Loess stratigraphy in Dutch and Belgian Limburg

Purpose of this paper is constructing a loessstratigraphical framework for a small region along the Albert Canal in Dutch and Belgian Limburg. Field ob­ servation and heavy mineral analysis are central focus of this study. The large-scale sediment trap environment, at scarpe foots or between river-terraces led to the conservation of thick and differentiated loess sequences. The exposed walls of the widened Albert Canal, loess sections in the adjacent loess-exploitation zone and excavations at four major archeological sites did provide important data. By means of characteristic lithological and paleopedological "marker" horizons, mineralogical and chrono­ logical data and correlation with loess sections in Ger­ many, a reliable loess-stratigraphical framework has been constructed. Beginning with the Holsreinian, it is match­ ing marine/ice isotope records and terrestrial lacustrine pollen records. It contains four glacial loess cycles with five intercalated interglacial paleosols on top of the Rothem river terrace of the Maas. [Lössstratigraphie in Niederländisch und Belgisch Limburg] Kurzfassung: Ziel dieser Veröffentlichung ist es, für einen Abschnitt entlang des Albert Kanals in Nieder­ ländisch und Belgisch Limburg eine Ubersicht zur Löss­ stratigraphie zu erstellen. Die Arbeit gründet sich vor allem auf Feldaufnahmen und Schwermineralanalysen. An Talhängen und zwischen Flussterrassen kam es zur Sedimentation und Erhaltung mächtiger, reich geglie­ derter Lössablagerungen. Die Wände des Albert Kanals, Lössaufschlüsse im an­ grenzenden Gebiet und vier große archäologische Aus­ grabungen wurden untersucht. Auf Grund von cha­ rakteristischen lithologischen und paläopedologischen Leithorizonten, mineralogischen und chronologischen Daten und deren Korrelation mit Lössaufschlüssen in Hessen wurde eine lössstratigraphische Übersicht für die Zeit seit dem Holstein-Interglazial erstellt, welche sich mit Ozeanund Eisisotopenabfolgen sowie mit *)Anschrift des Verfassers: Drs. E. P. M . MEIJS, Archeogeolab, Veulenerbank 3 3 , N L 6 2 1 3 JR Maastricht, The Netherlands. fesdändischen Pollenzonen korrelieren lassen. Es las­ sen sich vier glaziale Lösszyklen und fünf interglaziale Paläoböden auf der Rothem Terrasse der Maas ausma­ chen.


Introduction
Stratigraphical frameworks for loesses have had to be revised frequendy over the past 40 years (see HAESAERTS et al. 1981 andJUVIGNE et al. 1996).Because of the similarity of the sediments, sedimentological and cryogenic structures and pale osols originating from the different warm and cold stages of the Upper Pleistocene, few convincing stratigraphical "marker"-horizons can be detect ed.This is especially the case with regards to the pre-Weichselian loess-sediments, which until now, have been difficult to subdivide stratigraphically.In volcanic regions tephra layers can provide ex tensive lithostratigraphical markers.They often show a distinct mineral suite and are direcdy or indirecdy dated by means of absolute dating meth ods.
Apart from 14 C -dating on humic sediments, char coal or molluscs, different forms of TL-dating have been developed over the past decades.How ever, for both of the last-mentioned dating meth ods many methodological problems still exist, es pecially in loesses, leading to different chronostratigraphical interpretations.Since 1945 several researchers demonstrated the importance of heavy mineral analysis for litho stratigraphical studies on loess sediments.Over the past years intensified investigations have taken place in new and enlarged quarries in the study area.The results of heavy mineral analysis, in combination with lithological, paleopedological and chronological data from the extreme long loess sequences along the widened Albert Canal and in adjacent loess-pits, give new insight into the loess stratigraphy of the study area.

Research area
The research area is shown in Fig. 1.It is located in a part of the drainage basin of the river Maas.Since the middle Pliocene (4 ma) the Maas met the Rhine near Aachen (Germany).Afterwards the Maas moved westwards and started around 800 ka to erode the Cretaceous and Paleocene lime stones and Oligocene sands in the study area (van den Berg 1996).The continuous uplift of the area, the alternation of glacial and interglacial periods and the lateral westward shifting of the Maas were responsible for the preservation of an extraordi nary extended and complete river terrace sequence.Transition periods between interglacial and glacial times were favourable for river-erosion and inci sion, while gravel-accumulation took place under cold and severe conditions, on the base of widely observed cryogenic features within the gravel beds ' (van den Berg 1996).Interglacial regimes favoured the formation of river floodloams and lateral ero sion.After the retreatment of the Maas from the research area, the preserved river terraces were al ternately eroded by wind-and water-action and covered by eolian sands and loesses.The bounda ry between the coversand area in the north and the loess belt in the south is located near Gellik (Fig. 1).The present landscape is smoothly undu lating, with SW-NE orientated valleys, draining to the rivers Jeker and Maas in the east.Several brickyards are exploiting loess-pits near Kesselt and Veldwezelt (Nelissen pit "Kesselt" and Vandersanden pits "Veldwezelt-Hezerwater" and ^Eindhoven ^--v 777e NETHERLARY. BELGIUM 1 "Op de Sclians").In fact the area along the Albert Canal is an important loess-exploitation zone which will be further investigated in the future.The Nelissen pit "Kesselt" was, until 1996, the type-location for the so-called Hesbaye loess, which induced research by several universities.In the future this site will possibly be assigned as the type-location for the rare Upper Saalian "B-loess" (see chapter chronostratigraphy).The Vandersanden company started exploiting the loess-pit "Op de Schans" in March 2000 (Fig. 1).For the study of pre-Weichselian loesses, it is a promising loca tion because it is situated in line with the most complete and differentiated pre-Weichselian out crop along the widened Albert Canal (Fig. 2).

S/ Dusseft
Between 1975 and 1985 the Albert Canal was wid ened and detailed research of the exposed walls led to a thorough understanding of the geological and geomorphological genesis of this area.Ex posed sediments in neighbouring loess-pits were then correlated to the Albert Canal sequences.The discovery of four major archeological sites (Maastricht-Belvedere, Kanne, Kesselt and Veldwezelt-Hezerwater) in the research area led to ad ditional stratigraphical and chronological research.
In conclusion, the research area could be consid ered as a key-region for present and future loessstratigraphical research.

Terrace sequence of the river Maas
In Fig. 2  Four distinct river terrace sequences can be distinguished.According to gravel-base heights the Rothem terrace-level seems to be divided into two sublevels; representing an outward sliding meandercurve-complex (Fig. 1).In a wider con text, Felder & Bosch (1989) deduced from the presence of different gravel-bases and two distinct floodloam generations, the existence of two Rothem sub-levels: Rothem-1 and Rothem-2.As well for the Gravenvoeren level, a distinction be tween Gravenvoeren 1 and 2 seems justified, al though section F-H in Fig. 2 is the only example of a terrace-division in the Gravenvoeren sequence throughout the Maas terrace area between Liege in Belgium and Roermond in The Netherlands (Felder &Bosch 1989 andvan den Berg 1996).
According to Felder & Bosch (1989) the Pietersberg and Maarland terrace levels are present over a distance of 100 meters, in a SW direction from point A and in a NE direction from point K seen in the cross-section Lafelt-Oud Caberg, with gravel bases at about 98 m ASL and 38 m ASL (Fig. 1, Fig. 2).
It is clear that during long periods a sediment-trap existed between the heights of the Pietersberg and Gravenvoeren terraces, which resulted in an ex tremely thick loess cover in between.

Heavy mineral analysis
As stated above, lithological, sedimentological and paleopedological analyses of loesses are rarely able to discriminate between the different warm and cold stages of the Upper Pleistocene.However, correlations between different loess sections can be strengthened and stratigraphical conclusions can be drawn by means of heavy mineral analysis.
In combination with micromorphological and lithological research, the content of the green amphibole and volcanic minerals are used to cor relate individual loess sections and to construct a loess-stratigraphical framework for the research area.

Loess minerals:
The investigation of the mineralogical composi tion of the loess fraction (30-63 gm) in relation to the loess stratigraphy began in The Netherlands ( van Doormael 1945).This research was continu ed by Gullentops (1954), Lautridou (1968), Juvtgne (1978,1985), Balescu & Haesaerts (1984), Meijs (1985), Janus (1988) and Henze (1998) in the loess-belt of France, Belgium, The Nether lands and Germany.Juvigne (1978) proposed the ratio between "green amphibole + garnet / zir con + rutile" to distinguish Weichselian loess from pre-Weichselian loess.This approach was adopt ed in Germany by Henze (1998).Both Gullen tops (1954) and Meijs (1985) mentioned the im portance of green amphibole as a discriminating heavy mineral in stratigraphical research.Accord ing to Balescu et al. (1986) the West-European loess-belt is divided into separate regions with specific mineralogical properties.This implies that green amphibole percentages may differ slighdy between these regions.The overall tendencies how- ever are the same.Even in northern Germany drill ing data from HENNINGSEN & STEPHAN (2000) showed that the amount of green amphibole in creases from the Elsterian to the younger Saalian and Weichselian sediment series.
Requivenient for this kind of stratigraphical re search however, is that the samples should be of "real" eolian loess, as loess may have been redeposited and may be contaminated with older un derlying material, mixing the green amphibole con tent.
Wet sieving of chemically untreated samples was carried out to obtain the fraction 30-63 mm.The heavy and light minerals were separated in bromoform by using a centrifuge and the heavy min eral fraction was sucked up using a micropipet.
From each sample about 300 transparent heavy minerals were counted, using the ribbon-count ing method as described by VAN HARTEN (1965).
The percentage of green amphibole in the frac tion 30-63 gm was calculated based on the total amount of all counted transparent heavy miner als.
Volcanic minerals: Until now only three volcanic tephra layers have been found in the research area, all originating in the volcanic Eifel in Germany.Minerals from the Laacher See tephra have been detected on the archeological site of Kanne by PAULISSEN et al. (1981), just on top of the Holocene soil, that was burried by colluvium.The Laacher See tephra, 12,250 calibrated years BP, occurs in more than 50 pollen diagrams (Germany, Belgium, France and Italy) in the final phase of the Allerod oscillation (JUVIGNE 1993).
The Eltville tephra is the only tephra-layer visible by eye.It lies just beneath the Nagelbeek tongued horizon (NB) and is dated by 14 C and TL methods between 16 and 30 ka (JUVIGNE 1993).
Like the Laacher See tephra, the Rocourt tephra is only detectable by using microscopical equipment and occurs in secondary position in colluviated humic material, just above the Rocourt pedocomplex of the last interglacial (GULLENTOPS 1954, JUVIGNE 1977, MEES & MEIJS 1984).The estimat ed age on the base of indirect 14 C and 230 Th-234 Udatings lies between 61.5 and 106 ka (JUVIGNE 1993).
Stratigraphically the Eltville tephra is less interest ing than the Rocourt tephra because the Nagelbeek tongued horizon (NB), which lies in the same stratigraphical position as the Eltville tephra, is a very characteristic and almost always present mark er horizon from the French coast to eastern Eu rope (HAESAERTS et al. 1981, MEIJS et al. 1983).
For the mineralogical research of tephra-material, the sediment-fractions 63-106 mm (Laacher See tephra) and 106-300 mm (Rocourt tephra) were used.The further preparation and counting is sim ilar to the above-mentioned methods used for the green amphibole analysis.

Description and results
In this chapter four major litho-pedological units are described separately; namely "A-loess", "Bloess", "C-loess" and "D-loess" (Fig. 3).From each unit samples were taken for heavy mineral and/or micromorphological analysis.The results are pre sented and discussed beneath.Loess cycle A (max. thickness at location "Veldwezelt-Hezerwater": 24 meter) In lower topographical positions and in shallow paleovalleys this sequence often starts with the presence of white-greyish (10YR5/4) pale sedi ment, followed by darkbrown (10YR4/3) humic material, most often directly covering the redbrown coloured (10YR5/8) Rocourt pedocomplex of the last interglacial.The material predominantly con sists of redeposited loess-sediment, from the Ro court pedocomplex (JUVIGNE 1978, VANDEN BERGHE et al. 1998).The base of both horizons, especially that of the humic horizon, is of erosive origin (Fig. 5 bottom).The lower part is heavily bioturbated, leading to the formation of a mixed zone between the humic material and the pale se diment with numerous krotovinas.The humic ho rizon is often divided into three sublevels (MEES & MEIJS 1984, HAESAERTS et al. 1999); the bottom and top consisting of dark brown material and the middle part of lightbrown material that may laterally merge in a downslope direction into a grey ish pale horizon (Fig. 5 bottom).Just on top of this humic horizon a yellowish loess is sometimes present ("marker-loess"), representing the first eolian loess-supply since the last interglacial peri od.
Just after the outburst of the Eltville tephra, the climate was extremely cold and dry.As a result, a polar desert pavement (PL: patina layer) formed, discordantly lying upon the older deposits with windblown gravels at its base.The term "patina layer" has been chosen because of the clearly visi ble wind varnish (patina) present on the recov ered paleolithic artefacts from this horizon.The wind erosion created a totally new landscape, dis regarding the underlying valleys and ridges (Fig. 4  and 6).In fact it formed more or less the present landscape, because only a blanket of pure eolian loess was deposited on top of it afterwards, with no erosion phases in between (Brabant loess).The most important soil-formation phase which took place after the formation of the desert pavement resulted in the formation of the Nagelbeek tongued horizon (NB), lying just on top of the desert pavement and showing characteristic cryoturbation features at its base (HAESAERTS et al. 1981).
In general the underlying Eltville tephra is eroded during this wind erosion phase.Only in excep tional topographical positions, such as sedimenttraps, the tephra is preserved.
The present study and earlier studies ( VAN DOOR-MAEL 1945, GULLENTOPS 1954, LAUTRIDOU 1968, JUVIGNE 1978, BALESCU & HAESAERTS 1984, MEIJS 1985, JANUS 1988and HENZE 1998) reveal that the green amphibole percentages in Weichselian loess deposits range from 8 to 40%.In the study area the mean percentage for the loesses deposited, just above the humic horizon, is around 10%.Above these first loesses the percentage quickly reaches the mean values for the "A-loess" in this region, around 20% (Fig. 5 (1996) were able to detect the pres ence of volcanic glass shards in the heavy mineral section.
Recent mineralogical research reveals the presence of the Rocourt-tephra at three locations in the research area (Fig. 3, 5 and 6).Concentration val ues can decline dramatically in downslope as well in upslope directions in the same stratigraphical position (Fig. 5 bottom; see also Juvigne 1977).Fig. 5 shows that in extreme cases percentages are tending towards zero.The weakly developed hu mic horizon on the northern side of this section does not show any volcanic minerals at all.Loess cycle B (max. thickness at location "Kes selt": 12 meter) Thick loess sequences situated beneath the Rocourt pedocomplex are very rare.The relative abundance of these deposits in the research area is explained by the presence of geomorphologically defined se diment-traps between the different river terraces of the Maas.The most impressive "B-loess"-sequence is present in the Nelissen pit "Kesselt" which has been issue of smdying by several universities dur ing the past 50 years (Fig. 6).Here, even calcareous "B-loess" is found which has not been influenced by the soil-formation processes of the last intergla cial.In the Nelissen pit, the so-called "B-loess" se quence has been interpreted as middle Weichselian from the first description of the section (Gullen tops 1954) until 1996.In 1996, the uncovering of the Rocourt pedocomplex and the presence of the Rocourt tephra on top of it allowed Juvigne et al. (1996) to put the "B-loess" sequence into the time before last glaciation (see also chapter chronostratigraphy).
The "B-loess" sequence starts with a polar desert pavement, which in general is less pronounced than that of the "A-loess" and shows more features of water activity.Locally, the influence of running water is distinct, showing large and shallow gully systems.This wind and water erosion phase is re sponsable for an important hiatus in the different loess sections.The sequence on top of the desert pavement is representing an alternation of loesssedimentation and five tundrasol-formation phases (N1-N5).Fossil ice-wedge casts are present at three different levels, starting at about 10-20 cm above the paleo-surface of the middle three tundrasols (Mf.es & Meijs 1984).The greyish (10YR5/3) tun drasols, especially the lower two, are well devel oped and have a characteristic appearance some times showing a whitish to greenish (10YR7/1) thin layer on top and a strong orange-like, rusty (10YR5/8) horizon at the base.They significantly differ from the over-and underlying tundrasols, present in the "A-loess" and "C-loess".The dis tance between the lower two tundrasols (Nl and N2) varies from 2 to 4 meters and occasionally a faint weak developed tundrasol is present between them (see also Mees &Meijs 1984 andHuijzer 1993).
An analogue loess sequence, situated below an Eemian paleosol and containing up to six tundra sols, has been reported by Bibus (1974Bibus ( , 1980) ) at several locations in Germany and is indicated as the "Bruchköbeler Nassboden-Komplex".Count ing from below, the first and the third tundrasol have the most pregnant appearance.In compari son with the situation in the Nelissen pit "Kes selt" the similarity is striking, considering the fact that the second German tundrasol matches the occasionally appearing tundrasol of Kesselt (be tween the Nl and N2).
On top of this sequence the Rocourt pedocom plex is present at times, especially in sediment-traps like paleovalleys and scarp foots (Fig. 3, 5 and 6).
Here periods of colluviation alternated with sub sequent soil-formation finally produced a complex soil-profile showing several generations of clayilluviation, leaching and polygonal developement (Haesaerts & Van Vliet-Lanoe 1981, Mestdagh & Haesaerts 1998, Vanmontfort et al. 1998, Schirmer 2000and personal communication MOcher 1999).The issue of chronostratigraphical correlation with MIS-3 instead of MIS-5 will be discussed in the chapter dealing with chronostratigraphy.
The green amphibole percentages in the "B-loess" range from about 20% at its base to around 5% in the loess-sediments deposited at the end of this period, just prior to the onset of the last intergla cial.In general the "B-loess" is heavily truncated.
In such a situation only the lower amphibole-rich part is preserved from erosion and the overlying tundrasols and Rocourt pedocomplex are missing (Fig. 4).In that case distinction between "A-loess" and "B-loess" by means of their green amphibole content is not possible.Until 1984 loess sequenc es with high green amphibole percentages were assigned to the "A-loess".In the mid-eighties how ever Balescu & Haesaerts (1984) and Juvigne (1985) found in northern France and eastern Bel gium an interglacial paleosol between loess layers with a high green amphibole content and it lasted until 1996 that Juvigne et al. (1996) discovered the same situation for the sequence in the Nelis sen pit "Kesselt" (see Fig. 6).Taking into account these new insights, the stratigraphical framework had to be revised and a great part of the so-called Hesbaye loess ("A-loess") was interpretated as "Bloess" (Juvigne et al. 1996).
The consequent decrease of the green amphibole content from the older to the younger layers is striking (Fig. 6).The decrease of these percentag es towards the end of the "B-loess" period has also been reported in northern France and east ern Belgium by Balescu & Haesaerts (1984) and Juvigne (1985).Juvigne (1985) suggests that the chemical weathering of green amphibole could provide an explanation.However, as shown in Fig. 6 green amphibole percentages already start to diminish within the calcareaous loess, which has not been influenced by chemical weathering proc esses at all.Thin section, grain-size and sedimentological analyses all point to an upward coarsen ing, a decrease in mineral sorting, a diminishing soil-stability and an increasing stream power re gime in an upward direction within the "B-loess"sequence (Huijzer 1993, Vandenberghe et al. 1998).This implies that the climate was changing from dry towards more moist conditions.In ac cordance to the study area (Fig. 6) Bibus (1980) reported wide spread overlandflow activity in the above mentioned German loess sections, just above the sixth Bruchköbeler tundrasol, before the onset of the last interglacial.Thus climate change seems a more plausible explanation for decreas ing amphibole percentages than chemical weath ering.
As proposed by Juvigne (1978), the most proba ble provenance area for green amphibole is the outwash zone of the Scandinavian icecap.Possi bly dry climatic conditions were a dominant fac-  tor in the transport of green amphibole to the research area.During moist to wet climatic condi tions a long-lasting snow-cover probably prevent ed large-scale deflation from the braided outwash plains.
The uniformity of the green amphibole percent ages in the upper part of the "B-loess" and within the Rocourt pedocomplex (4-6%) in the different sections is striking (Fig. 3,4,5 and 6).This may mean that at the end of the "B-loess" period an uniform loess blanket was deposited, on which colluvial and soil-formation processes took place later on.
Loess cycle C (max.thickness at location "Albert Canal-west Kesselt": 7 meter) Sediments of the "C-loess" sequence are encoun tered at two locations in the research area (near Kesselt; see Fig. 4 and 6).
The sequence starts with a characteristic clayey loess (CL).In most cases a weak greyish tundrasol is present at the base as well as on top of this clayey loess.Above, laminated and heavily cryoturbated loess-sediment occurs, which is overlain by more homogeneous loess with another tundra sol transferring upwards into laminated and geliflucted sediment (Fig. 4).The sequence ends with the formation of a red-brownish (8.5YR5/6) inter glacial soil and overlying pale and humic horizons.Thin section analysis clearly shows the presence of in situ clay-cutans within the interglacial pale osol (MEES 1984).Micromorphological research also revealed that the overlying pale and humic horizons consist of eroded and colluviated mate rial with a lot of transported and rounded ferric nodules derived from the interglacial paleosol be low (MEES 1984).
According to the mineralogical results presented in Fig. 3 and 4 and on the base of lithological and paleopedological characteristics of the over-and underlying sediments, this interglacial paleosol can no longer be correlated with the Rocourt pedo complex.We therefore propose to name this pale osol after the neighbouring village of Hees in Bel gian Limbourg; the Hees soil.
The green amphibole percentages in this sequence are rather constant, ranging from 1-3%.In the Nelissen pit "Kesselt" (Fig. 6) it seems that the percentages here also decrease upthrough the stratigraphy.It may be possible that, given the more laminated and sandy character of the uppermost sediments, this is due to the mixing with older sed iment-material from below.
Mineralogical research of the Hees soil and over lying pale and humic horizons in cross-section Albert Canal-west Kesselt (Fig. 4) did not reveal the presence of volcanic minerals.
Loess cycle D (max.thickness at location "Albert Canal-east Op de Schans": 8 meter) In the research area this sequence represents the first eolian cycle after the formation of the wide spread and massive (sandy) floodloams belonging to the Rothem terrace-level of the river Maas.These sediments have been recognized at two lo cations along the widened Albert Canal (near Kes selt and Vroenhoven; Fig. 4 and 5).
The sequence starts with two different sediment fades.In the beginning, formation of broad, shal low gullys took place, filled up by redeposited sandy floodloam material.
The large number of charcoal pieces present in the pale horizon is striking.No micromorphologi cal indications were found for an existing erosion hiatus between the Montenaken soil and the over lying pale and humic horizons.However, glauconite grains were found within the Montenaken soil, but not in the sediments above (MEES 1984).
The percentages of green amphibole are very low (0.5-1%) in this sequence and are difficult to dis criminate between the sediments from below (Fig. 3 and 4).The Montenaken soil and the overlying pale and humic horizons, were checked for the presence of volcanic minerals.Here also the result was neg ative.

Rothem river terrace
As mentioned above the Rothem terrace is mould ed in the study area as an outward bending mean der-curve, with the deepest incision depth at the outerside of the curve (Fig. 1 and 2).During the process of lateral and vertical river erosion huge quantities of Oligocene sand were eroded and redeposited over short distances.The great sediment load probably lead to the accumulation of an im portant sequence of alternating clayey and sandy floodloams.
This sequence ended with the formation of an intense red-coloured (5YR5/8) interglacial pale osol, the so-called Pottenberg soil (FELDER & BOSCH 1988).In some sections of the widened Albert-canal another redbrown coloured luvisol is present beneath the Pottenberg soil, with inter calated floodloam in between (Fig. 2, 4).Probably local river activity of the "Caberg-1" Maas gave rise to a thin floodloam cover on top of this pale osol.It is representing the equivalent of the warm period between the formation of the Rothem-2 and Caberg-1 terraces.The luvisol is assigned to the neighbouring Dousberg and will be named Dousberg soil (Fig. 4).
The earlier results of BUSTAMENTE SANTA CRUZ (1973), which show negligible green amphibole percentages for the Maas terrace sequences, can be supported by the present study, where the per centages range from 0-0.5% (Fig. 4, 6).

Chronostratigraphy
By means of characteristic "marker" horizons, in combination with mineralogical and chronologi cal data and in correlation with the "Bruchköbel er" loess sections in Germany (Bibusl974, 1980) it is attempted to construct an extended chronostratigraphical framework for the research area, which is related to the river terrace sequence of the Maas, the standard Marine Isotope Stages (MIS) of the astronomical calibrated ocean records ODP-677 and MD900963 (BASSINOT et al. 1994), the dust and temperature successions in the Vostok ice core (PETIT et al. 1990(PETIT et al. ,1999) ) and the long continental pollen records from Massif Central, France (REILLE et al. 1998) (Fig. 7).
In this respect the Vostok dust record is of major importance, indicating continental aridity, dust mobilization and dust transport were most preva lent in the full glacial stages.This is also reflected by many other dust records from for instance Greenland (GRIP MEMBERS 1993), Africa (DEMEN-OCAL 1995) and China (VANDENBERGHE et al. 1997).
It is obvious that the extreme cold oscillations at the end of interglacial stages were too short to generate large-scale loess deflation (see also SCHIRMER 1999).Both the ice core sequence (PETrr et al. 1999) as well as the terrestrial pollen records (TZEDAKIS et al. 1997, REILLE et al. 1998) confirm the presence of four major climatic cycles during the past 400,000 years.Each climatic cycle is start ing with an extreme warm and humid stage (REILLE et al. 1998;MIS 11.3, 9.5, 7.5 and 5.5; Fig. 7) and ends with an extreme cold and dry stage (PETIT et al. 1999, BASSINOT et al. 1994, LOYER et al. 1995; MIS 10.2, 8.2, 6.2 and 2.2; Fig. 7).

Weichselian
For the Weichselian period (MIS 2-4) recent mulridisciplinary research at the archeological site Veldwezelt-Hezerwater is of great importance.Together with the standard sequences of Harmignies and Remicourt in Belgium (HAESAERTS 1974, HAESAERTS et al. 1999), this is one of the most complete Weichselian loess sequenc es in western Europe (GULLENTOPS et al. 1998).
The present study focuses however on MIS-stages 5-12.
Upper Saalian TL-dating of the "B-loess" sequence, containing the characteristic tundrasols beneath the (part of the) Rocourt pedocomplex with related Rocourt tephra, did provide mean ages of 130-140 ka in the Nelissen pit "Kesselt" ( VAN DEN HAUTE et al. 1998, WINTLE in VAN DEN HAUTE et al. 1998).These ages are in accordance with mean TL-dating re sults of around 140 ka from several loess sections in middle Germany, containing tundrasols of the "Bruchköbeler Nassboden-Komplex" present beneath a redbrown interglacial paleosol (ZÖLLER in FRECHEN 1991).At the archeological site Maas tricht-Belvedere, TL-dating of the first eolian loesses, present above Caberg-3 floodloams and influenced by soil-formation processes of the last interglacial, gives a mean minimum age of 115 ka (DEBENHAM 1993).In fact, these results clearly demonstrate that the "B-loess", containing the characteristic tundrasols and showing green am phibole percentages from 20-5%, perfectly fits into MIS-6 (Fig. 7).
However, loesses deposited during the period of strong overlandflow and gully-formation, just be fore the onset of the last interglacial period (see Nelissen pit "Kesselt", Fig. 6) yielded TL-ages of 50-60 ka (VAN DEN HAUTE 1998, WINTLE in VAN DEN HAUTE 1998).On the base of these TL-ages VAN DEN HAUTE et al. (1998) and MEIJS & GROENENDIJK (1999) assumed an important ero sion hiatus between the guUy-filling and underly ing tundrasol-sequence N1-N5.In contrary to JUVIGNE et al. (1996), VANDENBERGHE et al. (1998) did not correlate the interglacial-like paleosol to the Rocourt pedocomplex (MIS-5).On the base of numerous 14 C-dates of the N1-N5 tundrasols (30-37 ka) and the above mentioned TL-results of 50-60 ka they proposed for correlation with MIS-3 and assigned it as the Lafelt soil.According to new chronological data from a sim ilar sequence at the archeological site Veldwezelt-Hezerwater, it seems reasonable to accept the cor relation with MIS-5 after JUVIGNE et al. (1996).A thorough study is prepared on this subject in co operation with the universities of Gent (Belgium) and Amsterdam (The Netherlands).The chrono logical results from the archeological site Veldwezelt-Hezerwater will be published in a later pa per.
Middle Saalian TL-dating done on burnt flints and calcareous tufa and ESR-dating of molluscs, all found in relation to interglacial soil-formation on top of the Ca-Fig.7: Chronostratigraphical framework constructed by correlation of loess, river terrace, pollen, dust and marine/ice isotope sequences, on the base of lithological and paleopedological characteristics and mineralogical and chronological data.The dust and temperature records are adopted from PETIT et al.(1999), while the pollen and ice-volume sequences are acquired from REILLK et al. (1998) andBASSINOT et al. (1994).For legend see Fig. 2. berg-3 floodloams, show average ages ranging from 220 ka (ESR) to 250 ka (TL; with a mini mum age of 175 ka for molluscs: VAN KOLFSCHO-TEN et al. 1993).According to mean TL-data of around 245 ka for Middle Saalian loess sequences m middle Germany, present beneath two intergla cial paleosols (ZÖLLER et al. 1988) 1 and the aboveindicated data for the archeological site Maastricht-Belvedere, it seems most logical to assign the grav el-aggradation phase of the Caberg-3 terrace lev el and the "C-loess" with green amphibole per centages ranging from 1-3%, to MJS-8 (Fig. 7, see also VAN KOLFSCHOTEN et al. 1993).This implies that the Hees soil matches with MJS-7 and the Montenaken soil with MTS-9.
Lower Saalian For correlating the Weichselian and Saalian river terraces with the manne/ice isotopic records and the terrestnal pollen sequences, only two convinc ing reference points are present: the Maasmechelen terrace (first presence of Rocourt tephra; PAULISSEN 1973) and the Caberg-3 terrace (ESR and TL darings; see above).In downward direc tion only the Pietersberg-2 (800 ka), Geertruid-2 (1.1 ma) and Simpelveld-1 (2.1 ma) terrace levels form important fixing points by their paleomagnetical polanty and/or pollen content ( VAN DEN BERG 1996).Recent studies by FELDER & BOSCH (1988, 1989), JUVIGNE & RENARD (1992) and VAN DEN BERG (1996) reported no further chronostrati graphical data that can support more detailed cor relation.On the base of these fixing points and climo-tectomc modelling studies VAN DEN BERG (1996) found a correlation between the five lower river terraces of the Maas and the glacial manne isotopic stages 2/4, 6, 8,10 and 12, denoting that only long-lasting full glacial stages generate major river terrace systems (see Fig. 7).According to uranium senes dating and biostratigraphical re search MEIJER AND PREECE (2000) discovered ex actly the same situation for the lower terrace se quence of the nver Thames in England.On the base of these two studies the strongly developed, reddish Pottenberg soil, which has been found in many sections along the widened Albert Canal on top of the Rothem-1, Rothem-2 and Caberg-1 floodloams, should be attnbuted to MTS-11.Re cent chronometncal andpollen-stratigraphical re search on lacustnne dnlling cores in France, Italy and Greece and alluvial sediments in England, re veals strong evidence for assigning MIS-11 to the Holsteiman penod (TZEDAKIS et al. 1997, REILLE et al. 1998 and MEIJER AND PREECE 2000).In plac es the Pottenberg soil is underlain by a redbrown luvisol (Dousberg soil), which represents the warm period (MIS-13) between the formation of the Rothem-2 and Caberg-1 river terraces ( VAN DEN BERG 1996).This implies that the "D-loess" correlates with JMJS-IO and that the Pottenberg soil and Wolder soil are equivalent to respectively MIS-11.3 and MIS-11.1 (Fig. 7).

Conclusions
Thorough lithological, mineralogical and paleopedological examinations of numerous loess sec tions in the study area, the application of recent chronological data and extended correlation with loess sections in middle Germany, made it possi ble to construct a reliable loess-stratigraphical framework, which begins with the Holsteinian and matches with the manne/ice isotopic and terres tnal pollen records.It contains four glacial loess cycles with five intercalated interglacial paleosols on top of the Rothem nver terrace of the Maas.
GROENENDIJK and J-P.DE WARRIMONT for their positive cnticism and discussions in the field.Great thanks are also due to W FELDER and P.
BOSCH who supplied data from both sides of the widened Albert Canal, which was of emense val ue for this study.
Appreciation is extended to H. VANNIEUWENBORG of the company of Geolab for supplying dnllmgdata from the potential loess-exploitation zone between Kesselt and Lafelt.
SHAWN BUBEL, FREDERIC CLASSON and ELKE SCHLE-PÜTZ were so kind to correct the English text.
the roughly SW-NE and S-N orientated cross-sections clearly show the strong influence of river erosion and aggradation processes in the area.The cross-section Lafelt-Oud Caberg (A-K; Fig. 2 bottom) is constructed on recorded sections along the Albert Canal and adjacent loess-pits and additional drilling information (Halet 1932, Felder & Bosch 1989 and Geolab 1996), while the Vroenhoven-Veldwezelt cross-section (L-M; Fig. 2 top) is based on outcrop-information along the widened wall of the eastern side of the Albert Canal.

Fig. 2 :
Fig. 2: SW-NE and S-N cross-section showing the geomorphological and geological structure of the study area.The role of fluvial/eolian erosion and aggradation processes are pointed out.Abb.2: SW-NO und S-N Querschnitte, welche die geomorphologische und geologische Struktur des Gebietes darstellen.Auffallend sind die Hinweise auf fluviatile und äolische Abtragungs-und Sedimentationsprozesse.