Extensional basin geometry and

the elastic lithosphere

By C. J. E b i n g e r1 , J. A. Jackson2 , A. N. Foster2 a n d
N. J. Hayward1
1

School of Earth Sciences, University of Leeds, Leeds LS2 9JT, UK

2
Department of Earth Sciences, Bullard Laboratories,
University of Cambridge, Cambridge CB3 0EZ, UK

Although the morphology and dimensions of continental rift basins vary considerably worldwide, one aspect is similar; tectonically active rifts are bordered on one or
both sides by relatively long (tens of kilometres) normal fault systems (termed border faults) that largely control basin morphology. We compile data constraining the
geometry of border faults within the tectonically active East African Rift system, and
evaluate these results with respect to variations in thickness of the elastic lithosphere.
Border-fault lengths greater than 75 km occur in regions with deep crustal seismicity
and relatively high estimates of effective elastic thickness (Te ) derived from forward
and inverse models of gravity and topography data (Te > 25 km). Most East African
border faults cross-cut pre-existing structures and basement foliations, although segments of the longest faults (greater than 80 km) reactivate Precambrian shear zones
or structural fabrics. From observations in East Africa, comparisons with data from
the Aegean and Baikal Rifts, and considerations of the rheology of continental lithosphere, we propose that the elastic lithosphere determines the length, width and style
of faults within East Africa, and perhaps other continental rifts.
Keywords: border fault; continental rift; seismicity; rheology;
East Africa; elastic thickness

1. Introduction
Continental lithosphere undergoing extension develops basins bounded on one or
both sides by normal faults, but the dimensions and internal structure of these basins
vary along the length of rift systems, as well as between rift systems worldwide
(see, for example, Rosendahl 1987; Jackson & White 1989; Ruppel 1995). In crosssection, a majority of tectonically active continental rift basins shows a half graben
morphology, with most of the strain accommodated along the border, or master,
fault bounding the deep side of the basin (e.g. figure 1a). In plan view, displacement
decreases toward the tips of border faults where they interact with faults bounding
adjacent basins and/or transfer faults accommodating differential horizontal and
vertical displacements between rift basins (see, for example, Trudgill & Cartwright
Present address: Department of Geology, Royal Holloway College, University of London, Egham
TW20 0EX, UK.
Present address: Esso Exploration and Production UK, Mailpoint 25, Esso House, Ermyn Way,
Leatherhead KT22 8UY, UK.
Present address: Expro IS, Shell UK Exploration, 1 Altens Farm Rd, Aberdeen AB1 3FY, UK.
Phil. Trans. R. Soc. Lond. A (1999) 357, 741765
Printed in Great Britain
741

c 1999 The Royal Society

TEX Paper

742

C. J. Ebinger and others

04 km

rift flank uplift

basin width

07 km

10 >100 km

10 - 60 km
vvv

uplifted

border fault
v vv

flank

monocline

vv

Figure 1. Generalized form of tectonically active continental rift basins showing border faults in
cross-section (a) and plan form (b). (a) Most single-phase rift basins are half graben, with most
of the deformation accommodated by slip along the border fault. Flexural isostatic compensation
leads to uplift of the rift flanks, with the breadth and height dependent upon the strength of
the elastic plate, and, to a lesser extent, and the density of infilling material (see, for example,
Weissel & Karner 1989). The lower relief margin is commonly faulted, or is a monocline dipping
toward the basin. We measure the breadth of basins using the mean level of topographic relief
ca. 200 km away from the basin, as indicated by the open circles. (b) Border faults show maximum
displacement near the centre of the segment, with decreasing throws toward their tips usually
taken up by several smaller faults, or by monoclines. The approximate line of (a) is shown by
the light line. The elliptical area of depression is shown by shading.

(1994); and figure 1b). The geometry of border faults at mid- to lower crustal levels,
however, is poorly constrained.
Within the fault population of any one basin, the border faults are usually the
longest faults and have the greatest displacement. Shorter smaller offset faults often
are seen in the hanging wall of the border fault (see, for example, Hayward & Ebinger
1996). Seismicity data show that deformation is concentrated along the border-fault
system, since the largest earthquakes occur along the longest faults (see, for example,
Roberts & Jackson 1991). Thus, border faults are the principal tectonic component
of rift basins, and they provide insights into the deformational behaviour of the
continental lithosphere. As demonstrated in this comparison of representative border
faults in the East African, Aegean and Baikal Rifts, the lengths of border faults differ
by a factor of four between rift systems, a difference we attribute to variations in
thickness of the elastic lithosphere.
Jackson & White (1989) suggest a positive correlation between
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

743

(1) the thickness of the seismogenic layer and the maximum length of faults; and
(2) the effective elastic plate thickness and the maximum width of basins,
but their observations are restricted to fault lengths and seismogenic layer thicknesses less than 30 km. The unusually deep (ca. 35 km) seismicity and long (greater
than 50 km) border-fault segments in East Africa (see, for example, Ebinger et al .
1989; Jackson & Blenkinsop 1997) occur in relatively strong lithosphere (see, for
example, Bechtel et al . 1987; Ebinger et al . 1991). Hayward & Ebinger (1996) noted
a systematic decrease in the length and width of seismically active rift segments in
the Afar Rift with decreasing effective elastic thickness, suggesting that the rheology
of the upper continental lithosphere exerts a strong control on rift-basin geometry.
In this study, we report systematic variations in border-fault length and basin
width within parts of the East African Rift system, which encompasses extensional
segments of various ages (30 Ma to less than 2 Ma), progressive degrees of extension
( < 1.1 to > 2.0), and those with and without any surface expression of volcanism (see, for example, Ebinger 1989; Hayward & Ebinger 1996; and figure 2). We
then show that border-fault length and basin width in the East African, as well as
the Aegean and Baikal, Rifts increase with the effective elastic plate thickness and
seismogenic layer thickness, with implications for predictive models of rift basins and
the rheology of continental lithosphere.

2. East African rift basins

The seismically and volcanically active rift systems of East Africa have developed
atop two broad plateaux: the Afar Rift and Main Ethiopian Rift (MER) transect
the 1000 km wide Ethiopian plateau; the Eastern and Western Rifts cut the 1300 km
wide East African plateau (figure 2). The broad plateaux, their corresponding negative Bouguer gravity anomalies, the generally small degrees of extension, and the
geochemistry and large volume of eruptive volcanic products are evidence for one
or more mantle plumes beneath East Africa (see, for example, Mohr 1983; Ebinger
et al . 1989; Marty et al . 1996). The Turkana depression between the Ethiopian and
East African plateaux may be a topographic low between two discrete swells (Stewart & Rogers 1996), or the depression may be caused by isostatic compensation for
crust thinned during Mesozoic and/or Palaeogene rifting (see, for example, Bosworth
1992; Hendrie et al . 1994).
The 21 mm a1 , NESW-directed extension in the Red Sea, Gulf of Aden, and
eastern Afar determined from geodetic studies (see, for example, Ruegg et al . 1993),
changes to the much slower (38 mm a1 ) sub-EW-directed extension in the MER
(see, for example, Asfaw et al . 1992; and figure 2). The rate and direction of extension
determined from the MER geodetic observations are consistent with current plate
models and extension directions seen in earthquake focal mechanisms of teleseismic
events (Jestin et al . 1994; Foster & Jackson 1998). Within the Eastern and Western rifts where extensional velocities are poorly constrained, well-determined focal
mechanisms show predominantly dip-slip movement in directions nearly orthogonal to the approximately NS-striking border faults (see, for example, Nyblade &
Langston 1995; Foster & Jackson 1998).
Although the earliest record of rifting is buried by volcanic products in Afar, initial
volcanism and faulting along the length of the Red Sea commenced at ca. 32 Ma
Phil. Trans. R. Soc. Lond. A (1999)

744

C. J. Ebinger and others

RED
SEA

15N

L. Cenozoic
East African

Afar

Dobe
L. Addo-Do

Rift System

Gulf
ADEN

Tana

10N

Ethiopian

ER

Plateau

Tu

rka

5N

na
CB
L. Turkana
De
pre
ssi
on
Eastern

Albert
Western
0

Rift
5S

Lake
L.
Rutanzegi Victoria
L. Kivu

Rift
Natron

Eyasi
East
African
Plateau
L. Tanganyika

Manyara
INDIAN

L. Rukwa
Karonga
OCEAN
Malawi rift

10S

L. Nyasa

BililaMtakataka
15S
30E

35E

40E

Figure 2. Summary tectonic map of the East African Rift system showing location of the Tanzanian craton and specific rift basins described in more detail in the text. Bold shaded lines
enclose plateau elevations above 1000 m. Shaded arrows show extension direction in the Afar
and Main Ethiopian rifts (MER). CB is Chew Bahir Basin.

(see, for example, Omar & Steckler 1996; Menzies et al . 1997), and extension in the
Gulf of Aden had begun by 30 Ma (see, for example, Menzies et al . 1997). Sea-floor
spreading has propagated into the easternmost Afar depression during the past 5 Ma
(see, for example, Manighetti et al . 1998). Faulting and sedimentation began within
the northern Eastern Rift and the Main Ethiopian Rift (MER) by ca. 20 Ma (see, for
example, WoldeGabriel et al . 1990; Morley et al . 1992). Faulting and volcanism have
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

745

Table 1. Comparison of continental rift segment properties

(C is Te estimate from coherence analyses, F is Te estimate from forward modelling. See text
for references to fault length, basin width and -factor. Rift sectors shown on figure 2; MER
is Main Ethiopian Rift; ER is Eastern Rift; WR is Western Rift; AG is Aegean Rift; CBK is
Central Baikal Rift.)
rift
sector

basin

Afar
Afar
MER
ER
ER
ER
WR
WR
WR
AG
CBK

Dobe
Addo-Do
Chew Bahir
Natron
Manyara
Eyasi
Albert
Karonga
Bil-Mtaka
Corinth
South Basin

width
(km)
12
15
38
38
50
47
45
60
65
37
65

fault
length (km)

Te
z
}|
{
C (km) F (km)

34 1
27 1
65 5
67 + 2/ 2
82 + 10/ 5
97 + 10/ 5
75+
120 + 10/ 2
100
30 5
120 + 30/ 5

6a
6a
18a
24e
24e
> 40e
30c
35a
35c
10h
50j

5a
< 10b
ca. 15b
ca. 20b
ca. 35b
ca. 35b
24f
34d
ca. 35g
8h
40k

Tseis
(km)
< 10b
< 1.5
< 1.3
< 1.1
< 1.1
< 1.1
ca. 20b
ca. 35g
< 1.1
< 15h
ca. 40j

-factor
< 1.5

< 1.2
ca. 1.1
ca. 1.4
< 1.2

Data sources for Te and Tseis estimates:

a

Hayward & Ebinger (1996);

Ebinger et al . (1991);
g
Camelbeeck & Iranga (1996);
k
van der Beek (1997).

Foster & Jackson (1998);

Ebinger et al . (1997);
h
King (1998);

Ebinger et al . (1989);
Upcott et al . (1996);
j
Petit (1996);
f

propagated southward in the Eastern and largely avolcanic Western Rifts during the
past 12 Ma (see, for example, Ebinger et al . 1989; Foster et al . 1997; and figure 2).
Crustal thickness estimates from seismic refraction and receiver function studies
indicate thinning by a factor of two or more in northern Afar, but decreasing to a
factor of less than 1.5 in the northern MER, and in the Eastern and Western Rifts
(see, for example, Mechie et al . 1994; Last et al . 1997; and table 1).
East African Rift border faults have developed in metamorphic basement rocks
ranging in age from Archaean to Pan-African (more than 500 Ma), as well as in flood
basalt sequences of OligoceneRecent age. Numerous researchers have noted that
the East African extensional provinces formed along the margins of the Tanzanian
craton within Proterozoic orogenic belts, and suggest that border-fault geometries
are controlled by Precambrian upper crustal structures (see, for example, Wheeler
& Karson 1989; Theunissen et al . 1996; and figure 2). Below, we evaluate the local
relationship between MioceneRecent East African border faults and pre-existing
upper crustal structures to assess basement controls on border-fault geometry.

3. Rift segment dimensions

A brief summary of constraints on border-fault geometry, kinematics and seismicity
patterns within East African Rift segments is provided below and in table 1. We
use published cross-sections to illustrate the general form of East African basins and
Phil. Trans. R. Soc. Lond. A (1999)

746

C. J. Ebinger and others

E. Eyasi Basin L = 97 km
Yaida b.

Balangida b.

km

(a)

ODO fault

Natron Basin L = 67 km

km

(b)

km

50

Manyara Basin L = 87 km

3
(c)

Moinik
fault

km
3
3

(d)

Karonga Basin L = 120 km

km km
3

Livingstone fault

Albert Basin

3
(e)

Well

p C strata

km
3

Bunia fault

Omo
Basin

3
( f)

Toro-Bunyoro fault

Chew Bahir Basin L = 65 km

km
3

Western Afar
Escarpment
3
(g)

Addo-Do L = 24 km
Basin
"new" BF

"old" BF

Somali
Escarpment

km
3

(h)

(i)

1.5

Dobe Graben L = 27 km

1 1.5
Late Cenozoic
sedimentary rocks

km
0

km

50

Figure 3. For description see opposite.

Phil. Trans. R. Soc. Lond. A (1999)

Cenozoic volcanic rocks

Extensional basin geometry and the elastic lithosphere

747

uplifted rift flanks, but subsurface data are insufficient to characterize the internal
basin forms (figure 3). Basins are depressions bounded by one border-fault system
that comprises one or more linked fault segments (see below), but the along-axis
segmentation of basins may differ from that of border-fault segments, depending
on the geometry of accommodation zones (see, for example, Trudgill & Cartwright
1994; Anders & Schlische 1994). Normal fault systems may propagate and eventually link to form a continuous border-fault system during the earliest stages of basin
development, with the record of this along-axis propagation preserved in intrabasinal highs and relay-ramps (see, for example, Anders & Schlische 1994; Trudgill &
Cartwright 1994). The border-fault systems bounding East African basins described
below developed during the last 115 Ma, and some show signs of this propagation
and linkage process.
A comparison of rift-basin cross-sections along the length of the East African Rift
system shows the broad basins and uplifted flanks in the rifts that have developed
in ArchaeanProterozoic lithosphere (figure 3). The breadth of the uplifted flank
decreases with decreasing effective elastic thickness (see, for example, Weissel &
Karner 1989; Ebinger et al . 1991), and we see that the breadth also decreases with
fault length L, if we exclude areas of volcanic construction.
Descriptions are listed in an approximate evolutionary order, commencing with the
youngest and least extended regions. Border faults have been mapped along their
length in the field or in high resolution (less than 30 m) Landsat imagery (Ebinger
1989; Wheeler & Karson 1989; Hayward & Ebinger 1996; Jackson & Blenkinsop 1997;
Upcott et al . 1996; Foster et al . 1997). Wherever possible, we report the lengths of
other border faults within the same rift sector, and the relationship between metamorphic basement structures and Cenozoic faults. Estimates of maximum extension
factor () estimated from seismic data and, more commonly, reconstructions of fault
geometries are also listed for regional comparisons (table 1).
Fault segment length cannot be assessed without a definition of scale. Dawers &
Anders (1995) show that continuous faults at one scale are in fact segmented. Lateral
or en echelon offsets as small as 12 km across strike are known to halt earthquake
ruptures (see, for example, Yielding 1985), and we use 2 km as the maximum offset
allowed along a continuous fault segment (Foster 1997). Foster (1997) and LeTurdu
et al . (1999) examined sharp bends in fault traces in the Eastern Rift, and found no
offsets or changes in elevation of beds within the layered volcanic sequences forming
the escarpment, indicating that faults were continuous through these bends.
There are several sources of error in our fault length measurements. First, the resolution of remote sensing imagery and topographic data make it difficult to determine
the termination of fault segments where displacements approach zero. Given a low
recurrence rate of major earthquakes in East Africa (see, for example, Ambraseys
Figure 3. Representative cross-sections of East African Rift basins shown in figure 2. Elevations are from
1:50 000 topography maps and gravity station heights. The top diagram of the Dobe graben (h) is shown
at the same scale as the others, but (i) cross-section is a smaller scale in order to show details of the basin
form. L is the length of the border fault (table 1). In (a) ODO is Ol Donyo Ogol Fault; MF is Moinik
Fault. In (g) old BF is the initial OligoMiocene Border Fault and new BF is the PlioceneRecent
Border Fault. (a), (b), (c) from Foster et al. (1997), (d) from van der Beek et al. (1998); (e) from Upcott
et al. (1996), (f ) from Ebinger & Ibrahim (1994); (g), (h) and (i) from Hayward & Ebinger (1996). Note
the decrease in the breadth of the uplifted rift flank with decreasing fault length (and effective elastic
thickness, table 1).
Phil. Trans. R. Soc. Lond. A (1999)

748

C. J. Ebinger and others

1991a, b; Foster & Jackson 1998), alluvium and volcanic flows may cover the low
relief tips of fault segments. A second source of error comes in the interactions with
other faults, including intrabasinal and transfer faults (see, for example, Trudgill &
Cartwright 1994; Anders & Schlische 1994). Therefore, we restrict our data-set to
those border faults (comprising one or more fault segments) that are isolated and
show clear terminations. Basin widths are measured relative to the mean elevation
away from the uplifted rift flanks, as shown in figure 1a (figure 3, table 1).
Few historic earthquakes in East Africa have any known surface fault breaks to
compare with border-fault lengths, but Jackson & Blenkinsop (1997) present evidence showing that the entire length of border faults can slip in a single earthquake.
Although large earthquakes in East Africa are relatively infrequent, Ambraseys
(1991a, b) reports two historic earthquakes of magnitude greater than 7.0 within
the East African Rift.
(a) Southern end of the Eastern Rift: Eyasi, Natron, Manyara Basins
The southern end of the Eastern Rift is unique; faults have developed in the
Tanzania craton (Archaean), rather than the Late Proterozoic orogenic belts ringing
the craton (figures 2 and 4). Mantle xenoliths (see, for example, Dawson 1992) and
tomographic models (Ritsema et al . 1998) indicate that the thermal lithosphere is
thicker than 200 km beneath the craton, considerably more than the lithosphere
beneath the Eastern Rift north of Lake Natron, which is less than 100 km thick
(see, for example, Birt et al . 1997). There is no clear change in crustal thickness or
velocity structure across the ArchaeanLate Proterozoic surface contact (Last et al .
1997), and Archaean lithosphere may continue east beneath thin-skinned thrusts of
the Pan-African orogenic belt.
The NE-striking Eyasi border fault is the largest fault within the Eastern Rift, and
it cuts across the surface contact between Archaean and Pan-African (ca. 600 Ma)
crust (figures 4 and 6a). This 47 km wide half-graben is partly filled by 23 km of
sedimentary and volcanic strata as estimated from models of gravity and magnetic
data (Ebinger et al . 1997). The Eyasi border fault developed at about 3 Ma (Foster
et al . 1997).
The Natron Basin is a half graben bounded on its western margin by several long
normal faults, the longest of which is the 67 km long, east-dipping Moinik Fault
that cuts a parallel-bedded sequence of basalts, constraining the age of the fault to
less than 1 Ma (see Foster et al . 1997; and figure 3b). The ca. 85 km long Ol Donyo
Ogol Fault lies ca. 15 km west of the Moinik Fault, and this apparently inactive fault
marked the margin of an earlier (Mid-Pliocene?) basin (see, for example, Foster et
al . 1997).
The western margin of the ca. 50 km wide, ca. 100 km long Manyara Basin is
bounded by seismically active high-angle normal faults, the longest of which is at
least 82 km in length (figure 3c). The northern part of this segment cuts the less
than 1 Ma flood basalt sequence, whereas its southern part cuts Pan-African basement, making it difficult to assess its age and continuity there. South and west of
the Manyara Basin, seismically active, greater than 40 km long, normal faults are
morphologically the youngest faults in the region (see, for example, Balangida Fault,
figure 4 and Foster et al . 1997). They may mark the initial stage of development of
long border faults.
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

PlioceneRecent
sediments

PlioceneRecent
volcanics

o
o

Quaternary
Ash/Tuff

ARCHAEAN 35 PAN -AFRICAN

Pliocene-Recent
Normal Fault

v Monocline

o o
o
o

Volcano

o
o
o

o
o

o
o

o
o

Plains
B

o
o
o

MF

o o
o
o
o
o

C'

ga

ru

ka

o
o

Arusha

ida

Ya

b. MBULU
PLATEAU

vv

Balangida
fault

M
A
N
Y
A
R
A

vv

b.

A
AN
IA

3S

En
b.

vv
W
e
m
b
e
r
e

NY

NZ

E. EYASI
BASIN

W. EYASI B.

KE
TA

B.

o
o

N
A
T
R
O
N

Archaean crust

o
o

C
ODO
fault

o
o
o

Serengeti o o
o
o

o
o

749

MASAI

4S
B'

PLATEAU
Iramba
Plateau

Bubu
b.

A'

Singida
Mponde
basin

36E

Figure 4. Summary geological map of the EyasiNatronManyara Basins at the southern end
of the Eastern Rift, illustrating the predominance of long normal faults and wide basins. AA0 ,
BB0 and CC0 are lines of cross-sections shown in figure 3. Elevations are from 1:50 000 topography maps and gravity station heights. ODO is Ol Donyo Ogol Fault.

Although there is a parallelism between rift faults and NS to NESW striking

Pan-African structures at the regional scale, recumbent Pan-African folds are thin
skinned (Shackleton 1993) and are cross-cut by steeply dipping Late Cenozoic normal
faults. Basement foliations and shear zones within the Pan-African belt are locally
highly variable, and there is little correlation between basement structures and faults
along the eastern margins of the Natron and Manyara Basins (Foster et al . 1997).
Precambrian diabase dykes are subparallel or coincident with segments of the Eyasi
Border Fault, suggesting that some parts of the Eyasi fault zone reoccupy basement
shear zones.
Earthquakes detected both teleseismically (Nyblade & Langston 1995; Foster &
Jackson 1998) and locally occur throughout the 35 km thick crust beneath this region,
which is the most active seismically in East Africa (see, for example, Nyblade et al .
1996).
(b) Western Rift: BililaMtakataka, Karonga and Albert Basins
The ca. 100 km long BililaMtakataka Fault bounds the western side of the southern Malawi Rift where less than 1 km of sedimentary strata have accumulated (see
Phil. Trans. R. Soc. Lond. A (1999)

750

C. J. Ebinger and others

3445

3430

LAKE
NYASA

55

Mtakataka

14
15'S

45
70

Southern
Malawi
65

50

Rift

1430'S
0

km

16

85

70

1445'S

Basement
foliations
Basement
fracture
Cenozoic
normal
fault

70
50
45

Bilila

45
50

3430

3445

45

Figure 5. Summary structural map of the BililaMtakataka border-fault segment showing the
geometry of the ca. 100 km long border fault and representative strikes and dips of Late Proterozoic basement foliations and fractures (from Jackson & Blenkinsop 1997). Shaded area shows
the approximate extent of the sedimentary basin at the foot of the border fault.

Jackson & Blenkinsop 1997; and figure 5). The dip of the fault plane is generally
greater than 60 NE (Jackson & Blenkinsop 1997). Although there is a strong correlation between the strike of this fault and Late Proterozoic basement foliations at
the regional scale, foliations show more complex deviations in strike and dip at the
outcrop scale (Jackson & Blenkinsop 1997; and figure 5). Some segments of the fault
also show parallelism with sub-vertical WNW-striking fractures (figure 5).
An earthquake at a depth of 32 5 km occurred immediately to the north of the
fault in 1989, and field observations suggest that most or all of the 100 km long
BililaMtakataka fault slipped in a Late Quaternary earthquake (Jackson & Blenkinsop 1993, 1997). Jackson & Blenkinsop (1997) estimate that a Mw 8.0 earthquake
would be required to cause slip along this fault, which is much larger in magnitude
than typical continental rift events. Normal faulting earthquakes of comparable magPhil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

751

(a)
Ngorongoro
Oldeani

Pliocene tuffs

EYASI
BASIN

Sadiman

Quaternary
sediments
Pliocene basalts

Basement

SI

Southern
Eastern Rift
< 5 Ma

A
EY
E. .
B

b.

Ya

AN
HAE

OIC
OZ
TER

AR C

0 km 20

PRO

ida

SI B

A
. EY

(b)
Quaternary
sediments
Pliocene basalts
Miocene
felsites
Oligo-Miocene
basalts
Precambrian
Basement

CHEW BAHIR
BASIN
Southern
MER
< 18 Ma

5N

Arbore

0 km 20

Chew
Bahir
Basin

Omo
Basin

37E

t len
men
seg

ADDO-DO
BASIN
Southern Afar
Rift

gth

(c)

< 5 Ma
(~25 Ma)

Addo-do
Basin

ac
zoncommo
e
d

atio

n
Pliocene-Recent
sediments
Quaternary basalts
and rhyolites
Quaternary fissure
basalts
Pliocene basalts and
rhyolites

10 N

Major eruptive centres

0 km 20

41E

Normal fault

Figure 6. Fault patterns within three representative East African Rift basins showing the change
in scale of the border faults and basin dimensions as elastic plate thickness decreases, as estimated from seismogenic (Tseis ) and effective elastic (Te ) thickness. Approximate age of faults
indicated to the left of diagrams. All diagrams at the same scale. The less than 5 Ma old Addo-do
Basin developed in the hangingwall to an ca. 20 Ma old western Afar border fault (figure 3g).
Note the increase in number and displacement of intrabasinal faults as Te decreases.
Phil. Trans. R. Soc. Lond. A (1999)

752

C. J. Ebinger and others

nitude, however, have occurred in oceanic lithosphere seaward of trenches, causing

rupture at depths of 40 km or more (see, for example, Kikuchi & Kanamori 1995).
The ca. 60 km wide Karonga Basin at the northern end of the Malawi Rift is
bounded by the 120 km long Livingstone Fault. The northern part of the border
fault had developed by ca. 8 Ma (Ebinger et al . 1991). Depth to pre-rift basement is
greater than 4 km at the base of the Livingstone Fault, constraining throw along the
fault to greater than 6 km (Wheeler & Karson 1989; van der Beek et al . 1998; and
figure 3d). Only minor basement-involved faults occur across the basin, with nearly
all of the less than 10 km crustal extension accommodated along the Livingstone
Fault (Wheeler & Karson 1989; Ebinger et al . 1991; and figure 3d).
The central and southern parts of the Livingstone Border Fault are exceptional
in East Africa, in that they reoccupy Precambrian ductile shear zones (Wheeler &
Karson 1989; Theunissen et al . 1996). Here, brittle fractures and pseudo-tachylite
developed within the most highly strained parts of a steeply dipping Late Proterozoic
shear zone (Wheeler & Karson 1989).
Seismicity patterns recorded on local networks show a distribution throughout the
ca. 40 km thick crust (Camelbeeck & Iranga 1996; Zhao et al . 1997). Teleseismic data
show centroid depths of ca. 25 km for two earthquakes located to the north of the
Karonga Basin, but no teleseismic events have been recorded in this basin (Foster &
Jackson 1998). Seismically active, long (more than 120 km) normal faults north of
the Karonga Basin are excluded from the database because at least parts of these
border faults developed during a Permo-Triassic (Karoo) rifting episode, and they
show evidence for oblique-slip movement during one or both rifting episodes (see,
for example, Delvaux et al . 1998). One of these long faults, the 170 km long Kanda
Fault, may have been the site of the 1910 Rukwa earthquake, the largest recorded
in Africa (Ambraseys 1991a; Delvaux et al . 1998).
The northwestward-tilted North Albert Basin is bounded to its northwest by the
75 km long Bunia fault (Upcott et al . 1996; and figure 3e). The Bunia segment
could be up to 20 km longer, but access to the southern portions of the fault in
Zaire was impossible. Another long fault with smaller throw, the 58 km long Toro
Bunyoro Fault, bounds the opposite side of the basin, and its subsurface geometry
is well constrained by borehole, gravity and magnetic data (Upcott et al . 1996; and
figure 3e). No eruptive volcanic centres occur along the border faults, but a few
Neogene carbonatitic vents are located south of the rift (see, for example, Upcott et
al . 1996).
Foliations within Archaean to Early Proterozoic metamorphic basement strike NS
to NESW throughout the area, but dips are highly variable. Bends in the Toro
Bunyoro Fault near its southern termination cross-cut these basement structures,
arguing against reactivation of pre-existing upper crustal structures.
A 1966 dip-slip earthquake southwest of Lake Albert led to the development of a
20 km long fault scarp with 1.8 m displacement (Ambraseys & Adams 1986). Foster
& Jackson (1998) show centroid depths of less than 20 km for earthquakes in the
Albert Basin and Ruwenzori horst to its south.
(c) Main Ethiopian Rift: Chew Bahir Basin
The southern Main Ethiopian Rift (MER) developed in Pan-African lithosphere
that was affected by rifting in Cretaceous and probably Palaeogene time (see, for
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

753

example, Bosworth 1992; Hendrie et al . 1994). Thus, geothermal gradients were

probably higher than in adjacent rift sectors (Hendrie et al . 1994), and the initial
crustal thickness may have been less than the 3540 km found in most of East Africa
(see, for example, Mechie et al . 1994; Last et al . 1997).
Miocene basalts along the eastern margin of the Chew Bahir Basin are offset more
than 1000 m by displacement along the ca. 65 km long border fault (figures 3f and
6b). Its southern termination can only be estimated to within 5 km because it is
covered by landslide deposits of felsic ignimbrites (see, for example, Davidson 1983).
The 40 mGal negative Bouguer anomaly observed across the Chew Bahir Basin can
be explained by 12 km of sedimentary to volcanic rocks within the basin (see Ebinger
& Ibrahim 1994; and figure 3f ).
There is little evidence for reactivation of generally low angle gneissic foliations
in metamorphic basement or Mesozoic normal faults, where basement is exposed
(Davidson 1983). Several earthquakes were detected teleseismically in the region, and
aftershocks were recorded on a local network (Asfaw 1992). Focal depths of 9 and
14 km were determined from centroid moment tensor inversions for two earthquakes
with dip-slip mechanisms orthogonal to border faults, but the epicentral locations
are too inaccurate to determine if the border faults are still active (Foster & Jackson
1998).
(d ) Afar: Addo-do, Dobe Basins
Morphological evidence from the Western Fault line escarpment to the greater than
250 km wide Afar depression suggests that the short Addo-do and Dobe segments
are young, and developed within the hanging wall to originally longer border faults
(see Hayward & Ebinger 1996; and figure 3g). The Late Oligocene western Afar
escarpment has undergone considerable erosion, but drainage patterns into the Afar
and away from the rift along the westward-tilted uplifted flank indicate that Early
Miocene fault lengths were more than 50 km (see, for example, Weissel et al . 1995).
The centroid depth for an earthquake located near or along the escarpment is ca. 7 km
(Foster & Jackson 1998), but there is little evidence in the field or in remote sensing
imagery for active faulting along the eroded escarpment. Thus, we believe that active
extension within the highly extended Afar Rift occurs along the short fault segments
in the central Afar depression, and that long faults along the western escarpment are
inactive (figure 3a).
PlioceneRecent Rift basins in the central Afar depression developed in thinned
lithosphere, and total extension across these basins is the cumulative extension since
rift initiation. Extension estimates in the new basins of Afar shown in table 1,
however, represent extension from the PlioceneRecent stage, when the new alongaxis segmentation developed, to keep these estimates consistent with others shown
in table 1.
Faults bounding the Addo-do Basin cut PliocenePleistocene flood basalts, and
Quaternary eruptive centres occur near the tips of border-fault segments (e.g. figure 6c). The largest normal fault is 27 km long, whereas the mean length of fault
segments is 5 km, with fault length decreasing toward the Central Basin (see Hayward 1996; and figures 3 and 6c).
Strain in the 15 km wide Dobe graben is accommodated along border and intrabasinal faults that cut Quaternary basalt flows, producing stratal dips of 3035 (see
Phil. Trans. R. Soc. Lond. A (1999)

754

C. J. Ebinger and others

Hayward & Ebinger 1996; and figure 3i). The longest fault in the population of 273
faults measured from Landsat and space shuttle imagery is 35 km, the master fault
to the Addo-do Basin, whereas the mean length of faults is only 5 km (Hayward
1996).
Nowhere is metamorphic basement exposed in or near the Addo-do or Dobe Basins
to assess basement controls on PlioceneRecent faults. A large earthquake and 10
teleseismically detected aftershocks in 1989 were located along border faults bounding the Dobe Basin (Sigmundsson 1992). Braunmiller & Nabelek (1990) determine
depths of 47 km for these earthquakes.
(e) Summary
Although the data-set is small, consistent patterns are seen along the length of
the East African Rift system, regardless of proximity to, or volume of, magmatism
(table 1, figure 3). The border fault and basin scales also appear to be independent
of the age, metamorphic grade, or orientation of the Precambrian basement in which
the rift system has developed. Long border-fault segments bounding broad basins
with broad uplifted flanks characterize the young basins with small degrees of extension that developed in cratonic lithosphere. Basins within the southern MER (e.g.
Chew Bahir), where lithosphere was heated and stretched in Cretaceous time, show
similar patterns, but fault segments are shorter and basin widths narrower. Segments within the northern MER (e.g. Addo-do) and southern Afar (Dobe) preserve
the long border-fault phase along the eroded escarpments bounding the rift, but a
new, shorter segmentation has developed within the highly extended and intruded
central rift floor (see, for example, Hayward & Ebinger 1996; and figures 3g and 6c).
The consistent patterns in border-fault geometry within a number of Precambrian
and Archaean terrains with different structural trends argues against a basement
control on rift dimensions. From a regional analysis of border faults by Ring (1994),
border-fault segments may reactivate pre-existing basement structures, particularly
cataclastic strike-slip shear zones, if these structures are orientated within ca. 30
of the extension direction, similar to laboratory studies. Thus, the orientation of
a particular border-fault segment may be sub-parallel to pre-existing upper crustal
structure where the stress field is favourably orientated (e.g. Eyasi Basin), but the
actual fault lengths are more closely linked to the earthquake rupture process (see, for
example, Das & Scholz 1983). Some parts of the longest faults (BililaMtakataka,
Eyasi, Karonga), however, show reactivation of basement shear zones or pre-rift
dykes, suggesting that pre-existing fractures enhance fault development.

4. Correlation with effective elastic and seismogenic thicknesses

The lithospheric response to the faulting process is two-fold: there is a co-seismic
deformation of the lithosphere that occurs over periods of years; and a flexural isostatic response over periods greater than 104 105 years (see, for example, Stein &
Barrientos 1985; Weissel & Karner 1989). The longer-term isostatic response, as well
as erosion and deposition, modifies the co-seismic displacement field. Direct models of the co-seismic and isostatic response of the continental lithosphere, however,
are complicated by our poor understanding of its rheology and composition (see,
for example, Ma & Kusznir 1994; Burov & Diament 1995). If this rheology can be
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

755

represented by an equivalent elastic plate (effective elastic thickness, Te ), we can

predict the response of lithosphere with different thicknesses to faulting processes,
and compare models with observations. Clearly, though, more complicated models
are needed to explain the detailed stratigraphy within any discrete rift-basin segment
(see, for example, Contreras et al . 1997).
Scholz (1988) argues that the thickness of the seismogenic layer (Tseis ) is controlled by the depth-dependent frictional behaviour of the upper crust, which in turn
depends on geothermal gradient and crustal composition. Effective elastic thickness
(Te ) does not usually correspond to any depth horizon or the actual thickness of any
particular lithospheric layer, but it is believed to represent the combined effective
strength of the crust and upper mantle, which varies with composition, strain rate
and geothermal gradient (see, for example, Burov & Diament 1995). If the upper
mantle of continental lithosphere is weak, the seismogenic layer may be a proxy for
effective elastic thickness (see, for example, McKenzie & Fairhead 1997). Considering
these properties, the duration of heating from below from a mantle plume and/or
stretching, as well as the volume of magmatic material accreted to the lithosphere,
will modify elastic thickness, as measured by variations in Te and Tseis .
Te within rifts has been estimated from forward models of topography and gravity
anomaly profiles, and from two-dimensional inverse models of the coherence between
gravity anomalies and topography (see, for example, Weissel & Karner 1989; McKenzie & Fairhead 1997). It is commonly assumed that these rift gravity anomalies arise
from warpings of surface and Moho after faulting and isostatic compensation. Forward models require constraints on the density of infilling material, -factor (amount
of extension), and depth to detachment (necking depth). If models are required to fit
both gravity anomalies and topography, Te is usually well bracketed (see, for example, Weissel & Karner 1989; Ebinger et al . 1991). The Bouguer coherence method is
the most commonly used inversion technique, and it is a spectral method that may
be biased to stronger values due to the spatial averaging of data from rift basins and
flanks (see, for example, Ebinger et al . 1989; McKenzie & Fairhead 1997). We report
both estimates wherever possible (table 1).
Table 1 and figure 6 show that seismicity extends throughout the crust in the
relatively strong, weakly magmatic or amagmatic, mildly extended rift segments,
but that it decreases to less than 10 km in the central Afar depression. Clearly,
the period of seismicity observations is extremely short, but these south to north
variations are mirrored in the variations in Te (table 1). Estimates of effective elastic
plate thickness based on the coherence between the topography and Bouguer gravity
as a function of wavelength vary from ca. 6 to greater than 40 km, with the lowest
values found in the highly attenuated lithosphere of Afar.
McKenzie & Fairhead (1997) present an alternative approach to Te estimations
and their interpretations. A detailed comparison of techniques is outside the scope
of this study, but we provide a summary of arguments presented in King (1998).
McKenzie & Fairhead (1997) use the free-air admittance method to obtain Te estimates of ca. 15 km for the Eastern Rift, considerably smaller than estimates of greater
than 30 km obtained by Ebinger et al . (1989, 1997). They assume that topographic
relief is compensated within the crust, implying that the upper mantle has no longterm strength. The more traditional Bouguer coherence analysis, however, considers
the gravity anomaly to be the sum of surface and subsurface (approximately Moho)
density distributions, or topographic relief, at the base of the continental crust, as
Phil. Trans. R. Soc. Lond. A (1999)

756

C. J. Ebinger and others

Figure 7. Comparison of border-fault length and (a) seismogenic layer thickness (Tseis ); (b)
effective elastic thickness (Te ); and (c) basin width using observations listed in table 1. All of
the diagrams show a linear increase in fault length as Te , Tseis and basin width increase. Note
that is higher in the Corinth Basins.

imaged in seismic data (Ebinger et al . 1989). In both the free-air and Bouguer coherence techniques, sub-lithospheric density contrasts with length-scales of ca. 1000 km
(dynamic contributions) may lead to spectral aliasing (see, for example, McKenzie &
Fairhead 1997). Forward model estimates from several East African basins support
the Bouguer coherence results (table 1), and we use the Bouguer coherence estimates
as a comparative base throughout this study.
Figure 7 summarizes the correlations between border-fault segment length and
Te and Tseis , respectively. Both figure 7a and 7b show a linear increase in borderfault segment length with increasing thickness of the elastic lithosphere, as estimated
by Te and Tseis , respectively. The comparison of basin width and fault length also
shows an apparent linear trend, but this relation probably holds only for basins with
small extensional strains. Weissel & Karner (1989) model progressive stages of basin
development at constant Te , and show that basin width increases with stretching
factor, . The comparatively greater width for Te or fault length (L) seen in the
Gulf of Corinth outlier (table 1, figure 7c), could indicate that intrabasinal faults
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

757

accommodate significant extension, or reflect the shift in the location of border faults
with time (Ori 1989).

5. Discussion
If our relationships are correct, we should see similar fault length versus elastic thickness patterns in other rifts. Extensional basins within the Baikal and the Aegean Rifts
should represent the two end-members seen in East Africa: thick stronger lithosphere
of the Eyasi Basin compared to thin weaker lithosphere of the Addo-do Basin.
The Baikal Rift developed within and along the margin of the Siberian craton,
where geothermal gradients are low, seismicity occurs throughout the crust (Deverchere et al . 1991; Petit 1996), and estimates of effective elastic plate thickness
from Bouguer coherence analyses and forward models are 3050 km (Diament &
Kogan 1990; Ruppel et al . 1993; van der Beek 1997). Seismicity data indicate that
extension direction varies along the length of the Baikal Rift, and this variability
may contribute to its along-axis segmentation (Petit 1996). The broad Central Basin
of Lake Baikal, which contains ca. 7 km of sediments and water, is bordered to its
northwest by the more than 120 km long Primorskiy Fault, which was delineated from
Landsat imagery calibrated by field studies (Agar & Klitgord 1997). The subparallel
synthetic ca. 150 km long Morskiy fault defines the deep lake basin (see, for example, Agar & Klitgord 1995), but the continuity of this sub-lacustrine fault cannot be
evaluated in the same manner as other border faults used in this study. Antithetic
normal faults on the opposite side of the Central Basin show considerably smaller
throws and flank uplift, producing a faulted half graben cross-sectional morphology
(Petit 1996). The dimensions of the Central Baikal Basin and its location along the
margin of Archaean craton are similar to the tectonic characteristics of the Karonga
and Eyasi Basins of East Africa (table 1, figure 7).
Extensional basins within the Aegean region developed over the past 20 Ma in relatively hot lithosphere of a collapsing orogenic belt (see, for example, Roberts & Jackson 1991). Discrete basins within the Corinth Rift contain ca. 3 km of sedimentary
strata, and flanking uplifts rise more than 1000 m above the surrounding topography (King 1998; A. Hirn, personal communication). Border-fault lengths of 2530 km
within the Corinth and Evvia Rifts have been identified using spatial variations in
fault displacements and kinematic indicators detected in the field and remote sensing imagery (see Roberts & Jackson 1991; and table 1). Seismogenic layer thickness
determined from teleseismic and local arrays is 1015 km (see, for example, Roberts
& Jackson 1991). Te estimates are less than 15 km from the coherence model, and
less than 10 km from forward modelling (see King 1998; and table 1).
Are these values of Te and Tseis consistent and compatible with independent constraints on faulting processes and lithospheric rheology? The deep seismicity and
seismic velocities seen in the southernmost Eastern Rift and central Baikal Rifts,
which developed in Archaean lithosphere, suggest that the lower crust here has a
more mafic bulk composition (Nyblade & Langston 1995; Foster & Jackson 1998).
The high strength and deep seismicity are consistent with the low geothermal gradients found in the southern part of the rift system (see, for example, Nyblade et
al . 1990). In the EyasiNatronManyara Rift Basins of the Eastern Rift, there is no
clear evidence for lateral variations in crust and upper mantle velocity structure or
thickness across the Tanzania CratonPan-African orogenic belt boundary (Last et
Phil. Trans. R. Soc. Lond. A (1999)

758

C. J. Ebinger and others

al . 1997), suggesting that the deep seismicity within the Tanzanian craton primarily
results from low geothermal gradients, rather than compositional changes (Foster et
al . 1997).
Are Te and Tseis measuring similar lithospheric properties, and what can we infer
from their spatial variability? We can gain some information from numerical and analogue models. Forsyth (1992) estimates the stress required to cause a given amount
of extension in a finite-strength lithosphere. Expanding on these relations, Hayward
(1996) and Scholz & Contreras (1998) show that normal fault length and displacement increase with increasing Te . Models of seismicity data indicate that the maximum fault rupture length scales with the thickness of the strong layer(s) within
the deforming media (Das & Scholz 1983). The two- and three-dimensional numerical models of Gupta & Scholz (1998) suggest that elastic plate models can explain
many aspects of small-scale extensional fault arrays without requiring complicated
viscoelastic models (see, for example, Burov & Diament 1995; Petit 1996). Their
results may be applicable at the scale of basin-bounding faults (100 102 km) (Schlische et al . 1996), since the scaling relationships are similar (Gupta & Scholz 1998).
Comparing Te and Tseis , one could argue that the equivalent or slightly higher
Te estimates for a given border-fault length suggest that the continental mantle
lithosphere also deforms elastically and is relatively strong. Unfortunately, it is the
higher estimates of Te determined from coherence studies that are subject to the
largest errors, owing to data coverage and spectral averaging (Ebinger et al . 1989;
McKenzie & Fairhead 1997). Forward models of basin and flank morphology provide
Te estimates that are 220% less than Bouguer coherence results that are probably
biased to higher values, making Te and Tseis values roughly equivalent within the
errors of our observations (table 1).
Clearly, other processes will influence the border-fault length relations described
in this paper, but we consider these to be second-order effects. Probably the most
obvious influence is that of pre-rift basement fabrics and structures. The longest
border faults in our database have strikes that are sub-parallel to pre-existing shear
zones and dykes within metamorphic basement (e.g. Eyasi, BililaMtakataka), and
these zones of weakness may allow individual earthquakes to propagate a greater
distance for the same seismogenic layer thickness (see, for example, Kostrov & Das
1988). Thus, these border faults may be anomalously long due to the combined
effects of favourably orientated cracks, as well as a thick elastic lithosphere. Further
consideration of these factors requires a larger database of basin dimensions, and
better understanding of continental lithospheric rheology.
The consistent correlations between Te or Tseis and border-fault length suggest that
measurements of fault length distributions and geomorphology can be used to predict
initial rift-basin lengths and widths, as well as the spacing between accommodation
zones, in active rift zones. We would expect that both Te and Tseis will change after
rifting has ceased, but the border-fault length can still be used to predict the general
basin width and tectonic setting.

6. Conclusions
Tectonically active continental rift basins are bounded by long border-fault systems
that control the breadth and length of basins, the form of their uplifted flanks,
and, indirectly, depositional patterns within basins. Our comparison of border-fault
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

759

dimensions in the East African, Baikal and Aegean Rifts show that long (greater than
80 km) border-fault segments bound deep broad rift basins that develop in cold thick
lithosphere (Western and Eastern Rifts, Baikal); short (less than 30 km) border faults
bound narrow basins within initially weak lithosphere (Aegean), or develop as a new
segmentation within highly stretched and intruded lithosphere (Afar). The less than
30 km long, ca. 20 km wide rift segments in the Afar Rift overprint an older, longer
rift segmentation preserved in eroded escarpments bounding the western margin of
the Afar depression.
Border-fault length shows a linear increase with seismogenic layer thickness (Tseis )
and with effective elastic thickness (Te ), both comparative measures of elastic plate
thickness. These patterns observed within the East African Rift system, as well
as the Baikal and Aegean Rifts, indicate that the fault and flexural response of
the continental lithosphere to rifting processes is largely controlled by the elastic
lithosphere. Further studies in other active continental rifts worldwide, however,
are needed to extract information on the complicated rheology of the continental
lithosphere from variations in fault length, Te , and Tseis . Our results suggest that
we can assess the form of rift basins and flanks and tectonic setting during the synrift stage using measurements of border-fault length, improving predictive models of
rift-basin development.
The work of A.F. and J.J. was sponsored by grants from Shell and Amerada Hess. A.F. was
supported by NERC grant GR9/997A, N.H. by NERC grant GR9/496A. C.E. was, in part,
supported by a grant from Hunt United Exploration. Reviews by P. Cowie, R. Whitmarsh and
an anonymous reviewer, and discussions with T. King, D. McKenzie and C. Petit improved this
manuscript.

References
Agar, S. & Klitgord, K. 1995 Rift flank segmentation, basin initiation, and propagation: a
neotectonic example from Lake Baikal. J. Geol. Soc. Lond. 152, 849860.
Ambraseys, N. 1991a The Rukwa earthquake of 13 December, 1910 in East Africa. Terra Nova
3, 203208.
Ambraseys, N. 1991b Earthquake hazard in the Kenya Rift: the Sabukia earthquake of 13 December 1910. Geophys. J. Internat. 3, 203208.
Ambraseys, N. & Adams, R. 1986 Seismicity of the Sudan. Bull. Seism. Soc. Am. 76, 483493.
Anders, M. & Schlische, R. 1994 Overlapping faults, intrabasin highs, and the growth of normal
faults. J. Geol. 102, 165180.
Asfaw, L. M. 1992 Implication of shear deformation and earthquake distribution in the East
African Rift between 4 N and 6 N. J. African Earth Sci. 10, 745751.
Asfaw, L. M., Bilham, R., Jackson, M. & Mohr, P. 1992 Recent inactivity in the African Rift.
Nature 357, 457.
Bechtel, T., Forsyth, D. & Swain, C. 1987 Mechanisms of isostatic compensation in the vicinity
of the East African Rift, Kenya. Geophys. J. R. Astr. Soc. 90, 445465.
Birt, C., Maguire, P. K. H., Khan, M. A. & Thybo, A. 1997 A combined interpretation of the
KRISP 94 seismic and gravity data: evidence for a mantle plume beneath the East African
plateau. Tectonophys. 278, 211242.
Bosworth, W. 1992 Mesozoic and Tertiary rifting in East Africa. Tectonophys. 209, 115137.
Braunmiller, J. & Nabelek, J. 1990 The 1989 Ethiopia earthquake sequence. Eos 71, 1480.
Burov, E. & Diament, M. 1995 The effective elastic plate thickness (Te ) of continental lithosphere: what does it mean? J. Geophys. Res. 100, 39053927
Phil. Trans. R. Soc. Lond. A (1999)

760

C. J. Ebinger and others

Camelbeeck, T. & Iranga, M. D. 1996 Deep crustal earthquakes and active faults along the
Rukwa trough, eastern Africa. Geophys. J. Int. 124, 612630.
Contreras, J., Scholz, C. & King, G. C. P. 1997 A model of rift basin evolution constrained by
first-order stratigraphic observations. J. Geophys. Res. 102, 76737690.
Das, S. & Scholz, C. H. 1983 Why large earthquakes do not nucleate at shallow depths. Nature
305, 145148.
Davidson, A. (compiler) 1983 The Omo River Project. Bulletin 2, Ethiopian Institute Geological
Survey.
Dawers, N. & Anders, M. 1995 Displacement-length scaling and fault linkage. J. Struct. Geol.
17, 604614.
Dawson, B. 1992 Neogene tectonics and volcanicity in the north Tanzania sector of the Gregory
Rift Valley: contrasts with the Kenya sector. Tectonophys. 204, 8192.
Delvaux, D., Kervyn, F., Vittori, E., Kajara, R. & Kilembe, E. 1998 Late Quaternary tectonic
activity and lake level change in the Rukwa Rift. J. African Earth Sci. 26, 397421.
Deverchere, J., Houdry, F. & Diament, M. 1991 Evidence for a seismogenic mantle and lower
crust in the Baikal Rift. Geophys. Res. Lett. 18, 10991102.
Diament, M. & Kogan, M. 1990 Long wavelength gravity anomalies over the Baikal Rift and
geodynamic implications. Geophys. Res. Lett. 17, 19771980.
Ebinger, C. 1989 Tectonic development of the western branch of the East African Rift system.
Geol. Soc. Am. Bull. 101, 885903.
Ebinger, C. & Ibrahim, A. 1994 Multiple episodes of rifting in Central and East Africa: a reevaluation of gravity data. Geol. Rundsch. 83, 689702.
Ebinger, C., Bechtel, T., Forsyth, D. & Bowin, C. 1989 Effective elastic plate thickness beneath
the East African and Afar plateaux, and isostatic compensation for the uplifts. J. Geophys.
Res. 94, 28932901.
Ebinger, C., Karner, G. & Weissel, J. 1991 Mechanical strength of extended continental lithosphere: constraints from the Western Rift system, Africa. Tectonics 10, 12391256.
Ebinger, C., Poudjom-Djomani, Y., Mbede, E. & Foster, A. 1997 1991 Rifting the Archaean:
development of the NatronManyaraEyasi Basins, Tanzania. J. Geol. Soc. Lond. 154, 947
960.
Forsyth, D. 1992 Finite extension and low angle normal faulting. Geology 20, 2730.
Foster, A. 1997 Seismicity and tectonics of Africa. PhD thesis, University of Cambridge, UK.
Foster, A. & Jackson, J. 1998 Source parameters of large African earthquakes: implications for
crustal rheology and regional kinematics. Geophys. J. Int. 134, 422448.
Foster, A., Ebinger, C., Mbede, E. & Rex, D. 1997 Tectonic development of the northern Tanzanian sector of the East African Rift system. J. Geol. Soc. Lond. 154, 689700.
Gupta, A. & Scholz, C. H. 1998 Utility of elastic models in predicting fault displacement fields.
J. Geophys. Res. 103, 823834
Hayward, N. 1996 A quantitative comparison of oceanic and continental rift segmentation. PhD
thesis, University of Leeds, UK.
Hayward, N. & Ebinger, C. 1996 Variations in the along-axis segmentation of the Afar Rift
system. Tectonics 15, 244257.
Hendrie, D., Kusznir, N., Morley, C. & Ebinger, C. 1994 A quantitative model of rift basin development in the northern Kenya Rift: evidence for the Turkana region as an accommodation
zone during the Palaeogene. Tectonophys. 236, 409438.
Jackson, J. & Blenkinsop, T. 1993 The Malawi earthquake of 10 March, 1989: deep faulting
within the East African Rift system. Tectonics 12, 11311139.
Jackson, J. & Blenkinsop, T. 1997 The BililaMtakataka fault in Malawi: an active, 100 km long
normal fault segment in thick seismogenic crust. Tectonics 16, 137150.
Jackson, J. & White, N. 1989 Normal faulting in the upper continental crust: observations from
areas of active extension. J. Struct. Geol. 11, 1536.
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

761

Jestin, F., Huchon, P. & Gaulier, J. M. 1994 The Somali plate and the East African Rift system:
present-day kinematics. Geophys. J. Int. 116, 637654.
Kikuchi, M. & Kanamori, H. 1995 The Shikotan earthquake of October 4, 1994: lithospheric
earthquake. Geophys. Res. Lett. 22, 10251028.
King, T. 1998 Tectonics and isostasy in extensional provinces: the Aegean Rifts. PhD thesis,
University of Leeds, UK.
Kostrov, B. V. & Das, S. 1988 Principles of earthquake source mechanics. Cambridge University
Press.
Last, R., Nyblade, A., Langston, C. & Owen, T. 1997 Crustal structure of the East African
plateau from receiver functions and Rayleigh wave phase velocities. J. Geophys. Res. 102,
24 46924 483.
LeTurdu, C., Legall, B., Tiercelin, J.-J., Sturchio, N., Gente, P., Richert, J.-P. & Stead, D. 1999
A morphotectonic case study if a synrift extensional fault zone: the Baringo trachyte fault
system, Central Kenya Rift. Tectonophys. (In the press.)
Ma, X. Q. & Kusznir, N. 1994 Effects of rigidity layering, gravity, and stress relaxation on 3D
subsurface fault displacement fields. Geophys. J. Int. 118, 210220.
McKenzie, D. & Fairhead, F. 1997 Estimates of the effective elastic thickness of the continental
lithosphere from Bouguer and free air gravity anomalies. J. Geophys. Res. 102, 27 52327 552.
Manighetti, I., Tapponnier, P., Gillot, P., Jacques, E., Courtillot, V., Armijo, R., Ruegg, J.-C.
& King, G. C. P. 1998 Propagation of rifting along the ArabiaSomalia plate boundary: into
Afar. J. Geophys. Res. 103, 49474974.
Marty, B., Pik, R. & Gezahagen, Y. 1996 He isotopic variations in Ethiopian plume lavas:
nature of magmatic sources and limit on lower mantle convection. Earth Planet. Sci. Lett.
144, 223237.
Mechie, J., Keller, G. R., Prodehl, C., Gaciri, S., Braile, L. W., Mooney, W. D., Gajewski, D.
& Sandmeier, K.-J. 1994 Crustal structure beneath the Kenya Rift from axial profile data.
In Crustal and upper mantle structure of the Kenya Rift (ed. C. Prodehl, G. R. Keller & M.
A. Khan). Tectonophys. 236, 179200.
Menzies, M., Gallagher, K., Yelland, A. & Hurford, A. 1997 Volcanic and non-volcanic rifted
margins of the Red Sea and Gulf of Aden: crustal cooling and margin evolution in Yemen.
Geochim. Cosmochim. Acta 61, 25112527.
Mohr, P. 1983 Ethiopian flood basalt province. Nature 303, 577584
Morley, C. K., Wescott, W. A., Stone, D., Harper, R. M., Wigger, S. T. & Karanja, F. 1992
Tectonic evolution of the northern Kenya Rift. J. Geol. Soc. Lond. 149, 333348.
Nyblade, A. & Langston, C. 1995 East African earthquakes below 20 km depth, and their implications for crustal structure. Geophys. J. Int. 121, 4962.
Nyblade, A., Pollack, H., Jones, D., Podmore, F. & Mushayandebvu, M. 1990 Terrestrial heat
flow in East and southern Africa. J. Geophys. Res. 95, 17 37117 384.
Nyblade, A. A., Birt, C., Langston, C., Owens, T. & Last, R. 1996 Seismic experiment reveals
rifting of craton in Tanzania. Eos 77, 517521.
Omar, G. I. & Steckler, M. S. 1995 Fission track evidence on the initial rifting of the Red Sea:
two pulses, no propagation. Science 270, 13411344.
Ori, G. G. 1989 Geologic history of the extensional basin of the Gulf of Corinth. Geology 17,
918921.
Petit, C. 1996 Le Rift Baikal et la collision Inde-Asie. PhD thesis, Geosciences Azur, University
of Pierre et Marie Curie, Paris, France.
Ring, U. 1994 The influence of preexisting structure on the evolution of the Cenozoic Malawi
Rift (East African Rift system). Tectonics 13, 313326.
Ritsema, J., Nyblade, A., Owens, T., Langston, C. & VanDecar, J. 1998 Upper mantle seismic velocity structure beneath Tanzania: implications for the stability of cratonic roots. J.
Geophys. Res. 103, 21 20121 214.
Phil. Trans. R. Soc. Lond. A (1999)

762

C. J. Ebinger and others

Roberts, S. & Jackson, J. 1991 Active normal faulting in central Greece: an overview. In The
geometry of normal faults (ed. A. Roberts, G. Yielding & B. Freeman). Geol. Soc. Lond.
(Spec. Publ.) 56, 125142.
Rosendahl, B. R. 1987 Architecture of continental rifts with special reference to East Africa. A.
Rev. Earth Planet. Sci. 15, 445503.
Ruegg, J.-C. (and 13 others) 1993 First epoch geodetic GPS measurements across the Afar plate
boundary zone. Geophys. Res. Lett. 20, 18991902.
Ruppel, C. 1995 Extensional processes in continental lithosphere. J. Geophys. Res. 100, 24 187
24 215.
Ruppel, C., Kogan, M. & McNutt, M. 1993 Implications of new gravity data for Baikal Rift
Structure. Geophys. Res. Lett. 20, 16351638.
Schlische, R., Young, S., Ackermann, R. & Gupta, A. 1996 Geometry and scaling relations of a
population of very small rift-related normal faults. Geology 24, 683686.
Scholz, C. H. 1988 The brittle-plastic transition and the depth of seismic faulting. Geol. Rundsch.
77, 319328.
Scholz, C. & Contreras, J. 1998 Mechanics of continental rift architecture. Geology 26, 967970.
Shackleton, R. 1993 Tectonics of the lower crust: a view from the Usambara mountains, NE
Tanzania. J. Struc. Geol. 15, 663671.
Sigmundsson, F. 1992 Tectonic implications of the 1989 Afar earthquake sequence. Geophys.
Res. Lett. 9, 877888.
Stein, R. & Barrientos, S. 1985 Planar high-angle normal faulting in the Basin and Range:
geodetic analysis of the 1983 Borah Peak, Idaho, earthquake. J. Geophys. Res. 90, 11 355
11 366.
Stewart, K. & Rogers, N. 1996 Mantle plume and lithosphere contributions to basalts from
southern Ethiopia. Earth Planet. Sci. Lett. 139, 195211.
Theunissen, K., Klerkx, J., Melnikov, A. & Mruma, A. 1996, Mechanisms of inheritance of rift
faulting in the western branch of the East African Rift, Tanzania. Tectonics 15, 776790.
Trudgill, B. & Cartwright, J. 1994. Relay-ramp forms and normal-fault linkages, Canyonlands
National Park, Utah. Geol. Soc. Am. Bull. 106, 11431157.
Upcott, N., Mukasa, R. K., Ebinger, C. & Karner, G. 1996 Along-axis segmentation and isostasy
in the Western Rift, East Africa. J. Geophys. Res. 101, 32473268.
van der Beek, P. 1997 Flank uplift and topography at the central Baikal Rift (SE Siberia): a
test of kinematic models for continental extension. Tectonics 16, 122136.
van der Beek, P., Mbede, E., Andriessen, P. & Delvaux, D. 1998 Denudation history of the
Malawi and Rukwa Rift flanks from apatite fission track thermochronology. J. African Earth
Sci. 26, 363386.
Weissel, J. K. & Karner, G. 1989 Flexural uplift of rift flanks due to tectonic denudation of the
lithosphere during extension. J. Geophys Res. 94, 13 91913 950.
Weissel, J., Malinverno, A., Harding, D. & Karner, G. 1995 Erosional development of the
Ethiopian plateau of Northeast Africa from a fractal analysis of topography. In Fractals in
petroleum geology and earth processes (ed. C. Barton & P. LaPointe), pp. 127142. New York:
Plenum.
Wheeler, W. & Karson, J. 1989 Structure and kinematics of the Livingstone Mountains Border
Fault zone, Nyasa (Malawi) Rift, southwestern Tanzania. J. African Earth Sci. 8, 393413.
WoldeGabriel, G., Aronson, J. & Walter, R. C. 1990 Tectonic development of the Main Ethiopian
Rift. Geol. Soc. Am. Bull. 102, 439458.
Yielding, G. 1985 Control of rupture by fault geometry during the El Asnam (Algeria) earthquake. Geophys. J. R. Astr. Soc. 81, 641670.
Zhao, M., Langston, C., Nyblade, A. & Owens, T. 1997 Lower crustal rifting in the Rukwa
graben, East Africa. Geophys. J. Int. 129, 412420.
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

763

Discussion
D. McKenzie (Bullard Laboratories, University of Cambridge, UK ). Im intrigued
by the relationship between the faulting and the volcanism. Once the spreading really
gets going, it is perfectly reasonable that the scale of the segmentation is controlled
by the total thickness of the melting zone. But at the beginning, for instance in
East Africa, very small amounts of extension produce enormous volcanoes and long
segments. How does that fit together? How do you get these enormous volcanoes
when the amount of extension is so small?
C. J. Ebinger. There are large shield volcanoes along the length of the Eastern
(Gregory) Rift, and a few large shields located within two isolated volcanic provinces
of the Western Rift. One needs to have an extension, or beta, factor of about 1.5 to
generate basalts by adiabatic decompression melting, whereas < 1.2 in the Western
Rift. Both the chemistry and volume of eruptive products and the gravity evidence
for dynamic support of the East African and Ethiopian plateaux indicate that a
mantle plume underlies East Africa, despite the lack of evidence for lithospheric
thinning. The very small amount of extension with volcanism in the Western Rift
need not be a problem, if we consider lateral flow of plume material along relief at
the lithosphereasthenosphere boundary, followed by decompression melting some
distance from the plume source.
D. McKenzie. Well, thats not quite whats worrying me. In the southern part,
where the amount of extension is very small, there is still a relationship between the
individual volcanoes and the rift. Because the amount of extension is so small, it
can scarcely have much influence on melt generation. Which means that, if there is
a relationship, it has got to be the other way around, and the volcanoes control the
rifting. This is to me a strange notion.
C. J. Ebinger. In the Rungwe region, the southernmost volcanic province in the
Western Rift where 1.1, we can see that initial faulting was concurrent with,
or occurred immediately after, initial volcanism. If volcanism is driving faulting, the
ca. 600 km zone of extension without volcanism in the Western Rift is surely problematic. I would prefer to speculate that plume material rises and causes decompression
melting above previously thinned lithosphere (e.g. Karoo Rifts), or steep gradients
(e.g. craton boundaries), at the lithosphereasthenosphere boundary. In this scenario,
I would agree with you that magmatism locally is driving extension and contributes
to the along-axis segmentation of the rift system, but it is not a prerequisite for
generation of large normal fault systems.
N. Kusznir (University of Liverpool, UK ). Is there a possibility that the small
amount of rifting, while not enough to generate decompression melting, can actually
control the plumbing? At the end of the developing rift segments there may be some
tensile stresses that only affect the upper crust. However, these stresses may be
enough to develop dilatancy in a tensile environment which just makes it easier for
melts to get up there.
D. McKenzie. That is hard to do because the composition of the volcanoes Ive
looked at indicates that the melt is coming from the lithosphere and not from deeper
down. I can see that if you were to produce melt from somewhere below the lithosphere, and then have some sort of channelling all the way up to the surface caused
Phil. Trans. R. Soc. Lond. A (1999)

764

C. J. Ebinger and others

by extension, you could divert it away from the craton but, from the volcanoes I
have studied, it doesnt look as if the melt is coming from that deep.
K. E. Louden (Dalhousie University, Halifax, Nova Scotia, Canada). I have a simple question to ask. Some years ago I worked with gravity response functions to
investigate various regions of oceanic crust. Through the work of Tony Watts and
others, I felt that some understanding had been reached of the meaning of the effective thickness Te for oceanic lithosphere. But I never really did have a very clear
understanding of what Te means for continental lithosphere. Is it the same? What
does a Te of 5 km mean for an old lithospheric centre like Africa?
C. J. Ebinger. In continental, as in oceanic regions, we represent the rheological
properties of the compositionally layered lithosphere by an equivalent elastic thickness, the effective elastic thickness Te , as a means to compare response functions from
region to region. What the resulting effective elastic thickness then tells us about the
rheology of the continental lithosphere is very difficult to assess, but we can say that
variations in continental Te values are dependent upon geothermal gradient, crustal
composition, and strain rate.
A. B. Watts (University of Oxford, UK ). Basically, what you are saying is that
the range of elastic thickness values that you obtain reflects inheritance, in the sense
that youve got higher values reflecting colder lithosphere and lower values reflecting
warmer lithosphere. My point is that by the time we get to the case of oceanic crust
formed by sea-floor spreading, or if you look at fault lengths over passive margins
where you can actually measure the fault lengths, the fault lengths are already quite
short, similar to the fault lengths that correspond to the low elastic thickness case.
Also, the fault lengths on mid-ocean ridges themselves are very short. So, my question
is, what is your view about what seems to happen when we go the extra step of
forming an ocean? What is your view about what may happen to give us these very
low values of elastic thickness in the passive margins?
C. J. Ebinger. Why are passive margins perpetually weak?
A. B. Watts. Yes, and in the Biscay section they have short fault lengths and look,
elastically, like mid-ocean ridges.
C. J. Ebinger. To me, there are several possible interpretations of the patterns
you describe. First, we see in the East African Rift system that a new along-axis
segmentation can develop after considerable extension and magmatism has modified
and heated the continental lithosphere. That is, originally long rift segments are preserved along the outer margin of the extensional province and a short segmentation,
similar to mid-ocean ridge segmentation, is observed. One explanation, therefore,
for the short fault lengths is the Afar analogy - a new short segmentation develops
near the onset of sea-floor spreading, and is then perhaps a precursor of sea-floor
spreading segmentation.
Second, one needs to look landward of passive margins to assess the role of strength
during the initial stages of sea-floor spreading. Much of this record has been removed
by as yet poorly understood erosion processes, making it difficult to assess the original
border fault segment length during the initial stages of rifting. Studies of very young
passive margins that developed in thermally stable lithosphere (e.g. Red Sea margins)
may allow us to assess the temporal variability of border fault length implied by our
Afar Rift studies.
Phil. Trans. R. Soc. Lond. A (1999)

Extensional basin geometry and the elastic lithosphere

765

Third, some passive margins may have been weak, hot rifts from the outset, with
short fault segments, narrow syn-rift basins, and narrow rift flanks. In that case
segments will appear similar to segmentation styles along mid-ocean ridges where
the lithosphere was hot and weak at the time of fault initiation. Obviously, we need
combined land and marine studies of conjugate margins to be able to answer these
questions.
M. Menzies (Royal Holloway College, University of London, UK ). What mechanism
controls the orientation of the African Rift Valley and more importantly the Red
Sea? Is there evidence for reactivation of basement (crustal) lineaments or is it a
control from mantle structures (i.e. topography on the lithosphereasthenosphere
boundary)? The evidence from Yemen indicates that basement lineaments are at odds
with the orientation of the Red Sea and very little evidence exists for reactivation of
basement lineaments.
C. J. Ebinger. There are many papers comparing the trends of Precambrian lineaments detected in air photos and satellite imagery and the strikes of Late Cenozoic
normal faults in East Africa that conclude a strong control of basement structures
on rift faults. The few detailed studies of normal fault zones and basement structures
provide a different story. Much of the Red Sea and East African Rift system has developed within accreted terrains and fold belts of the Pan-African system, and Cenozoic
Rift faults often cut across Permo-Triassic and Cretaceous faults. Closer examination shows that steep Late Cenozoic Border Faults cross-cut Precambrian shear zones
or metamorphic foliations showing shallower or highly variable dips, and that they
cross-cut earlier border faults. Some segments of border faults may reactivate steep
shear zones (e.g. Karonga Basin), or make jogs to follow pre-existing fractures, but
the similarity of border fault and basin dimensions within sectors of the East African
Rift, and other rifts worldwide, argues against a basement control. I would argue that
favourably orientated pre-existing fractures and shear zones facilitate the propagation of border faults, rather than control their location and orientation.
A. Roberts (Badleys Earth Sciences Ltd, Lincs, UK ). If the high values of Te
reported in this study are correct, you should be observing a radius of footwall uplift
around individual faults of at least 100 km and possibly a lot more. I believe the
rule of thumb is that the radius of footwall uplift will be about 10 times the effective
elastic thickness. On seismic sections of buried basins, the radius of footwall uplift
at basin margins is typically a few tens of kilometres, indicating low values of Te
in these situations. What is the radius of fault-controlled footwall uplift that you
observe at the basin margins in this study?
C. J. Ebinger. The breadth of the uplifted rift flank, or radius of footwall uplift, is
dependent on the effective elastic thickness, as well as interactions between footwall
and hangingwall, beta-factor, and density of infilling material. Our forward models of
basin and flank morphology for a Te of 35 km predict an uplift radius of ca. 170 km,
whereas those for a Te of 10 km produce an uplift radius of ca. 60 km. These models
provide good fits to the observed basin and flank morphology and to the gravity
anomalies in the Western Rift, except in the immediate vicinity of the escarpment
where erosional processes significantly modify the fault response (see figure 3). Referring back to your first statements, I would be surprised if seismic data of buried basins
from the North Sea showed broad flanks, since these rifts did not develop in cratonic
lithosphere.
Phil. Trans. R. Soc. Lond. A (1999)