluvional processes from the end of the early Middle Ages to modern times.

5 - Conclusion

45Organic rich deposits at the valley bottom of small catchments in the hilly agricultural regions west of Strasbourg were unknown and unexplored before the beginning of the LGV Est excavations. Palaeoecological data are very scarce in the Upper Rhine valley and completely missing in the Kochersberg Region. Thus, the stratigraphy of silty and peaty sediments studied in the Ungerbruchgraben catchment is an important step to gain knowledge of the evolution of the Holocene environments in the Alsace region. At the scale of the Kochersberg region environmental data show some important relations of erosion and colluvional processes with archaeological sites and findings and more or less intense agricultural activities. Inside the Ungerbruchgraben catchment the sedimentation dynamics are directly related to Iron Age settlements and the intense land use from the beginning of the early Middle Ages up to modern times. Similar valley bottoms and catchments in other parts of the Upper Rhine valley should be investigated by interdisciplinairy studies in near future to gain more insight into the complex interactions between Man and his environment.