Stratigraphy , Geochemistry , and Paleomagnetism of Late Quaternary Bedrock and Paleosols , Karisoke Area , Virunga Mountains , Northwestern Rwanda

A surface soil and two buried paleosols, units II-III-IV (upper) and IV (lower)-V, provide important infor­ mation on paleoclimate/dtainage in the Vitunga Mountains ftom dtiet to wetter paleoclimatic episodes of the last glaciation and middle postglacial periods. In general, the sttatigra­ phy indicates a period of bedrock weathering, followed by episodic emplacement of tuffaceous sediments accompanied by weathering initially in a drier and perhaps colder paleoenvironment. Later, during the last glaciation (interstadial), incteased stteam activity emplaced alluvium, covered with a thin layer of tuff, all of which is weathered more vigorously than the underlying sediments. The surface soil (ground soil) shows comparatively less weathering and movement of mobile elements indicating the late middle to late Holocene paleoenvironment was somewhat drier and perhaps cooler compared with the eatly Holocene. The geochemistty of these three soils shows considerable depletion of Na, Ca, and K in the middle paleosol relative to the older buried paleosol and to the surface horizons. Iron is an important indicator of paleoclimate, suggesting that the middle II and III units of the upper paleosol formed under a paleoclimate wetter than today. The radiocarbon-dated middle paleosol indicates it could have been exposed to subaetial weathering from the Kalambo Intelstadial (*» 25,000 yr BP) to the middle Holo­ cene. The total weathering in the middle unit is too ad­ vanced to have occurred during the Holocene. The relatively high concentrations of Th in local bedrock, tuffs, alluvium and slope wash indicate that it is the soutce of tadioactivity reported by other workers in The Virunga Mountains. *) Adresses of the authors: Prof. Dr. WILLIAM C. MAHA­ NEY, Geomorphology and Pedology Laboratory, Department of Geogtaphy, Atkinson College, York University North York, Ontario, Canada M3J 1P3. Dr. R. G. V. HANCOCK, SLOWPOKE Reactor Facility and Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario, Canada M5S 1A4. Prof. RENE W. BARENDREGT, Department of Geography, Lethbridge University, Lethbridge, Alberta, Canada T l K 3M4. [Stratigraphie, Geochemie und Paläomagnetik spätquartärer Gesteine und Paläoböden im Karisoke-Gebiet der Virunga-Berge, Nordwest-Rwanda] Kurzfassung: Ein tezentei Boden und zwei begtabene Paläoböden, Einheiten II-III-IV (oberer) und IV (untetet)-V, liefern wichtige Informationen über Paläoklima und Abfluß­ verhältnisse in den Virunga-Bergen von trockeneren zu feuchteren paläoklimatischen Phasen des Spätglazials bis zu mittleren postglazialen Zeiten. Zusammenfassend betrach­ tet, weist die sttatigtaphische Abfolge zunächst auf eine Vetwitteiungspetiode des Festgesteins hin, der eine Phase folgt, in der tuffige Schichten abgelagert wutden. Die Zeit war von einer Verwitterung unter trockeneren und vielleicht kälteren Bedingungen begleitet. Spätet, während des letzten Gla­ zials, kam es bei veistärktet Aktivität det Flüsse zur Sedi­ mentation von Schwemmlandablagerungen, die von einer dünnen Tuff-Schicht überdeckt wurden. Diese Sedimente sind intensivei verwittert als die liegenden Schichten. Det tezente Boden zeigt eine vergleichsweise geringe Verwitterung mobiler Elemente, was datauf hindeutet, daß die Klimaverhältnisse im späten Mittel-Holozän und Spät-Holo­ zän etwas trockener und vielleicht kühlet waren als im FrühHolozän. Die Geochemie der drei Böden zeigt eine bedeu­ tende Abnahme von Na, Ca und K im mittleren Paläoböden im Vergleich zum älteren Paläoböden und zu den rezenten Böden. Eisen als ein wichtigei Indikator für das Paläoklima deutet daraufhin, daß die mittleten Einheiten II und III des oberen Paläobodens untet feuchteten Klimabedingungen als heute entstanden sind. Die Radiokarbon-Datietungen des mittleren Paläobodens zeigen, daß dieser Boden einer subaerischen Verwitterung vom Kalambo-Interstadial («• 25 000 Jahre vor heute) bis zum Mittel-Holozän ausgesetzt war. Die gesamte Verwittetung im mittleren Abschnitt ist zu weit fortgeschritten und wählend des Holozäns entstanden. Die relativ hohe Konzentration von Th in den anstehenden Festgesteinen, Tuffen, Schwemmlandsedimenten und Hang­ rutschmassen zeigt, daß Th die Quelle det Radioaktivität ist, von der andere Autoren aus den Virunga-Bergen berichten. Stratigraphy, Geochemistry, and Paleomagnetism of Late Quaternary Bedrock 135


Introduction
The surficial geology and geochemistry of the Virunga Mountains in northwestern Rwanda are very poory understood (ANTRUN and MAGNEE 1983;HARRIS 1982).Of the five Rwandan volcanoes, Karisimbi (4510m) is the highest and contains a complex series of lava flows which cover the largest area.Over 150 parasitic cones circle the summit area and presumably are the source of most pyroclastic sediments and rocks found around the mountain.One young lava flow (precise location unrecorded) from Karisimbi has been dated 56,000 yr by K/Ar (HARRIS 1982).The follow ing discussion concerns the stratigraphy, geochemistry and paleomagnetism of a section on the northeastern flank of Karisimbi (Figure 1) at 3100m (in the valley area between Karisimbi and Visoke).

Field Area
The Virunga Volcanoes in northwestern Rwanda (on the Zaire-Rwanda-Uganda border Figure 1), consist mainly of rocks of basaltic composition (MERCIER 1984).In some cases, thick cover over rocks of tufface-ous origin overlie the basalt; all are usually weathered to depths of 1.0 to 3.0m depending on age.Surficial deposits of loess, colluvium and alluvium are often weathered into thick (= 2.0m) brown (10YR 4/4) colored paleosols sometimes forming compound units (e.g.similar to IIA lib and IIA 12 b horizons in Figure 2) that appear to have formed during the last glaciation when the timberline was depressed several hundred meters.
The site is located in the upper portion of the Hagenia woodland dominated by Hagenia abyssinica (COE 1967;HEDBERG 1964).The presence of a dark moor land-like paleosol, at or near the surface, that can be traced to elevations as low as 2750m indicates a sub stantial vegetation depression.Because the thick moorland-like paleosol is similar to one observed on Mount Kenya (900 km to the east) (MAHANEY 1990), it is considered an important paleoclimatic indicator.The present-day climate is wet with mean annual pre cipitation at approximately 2000 mm (FOSSEY 1982).Evapotranspiration is unknown.The temperature in the upper Hagenia forest may approximate the mean annual temperature of 10 ± 6°C at a similar altitude on Mount Kenya (MAHANEY & BOYER 1986). 3

Methods
The KAR1 section was described using the nomen clature of the SOIL SURVEY STAFF (1951STAFF ( , 1975) ) and BIRKELAND (1984).Soil and sediment colors were estimated using the color chips of OYAMA & TAKE-HARA (1970).Samples were collected and dried after removing the granule and pebble size material (>2mm  (1965).The pH was measured by electrode using a soil-water ratio of 1:5.The < 2 mm samples were subsampled and approximately 1 to 2 g was placed in small flip-top polyvials for neutron activation analysis in the SLOWPOKE nuclear reactor at the University of Toronto (HANCOCK 1978(HANCOCK , 1984)).
To determine the concentrations of U, Dy, Ba, Ti, Mg, Na, V, Al, Mn, Ca and K, which produce short lived radioisotopes, the samples were irradiated se rially for 1 minute at a neutron flux of 1.0 x 10 11 n-cm-2> s-1 in the SLOWPOKE nuclear reactor at the University of Toronto.After a delay time of 19 minutes (which allows the very short-lived 28 Al to decay to acceptable levels), each sample was assayed using 5 minute counts with on-site gamma-ray spec trometers, as described by HANCOCK (1984).Appro priate gamma-ray peak areas were measured and the chemical concentrations calculated using the compara tor method.
The samples were then batch irradiated for 16 hours at a neutron flux of 2.5 X 10n-cm-2 -s -l , and after a waiting time of apptoximately 7 days the concentra tions of Sm, U, Yb, La, As, Sb, Br, Sc, Fe and Na were similarly determined, with the Na giving a cross check with the first analysis.After a two-week delay, the samples were recounted and the concentrations of Nd, Ce, Lu, Ba, Th, Cr, Hf, Sr, Cs, Ni, Th, Sc, Rb, Fe, Co, Ta and Eu were measured.Once again, the Sc and Fe were used to cross-check the second phase of the analysis.

Stratigraphy
A sequence of slope wash and alluvium, interbedded tuffs, and basalt that is presumably younger than • 50,000 yr (HARRIS 1982) provided an opportunity to study weathering and paleosol genesis in a humid tropical setting over a short span of time.The weath ered basalt and tuff (units V and IV) shown in Figure 2 are exposed in a gully along the lower flanks of Kari simbi Volcano.Because the two lower units were less highly weathered (based mainly on color differences observed in the field), it was inferred that they had formed over a short span of time followed by a longer period of weathering which produced hues as brown as 10YR6/3.Unit III appeared to be more highly altered with colors of 10YR4/6 indicating that Fehydroxides and oxides might make up the bulk of the weathering products.Because the lower two stratigraphic units consist of pyroclastic/volcanic rock materials that were at least altered to a moderate degree on a flow surface dated at = 50,000 yr, it is believed they might have weathered during the drier paleoclimate which accompanied the early stadial of the last glaciation ~ 50,000-35,000yr BP) (MAHANEY, 1990).The overlying units (I, II and III), consisting of slope wash, alluvium and waterlaid tuff, are considerably younger judging by a 14 C date of 6330 ± 200 yr BP (BGS -1214) on charcoal in the IIA lib horizon.
Because the precipitation is so high, the particle size so coarse (Table 1), and roots prevalent in the upper two units (Figure 2), it is highly likely that some downward movement of geobiochemical contami nants might have occurred.Units II and III appear to have formed through the interstadial of the last glaciation (Kalambo), the last glacial maximum and early to middle Holocene.
The particle size data in Table 1 show some important trends with respect to clay and silt that bear on the morphogenesis of the two buried paleosols.The lower paleosol [units IV (lower) and V] is considerably higher in silt compared with the weathered bedrock which might be related either to the airfall influx of tuffs and/or loess (glacially-crushed quartz grains indicate aeolian transport of sediments from the Ruwenzori Mountains or some other glaciated area; MAHANEY 1990) from the nearby Ruwenzori Moun tains or to reworking of locally derived silt follow ing a major vegetation belt depression during colder stadial climate at or near the end of the last glaciation.Clay translocation in this lower paleosol is only on the order of 1 % indicating a drier paleoclimate that was not conducive to the downward movement of finegtained particles.
The upper paleosol units [(II, III, and IV (upper)] also contain high silt relative to the surface slope wash unit and the weathered bedrock (V).As in the lower paleo sol this increase could be related to the emplacement of pyroclastic material or to sorting and reworking of older sediments by streams.The increase of percent clay in the IIA lib and HIB 2 b horizons indicates a paleosol with considerable power to form clay relative to the underlying paleosol.This increase of clay in the IIIB2b horizon supports the hypothesis of increased downward movement relative to the lower and upper units in the sequence.Because the surface soil is formed in slope wash reworked from nearby slopes covered with pyroclastic sediments it is possible only to point out that higher sand percentages suggest suffi ciently high surface run off to move coarse-grained sediments; increasing silt and clay with depth indicate the possibility of some downward movement.

Mineralogy
In order to test the hypothesis of a young (< 50,000 yr) age for the paleosols in the KAR 1 sequence we analyzed the mineralogy of the < 2/IM fraction in each horizon.In the oldest buried paleosol only a trace amount of illite was detected in the IVB2b.Quartz is present in all horizons with moderate amounts in the VCb and IVB2b horizons dimin ishing somewhat in the IVAb horizon.Volcanic feld spars (mainly anorthoclase) are present in small amounts throughout with somewhat lower reflections in the IV Ab horizon.Small amounts of hematite were also detected.
The mineralogy of the upper paleosol shows a similar absence of clay minerals with hematite appearing in small to trace quantities as the chief weathering pro duct.Quartz is highest in unit II and in the IVC2oxb (moderate amounts) horizons.Volcanics feldspars are highest in the IIAllb, IIA12b, and IIIBlb hori zons, declining somewhat in the lower horizons.The ground soil is noticeably different in that quartz (abundant amounts) and volcanic feldspars (small amounts) are higher than in the two buried paleosols.Colors follow Oyama and Takehara (1970) Also, pyroxene which exists in trace amounts in the buried paleosols is higher in the ground soil.Because quartz is relatively rare in the volcanic rocks of the Virunga Mountains it may signal airfall influx, per haps, from the nearby Ruwenzori Mountains.The plagioclase feldspars are noticeably higher in the A1 horizon when compared with the lower paleosols and with the Cox horizon indicting they are being contri buted from fresh outcrops up slope.Although it may be inappropriate to compare the < 2/tm fraction mineralogy with the < 2 mm fraction geochemistry, elevated Na and K concentrations are also observed in the Al and Cox horizons (see Table 3).

Soil Chemistry
Soil colors, pH and total salts (Table 2) were studied to determine whether they could be used to differen tiate the three units.The pH trends in the upper two units show slight increases with depth; the lower unit is uniform.These data indicate slight movement of H + ions in the upper two units and none in the lower unit.The data for total salts reveal only slightly higher concentrations in the A and Ab horizons relative to the other horizons.Based on the geochemical data discussed later the salts present must be NO and SO rather than CI.The soil colors shown in Table 2 are all in the 10 YR range suggesting younger material that is not highly altered.These yellowish red colors are com patible with the liberation of moderate amounts of goethite and minimal to trace amounts of hematite.
The color trends suggest that the HIB2b horizon might contain the highest concentration of liberated oxides, hydroxides and clay minerals.Table 3 con firms this prediction with its highly elevated Fe con tent in the IIIB2b horizon.

Geochemistry
The chemical analysis data for the KAR1 section, which are listed in Table 3, show considerable variab ility.Some of the variation may be associated with different lithological materials or with weathering, paleoclimate and /or drainage.Because Na and K are the most mobile of the major elements their depletion in the tuff/ash, alluvium, and lower tuff/loess hori zons indicates that weathering effects were especially prominent in forming these Inceptisols.This supposi tion is born out by a check of the inter-element corre lations using Pearson correlation coefficients, which show that Na and K are highly positively correlated.Both are positively correlated with Rb and are nega tively correlated with Sc, T, Cr, Ni, Br and Cs.
Four other correlation groups including major ele ments are also apparent.Magnesium, Ca, Mn, Co, V, Sr, CI, and Ba are all highly positively correlated and are negatively correlated with As, Sb, Br, the rare earth elements (REEs), Hf, Ta, Th, and U. Alumini um is positively correlated with As, Sb, the REEs, Hf, Ta, Th, and U, and is negatively correlated with Cr, Sc, V, Ti, Mn, and Co. Titanium, Sc, V, Cr, Ni, and Mn are all highly positively correlated and are nega tively correlated with As, Sb, REEs, Hf, Th, and U. Iron does not correlate with any of the other major elements, although it is negatively correlated with Sb, the REEs, Hf, Ta, Th, and U.If it is possible to relate these geochemical correlation findings to the specific mineralogy of the section, then there are significant quantities of at least five primary minerals combined to make up the sediments in the KAR1 section.
The Na and K are depleted in the tuff and ash, in the alluvium, in the lower two horizons (IVAb and IVB2b) in the tuff and loess, indicating major leach ing of these materials during or after their formation.The Mg, Ca, et al.-rich materials are highest in the slope wash material, in the IVC2oxb horizon in the tuff and loess, and in the weathered basalt horizon (VCb).Aluminium and its associated elements tend to increase in concentration downwatd through the section, with exceptionally high levels in the IVC loxb horizon in the tuff and loess.Titanium and the asso ciated transition metals tend to be depleted in the HIB2b horizon in the alluvium and in the next lower IVCloxb horizon in the tuff and loess, and in the IVAb horizon in the tuff and loess.They therefore tend to be depleted in the highest silt-containing units of the section, implying that they may be asso ciated with material in either the sand or the clay frac tions of the soils.
The distribution of Fe with depth shows trends that are probably related to paleoclimate and time of expo sure to subaerial weathering processes.The highest Fe content, by far is found in the IIIB2b horizon in the alluvium.This high Fe content could result from dif ferences in parent materials or from relative move ment (depletion/enrichment) caused by different paleoclimates.The Fe levels are next highest in the slope wash material, and in the IVC2oxb horizon in the tuff and loess.If one compares the two buried paleosols [(II, III, and IV (upper)) and (IV (lower) and V)], it appears that the lower one was exposed for either a shorter time or that it formed in a paleo climate with less available moisture than the upper one.It is likely that some of the Fe is organically complexed into a soluble form (see paleosol colors in Table 2), which allowed it to move downward in the upper buried paleosol, eventually enriching the buried B horizon.If this hypothesis is correct one is dealing with a considerably stronger paleoclimate in units II, III and IV (upper) compared with IV (lower) and V.
The Fe data tend to support the notion that the lower paleosol formed during one of the stadial intervals during the last glaciation, while the upper paleosol formed mainly during the Kalambo Interstadial, Last Glacial Maximum, and early postglacial period up to around 6330 ± 200 yr BP (BGS-1214).While paleoecological data are scarce from the East African moun tains, the period from 8000-4000 yr BP is considered to have been wetter and warmer than the early and later Holocene (MAHANEY 1988(MAHANEY , 1989)).A chondrite normalized plot of the rare earth ele ments for the sample suite shows relative uniformity both in curve shape and in magnitude (see Figure 4).In this figure, the outermost data points are shown with solid black lines and the profiles for representa tive samples Al, Cox, and IVAb are included for detail.
The relative uniformity of the pattern indicates that the < 2 mm soil (paleosol) fraction samples were formed from material of similar magmatic composi tion, probably over a short geological time span.ALTERNATING FIELD (mT)

Paleomagnetism
In July, 1987, samples for paleomagnetic analysis were collected between Visoke and Karisimbi volcancoes (Figure 1).In total, 73 polycarbonate cubes (2cm X 2 cm X 2cm) were inserted into the sediments using hand pressure, and where necessary through gentle tapping with a rubber mallet.In addition, several blocks of unweathered vesicular basalt were marked and collected in the field, and cored in the laboratory using a bench drill.Analysis was carried out using a DSM-1 spinner magnetometer in the paleomagnetic laboratory at the Pacific Geoscience Center.
Approximately one third of the pilot specimens (Figu re 5) collected from the Karisoke sediments can be characterized as a Type A magnetization, that is, they are single component (except for a small present earth's field component) and their magnetization values after stepwise demagnetization plot as a straight line that reaches nearly to the origin of an magnetization values which plot as a straight line that reaches nearly to the origin of an orthogonal projec tion.The remainder of the pilot specimens can be characterized as a type C magnetization, that is, they show incoherent magnetization values with stepwise demagnetization.The basalts beneath the Quaternary sediments show a single component (Type A) magne tization (Figure 6).
Type C magnetizations are recorded in sediments at 0-50 and 140-180cm below the surface (Figure 7).These are zones containing abundant grit and pebble size tuff particles.The coarser fraction of these sedi- ments is either too large for alignment in an ambient magnetic field and hence will show random magneti zation directions, or was only partially aligned.In either case, the coarser fraction of the sediments may entirely or partially mask the dipole magnetization of the finer matrix fraction.In these sediments, stepwise demagnetization did not produce a clear direction with an end point near the origin of the orthogonal plot.
As can be seen from Figure 7, the inclination and de clination profiles reveal a suite of normally magne tized sediments.The inclination profile shows a some what noisy record with values which in places are con siderably steeper than would be expected for an equa torial sampling latitude.Inclination values at the equator should essentially be zero.Some of the large swings in declination and inclination are associated with low K* values (Table 4), where K* is a measure of the degree of internal consistency within a sample, and are thus suspect.K* values greater than 10 are considered acceptable.The magnetization profile shows a rather consistent pattern of progressive waxing and waning of the sample intensity which in all likeli hood is a true depiction of field behavior.This record would seem to indicate that the 2.4 meters of uncon solidated sediments at Karisoke were deposited over a period of several thousand years.

Fig. 3 :
Fig. 3: Clay and silt distributions with depth in the KAR1 section.

Fig. 5 :
Fig. 5: Magnetization behaviot of pilot specimens from normally magnetized unconsolidated sediments following stepwise alternating field (af) demagnetization.Natural remanent magnetization is indicated by letters NRM.Normalized magnetization (M/MO) vetsus af demagnetization (a), and direction changes during a demagnetization projected on a lower hemisphere equal area stereographic projection (b) where open squares [Dl and solid squares [ • ] represent negative and positive inclination respectively, and otthogonal vectot projections (c), with tespect to present horizontal, where solid circles [] refer to N/S axis (horizontal plane) and open circles [00]to Up/Down axis (vertical plane).

Fig. 6 :
Fig. 6: Magnetization behaviot of a pilot specimen from the notmally magnetized vesicular basalt found directly beneath the unconsolidated sediments.Symbols as in Figure 1.
(FISHER, 1953) a 95 = radius of the cone of 95 % confidence about the resultant vector The Karisoke sediments are for the most part stably magnetized and record a normal dipole field.The zones of coarse sediment generally reveal incoherent magnetization directions which are nevertheless all normal.Mean remanent magnetization directions indicate a local field with an inclination of -0.5 degrees and a declination of 35 degtees indicating that a normal dipole field existed at the time of deposition of these sediments.The underlying vesicular basalt is likewise normally magnetized.

Table 1 :
Particle size ratios for horizons in the KAR1 paleosol sequence, Virunga Mountains, Northwestern Rwanda