glacial sedimentary processes and products (hambrey/glacial sedimentary processes and products) ||...

The Ordovician glaciation in Eritrea and Ethiopia, NE Africa

R.A. KUMPULAINEN

Department of Geology and Geochemistry, Stockholm University, 106 91 Stockholm, Sweden (e-mail: [email protected])

ABSTRACT

Ordovician (Hirnantian?) glacigenic deposits are described here for the first time from south-central Eritrea. These deposits rest on an almost peneplained Neoproterozoic basement and define,in Eritrea and Ethiopia, a depositional area measuring at least 200 km in an east-west directionand 170 km in a north-south direction. For this preliminary note, five sections through the glacigenicsuccession were logged in Eritrea and one in Ethiopia. Facies types are described and interpreted.An ice-proximal facies assemblage is located in the Tigray Province of northern Ethiopia, the typearea of the glacigenic Edaga Arbi Beds. These proximal deposits, c. 20 m thick, are characterisedby melt-out diamictites, with striated clasts, interlayered with sandstone beds displaying horizon-tal lamination and normal grading (sand-silt). The horizontal lamination in the section is transi-tional with climbing ripple beds. Ice rafted clasts in sand-granule grade are common in these sandybeds. This ice-proximal section also exhibits some, minor soft-sediment deformation, such as asym-metrically folded beds, south-dipping reverse faults and glacial grooves suggesting transport to thenorth. This proximal facies grades laterally into a cross-bedded arkosic sandstone, the EntichoSandstone, which probably represents deposition on subaqueous outwash fans. Cross-beds in thissandstone dip consistently to the north also in south-central Eritrea. Glacial striae and groovesare observed on top of the Enticho Sandstone in two localities in Eritrea. These proximal faciestypes are overlain by a distally deposited mudstone-dominated unit, 3–40 m thick, most probablydeposited from turbid overflow plumes, although it also contains ice-rafted clasts. Only this unithosts ice-rafted clasts in Eritrea. In Eritrea it also contains some diamictites. The name Edaga ArbiBeds is adopted for this unit in Eritrea. Icebergs were probably responsible for the deposition ofthe diamictites in south-central Eritrea. The development of this glacigenic succession was prob-ably related to a regular retreat of the ice margin from north to south. It is also probable, thatthis succession only represents one cycle of deglaciation, the last of the two (or three) recog-nised in other parts of North Africa. The post-glacial development is initially represented by thedeposition of a probable marine dune complex migrating from north to south. Fossil evidence andtrace fossils, particularly Arthrophycus alleghaniensis (Harlan) suggest that the age of these glacigenicdeposits is Late Ordovician, probably Hirnantian.

Keywords Glaciation, Ordovician, Eritrea, Ethiopia.

INTRODUCTION

Glacial deposits have been known since the 1970sfrom the Tigray Province of northern Ethiopia(Dow et al., 1971; Beyth, 1973). The presence of corresponding glacigenic rocks also in Eritrea was mentioned by Kumpulainen et al. (2002). Thepresent paper describes for the first time theserocks, known as the Edaga Arbi Beds, and their

relationship with the underlying and overlyingunits in Eritrea. Some complementary observa-tions were also made in the Edaga Arbi area ofnorthern Tigray.

These glacigenic deposits occur in a sandstone-dominated succession of the Adigrat Group(Garland 1980), which rests unconformably on apeneplained surface (Abul-Haggag, 1961) under-lain by Neoproterozoic schists and pan-African

GSP_4_C18.qxd 11/5/07 13:44 Page 321

Glacial Sedimentary Processes and Products Edited by Michael J. Hambrey, Poul Christoffersen, Neil F. Glasser and Bryn Hubbard © 2007 International Association of Sedimentologists. ISBN: 978-1-405-18300-0

322 R.A. Kumpulainen

granitoid rocks that are associated with the suturezone between the East- and West-Gondwana. In theEarly Palaeozoic Era, the area was clearly locatedwithin a very large continental block, allowing onlya thin veneer of sediments to accumulate over awide almost horizontal continental shelf (Fig. 1). A clastic sedimentary succession was deposited in Palaeozoic time on this flat-lying unconformity. In Ethiopia and Eritrea, this sedimentary cover is known as the Adigrat Group and is overlain by a marly limestone unit, the Antalo Limestone(Blanford, 1869; Bosellini et al., 1997), containing arich Oxfordian-Kimmeridgian marine fauna.

The lower part of the Adigrat Group is essenti-ally non-fossiliferous. The only body fossils weredescribed by Saxena and Assefa (1983) from thelower, glacigenic part of the clastic succession. Thefossil, Discophyllum peltatum Hall 1847, belongs to Siphonophorida and is of probable Ordovician-Silurian age. Otherwise, all the other parts of thesuccession could have been formed any time from Cambrian to Jurassic. The recent discovery oftrace-fossils particularly Arthrophycus alleghaniensis(Harlan), along with eight other types of traces,

from Eritrea also suggests a diagnostic Ordovician-Silurian age of the lower part of the sandstone formation resting on top of the glacigenic EdagaArbi Beds in Eritrea (Kumpulainen et al., 2006). In Ethiopia, the upper part of the possibly samesandstone exhibits Triassic dinosaur footprints(Assefa, 1987), contains Jurassic plant debris (vonzur Mühlen, 1931; Mohr, 1971; Bosellini, 1989) andgrades gradually into the overlying Jurassic AntaloLimestone. These conflicting palaeontological datalead to stratigraphic correlation problems. Newdata on stratigraphy in Eritrea has been discussedin some detail by Kumpulainen et al. (2006) andreviewed briefly below under the section on strati-graphy. The present paper will be limited to theLower Palaeozoic part of the Adigrat Group.

A large amount of data has accumulated the last few decades on the Late Ordovician glaciation in North Africa (Debyser et al., 1965; Beuf et al., 1971; Deynoux & Trompette, 1981; Ghienne, 2003),Arabia (Vaslet, 1990) and Turkey (Monod et al.,2003), and at least two major advances (Le Heronet al., 2004) and maybe more than two (Ghienne,2003) advances, separated by a recession, havebeen recognised in many areas. Interpretations ofthe extension of the ice cap have been presentedby e.g. Hambrey, 1985; Vaslet, 1990; Sutcliffe et al.,2000; Monod et al., 2003; Le Heron et al., 2004;Ghienne et al., this volume). According to theseinterpretations the margin extended from Africaeastwards beyond the successions discussed here(Fig. 1).

Stratigraphy

In northern Tigray, Ethiopia (Fig. 2), the AdigratGroup is composed of three formations (Garland,1980): (i) the glacigenic Edaga Arbi Beds, which inthe Edaga Arbi and Abi Adi areas rest on thenear-horizontal Neoproterozoic basement. Accord-ing to Dow et al. (1971), the Edaga Arbi Beds are divided into four informal stratigraphic units,Units 1–4 (Fig. 3). The coarse-clastic, Unit 1 gradeslaterally to the west, north and east into (ii) theEnticho Sandstone. The dominating upper part of the Edaga Arbi Beds in Ethiopia is composed ofmudstones (Units 2–4). Unit 2 contains striateddropstones, whereas units 3 and 4 are devoid ofany features suggesting a glacigenic origin. Unit 3carries trace fossils, yet to be described. The Edaga

Fig. 1 Palaeogeography of Gondwana at c. 440 Ma,exhibiting the inferred distribution of land areas, shelfseas and the Ordovician continental ice cap. The blackarrows indicate a radial ice-flow pattern. Limited amountof data is available in the east. The star indicates thelocation of the study area in Eritrea and Ethiopia.Modified from Vaslet (1990).

GSP_4_C18.qxd 11/5/07 13:44 Page 322

The Ordovician glaciation in Eritrea and Ethiopia, NE Africa 323

Fig. 2 Geological map of the area from Mekele inEthiopia northwards to south-central Eritrea. The ‘glacial and fluvial types’ of the Enticho Sandstone are not distinguished on the map. The scale does notallow the distinction of formations of the Adigrat Groupin the area north of the open arrow in Eritrea. In earlierliterature, the name ‘Eritrean sandstone’ was used forthese formations (Mohr, 1971). The thickness for theEdaga Arbi Beds (Eab) and Enticho Sandstone (Es),respectively, is given for a number of localities in Tigray (Beyth, 1973). The boundary between Eritrea and Ethiopia is approximate. Modified from Arkin et al.(1971), Garland (1980) and Hambrey (1981).

Arbi Beds and the Enticho Sandstone are bothcovered by (iii) the Adigrat Sandstone.

In the Enticho–Mekele areas (Fig. 2), the EdagaArbi Beds form a lensoidal body, which is c. 60 kmwide in the east-west direction (Fig. 4), and extendsfrom the Enticho area to the Mekele area in thesouth. The Edaga Arbi Beds are limited to the west,north and east by a ‘glacial type’ of the EntichoSandstone (Beyth, 1973), described as poorly sortedsandstone containing large clasts. To the east andnorth, the ‘glacial’ Enticho Sandstone grades later-ally over to a ‘fluvial’ cross-bedded Enticho Sand-stone. The facies transition from ‘glacial’ to ‘fluvial’remains to be described later and has not been indicated in Figure 2.

According to the previous information (Dow et al., 1971; Mohr, 1971; Garland, 1980), the EdagaArbi Beds wedge out northwards, never reach-ing Eritrea. However, new information revealsthat this glacigenic succession occurs in Eritrea(Fig. 2), although it has not been traced, so far, inoutcrop across the border area between Ethiopia and Eritrea. In the Mai Aini–Adi Keyih–Genzeboareas, three distinctly different formations (Fig. 5)have been identified (Kumpulainen et al., 2006): (i) the arkosic Enticho Sandstone with palaeo-currents trending approximately to the north, restson the peneplained Neoproterozoic basement, (ii)a maroon, mudstone-dominated formation withoutsized clasts, correlated with Unit 2 (Fig. 3) in theEdaga Arbi type section in Tigray, but for whichthe name Edaga Arbi Beds is applied in this paper,and (iii) the Adigrat Sandstone formation. The tracefossil information (Kumpulainen et al., 2006) inthe lower part of the Adigrat Sandstone (sensuGarland, 1980) suggests an Ordovician to Silurian

GSP_4_C18.qxd 11/5/07 13:44 23


age for the lower part of the formation, whereasthe upper part elsewhere is Triassic to Jurassic(see above). The informal Adigrat Sandstone forma-tion was introduced by Kumpulainen et al. (2006)for the lower part of the Adigrat Sandstone (sensuGarland, 1980) in order to circumvent the requiredredefinition of the lithostratigraphical nomenclaturein the study area, for which the amount of data currently available is insufficient (cf. Salvador,1994). Further work may prove that two or more

sandstone formations of very different ages andcharacteristics instead of one Adigrat Sandstonemay be present in the Horn of Africa.

The presence of trace fossils in the successions ofthese two areas may also be significant for strati-graphical correlations between Eritrea and Tigray.In Eritrea, a rich trace fossil fauna occurs particu-larly in the Adi MaEkheno Member, which rests on top of the Edaga Arbi Beds. In Tigray, trace fossils occur in Unit 3 (sensu Dow et al., 1971) of the

Fig. 3 Stratigraphy of the Edaga Arbi Beds in their type area, compiledfrom the description of Dow et al.(1971). Four informal units (1–4) havebeen identified in this succession.Unit 2 contains outsized clasts andmay represent the youngest part ofthe glacigenic succession in this area.In contrast to the original suggestion,units 3 and 4 do not belong to theEdaga Arbi Beds, but representyounger events of sedimentation.

Fig. 4 Two profiles describing thefacies relationships between theEdaga Arbi Beds and the EntichoSandstone in Tigray, representing the Enticho–Adigrat area and theEnticho–Mekele area. The Edaga Arbi Beds are a lensoid body limitedlaterally by the ‘glacial’ facies of theEnticho Sandstone. The western end,left of the arrow, of the east-westprofile projects outside the left marginof the map (Fig. 2). The AdigratSandstone and the Antalo Limestoneare omitted in these profiles. Redrawnfrom Beyth (1973).

GSP_4_C18.qxd 11/5/07 13:44 Page 324


Edaga Arbi Beds, which does not exhibit any evid-ence of glacigenic origin, but rests on top Unit 2containing dropstones. It is therefore suggestedhere that only Units 1 and 2 (sensu Dow et al., 1971)belong to the Edaga Arbi Beds in Tigray, whereasUnits 3 and 4 belong to another lithostratigraph-ical unit, i.e. the Adigrat Sandstone formation orconstitute a new stratigraphic unit of their own.

The Adigrat Sandstone formation in Eritrea isdivided into two members, the lower, Adi MaEkhenoMember and an informal upper member, Member2. The Adi MaEkheno Member is dominated by one tabular, cross-stratified sandstone bed, up to20 m thick. The attitude of the clinoforms sug-gests bed-form migration towards the south orsouthwest. Member 2 rests with an erosional contact on the Adi MaEkheno Member and iscomposed of sandstone beds with cross-beddingsuggesting palaeocurrents trending to the north.Hence, with their opposing palaeocurrent polar-ities, these two sandstone members of the AdigratSandstone formation are readily delimited in theoutcrop. The upper boundary of Member 2 has notbeen studied by the present author. A schematicnorth-south correlation profile between the Edaga

Fig. 5 The Lower Palaeozoicstratigraphy of south-central Eritrea.The Jurassic Antalo Limestone andyounger units are omitted. Modifiedfrom Kumpulainen et al. (2006).

Arbi–Enticho area and Mai Aini–Adi Keyih area isprovided in Figure 6.

A similar three-fold stratigraphy was identifiedby Hutchinson & Engels (1970) eastwards in theAdeilo area, centred along the Eritrean coast (Fig. 2, insert map). They found a lower sandstone,a middle mudstone and an upper sandstone, and ascribed these units to the Adigrat Sandstone.New work in that area (Fig. 7) reveals that the mud-stone unit carries outsized clasts up to cobble size,and is texturally and lithologically very similar to the mudstones of the Edaga Arbi Beds in south-central Eritrea (Fig. 3); this concerns also the colourof the rocks. This new information shows that the Edaga Arbi Beds extend from the Mai Aini area c. 200 km eastwards to the Adeilo area alongthe Eritrean coast and, indeed, further arcoss theRed Sea to Yemen. This stratigraphy probably also extends westwards from the Mai Aini area(Zanettin et al., 1999), but it has not been studiedthere. In the north-south direction, the glacigenicsuccession extends from the Mai Aini area for atleast 170 km southwards to the Abi Adi area westof Mekele in Tigray, and its extension further southinto other parts of Ethiopia is highly probable.

GSP_4_C18.qxd 11/5/07 13:44 Page 325


It is worth noting that this three-fold strati-graphy, particularly in the highland areas ofEritrea and Ethiopia, produces a very impressiveerosional landscape of table mountains and buttes

with steep-sided valleys and flat valley floors (Fig. 8). The Precambrian schists exhibit an ero-sional, smooth hilly landscape. The top of thePrecambrian sequence is close to a peneplain andparticularly in Eritrea, this surface is covered by the Enticho Sandstone. The Edaga Arbi Beds aresoft and easily removed by erosion. They commonlyform an inclined slope covered by scree, but also occasionally providing a complete exposureacross the whole Edaga Arbi Beds. The AdigratSandstone formation on top of the mudstonesforms vertical cliffs. These circumstances providean excellent opportunity to trace the individual formations across these regions.

Fig. 6 A cross-section from EdagaArbi to south-central Eritreasuggesting the presence of a three-fold stratigraphy in both areasof Ethiopia and Eritrea. The EdagaArbi Beds grade laterally into theEnticho Sandstone. Modified fromKumpulainen et al. (2006).

Fig. 7 Geological map of the Adeilo area (indicated withA in the insert map, Fig. 2), along the Danakil coast ofEritrea. Maroon ice-rafted debris-bearing glacigenicmudstone with sand to cobble-sized dropstones werefound in the Aitos River valley (encircled). Anotherlocality, c. 5 km NE of Adeilo, is indicated by an arrow.The extension of that mudstone unit within theunspecified Adigrat Group is inferred with a brokencurve in other parts of the Adeilo area. Modified fromSagri et al. (1998).

Fig. 8 The Lower Palaeozoic stratigraphy exposed in a steep hill-side north of Mawray. The variousstratigraphic units are indicated. AMM – Adi MaEkhenoMember, M2 – Member 2 of the Adigrat Sandstoneformation.

GSP_4_C18.qxd 11/5/07 13:44 Page 326


THE GLACIGENIC SUCCESSION

General

Glacigenic successions may be formed (e.g.Hambrey, 1994; Miller, 1996) in a variety of deposi-tional settings, either (i) by melt-out or lodge-ment of sediment in contact with the depositingglacier or ice cap, or (ii) transport of sediment tothe depositional site on land, or into lakes or seasby traction currents, gravity flows, in suspensionor by ice-rafting, followed by possible subsequentredeposition.

Boundary relationships

Around Edaga Arbi and Abi Adi in Tigray, theEdaga Arbi Beds rest unconformably, with a sharp boundary, on top of the Neoproterozoicgreenschist facies volcanic sedimentary rocks (e.g. Dow et al., 1971). Locally, glacial striae have beenreported on top of the Neoproterozoic basement.The boundary of the Edaga Arbi Beds towards the Enticho Sandstone was interpreted by Dow et al. (1971), Beyth (1973) and Garland (1980) as alateral facies change.

In Eritrea, the base of the Edaga Arbi Beds is commonly sharp or gradational over a distance ofa few centimetres, and they always rests on theEnticho Sandstone. Glacial striae or grooves areobserved on the top of the Enticho Sandstone intwo localities in Eritrea: one in the village AdiMaEkheno (Fig. 9A) with the striae oriented 45°-225° and the other in the village Zeare (Fig. 9B),where the grooves trend 320°-140°. In AdiMaEkheno, the outcrop is a striated surface of theEnticho Sandstone, but it provides no clear rela-tionship with the overlying rocks. In Zeare, thegrooved surface is covered by mudstones of theEdaga Arbi Beds. Glacial grooves are also observedon top of Unit 1 (Fig. 3) of the Edaga Arbi sectionin Tigray (Fig. 9C). In Adi Mekeda in Eritrea, thecontact between these two units exhibits strikingevidence of liquefaction. The boundary is a ‘mixingzone’ (Fig. 10a), about 0.5 m thick, composed of sand clasts from the underlying Enticho Sandstone andmud clasts from the overlying Edaga Arbi Beds.Irregular sandstone dykes from the underlyingEnticho Sandstone cut this mixing zone and thelower 2–3 metres of the Edaga Arbi Beds.

The upper boundary of the Edaga Arbi Beds in Tigray (Dow et al., 1971; Garland, 1980), i.e. thetop of Unit 4, has been described as an undulat-ing unconformity, whereas the boundary betweenUnit 2 and 3 has not been described. The upperboundary of the Edaga Arbi Beds in Eritrea is com-monly depositional and conformable, but locally

Fig. 9 Glacial striae and grooves. (A) Glacial striaepointing to 45° on top of the Enticho Sandstone, AdiMaEkheno, (B) Glacial grooves, trending 320°, on top ofthe Enticho Sandstone, which is overlain by the EdagaArbi Beds (mudstone); hammer for scale, (C) Glacialgrooves on top of a diamictite, with a cobble projectingout of the bed top, west of Edaga Arbi (cf. Fig. 11,section 1).

GSP_4_C18.qxd 11/5/07 13:44 Page 327


Fig. 10 (A) The boundary zone between the Enticho Sandstone and the overlying Edaga Arbi Beds is a mixing zonecomposed of soft-sediment deformed sand clasts from the underlying unit and mud clasts from the overlying unit;hammer shaft c. 25 cm; Adi Mekeda; (B) Massive matrix-supported diamictite (Dmm), west of Edaga Arbi (Fig. 2), scale increments in centimetres; (C) Facies diamictite (Dmm) carrying itself a diamictite clast, Zeare; (D) A striated clast; (E) Facies diamictite (Dms) resting on a mudstone unit in the lower part of the Edaga Arbi Beds with inclined bedding,a possible glacitectonite or a creep deposit; (F) Climbing ripples (Sr) of type-A developing into type-B, west of EdagaArbi (Fig. 2), (G) Graded sand-silt couplets, which are slightly rippled; note the ice-rafted debris in the centre, west ofEdaga Arbi (Fig. 2); (H) Massive mudstone (Mm) with ice-rafted debris up to granule grade.

GSP_4_C18.qxd 11/5/07 13:45 Page 328


erosional. In the latter case, the boundary is coveredby a thin, less than 15 cm thick, conglomerate on topof the glacigenic unit. This is the case at Mawrayand Mendefera in Eritrea (Fig. 2). At Mendefera,the Adi MaEkheno Member is missing, so the con-glomerate rests on top of the mudcracked EdagaArbi mudstones and the cracks penetrate c. 30–35 cm into it. However, the cracks could also be seasonal frost cracks. They are filled with materialsimilar to that in the overlying thin granular to pebbly conglomerate.

Thickness variations

The thickness of the Edaga Arbi Beds varies fromone area to another, probably reflecting changes in palaeobathymetry and depositional conditions. In northern Ethiopia, the thickness of the Edaga Arbi Beds has been estimated by Dow et al. (1971)and Garland (1980) at c. 150 to 180 m. If the newcorrelation (see above), based on trace fossil evid-ence is correct, then the Edaga Arbi Beds in Tigraymay be limited to Units 1 and 2 (sensu Dow et al.,1971), and the thickness reduced to about 50 m.According to Beyth (1973), the Edaga Arbi Beds inTigray were accumulated in a depression, beingthickest along the axis of the depositional troughand wedging out laterally.

Earlier workers reported that the glacigenic succession thinned out towards the north beforereaching Eritrea. Work in Eritrea (Kumpulainen et al., 2006), demonstrates that the largest thicknessof the Edaga Arbi Beds, measured thus far, is about40 m at Mendefera village. From there, it thins west-wards and southwards to 10–15 m. Eastwards, inKohayto (Fig. 3) the unit is no more than 3 m thick.Further eastwards, in the Adeilo area along theEritrean coast, the Edaga Arbi Beds are probablyagain much thicker, perhaps several tens of metres,but the succession there is heavily dissected by faultsprecluding reliable thickness estimates.

Facies types

There are a number of important factors that influ-ence interpretations of various facies types and theprocesses that are responsible for their formation(Reading, 1996). Such factors include topographicaland bathymetrical conditions, circulation patternsand wave energy in marine and lacustrine settings.

Geomorphological studies in Eritrea and Ethiopia(Abul-Haggag, 1961) suggest, that the top of theNeoproterozoic schists is close to the peneplain over the area of study. The thickness of the EntichoSandstone (up to 300 m) has been proposed tocorrespond with irregularities in the substratumfilling of basement depressions (Mohr, 1971). Nosystematic morphological studies have been carriedout concerning the upper surface of the EntichoSandstone, but tracing the contact between theEnticho Sandstone and the Edaga Arbi Beds alongthe steep hillsides in Eritrea suggests that it isclose to planar (Kumpulainen et al., 2006). Hence,the Edaga Arbi Beds were deposited on a near-horizontal surface, limiting the types and import-ance of transportational and depositional processesresponsible for the formation of the facies assem-blage of the Edaga Arbi Beds, as well as the envir-onment in which they could have accumulated, the probable setting being a glacially influencedshallow continental shelf.

For this preliminary study six sections across the Edaga Arbi Beds were logged graphically; five sections in south-central Eritrea and one inTigray. Logging was generally centred on a narrow,1–2 m wide, profile perpendicular to the beddingin the succession. Occasionally, some beds weretraced laterally in order to characterise the geo-metry of the bed. About ten different facies typeswere encountered and described (Table 1); they aresummarised below.

Massive, matrix-supported diamictite (Dmm): Thisfacies is a massive, poorly sorted, commonlymaroon or greenish grey diamictite with sandy-muddy matrix (Fig. 10b, c). The framework clastsare commonly less than 0.5 m in diameter, but mayrange up to several metres. They are sub-roundedto rounded in shape and occasionally striated(Fig. 10d). The content of framework clasts variesfrom less than 10% to more than 30%. The lowerpart of a diamictite bed may exhibit deformationstructures, such as flow structures and isoclinal soft-sediment folds. The bed boundary may beslightly undulating, and indeed in some localitiesthis facies is loaded into the subjacent mudstone.In other places, it rests on laminated (Sh) and orripple-cross bedded sandstones (Sr), which in those cases may display reverse, soft-sediment faults. Thethickness of the beds of the Dmm facies may reachc. 3 m. It rests on various kinds of substratum, which

GSP_4_C18.qxd 11/5/07 13:45 Page 329


occasionally may be grooved or striated. This faciesis associated with Dms, Sd, Sr, Sm, Mm, Mh.

Interpretation: Three possible interpretations may be presented as to the origin of this facies. In case (i), the sequence displays essentially ‘vertical’deformation, such as diapiric rise of mud andconcomitant loading of the Dmm facies, and thelikely interpretation of formation would be dump-ing of the coarse- to fine-clastic melt-out load of a tilting iceberg and the deposition of a melt-outdiamictite. In case (ii), oblique deformation, suchas reverse faults or asymmetric folds, is present in the substratum of the facies. Accepting these data in isolation, the original site of depositionwould have been close to the grounding line of atidewater glacier. The probable deposits could be

interpreted as a waterlain till, a basal melt-out tillor a debris-flow (cf. Eyles et al., 1983). In case (iii), deposition of a diamictite and later deformation by an overriding glacier could have produced theoblique deformation structures, but recognised asfacies Dms (below).

Poorly stratified and poorly sorted diamictite (Dms):This facies is a poorly sorted and poorly stratified,maroon to grey rock with a sandy to muddy matrix,in which clasts of various lithologies occur in analigned fashion. These rocks are commonly folded(Fig. 10e) and display flow structures. Clast sizesvary from granule to pebble. The clasts are com-monly rounded to well-rounded. The content offramework clasts is low. These units may displayreverse grading, normal grading or no grading. In

Table 1 Facies types of the Edaga Arbi Beds, Eritrea and Ethiopia

Code

Dmm

Dms

Sm

Sr

Sh

Sd

SMg

Mh

Mm

IRD

Description

Massive, diamictite with sandy-muddy matrix, inwhich stones up to metre-size occur. The contentof frame-work clasts varies from less than 10% tomore than 30%

Diamictite with poor stratification. Sandy to muddymatrix, in which stones of various lithologies occur.Stone sizes vary from centimetre-size to10–20 cm.The content of framework clasts is low

Massive sand with possible dish structures andpoorly preserved bedding

Ripple-cross bedded and ripple-laminated sandstone.The ripples are commonly climbing

Horizontally laminated sandstones

Soft-sediment deformed sandstones

Normally graded sand-mud couplets commonly lessthan 1 cm thick. May be combined with Sr

Horizontally laminated mudstone

Massive, structureless mudstone

Outsized clasts. Grain sizes vary from sand toboulder. Occur in facies types Sr, SMg, Mm, Sh

Interpretations

Waterlain or basal melt-out diamictite

Waterlain or basal melt-out diamictite, a possiblecreep deposit or glacitectonite

Liquifaction or fluidisation and possibleredeposition of a stratified sand bed

Traction currents and, in case ripples climb, high-suspension clastic input

Lower or upper flow regime traction

Sediment instability, ice-push or earthquake

Possible cyclopsams, cyclopels, turbidites

Distal supension deposit influenced by tidalcurrents

Distal suspension deposit from suspension plumes;rain-out from icebergs; homogenisation byliquifaction, fluidisation

Ice-rafted debris melt-out from glacier base oricebergs. Difficult to recognise in coarse grained orhomogenised units

GSP_4_C18.qxd 11/5/07 13:45 Page 330


the lower part of some diamictite beds, the frame-work clasts (less than 10 cm in diameter) are rotatedand together define an overturned fold. Bed thick-nesses range up to 1–2 m. This facies may be associated with Dmm, Sr, Sm, Mm, Mh, SMg.

Interpretation: The most likely origin is as a water-lain or basal melt-out till with subsequent rework-ing by traction or gravity (cf. Eyles et al., 1983), butmay also be a creep deposit or a glacitectonite(Isbell et al., 2001).

Massive sandstone (Sm): This sandy facies containsdish structures and poorly developed lamination.The grain size is medium to coarse sand and thegrains are sub-angular to rounded. The content ofmatrix is very low. The lower bed boundary maydisplay loading structures. The thickness of bedsmay reach 4 m. This facies is associated with faciesMh, Mm and Sm.

Interpretation: The probable mode of deposi-tion was liquefaction or fluidisation and possibly redeposition of an originally stratified, subaqueoussandstone bed(s) or a rapidly deposited sand-flow or other density-flow unit (cf. Middleton &Hampton, 1973).

Ripple-cross laminated sandstone (Sr): Ripple-crosslaminated sandstone forms bedsets up to morethan a metre in thickness or, less commonly, thinsingle beds. Climbing ripples of both type-A- and type-B (sensu Ashley et al., 1982) are common(Fig. 10f) and may develop in the section from one end member to the other, particularly in thesuccession in Tigray. Successive cross-beds indicateunimodal as well as opposing transport polarities.This facies is associated with Dmm, Dms, Sh, SMgand commonly contains outsized clasts.

Interpretation: Ripple-cross bedding is formedby traction current in the lower flow regime (cf.Middleton & Southard, 1984). Formation of climb-ing ripples depends on the interplay betweenclastic input and current velocity. At high clasticinput and low current velocity, deposition takesplace also on the stoss side of a ripple (cf. Ashleyet al., 1982) and may cause a steep climbing angle(type-B). Increasing current velocity and low clastic input produces asymmetrical ripples withlower climbing angle (type-A). Climbing ripples are common in ice-proximal subaqueous sites,where icebergs provided the clastic contribution to the ice-rafted debris (Miller, 1996; Ó Cofaigh &Dowdeswell, 2001).

Horizontally laminated sandstone (Sh): Only a fewthin laminae, from less than a millimetre to bedsa few centimetres thick of horizontally laminatedsandstones, are recorded in the measured sections.The grain sizes are medium to coarse sand. Theboundaries with other facies are commonly sharpand abrupt. No grading is observed in this facies.This facies is associated with Mm, Mh, SMg andis transitional to SMg.

Interpretation: Horizontally laminated sandstonesmay be deposited from traction currents in thelower and upper flow regime (cf. Middleton &Southard, 1984). In a subaqueous, glacigenic setting,very thin sand laminae may form when sandymaterial rains-out from an iceberg and is depositedunder the influence of sea currents (Ó Cofaigh &Dowdeswell, 2001).

Soft-sediment deformed sandstone (Sd): Some sand-stone beds may display deformation features, suchas symmetric or asymmetric folds, and normal orreverse faults. The original depositional features,particularly lamination are still partly preserved.This facies is associated with Sh, Mm, SMg, Dmsand Dmm.

Interpretation: The most likely explanations aresediment instability resulting from deposition onover-steepened depositional slope, excess porepressure within the sand bed induced either by ice-push, seismic or tectonic activity or glacial loading(cf. Middleton & Hampton, 1973; Allen, 1984)

Normally graded sand-mud couplets (SMg): Norm-ally graded sand-mud couplets in this area arecommonly less than 1 cm thick and form units upto c. 0.5 m thick. A set of normally graded laminaemay develop into climbing ripples, where the indi-vidual laminae are also normally graded (Fig. 10g)and contain outsized clasts commonly in the sand-granule grade. This facies is associated with faciesMh, Mm, Sr, Gms.

Interpretation: The normally graded sand-mudcouplets may be glacial cyclopsams and cyclopels,deposited from overflow plumes, or thin turbidites(cf. Eyles et al., 1983; Mackiewicz et al., 1984;Cowan & Powell, 1990; Ashley 1995; Ó Cofaigh &Dowdeswell, 2001).

Horizontally laminated mudstone (Mh): Horizont-ally laminated mudstone facies comprises mudinterbedded with thin laminae and beds of finesandstone. It forms beds from a few centimetres up to c. 4 m in thickness, that extend laterally for

GSP_4_C18.qxd 11/5/07 13:45 Page 331


several metres or more. Boundaries to beds of otherfacies types, Mm, SMg, Sh, Sm, are commonly sharpand horizontal and only occasionally deformed. This facies contains outsized clasts.

Interpretation: The responsible mechanism forformation of horizontally laminated mudstone istraction currents in lower flow regime (cf. Middleton& Southard, 1984) or, in a subaqueous glacial set-ting, deposition from turbid overflow plumes (ÓCofaigh & Dowdeswell, 2001). The outsized clastsrepresent ice-rafted debris.

Massive, structureless mudstone (Mm): Massive,structureless mudstone is an important compon-ent, particularly in the Eritrean successions of theEdaga Arbi Beds. Some of these massive mud-stones contain outsized clasts (Fig. 10h), commonlyof sand or granules, whereas pebbles or largerclasts are less common. The outsized clasts arecommonly concentrated in certain, laterally per-sistent intervals of varying thickness and varyinglevels within this facies. The thickness of massivemudstone units may range up to c. 15 m. No bio-turbation has been observed in this facies, whichis commonly intercalated with facies Mh, Sh andSd; bed boundaries are sharp or diffuse.

Interpretation: The massive character of this faciesmay be interpreted in three different ways, by (i)rain out of fine-clastic material from icebergs (cf.Miall, 1983; Ó Cofaigh & Dowdeswell, 2001), (ii)homogenisation due to liquefaction or fluidisa-tion of mud (Middleton & Hampton, 1973) or (iii)bioturbation (Singer & Anderson, 1984). Some of themassive sandy mudstones may be fine-graineddiamictites (Ó Cofaigh & Dowdeswell, 2001). Notrace fossils have been observed in this unit so far,hence, bioturbation is an unlikely interpretation forits formation.

Outsized clasts: This is not a facies type of its own, but it contributes to the clastic component of other facies types. Outsized clasts (Fig. 10g) are common in the Edaga Arbi Beds in Eritrea and Ethiopia and they vary from sand to boulder, but the dominating fractions are sand to smallpebbles. Outsized clasts, which also may displaystriae, have been observed in facies types Sr, SMg,Mh and Mm.

Interpretation: Outsized clasts are commonly, butnot exclusively deposited as glacier ice melts andreleases the trapped clastic particles, which may be rain-out in a variety of settings and environ-

ments (Miller, 1996). Alternatively, and particularlyin terrestrial settings, larger clasts may roll, or betransported by wind into low-energy environments.Generally, the outsized clasts may be recognised inclearly finer-grained deposits. In the present case, theoutsized clasts are interpreted as ice-rafted debrisand the code IRD (ice-rafted debris) is thereforeapplied here. Striated ice-rafted clasts are typicalof glacilacustrine and glacimarine environments.

Studied sections

Interpretations of the various facies types encoun-tered in the study area are presented above. Theselected sections through the Edaga Arbi Beds aredescribed below with respect to the various faciestypes and the individual stratigraphies that theyexhibit. The interpretation of the depositional set-ting of the Edaga Arbi Beds in the study area isthen discussed, based on the information from thevarious sections collectively.

The Edaga Arbi section. The Edaga Arbi Beds havetheir type section near the village Edaga Arbi inTigray, northern Ethiopia (Fig. 2) and the sectionwas described by Dow et al. (1971). Depending on the interpretation outlined above, the Edaga Arbi Beds reach c. 50 m, as proposed in this paper,or 180 m in thickness, and may be divided into twoor four informal subunits (Fig. 3). The lowermostunit, Unit 1, rests on a striated Neoproterozoicbasement surface and has been described (Dow et al., 1971) as massive diamictite containing clastsup to c. 6 m in diameter, some of the clasts beingstriated.

For this paper, a section of the lowermost 20 mof the Edaga Arbi Beds, is described from a rivergorge c. 4 km west of the type section (Fig. 11, section 1; 14° 02′ 20″ N, 38° 59′ 15″ E). This sectionprobably corresponds to Unit 1 sensu Dow et al.(1971). Here, the succession is deposited in a minornorth-south trending valley, a few metres deep. Thissection apparently deviates from that describedby Dow et al. (1971), since it comprises nearlyequal proportions of diamictite and sand-dominatedbeds, not only ‘massive diamictite’. Many of thesand-dominated beds display ripple-cross bedding,including type-A and type-B climbing ripples (Sr), sensu Ashley et al. (1982), which in a sectionmay develop from one end-member to the other(Fig. 10f). Normally-graded varve-like laminae or

GSP_4_C18.qxd 11/5/07 13:45 Page 332


cyclopsams (SMg, Fig. 10g) are also common. In avertical section, the graded laminae may gradu-ally develop into climbing ripples, where the individual laminae retain normal grading fromsand to silt. Occasionally, these facies types dis-play soft-sediment deformation. Several of thesand-dominated units also contain outsized clasts.Four diamictite beds (Dmm, Dms), each about 2 mthick and three thinner diamictite beds (Dmm,Dms), less than 0.5 m thick are interlayered withthe sandy beds in this section. Three of the thickerbeds have a gradual, transitional boundary witheither the overlying or underlying sand-dominated

Fig. 11 Six logged sections, five from Eritrea and one west of Edaga Arbi in Tigray, Ethiopia, where only the lowermostpart of the glacigenic unit was studied. For facies codes see the text. Scale increments are in metres. For locations, seeFig. 2.

unit. The uppermost diamictite exhibits a sharplower and upper boundary. The top of this upper-most diamictite bed displays parallel groovestrending approximately north-south. Clast litho-logies in this upper diamictite bed include blackmarble, white quartzite, black schist, mica schist,chlorite schist, quartz, metabasalt, metarhyolite,quartz-K-feldspar-porphyry, K-feldspar-granite.Some of the diamictite beds or their boundaries display minor soft-sediment folds. Additionally,south-dipping, reverse soft-sediment faults, areencountered particularly on top of the second,from bottom, diamictite bed.

GSP_4_C18.qxd 11/5/07 13:45 Page 333


The Mawray section (14° 42′ 43″ N, 39° 15′ 38″ E).In the Mawray area (Fig. 11, section 2), the Edaga ArbiBeds is about 20 m thick. The section on the hillsidejust north of the village of Mawray exposes c. 6.5 mof the upper part of the formation. The lower partof the exposed section is dominated by massivemudstone (Mm), which is overlain by a structurelesssandstone bed (Sm), 15 cm thick. The uppermostmetre of the Edaga Arbi unit is composed of sandyto granular mudstone (Sm + Sh + IRD), interruptedby a thin, 10 cm, and distinctly laminated mud-stone bed (Mh). This glacigenic unit is overlain bya mud-clast dominated conglomerate, 10 cm thick,in the base of mudstones of the Adi MaEkhenoMember of the Adigrat Sandstone formation. TheAdi MaEkheno Member displays here poorly pre-served trace fossils similar to Palaeophycus tubularis(Hall, 1847). The contact between the Edaga ArbiBeds and the Adigrat Sandstone formation is con-formable, but may be slightly erosional.

The Betbey section (14° 51′ 13″ N, 39° 07′ 01″ E).The Betbey section on the hillside next to the oldvillage Betbey exhibits a complete section throughthe Edaga Arbi Beds, which here attain the thick-ness of 10 m (Fig. 11) and rest with a sharp yet conformable contact upon the Enticho Sandstone.The upper boundary is a diffuse transition intotrace-fossil-bearing sandy, mudstone of the AdiMaEkheno Member of the Adigrat Sandstone formation. The succession is characterised by reddish, laminated mudstone (Mh), which in someintervals displays normal grading (SMg) and contain occasional laminae in sand grade withsharp boundaries (Sh). Beds of massive sandstone(Sm; each <10 cm thick) are interbedded with themudstone. Some sandstone beds, (e.g. at 7–8 m onthe section), are lenticular and partly undulating.A few isolated outsized clasts are scattered through-out the section, but are particularly concentrated ina 1 m interval at about 8 m on the logged section(Fig. 11, section 3), where the largest clast is 18 cmin diameter. The uppermost sandstone bed, whichdisplays a normal coarse-tail grading, is covered by trace-fossil-bearing mudstone (c. 3 m thick)where the carbonate component increases upwards. The section ends in cross-bedded friable sand-stones with a rich Cruziana-type, trace fossil fauna(Kumpulainen et al., 2006).

The Adi Mekeda section (14° 53′ 14″ N, 39° 03′ 22″E). The Adi Mekeda village exposes an almost

complete section through the Edaga Arbi Beds(Fig. 11, section 4). The unit rests on the EntichoSandstone, and it is overlain by nodular mud-stones of the Adi MaEkheno Member of theAdigrat Sandstone formation. The top of theEnticho Sandstone and the base of the Edaga Arbi unit are both liquefied and form together a mixing zone (Sm + Mm), which is c. 0.5 m thick,and composed of clasts from both sedimentaryunits (Fig. 10a). Additionally, sandstone dykes cut this mixing zone and also intrude into theapparently liquefied mudstone (Mm). At about 2 m above the base, a probable sandstone bed is also liquefied (Sm) and forms a sill-like unit ofirregular thickness, 5–50 cm, and displays minorsandstone dykes in the base and the top of theliquified bed. The Adi Mekeda section is domin-ated by massive mudstone (Mm), although a 1 m thick interval at c. 4 m on the section exhibitsa horizontal lamination (Mh). Other interbeds are a single discontinuous, massive sandstone bed(Sm), about 5 cm thick, in the upper part of the section. Sand-grade ice-rafted debris is particu-larly common in two intervals, which are about 1 m thick each, one at 6 m and the other 11–12 m.A few clasts of diamictite are also observed.

Zeare section (14° 53′ 30″ N, 39° 06′ 25″ E). TheEdaga Arbi Beds are exposed almost continu-ously for at least 200 m along the hillside in theZeare village. The measured section, is located c. 50 m north of the church, and exposes an almostcomplete sequence, about 9 m thick, through theEdaga Arbi Beds (Fig. 11, section 5). Mudstone of this glacigenic unit rest conformably, with asharp contact on the presumably glacier-groovedtop of the Enticho Sandstone. Apart from thegrooves, no other deformation is observed in theunderlying sandstone. The upper boundary of the Edaga Arbi Beds is hidden in an unexposed gap, perhaps a metre wide. The unit is overlain by friable, cross-bedded sandstones of the AdigratSandstone formation, which carry a rich trace-fossil fauna. The basal c. 2 m of the Edaga Arbi Beds along the hillside is a mudstone-dominatedpart of the section. This lower mudstone unit maybe sub-divided into beds with characteristicallydifferent features. It begins with a laminated mudstone (Mh), c. 25 cm thick, which, except forthe distinctly-laminated lowermost 5 cm, containssandy ice-rafted debris. This basal unit is overlain

GSP_4_C18.qxd 11/5/07 13:45 Page 334


by a massive claystone (Mm) without ice-rafteddebris, containing irregularly shaped clasts of mud-stone with ice-rafted debris (Fig. 12A). The next bed,about 10 cm thick, is a laminated mudstone withdistinct sand laminae (Sh), whereas the upper-most 20–30 cm are again rich in mud clasts andcharacterised by some soft-sediment deformation.A mixed diamictite-dominated succession withlaterally discontinuous diamictite beds (Dmm)rests on top of this basal, mudstone-dominatedpart of the Zeare section. The diamictite beds are

separated by laminated and partly soft-sedimentdeformed mudstone, which rise diapirically upinto, and wrap around, the lower parts of theloading diamictite pillows. The diamictite has amuddy-sandy matrix, in which stones of granuleto cobble size are distributed. Stone lithologiesinclude granitoid rocks, schist and also diamictite(Fig. 10c).

Mendefera section (14° 54′ 25″ N, 39° 19′ 20″ E).The hillside north of the village of Mendefera northof Adi Keyih exposes a section, about 40 m thick,through the Edaga Arbi Beds. The lower bound-ary of the unit towards the underlying EntichoSandstone is sharp and conformable. The upperboundary of the unit is sharp and mudcracked. Theunit is overlain by the basal conglomerate, c. 10 cmthick, of Member 2 of the Adigrat Sandstoneformation. Laminated, ice-rafted debris-bearingmudstones (Mh + IRD) form the lowermost part of this glacigenic unit. Otherwise the lowermost 18 m are dominated by beds of homogenisedsandstone (Sm), 1.5 m and 4 m thick, respectively,and deformed muddy sandstone which forms units2.5 m and 7 m thick, respectively. The sandstonebeds display diffuse lamination and dish struc-tures; loading structures occur in the base of thebeds. The muddy sandstone units are pervasively soft-sediment deformed and contain a mixture ofirregular clasts of mudstone and sandstone, andadditionally are cut by a network of thin sandstonedykes (Fig. 12B). The proportions between sand and mud vary through these two deformed units.The upper half of the section is dominated bylaminated mudstone (Mh). The lower 4 m of theseupper mudstones are laminated, contain some ice-rafted debris and display only minor soft-sediment deformation. A diamictite bed (Dmm), c. 3–4 m thick, rests on the mudstones at c. 22 mon the section. This diamictite has a sandy matrixand a low content of pebbly outsized clasts. Itexhibits normal grading and a southeast-dippingfissility. The lower part of the mudstone above thediamictite is laminated (Mh) and contains some carbonate; the carbonate content increases upwardsand the mudstone becomes apparently massive in the upper part of the unit. The top of the mud-stone is cut by a network of mud cracks filled bymaterial in sand and granule grade, resembling the material in the basal conglomerate bed ofMember 2 of the Adigrat Sandstone formation.

Fig. 12 (A) A soft-sediment deformed massive sandymudstone (Mm) clast including ice-rafted debris in amassive claystone, Zeare; (B) Pervasively soft-sedimentdeformed sandstone-mudstone; m – mud clast, s – sandclast, d – sandstone dyke, Mendefera (Fig. 11, section 6.)

GSP_4_C18.qxd 11/5/07 13:45 Page 335


DISCUSSION

Pre-glacial morphology

Eritrea and Ethiopia were part of the pan-Africanorogenic mountain chain in early Cambrian timeand developed to a large peneplained landscape in Late Ordovician time. The reason for this pene-planation is uncertain. It could have been relatedto earlier events of the Ordovician glaciation. Varia-tions in the thickness of the Enticho Sandstone, inthe range of 0–300 m, have been proposed (Mohr,1971) to indicate filling of basement depressions by those sandstones, but those variations could alsodepend on depositional conditions (Fig. 4) andneed be assessed later.

In other parts of North Africa that were coveredby the Late Ordovician glaciers, sinuous channels,several tens of kilometres wide, a few tens ofmetres deep were cut in the subjacent rocks andsoft sediments. Much of the later glacial depositswere accumulated in these channels (Le Heron et al., 2004; Ghienne et al., this volume). Channeldirections coupled with information on deforma-tion structures together with glacial striae andgrooves in North Africa are indicated in Figure 13.The present study area in Eritrea and Ethiopia hasa limited extension, and the succession, particu-larly that in south-central Eritrea, can be accom-modated in a glacial channel of dimensions similarto those in other parts of northern Africa. A channel-like feature in which the Edaga Arbi Beds weredeposited was described by Beyth (1973) in Tigray

(Fig. 4). It is not an erosional valley, however, but depends on localized deposition of the EdagaArbi Beds and successive accumulation of traction-current deposits, i.e. the Enticho Sandstone. Thusfar, the data do not support the presence of largeglacial valleys in the Horn of Africa.

Depositional environments

A proximal depositional setting is inferred for thesuccession in the Edaga Arbi–Enticho areas of Tigray(Dow et al., 1971; Garland, 1980). Similarly, the newdata (Fig. 11, section 1) from the lowermost part ofthe Edaga Arbi Beds (i.e. Unit 1, sensu Dow et al.,1971; Fig. 3), west of the Edaga Arbi type section,supports this interpretation. Here, Unit 1 containsseveral coarse-clastic diamictite beds, which prob-ably were deposited sub-aqueously as melt-outdiamictite (Dmm) released from the ice-margin, andwere partly reworked by traction currents, par-ticularly melt-water currents from subaqueous icetunnels or by gravity transport (Dms). The dia-mictites are interbedded with sandstone beds com-posed of horizontally laminated sand (Sh), whichmay be graded (SMg) and form cyclopsams ofsand-silt (Fig. 10g). These facies types were prob-ably deposited from turbid overflow plumesunder the influence of tidal currents (Ó Cofaigh &Dowdeswell, 2001). The graded laminae couldalso be turbidites. The sandstone beds additionallydisplay cross-bedding (Sr), with climbing ripples oftypes A and B (sensu Ashley et al., 1982) developingin one case (Fig. 10f) in a vertical section, suggesting

Fig. 13 Distribution of the knownoccurrences of Upper Ordovicianglacigenic deposits in North Africa,Arabia and those in Eritrea andEthiopia. Compiled from Vaslet(1990), Klitzsch and Wycisk (1999)and Monod et al. (2003).

GSP_4_C18.qxd 11/5/07 13:45 Page 336


in this case waning current velocity. The observedcharacteristics support the interpretation that thestudied section through the lower part of the EdagaArbi Beds in Tigray was deposited in a subaqueousenvironment near a tidewater glacier margin.

The succession comprising Unit 1 displays soft-sediment deformation, such as asymmetric north-verging folds, south-dipping reverse faults andglacial grooves trending to the north on top of the uppermost diamictite bed of the logged section (Fig. 9C, Fig. 11). As this section was deposited ina south to north-trending minor valley, a generalglacier transport to the north is inferred. However,all these deformation structures are considered to be minor, and most sedimentary structures arewell preserved (Fig. 10F & G), which is unlikely tobe consistent with an ice-contact setting, unless the margin was retreating and exerted little or nostress on the deposits along the ice margin. Thedeformation structures could have been formed by the collapse of the ice-margin or from icebergs.In the event of prograding ice, much more severetypes of deformation, such as large-scale folds andassociated large faults, could have been expected,but such features have not been observed. How-ever, this does not preclude their presence in otherparts of Tigray and Ethiopia.

The evidence presented above suggests a subaqueous setting for Unit 1 of the Edaga Arbi Beds (Dow et al., 1971). As this coarse-clastic facies appears to grade laterally into the EntichoSandstone, which is consistent with models of ice-marginal sedimentation (Hambrey, 1994; Miller,1996; Le Heron et al., 2004), then the EntichoSandstone must also be of subaqueous origin andglacigenic, rather than fluvial as suggested pre-viously. The likely depositional site would have beensubaqueous outwash fans along the retreating icemargin, where the melt-out debris was continuallyreworked by melt-water currents, obliterating themassive sediments. Melt-out diamictites were firstpreserved in Tigray as the ice had already recededfrom Eritrea. This hypothesis needs be tested byfuture work.

In Tigray, Unit 1 is overlain by ice-rafted debris-bearing mudstones of Unit 2 (Fig. 3). According tothe earlier descriptions (Dow et al., 1971), Unit 2 in Tigray resembles the mudstone-dominatedglacigenic succession in Eritrea. These two succes-sions most probably constitute a mudstone unit

extending across the entire study area, covering the coarse-clastic glacial facies in Tigray and theEnticho Sandstone in Eritrea. This extensive mud-stone blanket is dominated by two facies types: (i) distinctly laminated mudstones (Mh) and (ii)massive mudstones (Mm). Beds of these faciestypes occur in various proportions in all sections,commonly exhibiting sharp bed boundaries, whichare conformable and sub-horizontal. The mudstonescommonly host ice-rafted debris, from sand togranule, rarely to boulder size, and also diamictiteclasts, whereas other intervals are devoid of ice-rafted debris. In a glacially-influenced environment,mudstones are deposited sub-aqueously, distal tothe contemporaneous ice margin, e.g. from turbidoverflow plumes (Mh) or by melt-out from ice-bergs (Mm). Alternatively, the massive mudstone(Fig. 10H) could have been deposited first as a laminated mudstone and then homogenised as a result of a glacier over-riding, or from an earth-quake chock.

Diamictite is a minor facies component in theEritrean succession and has, thus far, been encountersin three localities. Two of them are the Zeare andMendefera sections (Fig. 10C; Fig. 11), but they dis-play different characteristics. In Zeare, the diamic-tites are down-loaded into the underlying partlylaminated, partly reworked mudstone, while themudstone beds locally diapirically intrude the dia-mictite. No oblique deformation is observed here,and the likely interpretation is that the diamictitebeds were dumped from an iceberg that may havegrounded close to the measured section. The dia-mictite in the Mendefera section, which here occurs(at c. 22–26 m in the section) on top of an obviouslysoft-sediment deformed lower part, displays a clearsouth-dipping preferred clast orientation, suggest-ing either ice push or slow creep on a depositionalslope. In the third locality, half way between Zeareand Adi Keyih, the diamictite rests, as in the Zearesection, on a lower, but here massive, mudstone-dominated unit. The contact between these faciestypes is sharp and on the large scale planar, but in detail slightly undulating with the wavelengthless than 15 cm and the amplitude less than 5 cm.The lower part of the diamictite bed displays (Fig. 10E) rotated framework clasts (less than 10 cmin diameter) that together define an overturned fold. Again, this could be the product of ice pushor creep.

GSP_4_C18.qxd 11/5/07 13:45 Page 337


Cross-bedded sandstones are a minor componentin the Eritrean succession, whereas two beds of massive sand (Sm) are important components in thelower part of the Mendefera section. These two bedsrepresent a rapid input of fairly clean sand into thedepositional system probably by a gravity-flowmechanism or by homogenisation of a sequence ofsand beds. These sand beds occur in the similarlysoft-sediment deformed (Fig. 12B) mixed, sand-mudfacies (Sm + Mm) that was originally interlayeredthinly bedded sand and mud.

The distribution of the various facies types in theEritrean glacigenic succession presents a problem,since the coarse-clastic and less organised faciestypes (Dmm, Dms, Sm) occur in the northernmostknown successions (Fig. 2), that are interpretedhere as the most distal, as opposed to proximal inTigray. The presence of the coarse-clastic and lessorganised facies types (Dmm, Dms, Sm) in thedistal succession in south-central Eritrea may beinterpreted in two different ways: (i) as initial iceretreat towards the south, followed by a glaciergrowth in the north before the final retreat backtowards the south or (ii) deposition by dumpingfrom icebergs. In the first case, the growing glacierwould flow over earlier deposits and most likelyproduce a compacted basal till (cf. Evans et al., 2004). A rapid glacier advance in the Horn ofAfrica would be consistent with the interpreta-tions presented for some other successions of theLate Ordovician glacial record of North Africa (LeHeron et al., 2004; Ghienne et al., this volume).

Evidence that is in conflict with the hypothesisof a second glacier advance and over-riding of the entire Eritrean glacigenic succession derives from the measured sections. They, indeed, displayvarious degrees of soft-sediment deformation, yetin most sections distintcly laminated mudstoneforms the lowermost beds (e.g. Betbey, Zeare andMendefera), or that they are also interlayered withbeds of massive, sandy mudstone (e.g. Fig. 10H)exhibiting very little deformation. Neither basal till nor other features such as large-scale isoclinalfolding of the beds have been recognised in these sections. The most deformed section is thatat Mendefera, whereas Mawray and Betbey are least deformed. It is perhaps more likely that thecoarse-clastic facies were transported to the distalenvironment, e.g. Zeare, by icebergs. Likewise,the soft-sediment deformation, including that at

Mendefera, was produced by grounded icebergs.The diamictite in the middle of the section (Fig. 11)maybe also have been deposited from grounded icebergs.

Development of the glacigenic succession

Interpretation of the development of the glacigenicsedimentary succession in Eritrea and Ethiopiarelies on the following evidence: (i) an ice-proximalfacies in Tigray, which (ii) grades laterally over to the traction-current deposited Enticho Sand-stone, exhibiting (iii) cross-bedding with consistentlynorth-dipping foresets, (iv) palaeocurrents in Tigrayindicate transport to a general north.

Assuming retreat of a single ice-margin fromnorth to south, whilst still in Eritrea, the margin produced melt-out diamictite, which were con-tinually reworked by jet currents from subglacial,subaqueous melt-water tunnels. The melt-watertunnels also issued clastic debris into the sediment-ary system, forming underflow-fans (the EntichoSandstone). The ice-margin migrated to Tigray, per-haps becoming stationary there for some time andreleasing the coarse-clastic deposits that, more thanin other parts of the succession, display featuresindicating an ice-margin. Fine-clastic sedimentsaccumulated from turbid overflow plumes andproduced successively a mud-dominated blanket(Edaga Arbi Beds in Eritrea) on top of the older,gravelly and sandy deposits. Icebergs were dis-charged into this water body, rafting clastic mater-ial, single clasts and diamictite clasts as well aslarger quantities of debris to form beds of melt-outdiamictite in the distal environment. Locally, the icebergs probably came into contact with the softdeposits causing local deformation in the oldernon-lithified sedimentary succession. The deposi-tional environment is depicted in a cartoon inFigure 14.

Post-glacial development

The sections studied and earlier work byKumpulainen et al. (2006) also provide usefulinformation about sedimentation spanning theglacial and post-glacial times. Interestingly, whilstthe glacier in the south fed the basin from south tonorth, the subsequent clastic input in post-glacialtimes was from approximately north to south and

GSP_4_C18.qxd 11/5/07 13:45 Page 338


deposited on top of the glacigenic succession as thelarge-scale cross-bedded, Adi MaEkheno Member,which is probably a sub-marine dune complex.The source area for that northern material is notknown. Conditions for life improved greatly, so that a Cruziana type trace-fossil fauna was estab-lished, and is now preserved in the Adi MaEkhenoMember, maybe also in Member 2 of the AdigratSandstone formation.

The sub-marine dune complex migrated south-wards; it reached as far as the Mendefera area,where the thickest known glacigenic succession in south-central Eritrea forms a depositional andbathymetric high. Engulfing first that high, thedune complex migrated further south to at least the Genzebo area, a total distance of more than 40 km. It is possible that the mudstones of Unit 3(and maybe also Unit 4 of the Edaga Arbi Beds,sensu Dow et al., 1971) constitute the distal faciesof the Adi MaEkheno Member in Tigray. Furtherwork is required to assess these relationships. The Adi Maekheno Member wedges out towardsthe Mendefera depositional high. Otherwise, thelower boundary of the Adi MaEkheno Member iscomformable, whereas its upper surface is a dis-conformity overlain by a local thin conglomeratein the base of Member 2, suggesting a possiblyminor erosional event before deposition of thatmember.

At the start of deposition of Member 2, the trans-port polarity again reversed towards the north andtabularly cross-bedded sandstone beds accumulatedon top of the minor disconformity. It is interestingto note that Member 2 rests on top of the EdagaArbi Beds in the Mendefera high, where the muddy

Fig. 14 A schematic profile of glacialdepositional environment in south-central Eritrea and northern Tigray.

top of the high emerged for a certain time, and beingsubject to some erosion and desiccation, giving riseto a mud-crack network. The mud cracks were then filled by material similar to that in the thinconglomerate at the base of Member 2, indicatingthat this member began accumulating before theEdaga Arbi Beds were lithified. Alternatively, themud cracks may represent seasonal frost cracks (cf. Berg & Black, 1966; Péwé, 1966; Washburn,1970). Frost action may also have influenced the solidrocks, but in that case the cracks would probably bestraight, because if lithified, the mudstone wouldhave split into blocks. The cracks are not straight,however, but undulating and irregular, suggest-ing that they were formed in a non-lithified mud.Also frost action requires that the top of the EdagaArbi Beds had emerged above the waterline.

CONCLUSIONS

A glacigenic succession has been located in Eritreaand Ethiopia within the greatest limit of the Ordo-vician glaciation in the Horn of Africa and Arabia.The succession rests on a nearly horizontal non-conformity, underlain by Neoproterozoic low-grademetamorphosed volcanic sedimentary and intrusiverocks. This succession exhibits facies assemblagesthat indicate sedimentation in both proximal anddistal glacial environments, and was probablyformed during one cycle of deglaciation towardsthe end of the Ordovician Period. The glacigenicsuccession began accumulating in Eritrea with thedeposition of cross-bedded, subaqueous under-flow outwash fans, whilst at the same time most

GSP_4_C18.qxd 11/5/07 13:45 Page 339


of the massive melt-out debis was reworked andbecame incorporated in the cross-bedded succes-sion, the Enticho Sandstone that displays con-sistent transport polarity to the north. The glaciermargin retreated slowly towards Tigray, where an ice-proximal succession has been preserved,and where the lateral facies change from traction-current deposits to ice-proximal melt-out diamictitehas been reported previously. Transport-directionindicators, south-dipping reverse soft-sedimentfaults, north-verging asymmetric folds, cross-bedsand glacial grooves in Tigray suggest transporttowards the north.

In the distal setting, mud-dominated sedimentwas deposited from turbid overflow plumes underthe influence of currents. Massive and laminatedmudstone was formed. Icebergs rafted clastic debrisinto this distal setting, depositing clastic materialeither as single grains or as aggregated (diamictiteclasts) and also as beds of diamictite. Locally, ice-bergs disturbed the already deposited non-lithifiedmuddy sediments.

In other parts of North Africa and Arabia, twoice sheet advances have been recognised. The present interpretation for the succession in Eritreaand Ethiopia assumes only one retreat and no second advance, although it is possible that the areaexperienced two ice sheet advances, the secondadvance removing the deposits of the first retreat(interglacial). In that case, the deposits in Eritreaand Ethiopia would represent the second retreat.

Post-glacial deposition in the area is representedby a large-scale cross-stratified bed (Adi MaEkhenoMember; c. 20 m thick), a probable sub-marinedune complex. The south- or southwest-dipping clinoforms suggest that this complex has itssource area in the north and it migrated to the southor southwest. This unit rests conformably on theglacigenic succession although minor erosion islocally observed. Fossil evidence, particularly thetrace fossil Arthrophycus alleghaniensis (Harlan) sug-gests that the glacigenic unit is of Late Ordovician,possibly Hirnantian, age.

ACKNOWLEDGEMENTS

This study was financed by the University ofAsmara, Eritrea; Uppsala University, Sweden; theSwedish International Development Cooperation

Agency and Stockholm University, Sweden. Theideas presented here were gained from field visits and discussions with E. Ferrow, S. Ghirmay,T. Kreuser, A. Kumar, J. Shoshani, M. Teklay, B.Woldehaimanot, B. Zerai and the field course students (of January 1998). R. Bussert borrowed his copy of the geological map sheet Mekele. The author is grateful for all this support. Theauthor is also indebted to the two reviewers, D.P.LeHeron and O.E. Sutcliffe and the volume editorM.J. Hambrey, who all provided many, usefulcomments on an earlier version of the manuscript.

REFERENCES

Abul-Haggag, Y. (1961) A contribution to the physiographyof northern Ethiopia. University of London, The AthlonePress, 153 pp.

Allen, J.R.L. (1984) Sedimentary structures: their charac-teristics and physical basis. Elsevier, Amsterdam, 663 pp.

Arkin, Y., Beyth, M., Dow, D.B., Levitte, D., Haile, T. and Hailu, T. (1971) Mekele map sheet, ND 37–11, 1:250 000, Geol. Surv. Ethiopia (preliminary version).

Ashley, G.M. (1995) Glaciolacustrine environments. In:Modern Glacial Environments: Processes, Dynamics andSediments (Ed. J. Menzies), pp. 417–444. Butterworth-Heinemann. Oxford.

Ashley, G.M., Southard, J.B. and Boothroyd, J.C. (1982)Deposition of climbing-ripple beds: a flume simula-tion. Sedimentology, 29, 67–79.

Assefa, G. (1987) First record of Triassic Dinosaur foot-prints from the lower sandstone unit of Kersa area,eastern Ethiopia. Riv. Ital. di Paleontol. Stratigr., 93,171–180.

Berg, T.E. and Black, R.F. (1966) Preliminary measure-ments of growth of non-sorted polygons, VictoriaLand, Antarctica. Am. Geophys. Union Publ., 1418,61–108.

Beuf, S., Biju-Duval, B., De Charpal, O., Rognon, P.,Gabriel, O. and Bannacef, A. (1971) Les grès duPaléozoïque inférieur au Sahara. Sédimentation etdiscontinuités, évolution structurale d’un craton. Sci.Techniq. Pétrole, 18, Paris, 464 pp.

Beyth, M. (1973) Correlation of Paleozoic-MesozoicSediments in Northern Yemen and Tigre, NorthernEthiopia. AAPG Bull., 57 (for 1972), 2440–2446.

Blanford, W.T. (1869) On the geology of a portion ofAbyssinia. Q. J. Geol. Soc. London, 25, 401–406.

Bosellini, A. (1989) The continental margins of Somalia:their structural evolution and sequence stratigraphy.Mem. Sci. Geol., 41, 373–458.

GSP_4_C18.qxd 11/5/07 13:45 Page 340


Bosellini, A., Russo, A., Fantozzi, P.L., Assefa, G. andSolomon, T. (1997) The Mesozoic succession of theMekele Outlier (Tigre Province, Ethiopia). Sci. Geol.Mem., 49, 96–116.

Cowan, E.A. and Powell, R.D. (1990) Suspended sedi-ment transport and deposition of cyclically inter-laminated sediment in a temperate fjord, Alaska, USA.In: Glacimarine Environments: Processes and Sediments(Eds J.A. Dowdeswell and J.D. Scourse), Geol. Soc.London Spec. Publ., 53, 75–89.

Debyser, J., De Charpal, V. and Meralet, O. (1965) Sur lecaractère glaciaire de la sédimentation de l’Unité IV auSahara Central. C. R. Acad. Sci., Paris, 261, 5575–5576.

Deynoux, M. and Trompette, R. (1981) Late Ordoviciantillites of the Taodeni Basin, West Africa. In: Earth’spre-Pleistocene glacial record (Eds M.J. Hambrey andW.B. Harland), pp. 89–96. Cambridge Univ. Press.Cambridge.

Dow, D.B., Beyth, M. and T. Hailu, (1971) Palaeozoicglacial rocks recently discovered in northern Ethiopia.Geol. Mag., 108, 53–59.

Evans, J., Pudsey, C.J., Ó Cofaigh, C., Morris, P. andDomack, E. (2004) Late Quaternary glacial history, flow dynamics and sedimentation along the easternmargin of the Antarctic Peninsula Ice Sheet. Quatern.Sci. Rev., 24, 741–774.

Eyles, N., Eyles, C.H. and Miall, A.D. (1983) Lithofaciestypes and vertical profile models; an alternativeapproach to the description and environmental inter-pretation of glacial diamict and diamictite sequences.Sedimentology, 30, 393–410.

Garland, C.R. (1980) Geology of the Adigrat Area. Geol.Surv. Ethiopia, Mem., 1, 1–51.

Ghienne, J.-F. (2003) Late Ordovician sedimentary envir-onments, glacial cycles, and post-glacial transgres-sion in the Taoudeni Basin, West Africa. Palaeogeogr.Palaeoclimatol. Palaeoecol., 189, 117–145.

Ghienne, J.-F., Heron, D., Moreau, J. and Deynoux, M.(this volume) The Late Ordovician glacial sediment-ary system of the North Gondwana platform.

Hall, J. (1847) Paleontology of New York, vol. 1. C. VanBenthuysen, Albany, 338 p.

Hambrey, M.J. (1981) Palaeozoic tillites in northernEthiopia. In: Earth’s pre-Pleistocene glacial record(Eds M.J. Hambrey and W.B. Harland), pp. 38–40.Cambridge Univ. Press. Cambridge.

Hambrey, M.J. (1985) The Late Ordovician–Early Silurianglacial period. Palaeogeogr. Palaeoclimatol. Palaeoecol.,51, 273–289.

Hambrey, M.J. (1994) Glacial environments. UCL Press Ltd.London, 296 pp.

Harlan, R. (1831) Description of an extinct species of fossil vegetable, of the family Fucoides. J. Acad. Nat.Sci. Philadelphia, 6 (for 1830), 289–295.

Hutchinson, R.W. and Engels, G.G. (1970) Tectonicsignificance of regional geology and evaporite litho-facies in Northeasten Ethiopia. Phil. Trans. Roy. Soc.London, A 267, 313–329.

Isbell, J.L., Miller, M.F., Babcock, L.E. and Hasiotis, S.T.(2001) Ice-marginal environment and ecosystem priorto initial advance of the late Palaeozoic ice sheet inthe Mount Butters area of the central TransantarcticMountains, Antarctica. Sedimentology, 48 (5), 953–970.

Klitzsch, E. und Wycisk, P. (1999) Beckenentwicklung undSedimentationsprozesse in kratonalen BereichenNordost-Afrikas im Phanerozoikum. In: Nordost-Afrika: Strukturen und Ressourcen: Ergebnisse aus demSonderforschungsbereich ‘Geowissenschaftliche Probleme in ariden und semiariden Gebieten’. (Eds E. Klitzsch and U. Thorweihe), pp. 61–108. Wiley-VCH Verlag,Weinheim.

Kumpulainen, R.A., Kumar, A. and Woldehaimanot, B.(2002) The Horn of Africa in the cross road of twoglaciations; new evidence from Eritrea and Ethiopia.16th Int. Sed. Congr., Johannesburg. Abstr., 204–205.

Kumpulainen, R.A., Uchman, A., Woldehaimanot, B.,Kreuser, T. and Ghirmay, S. (2006) Trace fossil evid-ence from the Adigrat Sandstone for an Ordovicianglaciation in Eritrea, NE Africa. J. Afr. Earth Sci., 45,408–420.

Le Heron, D., Sutcliffe, O., Bourgig, K., Craig, J.,Visentin, C. and Whittington, R. (2004) Sediment-ary architecture of Upper Ordovician tunnel valleys,Gargaf Arch, Libya: Implications for the genesis of ahydrocarbon reservoir. GeoArabia, 9 (2), 137–160.

Mackiewicz, N.E., Powell, R.D., Carlson, P.R. andMolnia, B.F. (1984) Interlaminated ice-proximalglacimarine sediments in Muir inlet, Alaska. Mar.Geol., 57, 113–147.

Miall, A.D. (1983) Glaciomarine sedimentation in theGowganda Formation (Huronian), Northern Ontario.J. Sed. Petrol., 53, 477–491.

Middleton, G.V. and Hampton, M.A. (1973) Sedimentgravity flows: mechanics of flow and deposition. In: Turbidites and Deep-Water Sedimentation (Eds D.J. Stanley and D.J.P. Swift). SEPM Short Course, 1,1–38.

Middleton, G.V. and Southard, J.B. (1984) Mechanics ofsediment movement. SEPM Short Course 3, 1–401.

Miller, J.M.G. (1996) Glacial sediments. In: SedimentaryEnvironments: Processes, Facies and Stratigraphy (Ed. H.G. Reading), pp. 454–484. Blackwell. Oxford.

Mohr, P. (1971) The Geology of Ethiopia. Haile Selassie IUniversity Press. Addis Ababa. 268 pp.

Monod, O., Kozlu, H., Ghienne, J.-F., Dean, W.T.,Günay, Y., Le Hérissé, A., Paris, F. and Robardet, M.(2003) Late Ordovician glaciation in southern Turkey.Terra Nova, 15, 249–257.

GSP_4_C18.qxd 11/5/07 13:45 Page 341


Ó Cofaigh, C. and Dowdeswell, J.A. (2001) Laminatedsediments in glacimarine environments: diagnostic criteria for their interpretation. Quatern. Sci. Rev., 20,1411–1436.

Péwé, T.L. (1966) Ice-wedges in Alaska – classification,distribution, and climate. Nat. Acad. Sci., Nat. Res.Council Can., 1287, 76–81.

Reading, H.G. (1996) Sedimentary Environments: Processes,Facies and Stratigraphy. Blackwell. Oxford. 688 pp.

Sagri, M., Abbate, E., Azzaroli, A., Balestrieri, M.L.,Benvenuti, M., Bruni, P., Fazzuoli, M., Ficcarelli, G.,Marcucci, M., Papini, M., Pavia, G., Reale, V., Rook,L. and Tecle, T.M. (1998) New data on the Jurassicand Neogene sedimentation in the Danakil Horstand Northern Afar Depression, Eritrea. Mém. Mus.natn. Hist. Nat., 177, 193–214.

Salvador, A. (1994) International stratigraphic guide. 2nd Ed.Int. Union. Geol. Sci. and Geol. Soc. Am., Boulder,Colorado. USA. 214 pp.

Saxena, G.N. and Assefa, G. (1983) New evidence on the age of the glacial rocks of northern Ethiopia.Geol. Mag., 120, 549–554.

Singer, J.K. and Anderson, J.B. (1984) Use of total grain-size distribution to define bed erosion and transportfor poorly sorted sediment undergoing simulatedbioturbation. Mar. Geol., 57, 335–359.

Sutcliffe, O.E., Dowdeswell, J.A., Whittington, R.J.,Theron, J.N. and Craig, J. (2000) Calibrating the Late Ordovician glaciation and mass extinction by the eccentricity cycles of Earth’s orbit. Geology, 28,967–970.

Vaslet, D. (1990) Upper Ordovician glacial deposits inSaudi Arabia. Episodes, 13, 147–161.

von zur Mühlen, L. (1931) Zur Geologie der Gegend vonHarrar und Deder in Ostabessinien. Z. Deut. Geol. Ges.,83, 625–634.

Washburn, A.L. (1970) An approach to genetic classi-fication of patterned ground. Acta Geogr. Lodz., 24,437–446.

Zanettin, B., Bellieni, G., Justin-Visentin, E. and Haile, T. (1999) The volcanic rocks of the EritreanPlateau; stratigraphy and evolution. Acta Vulcanol., 11,183–193.

GSP_4_C18.qxd 11/5/07 13:45 Page 342

glacial sedimentary processes and products (hambrey/glacial sedimentary processes and products) ||...

Documents