Content uploaded by Clive Oppenheimer
Author content
All content in this area was uploaded by Clive Oppenheimer
Content may be subject to copyright.
int. j . remo te sensing , 1 997, vol. 18, no. 8,1661± 169 2
Remote sensing of heat, lava and fumarole emissions from Erta `Ale
volcano, Ethiopia
C. OPPENHEIMER
Department of Geography, Downing Place, Cambridge CB23EN, England,
U.K.
P. FRANCIS
Departm ent of Earth Scienc es, The Open U niversity, M ilton Keyne s MK7 6AA,
England, U.K.
(Rec eive d 8 May 1996; in ® nal fo rm 3 Se ptembe r 1996 )
Ab stract. Erta `Ale volcano, sited within the Afar Triangle of Ethiopia, is one of
the least frequented, perennially active, subaerial volcanoes. By compiling a time
series of Landsat MSS and TM, JERS-1, SPOT, and AVHRR digital imagery,
and space-borne photographic data, we have been able to constrain the activity
of this volcano, important for its geodynamic setting, during the long period since
1974 when volcanological investigations e ectively ceased. Existing techniques
for infrared thermometry have been modi® ed to cope with saturation of the short
wavelength infrared Landsat TM band 5 and 7 sensors, enabling derivation of
thermal ¯ uxes from Erta `Ale’s active lava lakes. Lake levels have been estimated
from measurements of shadow lengths cast by the crater rim. Changes in caldera
and ¯ ank re¯ ectances identify new lava ¯ ows whose areas and volumes we have
constrained in order to derive eruption magnitudes and e usion rates. Between
1968 and 1974, approximately 3Ö1010 kg of lava was erupted at peak discharge
rates exceed ing 400 kg sÕ1, though much of this subsequently drained back or
subsided rigidly. No post-1974 over¯ ows were detected in the imagery, although
thermal ( 100± 400 MW) output from the lava lakes, and gas emissions appear to
have been sustained up to the time of writing. The longevity of the lava lakes
provides evidence for convective circulation between the lakes and a deeper
magma reservoir. The high heat ¯ ux and low exogenous growth rates (#10 kg s Õ1
integrated over the last thirty years) are indicative of a volcano that `grows’ largely
by magmatic intrusion, which is consistent with the formation of new igneous
crust in the extensional tectonic environment of northern Afar.
1. I nt rodu ct io n
Erta `Ale is unique Ð a persistently active, subaerial volcano situated on a
spreading axis above a mantle plume (White and McKenzie 1989 ). While docu-
mented observations of the `smoking mountain’, as the Afar name signi® es, have
been few and far between, it has been suggested ( Barberi et al. 1970) that lava lake
activity and magmatic degassing have prevailed there for at least the last century.
The volcano is located about halfway along, and at the maximum width of, the Erta
`Ale range ( ® gu re 1) which is built largely of subalkaline to transitional basaltic ¯ ows
erupted fro m axial ® ssures d uring the Quaternary ( Barberi et al. 1 973 ). Ert a `Ale
range, the normal faults and gaping ® ssures that slice it, and the summit caldera of
Erta `Ale volcano, all share the same NNW± SSE (Eritrean) orientation, sub-parallel
to the Red Sea rift.
0143± 1 161/ 97 $12.0 0 Ñ1997 Taylor & Francis Ltd
C. Oppenheimer and P. Francis1662
(b)
(a)
Figure 1. (a) Apollo-Saturn 9 astronaut near-vertical photograph of the 95 km Ö42 km Erta
`Ale range composed predominantly of dark basalts, acquired with an 80 mm lens.
Date and altitude are not recorded but the mission ran between 3 and 13 March 1969.
The summit of Erta `Ale volcano is located on the axis of the range, at the intersection
of the grid marks shown. The arrow points approximately due north. (b) Declassi® ed
intelligence satellite image (CORONA KH-4A mission) of Erta ’Ale, recorded on
19 May 1965. The caldera dimensions are #16 00 mÖ700 m and north-northwest is
approximately up the page.
Remote sensing of Erta `Ale volcano 1663
The volcano, approximately 30 km across at its b ase, rises some 700 m above t he
Danakil Depression. Very gentle concave-upwards slopes meet at an elongated
summit caldera, 1600 m Ö700 m in dimensio ns. Th e caldera i s thought to have
formed from the coalescence of three collapse structures ( Barberi and Varet 1970),
the n orthern and central o f which have been occupied in recent years by lava lakes Ð
one in each. Open fractures on t he outer ¯ anks mimic the Eritrean trend and are
not radial; some of these on the northern ¯ ank have fed ® ssure eruptions. Less than
1 km from the southern rim of the caldera is another, larger, collapse caldera,
3´1 km Ö1´8 k m in dimensio ns, likewise aligned with t he Eritrean trend. There is no
record of witnessed eruptive activity from this centre.
Foreigners have tended to fare badly in the Danakil desert, and today the area
remains one of t he more inhospitable parts of the world. Indeed, Dainelli concluded
that `taking into account the general condition of the area, it is quite improbable
that the geology (of Afar) will ever be systematically studied’ (CNR± CNRS Afar
team 1973). The Afars themselves steer clear of Erta `Ale’s summit on account of the
spirits of departed herdsmen who encircle it on ¯ ying horses (Hildebrandt 1875 ). In
order to bridge t he long gaps between ® eld visits to the volcano, we have studied
image data from several instruments. Image archive searches unearthed numerous
Landsat Multispectral Scanner (MSS), Landsat Thematic Mapper ( TM ), Systeme
Probatoire pour l’Observation de la Terre (SPOT ), Japanese Earth Resources
Satellite (JERS-1 ) and NOAA Advanced Very High Resolution Radiometer
(AVHRR) digital imagery, as well as intelligence satellite and astronaut photographs
of Erta `Ale, spanning three decades. We have selected a signi® cant proportion of
these images (table 1).
Rothery et al. (1988) made an important contribution to refocusing attention on
Erta `Ale with their interpretation of two Landsat TM scenes of the mid-1980s (one
digital, one hardcopy). Here, we integrate our new ® ndings of the volcano’s activity
based o n a much larger dataset, with known ® eld observations. Some volcanological
implications of the work are raised but these are developed in greater detail elsewhere
(Oppenheimer and Francis, 1997).
1.1. Review of ® eld observations
Because on-site experience of Erta `Ale is rather limited , the few observations
that do exist are all the more important in aiding interpretation of our remotely
sensed data. We therefore give a quite detailed account of documented activity of
the volcano in this section. Some confusion in the earliest accounts arises from
toponymy, both the range and the volcano sharing the same name, and imprecise
navigation. We consider it possible that some observations apply to Ale Bagu, south-
west of Erta `Ale volcano. This cone is nearer to the routes traversing the western
side o f the Erta `Ale range, and is considerably steeper. Bein g closer, and looking
much more like a volcano should, this edi® ce may have attracted the attention of
the earliest European explorers more than Erta `Ale’s summit.
Probably the ® rst European to hear of the volcano was d’Abbadie who reached
the Depression in 1841, and reported that `near Lake Dagad is a mountain which
is always smoking’ (d’Abaddie 1890 ). Munzinger, crossing the salt plain within
several kilometres of Erta `Ale in 1867, also noted `continuous fumes’ (Munzinger
1869 ) . Much of his journey was made at night but he makes no mention of summit
glow. Hildebrandt seems to h ave been the ® rst European to climb t o the top in 1873,
and his travelogue exquisitely understates the rigours o f the journey. He wrote of
C. Oppenheimer and P. Francis1664
Table 1. Data examined in this study, including details, where known, of sun elevation, h,
and azimuth, w. Lava lake levels, d, estimated from shadow lengths measured in
Landsat TM and SPOT data, are given for the central lake (§2.1 ). Errors are based
on assumed accuracy in shadow length measurements of Ô15 m. For JERS-1 data,
OVN represents visible and near infrared, and OSW short-wavelength infrared
channels.
h w d
Date Platform Sensor ID (ß) (ß) (m)
19 May 1965 CORONA KH4A DS1021-1009DF049
Mar 1969 Apollo 9 AS09-23-3535
8 Sept 1972 Landsat 1 MSS 180-51
30 Jan 1973 Landsat 1 MSS 180-51
24 Mar 1975 Landsat 2 MSS 180-51
1 Jun 1975 Apollo-Soyuz AST14-943
3 Mar 1979 Landsat 2 MSS 180-51
1 Feb 1984 Space Shuttle STS41B-42-2532
21 Apr 1984 Landsat 5 TM 168-51 59 089 96Ô23
24 Jun 1984 Landsat 5 TM 168-51 57 066 92Ô17
27 Aug 1984 Landsat 5 TM 168-51 58 091 114Ô22
5 Oct 1984 Space Shuttle STS41GH-32-14
30 Oct 1984 Landsat 5 TM 168-51 52 133 76Ô15
10 Nov 1984 Space Shuttle STS51A-33-88
1 Dec 1984 Landsat 5 TM 168-51 45 141 82Ô11
2 Jan 1985 Landsat 5 TM 168-51 42 138 95 Ô11
18 Jan 1985 Landsat 5 TM 168-51 43 135 100Ô11
7 Mar 1985 Landsat 5 TM 168-51 51 117 88 Ô17
23 Mar 1985 Landsat 5 TM 168-51 55 108 102Ô20
5 Jan 1986 Landsat 5 TM 168-51 41 136 93 Ô10
30 Sept 1988 Space Shuttle STS26-39-78
14 Mar 1989 Space Shuttle STS29-78-69
7 May 1989 Space Shuttle STS30-95-89
14 Jan 1990 Space Shuttle STS32-101-23
21 Oct 1990 SPOT Pan 139-323 120
9 Jun 1992 JERS-1 OVN+OSW 236-278
2 Aug 1992 Space Shuttle STS46-81-92
13 Sept 1992 Space Shuttle STS47-72-67
25 Oct 1992 Space Shuttle STS52-75-44
29 Apr 1993 Space Shuttle STS55-151B-149
14 May 1994 JERS-1 OVN 236-278 72 072
4 Nov 1994 Space Shuttle STS66-150-29
4 Nov 1994 Space Shuttle STS66-150-31
4 Nov 1994 Space Shuttle STS66-154-5
6 Feb 1996 NOAA14 AVHRR AL14020796230341
the summit crater (Hildebrandt 1875): `It is as if a pitch black sea churned up by a
mighty hurricane has broken against cli s, forming towers of foam that run and
twist and suddenly freeze sti . Thus lies the deserted mass of rock like a tomb stone
to some great primeval force’. Nevertheless, he says nothing that unambiguously
identi® es the presence of active lava. Dainelli and Marinelli ( 1906 ) believed that
Bianchi made the same journey in 1884. He reported four cones with open craters,
known for emissions of smoke and ® re, which, according to the elders, were the
cooking pots of demons who had `run out of wood, turned o the ® res, and left to
do their cooking elsewhere in Rorom’ ( Dainelli and Marinelli 1906 ).
Remote sensing of Erta `Ale volcano 1665
Furth er indications of the volcano’s behaviour came from Rossini in 1903± 1904,
who noted that `the summit was constantly wrapped in smoke which was more
intense in the winter months, but never touched by ¯ ames; nor were rumblings
heard ’ ( Dainell i a nd Mari nelli 1906 ) . O n the oth er h and, Dant e Odori zzi, w ho
crossed the northern part of Afar in 1906 ( Dainelli and Marinelli 1907), was told by
local people of `¯ ames’ visible at night. In the same year, Pastori climbed t o, or close
to, the crater rim; his observations of red crater glow at night re¯ ected in the plume
are consistent with ongoing lava pool activity ( Barberi et al. 1970). Further observa-
tion s were made by Tancredi ( 1912), Dainelli and Marinelli (1912 ), and Nesbitt who
travelled to the west of the range with Pastori in 1928. Nesbitt sighted fuming from
several of the volcanoes in the Erta `Ale range, apparently including Erta `Ale volcano,
and noted red glow but without specifying from which crater (Nesbitt 1934). From
the 1940s onwards, there h ave b een intermit tent air pilots’ reports of fuming and
incandescence at the summit.
In 1960, Paul Mohr observed a single active lava lake `® lled with red-hot pasty
basalt from the centre of which projected a small cone of sulphur-rich material’,
intense fuming, and `extremely fresh, steaming basalt ¯ ows’ erupted from radial
® ssures on both the north and south ¯ anks (Mohr 1962, personal communication
1996 ) . Some lateral ® ssure ¯ ows evident on the south-west ¯ an ks may have resulted
from de¯ ation of a lava lake. Hildebrandt’s and Pastori’s failure to en ter the crater
may indicate subsequent episodes of lava e usion because the teams visiting the
volcano in the 1960s found the crater ¯ oor close to the rim, allowing easy access.
Qualitative observations of Erta `Ale were made during several brief helicopter
stopovers in the late 1960s but it was not until December 1971, when the joint
CNR± CNRS team led by Haroun Tazie and Giorgio Marinelli spent four days in
the caldera, that more focused volcanological studies were undertaken (CNR± CNRS
Afar team 1973). The volcano greeted several further ® eld parties in the early 1970s.
Active lava lakes were observed each time, though t heir levels and areal extent
¯ uctuated between and during visits.
In 1968, there were two lava ponds, one in the northern lobe of the caldera,
boasting a number of ® re fountains. The sub-circular pond, approximately 100 m in
diameter, lay 160 m below the rim of a three-tiered crater, 350 m wide at the top
(® gure 2; Tazie 1973). About 400 m away in the northern part of the middle lobe
was a second pit crater, 65 m in diameter, from which d ense whitish fumes issued
( Varet 1971 a). Though its ¯ oor was not observed, red n ight-time glow indicated t he
proximity of lava. Inspection of air photographs taken in 1963 gave no indication
of ¯ ank activity or modi® cation to the summit region in the intervening period
( Varet 19 71 b) . This is co rrobo rated b y comp arison of th e i ntelli gence sat ellite p hoto -
graph ( ® gure 1 (b)) with subsequent imagery. By 1969, the level of the north lake had
risen to a depth b elow the rim of 100 m; fountain ing was observed in November
(Barberi et al. 1970 ) but not in December ( Varet 1971 b).
Vigorous activit y was observed in December 1970, with incandescence visible
from t he air in full daylight. The lakes overturned rapidly and strong degassing
accomp anied energetic ® re fountains. Modi® cations to t he northern pit morphology
were evident: its lava lake had risen to within 40 m of the rim and had swollen to a
diameter of 150 m, the periphery solid i® ed and composed of a tidewrack of fountain
pyroclasts, the centre occupied by active lava ( Tazie 1973, Le Guern et al. 1979).
A new collapse pit, 80 m in diameter, cut the northern ¯ ank of the nort hern p it. Th e
lava level in the central pit matched that in the northern.
C. Oppenheimer and P. Francis1666
Figure 2. Fluctuation of lava lake levels at Erta `Ale from ® eld observations; after Le Guern
et al. (1979 ) and J.-L. Chemine
Âe (personal communication, 1994 ).
In 1971, the summit was observed in February, March and December. By
December 1971 the northern la va lake w as 1 2 m b elow t he r im, and only 20 m b elow
the ¯ oor of the central lobe of the caldera, its level ¯ uctuating by up to a metre over
a period of minutes (Tazie 1973 ). A thin pahoehoe crust surrounded a circular 80 m
diameter area of upwelling in which fountains played to heights of 10± 15 m. Lava
¯ owed south-east in the lake at about 7 m minÕ1, transporting some of the foun-
tains (Le Guern et al. 1979). Spatter cones had been erected 100± 150 m from the
nort hern p ond.
Lava began over¯ owing both pits in February 1972, raising the ¯ oor of the
nort hern lobe 12 m during t he year, bringing it within a few metres of the level of
the central lobe ( Barberi et al. 1973). The diameter of the central pit was 80 m. In
January 1973, the northern pit was 25 m deep, and the central pit had also raised
Remote sensing of Erta `Ale volcano 1667
its rim, about 8 m, by over¯ ows. Lava was then at brim-full stage in the central pit,
feeding 1 km long overspills. By March 1973, these ¯ ows had spread across both the
middle and southern lobes of the caldera, raising their level by at least four metres,
and had even spouted on to the south-eastern ¯ anks of t he volcano (Tazie 1973,
Barberi et al. 1973). As before, some fountains were ® xed above feeding vents while
others drifted with the ¯ owing lava. Flows moved south-east from both ponds at
speed s of up to 1 m sÕ1. Spatter cones, built in part of Pele’s hair, formed o n the
crusted lake surface. Lava also crossed a spillway at the northern end o f the caldera.
Countless small transparent ¯ ames ¯ ickered above the surface of the central lava
lake ( Tazie 1994 ) .
In January 1974, the two lava ponds were of similar area but over¯ ows had
raised their rims further, and a ¯ ow had overtopped the north rim of the caldera
(Le Guern et al. 1979). The two lava lakes were sighted again in February 1976
according to Le Guern et al. ( 1979) and Rothery et al. (1988 ), though the diameter,
given as 20 0 m, was probably overestimated.
Erta `Ale was over¯ own in September 1992 (Smithsonian Institution 1992 a, b),
and ® nally revisited in November 1992 (Smithsonian Institution 1992 c). The central
pit held a lava lake at 100 m depth , with diameter 40 m Ö70 m; 5 m high lava
fountains played at four locations in the rapidly convecting lake (Smithsonian
Institution 1992 c). Active lava was absent from the crater, and its ¯ oor was covered
with talus; two strong fumaroles were located on its southern rim. In February 1994,
the n orth crater had an estimated diameter of 300 m and depth of 200 mÐ the lava
lake had not reappeared but there were strong fumaroles on the southern rim. About
half the ¯ oor of a 100 m deep, 100 m diameter pit in the central caldera lobe was
occupied by active lava (® gure 2, J.-L. Chemine
Âe, personal communication 1994 ). A
further visit to the summit in December 1995 revealed little change (Smithsonian
Institution 1995).
2. Lava ¯ uxes a nd m o rpho lo gi ca l cha nges
From the foregoing summary it is evident that a period of signi® can t lava e usion
occurred in the early 1970s. We turn next to analysis of the remotely sensed dataset
in order to constrain eruption magnit udes and mass ¯ ux rates during this episode,
and to determine whether o r not subsequent lava ¯ ow eruptions occurred. The ® rst
stage involved geometric recti® cation and geocoding of all the images such that user-
de® ned line lengths and polygon areas could be measured. Image processing was
carried out with ERDAS IMAGINE software.
2.1. L ava lake levels
Fluctuations in the level of active lava lakes are commonly observed (e.g., Tilling
1987 ) and re¯ ect pressure changes in the lake plumbing system and feeder reservoir.
Short term (minutes to hours) variations in height, with typical amplitudes of a few
metres, sometimes termed `gas-piston action ’ re¯ ect degassing cycles. Longer-term
¯ uctu ation s ( t ens to hun dreds of d ays p eriod ) wi th generall y larger a mplitu des of
tens or hundreds of metres, can be related to in¯ u x of new magma, outbreaks,
intrusions, and crystallization and cooling (Tilling 1987 ). The ® eld observations at
Erta `Ale from 1968 to 1973 charted an approximately 160 m rise of the lake levels,
synchrono us in both pits, preceding the overspills. The concurrent rise in both p onds
implies their connection to a deeper, common reservoir which may have received a
batch of new magma from the mantle.
C. Oppenheimer and P. Francis1668
Figure 3. Cartoon showing geometry of crater shadow in central pit and parameters required
to determine pit depth (equation (1 )).
We have found it possible to estimate subsequent heights of the magmatic column
by measurement of shadows cast in t he vertically walled central pit. For a given sun
elevation, t he pit ¯ oor should be higher, the more of it t hat is illuminated by the
sun. Using solar azimuth and elevation data given in ancillary ® les of the Landsat
TM data, we have inspected each image closely for evidence of crater shadow.
Measurements of horizontal shadow length yield pit depth, d, (and, by implication,
lava lake level ) according to the following relationship ( ® gure 3 ):
d=[r +(nÕr) sin(wÕv)] tan h( 1 )
where r is the radius of the crater, n is the length of the shadow measured along the
cross- track diameter of the crater (number of pixels multiplied by pixel size), wis the
solar azimuth, vthe angular di erence between true- and image-north, and h
the solar elevation. The shadow termination was de® ned by the steepest gradient in
band 3 digital numbers (DN ) within the pit; the east rim of the pit was similarly
determined; n was measured between these two points making Ô0´5 pixel allowances
for mixed pixels. The accuracy of the estimated depth to lava is limited since the
distances that need to be measured (<120 m) little exceed the spatial resolu tion o f
the TM sensors. Furthermore, if no shadows can be discerned, only a minimum pit
depth can be obtained; strict ly this is true for all measurements, since even when
shadows are apparent, they may b e cast on a perched terrace above a deeper nested
crater that contains the lava. A further p oten tial di culty arises from spectral
re¯ ectance di erences across the crater ¯ oor when only part of it is occupied by
active lava.
Despite these complications, the shadows are surprisingly clear and we are certain
of their interpretation as such since they make appropriate geographic shifts with
Remote sensing of Erta `Ale volcano 1669
changes in solar azimuth. Further con® dence can be derived from the fact that
although n varies from #30± 60 m, d is quite uniform. Table 1 summarizes depth
estimates which were made after inspection of digital number pro® les in TM band 3
across the centre of the central p it. Th e TM data suggest a stable lava level through
the entire period and, importantly, that lava was always some distance below the
rim of the containing p it between at least April 1984 and January 1986. Thus, the
air± magma surface had dropped signi® cantly since 1974. A further estimate of lake
level h eight from a hardcopy image of a SPOT scene recorded in October 1990, and
the o bservations and measurements of the fearless abseilers in the 1990s (Smithonian
Institution 1992 c, 1995 ) yield no evidence for higher lake levels. The spatial resolution
of Landsat MSS data is inadequate to measure the crater shadows here.
2.2. Eruption magnitudes
Visual examination of the time-series of images reveals the spectral changes
associated with resurfacing of the caldera ¯ oor and outer ¯ anks by lava ¯ ows. Some
of the most obvious changes are apparent in ® gure 4, such as the ® llin g of part of
the caldera ¯ oor between 1972 and 1973, and further caldera ¯ ooding and caldera
overspills by 1975. Timing of events, and thicknesses of lava can be further con-
strained from ® eld reports. Flow areas were estimated mainly from geocoded Landsat
TM data in preference to the Landsat MSS imagery to optimize accuracy. Volumes
were t hen determined for assumed ¯ ow thicknesses, and masses for an assumed
density. By inference, all mass and volume ® gures are relative to the 1968 topography
and lake depths (® gure 5 ) .
Table 2 presents the eruption magnitude estimates. Mean ¯ ux rates are deter-
mined simply by dividing magn itude by length of interval b etween observations. The
high est in tegrated ¯ ux r ates exceed 400 k g sÕ1, though these do not necessarily re¯ ect
peak discharge rates. During the 1968 ± 1974 eruptive period, the mean discharge
rate was about 160 kg sÕ1. The main sources o f uncertainty are the t hree-dimensional
shape of the pre- and post-eruption north crater, and ¯ ow-® eld thicknesses. Estimates
of magnitudes and eruption rates depend strongly on the assumed average thickness
of lava covering the central and southern caldera lobes which were based on informa-
tion g iven by Tazie ( 1973 ) , Ba rberi et al. (1973 ) and Le Guern et al. (1979 ), and
accuracies cannot be reliably stated. Between 1968 and 1975, we ® nd a total eruption
magnitude of #2´6Ö1010 kg (equivalent t o a sphere of magma of radius 130 m).
Eruption rates between December 1971 and March 1973 averaged around 300 kg sÕ1.
The longest ¯ ow ran app roximately 2´7 km from the central pit, across the southern
lip of the caldera, and down the southern ¯ ank of the volcano.
Although the mismatch in spatial scales of TM and MSS data precludes a
de® nitive statement, similarities in patterns of light and dark caldera in® ll suggest
strongly that little or no lava overspilled the pits between 1975 and 1984 (® gure 6).
Throughout the period o f TM and subsequent JERS, SPOT and ® eld observations,
there is no evidence for lava ¯ ow emplacement. Indeed, the deep (#100 m) level of
magma within the central pit con® rms that magma was not in a position to generate
new ¯ ows. Comparison of air photographs acqu ired in December 1995 with the
Landsat TM data of the mid-1980s indicates that lava remained below the rims of
the pits in the intervening period. Therefore, we can state that from 1975 to 1995
little or n o lava over¯ owed the pit craters.
C. Oppenheimer and P. Francis1670
(a) (b)
(c) (d)
Figure 4. First principal component images of Landsat M SS scenes of Erta `Ale acquired on
(a) 8 September 1972; (b) 30 January 1973; (c) 24 M arch 1975; (d) 3 March 1979. The
images h ave been resampled to 28´5 mÖ28 ´5 m pixe ls. T he cal dera dimensi ons are
#1600 m Ö700 m and north is approximately up the page.
2.3. Crater `collapse’
A signi® cant fract ion (as much as a third) of the eruption magnitude calculated
in the preceding section is represented by the late 1960s± early 1970s in® ll of the
350 m diameter crater in the north lobe by progressive overspills from the north
pond. Eventually lava streamed over a spillway on the northern lip of the caldera,
and ¯ owed 700 m downslop e. Th e 1992 ® eld party (S mithson ian Institution 1992 c)
foun d th is north lobe reo ccupied b y a 300 m d iameter, 200 m deep crater ( i.e., similar
to the pre-1968 morphology, ® gure 2) . This redevelopment of the crater is indicative
of density or mass ( pressure) changes in a deeper reservoir. Possible causes includ e
drain back of still ¯ uid magma in the crater soon after the eruptive episode, perhaps
as a result of dyke intrusion and withdrawal of magma from the reservoir.
Alternatively, the crater may have subsided episodically over the two decades since
1973 as a result of reduction of reservoir magma volume by crystallization and
Remote sensing of Erta `Ale volcano 1671
Figure 5. Cartoon (not to scale) showing ® eld measurements used to estimate volumes of
lava erupted between 1968 and 1974. North crater is represented by truncated and
nested right-cones; central crater by cylinders.
degassing. Walker (1988 ) h as discussed the ratio of subsidence to erupted volume of
calderas, noting several cases (including Hawaii, Askja, and Fernandina) where the
ratio exceeds unity.
Unfortun ately, it is di cult to estimate the depth of the north crater from the
Landsat TM images because of the presence of fumes. However, if we assume that
the magma levels in each crater were similar during the mid-1980s, then the shadow
measurements of central pit depth suggest t hat the north crater was already around
100 m deep by this t ime. If this is correct, then the estimated 200 m depth of the
north crater in the 1990s indicates further collapse (assuming the 200 m is estimated
from a point at similar altitude to t he rim of the central pit ). An embayed scar,
visible in the April 1984 TM scene (and all subsequent images), intersects the east
rim of the north crater. This may be a re-expression of the uppermost bench in the
1960s north crater, and lends support t o the idea that some drainback of lava from
this crater occurred in the decade after 1974. Accounting for the total `lost’ volume,
we arrive at an integrated surface mass ¯ ux rate of approximately 10 kg sÕ1between
1963 ± the acquisition date o f air photographs studied by Varet ( 1971 b)± and 1995.
This value is sensitive to the estimated 200 m depth of the north crater in 1994
(which could do with veri® cation), and the assumption of a funn el-sh aped crater.
3. Hea t ¯ uxes
A notable feature of the Landsat TM data (® gure 7) is the strong thermal emission
eviden t in bands 5 ( 1´55± 1´72 mm FWHM), and 7 ( 2´08± 2´35 mm). There are two
endmember types of lava lakesÐ active ones which are typically long-lived and
plumbed into a deeper reservoir of magma, and passive ones which are not. Erta
`Ale’s longevity indicates that it is of the former type. The surface of a typical active
lava lake is composed of orange- to red-hot incand escent cracks and bubbling zones
separated by substantially cooler skin or lava crust ( ® gure 8 ). The surface temperature
distribution across such a lake is therefore markedly heterogeneous Ð at the
C. Oppenheimer and P. Francis1672
Table 2. Lava ¯ uxes estimated from ® eld reports and satellite data . Erupted vol ume estimated for funnel-shaped north crater and cylinder-shaped
cent ral pit, using documented and satellite- measured lake levels. Al so given are eruption magnitude (m), cumulativ e ma gnitude (Sm) , magma
discharge rate (m
Ç) and net magma discharge rate (m
Çintegrated sin ce Decemb er 1968 ). Lava ¯ ow volumes beyond the p it craters are determined
from assumed thicknesses and satellite-measured p lan imetric areas (*co mprised o f 5 m thick ¯ ows over the we dge -shaped n ecks NW an d NE
of the cen tral pit, between n orth and central caldera lobes, 5 m depth in® ll of the central and south caldera lobes, and 2 m thi ck overspill ¯ ow
on SE ¯ ank; **includes 2 m thi ck ¯ ows over NW and SE cald era rims). Lava density of 2700 kg mÕ3assumed. Negative values represent crater
coll apse or lava d raining from the north cra ter, estimated fo r a righ t-co ne of basal radius 150 m and height 200 m.
Lava volu me ( Ö105m3)mSm m
Çnet m
Ç
Date s (no rth p it) (centra l pit ) (¯ ows) ( tota l ) ( Ö109kg ) ( Ö109kg ) ( kg sÕ1) ( kg sÕ1)
Dec 1968± Dec 1969 6´0 2´0 0´0 8´0 2´2 2´2 68´3 68´3
Dec 1969± Dec 1970 8´9 2´0 0´0 10´9 2´9 5´1 93´5 80´9
Dec 1970± Dec 1971 9´6 0´9 0´0 10´5 2´8 7´9 89´8 83´9
Dec 1971± Feb 1972 6´3 0´5 0´0 6´8 1´8 9´8 348´5 97´8
Feb 1972± Mar 1973 8´8 0´6 43´9* 53´3 14´4 24´2 42 1´5 18 0´3
Mar 1973 ± Jan 1974 2´3 0´3 4´6** 7´1 1´9 26´1 73´0 162´7
Jan 1974± Jan 1986 Õ45´7 Õ14´0 0´0 Õ59´7 Õ16´1 10´0 Õ42´6 18´5
Jan 1986± Nov 1992 Õ1´5 0´0 0´0 Õ1´5 Õ0´4 9´6 Õ1´8 12´6
Nov 1992 ± Dec 1995 0´0 0´0 0´0 0´0 0´0 9´6 0´0 11´3
Remote sensing of Erta `Ale volcano 1673
(a) (b)
(c) (d)
(e) (f)
(g) (h)
(i) (j)
(k) (l)
Figure 6. Images of summit region recorded by Landsat TM band 1 on (a) 21 April 1984;
(b) 24 June 1984; (c) 27 August 1984; (d) 30 October 1984; (e) 1 December 1984;
(f) 2 January 1985; (g) 18 January 1985; (h) 7 M arch 1985; (i) 23 March 1985; ( j) 5
January 1986; and by JERS-1 OPS band 2 on (k) 9 June 1992; (l) 14 May 1994. The
JERS- 1 images have be en resampled to 2 8´5 m Ö28´5 m pixels. The caldera dimensions
are #1600 m Ö700 m and north is approximately up the page.
C. Oppenheimer and P. Francis1674
(a)
(c)
(b)
(d)
(e) (f)
(h)
(j)
(l)
(n)
(g)
(i)
(k)
(m)
Remote sensing of Erta `Ale volcano 1675
centimetre scale, incandescent crack and cool crust can be in close proximity ( Flynn
et al. 1993 ) . This, of course, leads to strongly mixed Landsat TM pixels in the
thermal sense. Thermometry of such high-temperature surfaces has nevertheless been
tackled u sing multi-wavelength techniques ( Rothery et al. 1988).
The radiative heat ¯ ux from Erta `Ale’s lakes is su ciently great that it results
in saturation of both TM bands 5 and 7 in every image (appendix) . This situation
has been encountered repeatedly in previous investigations of volcano hotspots (e.g.,
Oppenheimer et al. 1993, Rothery et al. 1992) and seriously limits attempts at
radiometry. Oppenheimer et al. ( 1993) avoided making ¯ ux estimates altogether
from time-series Landsat TM data of Lascar volcano (Chile), preferring instead
simply to chart variations in spectral radiance. A rare case where spectral radiance
from an active lava body was largely within the dynamic ranges of the TM band 5
and 7 sensors is found in the analysis of an active andesite ¯ ow erupted from
Lonquimay volcano, Chile (Oppenheimer 1991).
3.1. Short wavelength inf rared radiometry
Here, we attempt to estimate the radiative and convective heat ¯ uxes from Erta
`Ale’s lakes by following the techniqu es presented in Oppenheimer (1991 ) and
Oppenheimer et al. (1993) but making some additional assumptions concerning
surface temperatures in order to constrain radiative and convective heat ¯ uxes. We
utilize t he two-thermal component model for hot volcanic surfaces originally
develo ped by Rothery et al. ( 1988 ) and discussed from a thermodynamic standpoint
by Crisp and Baloga (1990) but follow Oppenheimer (1991 ) in assuming that the
temperature of the hotter component ( Tc), which occupies a fraction f of a pixel’s
representative ground area, is more accurately known than that of the cooler compon-
ent ( Ts) . Rothery et al. ( 1988) assumed that the background component ( Ts) was
negligible. In their interpretation of the 5 January 1986 Landsat TM image of Erta
`Ale they co ncluded t hat the no rth l ake had incand escent lava at 800± 100 ßC occupy-
ing 0´04± 0´2 per cent of a pixel. For the central lake, they gave values of Tcof 570±
750 ßC occupying 0´3± 0´8 per cent of a pixel for the non-saturated parts of t he
anomaly. In the saturated core, they used the observation of no thermal radiance in
ban d 4 ( 0´76 ± 0 ´90 mm) to constrain f for various possible values of Tc. By neglecting
crust temperatures in this way, heat ¯ ux estimates based on these values are likely
to considerably underestimate true valu es (® gure 7 in Oppenheimer et al. 19 93 ). Th is
is because the fractional areas of the hot component (f ) are typically small (<1 per
cent ) such that the cooler crust dominates the heat loss.
We start with a value of Tcof 1100 ßC based on the measurements of Le Guern
et al. (1979 ) . Using appropriate radiometric calibration coe cients we can then
calculate the value of Tsfor any TM band and any given value of DNx( DN in band
Figure 7. Images of summit region acquired by Landsat MSS bands 4, 2, and 1, in red, green
and blue, respectively, on (a) 8 September 1972; (b) 30 January 1973; (c) 24 March
1975; (d) 3 March 1979; and by Landsat TM bands 7, 5, and 4, in red, green and blue,
respectively, on (e) 21 April 1984; ( f) 24 June 1984; ( g) 27 August 1984; (h) 30 October
1984; (i) 1 December 1984; ( j) 2 January 1985; (k) 18 January 1985; (l) 7 M arch 1985;
(m) 23 March 1985; (n) 5 January 1986. The MSS data have been resampled to
28´5 mÖ28 ´5 m pixe ls. The cal dera dimensi ons are #16 00 mÖ700 m and north is
approximately up the page.
C. Oppenheimer and P. Francis1676
(a)
(b)
Figure 8. (a) Field photograph of the Erta `Ale lava lake in December 1995 from the east
rim of the central pit (courtesy of Pierre Vetsch, SocieÂteÂde Volcanologie Gene
Áve);
(b) sketch map of lake position within the central pit (courtesy of Gerald Favre and
Pierre Vetsch, personal communication, 1996).
x) and f:
Ts=c2/{
l
ln [ 1+c1(1Õf )
l
Õ5(Rx/eltlÕfBl,Tc)} (2)
where c1and c2are 1´19Ö10Õ16 W m2and 1´44Ö10Õ2m K, respectively, Rxis the
spectral radiance (obtained by radiometric calibration of DNx, and corrected for
path radiance, atmospheric absorption, and surface re¯ ectance) in band x whose
centre wavelen gth is
l
,eland tlare the spectral emissivity of lava, and the atmo-
spheric spect ral transmittance, respectively, and Bl,Tcis the Planck spectral radiance
at wavelength
l
from a black body at temperature Tc. The curves in ® gure 9 represent
the loci of solutions for Tsand f determined from the Planck function for the Tcof
1100 ßC (approximate recorded temperature of active lava in t he Erta `Ale ponds, Le
Guern et al. 1979 ) and also of 900 ßC. So-called `dual-band’ solutions are possible
Remote sensing of Erta `Ale volcano 1677
(a)
(b)
Figure 9. Theoretical curves for two-thermal-component models based on Landsat TM
radiometry, for (a) Tc=1100 ßC, and (b) Tc=900 ßC. Solid lines represent band 7,
dashed band 5, dash-dotted band 4, and dotted band 3. Numbers alongside curves
refer to DNxbased on radiometric calibration coe cients given in appendix. The open
circles show dual-band solutions determined for all non-saturated pixel pairs of DN5
and DN7(taken from 5Ö5 pixel grids centred on the thermal anomalies) for all the
Landsat TM images.
where non-zero and non-saturated responses to thermal emissio n are detected in
two co-registered spectral chan nels.
Overlaid on ® gure 9 are all TM bands 5 and 7 solutions for Erta `Ale lava lake
C. Oppenheimer and P. Francis1678
(a)
(b)
(c)
Remote sensing of Erta `Ale volcano 1679
pixels. Re¯ ectance, path radiance and atmospheric transmission were all corrected
for in the way described in Oppenheimer et al. ( 1993). Whether these dual-band
solutions provide useful information or not is questionable because it may be the
case that these pixels merely represent `blur’ of the thermal signature arising from
the sensor point spread function and resampling artefactsÐ all the dual-band solu-
tion s pertain to the peripheries of 5 Ö5 pixel grids centred on the hot spots, and the
thermal anomalies extend spatially beyond the pit wall (® gure 10 ). What is clear is
that the important information is obscured by sensor saturation over the core of the
anomaly. To tackle this problem we use two important pieces of information: (i) the
response of Landsat TM band 4, and ( ii ) the estimated Tso f around 550± 650 ßC ( Le
Guern et al. 1979), and one assumption: that the dual-band solutions are meaningful
and as Tsincreases, so should f. In most of the Erta `Ale Landsat TM images, there
is no clear response in TM band 4. We can say that it is probab ly less than or equal
to about 4 DN as this would be a reasonable detect ion limit, t, for these particular
data (the threshold, tx, in band x, depends on sensor noise, variation in spectral
re¯ ectances across the crater, and illumination levels and angles). The 1 December
1984 image is the clearest exception where an estimated 17 DN in band 4 derive
from thermal emission (® gures 7 i, 10 c). The 7 March 1985 image also suggests
elevated band 4 DN of about 4± 5 over the lava lake.
To estimate ¯ uxes, we have separated `hot’ pixels for lava lake extracts (maximum
dimensions set to correspond with the present pit diameter) into ® ve classes according
to the following criteria:
(i) DN7>t7; DN5<t5
(ii) 255 >DN7>t7; 255>DN5>t5
(iii ) DN7=255; 255>DN5>t5
(iv) DN7=DN5=255; DN4<t4
(v) DN7=DN5=255; DN4>t4
where the subscripts represent TM band numbers, and DNxrepresents the thermal
response only. For the given Tcwe can use curves such as those in ® gure 9 to ® nd
representative values of Tsand f for each of the above classes. Only for class (ii ) are
dual-band solutions actually feasible, though not always so because of the possibility
of non-convergence. Class ( iv) is the crucial class for our study since the cores of all
the Landsat TM thermal anomalies are composed of numerous saturated pixels in
bands 5 and 7 without corresponding band 4 response. Taking t4=5, inspection o f
® gure 9 shows that there is a wide range of possible Tc, Tsand f solutions. Using
our assumption that f increases as Tsincreases, and t hat Tsis in the range 550±
650 ßC, signi® cantly narrows the space. For Tc=1100 ßC, we selected Ts=550 ßC,
and f=10Õ3for class (iv). For Tc=900 ßC, we set Tst he same but increased f to
10Õ2. Representative values for classes ( i ), (iii) and (v) were foun d in a similar manner.
Systematically assigning each multi-spectral `hot’ pixel to one of t hese classes
Figure 10. Approximately west-east cross-track DN pro® les, normalized as ( DNx-
DNx,m in )/(DNx,max-DNx,m in ) where DNx,max and DNx,min are the maximum and min-
imum DNx, respectively, in the given line, traversing the core of the thermal anomaly
in Landsat TM data of (a) 21 April 1984; (b) 24 June 1984, and (c) 1 December 1984.
Circles represent band 3, squares band 4, triangles band 5, and diamonds band 7.
Thick vertical lines indicate edges of central pit. High DN in bands 3 and 4 west of
the pit result from higher re¯ ectances of pre-1970 lavas preserved on a kipuka.
C. Oppenheimer and P. Francis1680
determines the lake’s thermal distribution in each image. Radiative ¯ uxes can then
be calculated from the Stefan± Boltzmann relationship, and convective ¯ uxes from
empirical formulae (see Oppenheimer 1991). The results for Tc=90 0 ßC are given in
table 3. We feel this value of Tcis more realistic since it pushes up class (iv) pixels
into the 1± 10 per cent range for f. This is compatible with the estimate of f as
approximately 7 p er cent in 1973 ( Le Guern et al. 1979), although there is no reason
to assume that this is representative of long-term ( tens of years) trends (see Flynn
et al. 1993, for a discussion of variability in f for Kupaianaha lava lake, Hawaii).
The apparent variations in heat ¯ ux indicated by these satellite snapshots must be
considered in the context of the short-term (minutes to hours) variability in thermal
output from lava lakes as they go through cycles of quiescence, steady overturning
and fountaining ( Tilling 1987, Flynn et al. 1993). Table 4 indicates lava surface
conditions compatible with f of 10Õ1, 10Õ2and 10Õ3.
`Hot’ pixels are also evident in band 4 (0´8± 1´1 mm) of Landsat MSS data of 1972,
corresp ond ing with the north pit, and of 1973, with the central pit ( ® gure 7a,b).
These indicate vigorous activity compared with the Landsat TM era data in which
few Landsat TM band 4 pixels reveal thermal signatures despite the similar bandpass,
smaller instantaneous ® eld of view ( IFOV) , and higher radiometric reso lution of the
sensor. For the anomalous Landsat MSS pixels, heat ¯ ux estimates were based on
pixel-integrated temperatures. The 1992 JERS-1 scene (® gure 11 ) reveals strong
thermal emission from the central pit in the short-wavelength infrared bands (centre-
wavelen gths: band 5, 1´655 mm; ban d 6, 2´065 mm; ban d 7, 2´19 mm; b and 8, 2´335 mm)
but severe sensor n oise precludes radiometry.
It is di cult to assess the veracity of our ¯ ux estimates, since they are entirely
model-dependent. Overlap of IFOVs of adjacent pixels, and the temporary downscan
damage seen in most of the band 5 and 7 data, are both likely to result in overestima-
tion of the intensity (size) of anomaly. However, our resu lts are in general agreement
with t he estimated ¯ ux of 226 MW (113 MW for each pond ) based on ® eld o bserva-
tion s in January 1973 by Le Guern et al. (1979 ), and we are co n® dent that, at the
least, they give `ballpark ® gures’ and indicate qualitative trends. Two observations
pertaining to accuracy are worth emphasizing, however: (i) inclusion of peripheral
pixels, i.e., classes ( i )± (iii ), makes a relatively small (#20± 40 per cent) di erence to
estimated heat ¯ uxes; ( ii ) setting Tcto 90 0 or 1100 ßC has even less impact on the
result s (<1 per cent ). We can be reasonably con® dent that lava lake activity has
persisted at Erta `Ale throughout the observation period, though there remain some
lengthy gaps in short-wavelength infrared data acquisitions (notably between 1974
and 1984 ).
NOAA AVHRR image data have also been utilised for volcano thermal mon-
itoring notably through analysis of the channel 3 ( 3´55± 3´93 mm) response (e.g., Harris
et al., 1995 ) . We have obtained a night-time AVHRR image covering Erta `Ale,
recorded on 6 February 1996, which shows a clear thermal signal in channel 3 above
the lava lake(s) con® rming ongoing activit y (® gure 12 ).
3.2. Spatial patterns of heat ¯ ux
In all the Landsat TM imagery studied, the central lava lake appears active
(® gure 7 ). The thermal signature is approximately equidimensional. De® ned by
band 7 saturated pixels, it varies from 3 Ö4 t o 5 Ö5 pixels in size (appendix, the
5 January 1986 image reveals a 3 Ö2 anomaly but this is partly a result of the cubic
convolution resampling of the image Ð all the others have been processed with a
Remote sensing of Erta `Ale volcano 1681
Table 3. S ummed radiative and co nvectiv e heat ¯ uxes from Erta ’Ale lava ponds determined fro m infrared satellite data. For Lan dsat TM observations,
values are given for d i erent sets o f pixel cla sses as d e® ned in §3´1. Excludin g periph eral pixels (e.g., classes ( iii ) ± (v), and ( iv )± (v)) has a mo re
pron ounced e ect on the less intense the rmal featu res. Senso r damage, characterized by down-scan lines of saturated band 5 and 7 pixels will
also result in overestimati on of ¯ uxes for bigger/hotter lava lakes. The Landsat MSS-derived ¯ uxes (asterisked ) were deriv ed from pixel-
integrated tempera ture s in MSS band 7 (0´8± 1´1 mm).
Heat ¯ u x/MW
North lak e Central Lake
Date (classes ( i )± (v)) (classes ( i )± (v)) (classes ( ii )± (v) (classes ( iii)± (v)) (cl asses ( iv)± (v))
8 Sept 1972 330*
30 Jan 1973 160*
21 Apr 1984 140 130 110 80
24 Ju ne 19 84 190 180 160 130
27 Aug 1984 200 190 170 140
30 Oct 1984 4 310 310 280 260
1 Dec 1984 420 420 400 390
2 Jan 1985 200 190 180 140
18 Jan 1985 270 270 250 210
7 Ma r 1985 9 280 280 260 240
23 M ar 1985 2 250 250 220 190
5 Jan 1986 16 130 120 90 80
C. Oppenheimer and P. Francis1682
(a)
(d)(c)
(b)
Figure 1 1. R aw JERS- 1 l evel zero O SW (short- wavelen gth in frared ) images o f summit re gion
acquired on 9 June 1992: (a) band 5, (b) band 6, (c) band 7, and (d) band 8. The caldera
dimensions are #1600 Ö700 m and north is approximately up the page.
nearest-neighbour kernel ). What is prominent in every image, when bands 5 or 7
are compared with bands 1± 3, is the displacement of the thermal signature to the
western or south-western part of the central pit ( ® gures 7, 10). This is very probably
real, rather than an artefact of interband misregistration, since this position conforms
very closely with that observed by ® eld parties in t he 1990s (® gure 8 ), and there are
no indications of signi® cant interband misregistration elsewhere in the images. Also,
Table 4. Representative dimensions of lava lake surface features (incandescent cracks and
lava fountains) compatible with f values of 10Õ1, 10Õ2, and 10Õ3, f or 28 ´5 m Ö28´5 m
pixels.
Crack length (m)Öwidth Diameter of fountain
f (m) (m)
0´001 25Ö0´032 1´0
50 Ö0´016
75 Ö0´011
100Ö0´008
125Ö0´007
0´01 25Ö0´32 3´2
50 Ö0´16
75 Ö0´11
100Ö0´08
125Ö0´07
0´1 25Ö3´2 10´0
50 Ö1´6
75 Ö1´1
100Ö0´8
125Ö0´7
Remote sensing of Erta `Ale volcano 1683
Figure 12. NOAA AVHRR Channel 3 image of the whole Erta `Ale range ( 95 kmÖ42 km)
recorded at 0436 GMT on 6 February 1996. Note that the image is not geometrically
corrected and that the DN scale has been inverted for display. Hot spot is arrowed.
taking into consideration the optical and resampling e ects that act to smear out
ther mal an omalies in La ndsat TM data ( e.g., Roth ery and Oppe nheimer 1994 ) , th e
apparent dimensions of the lake (as de® ned by saturated band 7 pixels) are quite
similar to those recorded in the ® eld. Only one of the Landsat TM images reveals
a strong thermal signature in ban d 4, that of 1 December 1984, in which one pixel
at the western edge of the pit has a corrected DN4of 17. This position coincides
with t he dominant fountaining site recorded b y the 1995 ® eld party (S mithsonian
Institution 1995), perhaps indicating a long-lived phenomenon.
While the central lava lake has persisted, activity of t he northern lake appears
to have been intermittent. Thermal emission is clearly evident only in the images of
30 Octo ber 1984, 7 and 23 March 1985, and 5 January 1986, and even then, the
feature is very much less intense than that in the central pit. Two explanations can
be o ered for this variability in thermal signature: ( i) the north lake remained active
but thermal radiation from it was intercepted by absorption and scattering in the
dense fumes often seen emanating from the north crater, or (ii ) the lake was truly
ephemeral. If the latter is the case then this may re¯ ect intermittent collapse of the
nort h crater as discussed in §2.3, covering the lake surface with t alus. When present,
the t hermal anomaly is consistently located in the centre of the north lobe. The
comparatively high TM band 7: band 5 radiance is indicative of an active lava body
rather than hot fumarole vents (Oppenheimer et al. 1993 ).
3.3. L an dsat T M band 6
For all of the Tc, Ts, f solutions we discuss in §3.1, the Landsat TM thermal
ban d 6 ( 10´42± 12´45 mm; nominal 120m spatial resolution) should be easily saturated
if the lake occupies substantially all of the sensor instantaneous ® eld of view (IFOV).
C. Oppenheimer and P. Francis1684
We ® nd band 6 saturat ed in the images of 21 April and 30 October 1984, 2 and
18 Janu ary and 23 March 1985, and 5 Janu ary 1986. However, the scenes of 24 June,
27 August and 1 December 1984, and 7 March 1985 indicate maximum DN6of 252,
226, 233, and 251, respectively, over the central lake. These non-saturated data could
indicate that crust temperatures ( Ts) are much lower than we assumed, that there is
signi® cant absorption in band 6 of thermal emission from the lake surface by fumes,
and/or that the lake occupies a smaller fraction of the sensor IFOV. Given that one
of the non-saturated TM band 6 pixels represents the lake at its apparent maximum
activity ( 1 December 1984 ), we feel that the latter t wo factors are more likely to be
relevant.
4. G as/aerosol plume s
Magmatic degassing takes place directly from Erta `Ale’s lava ponds. Gases
collect ed above active lava in spatter cones and ponds in the early 1970s revealed a
very st able co mpositio n, with typ ically 80 mol% H2O, 10% CO2, 7% SS (most ly as
SO2), 1´5% H2, 1% HCl, and 0´5% CO (Le Guern et al. 1979, Gerlach 1989 ). The
comparatively high CO2concentrations led Gerlach ( 1989 ) to conclude that magma
feeding the active lava ponds did not reside for long in shallow crustal reservoirs.
Le Guern et al. ( 1979) estimated a sulphur dioxide ¯ u x fro m Erta `Ale of 0´6 kg sÕ1
based on data collected in 1973.
In most of the images stud ied, condensed steam plumes are evident (® gures 4,
6). The interpretation of these small aerosol clouds (apparent in the visible and near-
infrared bands) is limited since their visibility re¯ ects meteorological as well as
volcanic factors. It is reasonable, nevertheless, to assert that they represent emissions
from moderate temperature (#200± 300 ßC) fumaroles sited on or near the pit
walls. High (magmatic) temperature gases emitted directly from the lava ponds have
generally been reported as invisible to the eye.
In the late 1960s, dense fumes emanated from the central pit. Such emission is
eviden t from our earliest space data: both the declassi® ed intelligence satellite and
Apollo 9 astronaut photographs (® gure 1) show the central pit ® lled with condensed
steam. As the lava level rose in both p its, emission of dense fumes seems to have
subsided, as indicated by the Landsat MSS images of 1972± 1975. Subsequently,
dense fume emission switched to the north pit. In the Landsat MSS image of 1979
a 30ßwedge- shaped plume made of extended pu s is seen, visible for at least 5 km
ENE from the centre of the north crater from where it issues. It is not clear whether
this unusually prominent plume re¯ ects atmospheric or volcanic conditions but one
possibility is that it was a consequence of a collapse event within the north pit.
In the Landsat TM data of the mid-1980s, the densest fumes are seen on the
insid e of the southern rim o f the north lobe ( ® gure 13 ). This corresponds closely
with the disposition of fumarole vents issuing dense white fumes on the south and
sout h-west border and inside wall of the north crater observed in the early to
mid-1990s (S mithsonian Institution 1992 b, 1995) . A fumarole t emperature o f 230 ßC
was recorded here in 1994 ( J.-L. Chemine
Âe, personal communication, 1994 ). In
several Landsat TM scenes, discrete p u s are seen drifting away from the north lobe.
This is clearest in the Landsat TM image of 18 January 1985, where at least ® ve
pu s can be seen reaching more than 6 km from the vent. For windspeeds in the
rang e of 2± 10 m sÕ1, the crater is pu ng every 2± 10 min. This behaviour doubtless
result s from aerodynamic e ects, not from discontinuou s emission fro m the vents.
The January 1986 Landsat TM scene indicates a particularly fume-free crater, coin -
Remote sensing of Erta `Ale volcano 1685
(a)
(i)(h)(g)
(f)(e)(d)
(c)(b)
Figure 13. Landsat TM band 1 images, contrast stretched to emphasize fumarole emissions
from the north crater, acquired on (a) 21 April 1984; (b) 24 June 1984; (c) 27 August
1984; (d) 30 October 1984; (e) 1 December 1984; ( f) 2 Jan uary 19 85; ( g) 18 January
1985; (h) 7 March 1985; (i) 23 March 1985; ( j) 5 January 1986. The caldera dimensions
are #1600 m Ö700 m and north is approximately up the page.
cident with resurgence of active lava in this pit. The coincidence of lava lake activity
and absence of dense clouds of low temperature fumarole vapours from the same
crater may be causally related, though as discussed in §3.2, this could be simply
because dense fumes will tend to obscure the crater ¯ oor. Resumed pu ng from the
nort h crater is evident in photographs taken from the Space Shuttle on 30 September
1988, 14 March 1989 and 7 May 1989, and in the subsequent SPOT and JERS scenes.
C. Oppenheimer and P. Francis1686
5. S yn th es is o f E rt a `Al e’ s ac ti vi ty
Integratin g all the information available to us, we can summarize Erta `Ale’s
activity o ver the last 30 years as follows (® gure 14). During the late 1960s the levels
of magma within two conduits, approximately 400 m apart, rose in tandem, rapidly
at ® rst but slo wing as the wide north crater began ® lling up. By 1973, both north
and central lava lakes were substantially in® lling the north, central and southern
caldera lobes. Over a period of approximately one year, the caldera ¯ oor was
gradually resurfaced and some thin ¯ ows overtopped both the north and south lips
of the caldera. The average e usion rate over this period was around 160 kg sÕ1,
similar to that estimated for Halemaumau lava lake (Hawaii) averaged over its
century-long life (Francis et al. 1993). There is no evidence for subsequent lava ¯ ow
emplacement, and for at least the period 1984± 1995, the lava lake level appears to
have achieved a steady-state around 100 m below the caldera ¯ oor. This implies that
the magma level over this time has remained 60 ± 80 m higher than it was in 1968
(accou nting for the increased height of the central crater rim due to overspills in
the early 1970s). Heat ¯ uxes of the order of 100± 400 MW were probably sustained
thro ughout the period, indicative of convective circulation between the active lava
lake (s) and a deeper reservoir. High thermal emission may have resulted from more
vigorous convection and/or lava fountain activity associated with the in¯ ux of
gas-rich reservoir magma. The apparently steady-state magma level may indicate
that, over at least the last decade, caldera ¯ oor subsidence has kept pace with the
reservoir de¯ ation that arises from crystallization and degassing. If this
is the case, the underlying magma body should be at shallow depth and laterally
extensive.
Dense white fumes have been emitted from one of the two pit craters for most
of the observation period. These moderate temperature fumarole emissions suggest
that, despite the arid environment, some kind of hydrothermal system exists at
shallow levels in the volcano. This may result from sub-surface condensation of
magmatic ¯ uids when crater collapses cover the lava pond(s) with talus. Considerable
subsidence or magma withdrawal has occurred since the in® lling of the northern
lobe of the caldera in t he early 1970s. When observed in 1992, it had a morphology
not dissimilar to that which it had before the eruption episode. From our dataset it
is di cult to constrain precisely the timing of `collapse’ but there is evidence for
episodes between 1974 and 1984 ( possibly drainback of ¯ uid ), and between 1984
and 1992 ( possibly subsidence of rock).
Francis et al. (1993 ) estimated that a magma in¯ ux of around 500 kg sÕ1was
required to support the SO2degassing rate from Erta `Ale’s lava ponds, and conjec-
tured that this was unlikely to be matched by the actual eruption rate. Integrating
the net lava e usion over the last t hirty or so years, we ® nd a rate of only around
10 kg sÕ1, supporting the hypothesis that the volcano `grows’ largely by intrusion of
dykes and sills, and formation of cumulates, consistent with, and probably promoted
by, the extension of the north Afar crust.
6. C o nc lu sions
(1) We believe this is the ® rst time that satellite data have been utilized in order
to study the ¯ uctuations of magma level within a conduit. These are important since
they are related to magma pressures in connected reservoirs and hence say something
about sub-surface processes. Magma levels could be particularly relevant in estab-
lishing the trigger of some eruptions. For example, it has been suggested that cycles
Remote sensing of Erta `Ale volcano 1687
(a) (b)
(c)(d)
Figure 14. Graphs o f (a) summed radiative and convective heat ¯ ux, (b) exogenous growt h, (c) lava lake levels, summarizing external phenomena at
Erta `Ale between 1968 and 1995, and (d) lake levels (solid line) and heat ¯ ux (dotted line) between 1984 and 1985.
C. Oppenheimer and P. Francis1688
of withdrawal of magma and consequent lava dome subsidence presaged explosive
eruptions of La
Âscar volcano, Chile (Matthews et al. 1996 ).
(2) Given the dynamic ranges of exist ing spaceborn e short-wavelength infrared
channels, sensor saturation will remain a problem for high temperature thermometry.
This is likely to be the case even for future sensors with optional gain settings such
as the Landsat Enhanced Thematic Mapper ( ETM +) due for launch in 1998.
Radiometry of phenomena such as lava bodies, hot fumarole vents, ¯ ares and ® res,
requires therefore a strategy for coping with saturation. The methodology we have
employed here could be adapted to other sensors and applications.
(3) Numerous e orts have been made to develop and re® ne algo rithms for
radio metry of hotspots (volcanoes, ® res, etc.). With archive imagery becoming more
cheaply and readily available, there is considerable scope for applying the
techniques to longer time-series datasets. We are particularly grateful for the
continuity of the Landsat programme, and t he archives of the Eros Data Center
and EOSAT. Many valuable data reside in t he archives of ground stations
worldwide that are less seldom tapped intoÐ easier, cheaper access to some of
these data archives would be welcome. Clear information regarding radiometric
calibration is essential.
(4) We have shown how remotely sensed data provide simultaneous information
on quite di erent parameters: in this case, lava e usion, lake levels, heat ¯ uxes,
fumin g and other visible p heno mena. Previous studies have sometimes focused on
retrieval of just one parameter, e.g., thermal output, possibly overlooking relevant
detail that could be provided by t he neglected bands. We encourage multi-sensor
remote sensing investigations of active volcanoes, and note for example, that Erta
`Ale should be an excellent target for di eren tial interferometric synthetic aperture
radar studies on account of its likely high rates of deformation, low angle slopes and
arid, barely vegetated environment. A combination o f interferometric SAR, t hermal
and lake level studies, all by satellite remote sensing, could provide further insights
into the magma storage and supply to this fascinating volcano.
Acknowledgments
We thank the Royal Society and the Ethiopian Institute of Geological Surveys
for supporting work in north Afar, the Earth Observation Satellite Company for the
Landsat data (obtained through data grant 134), the National Space Development
Agency of Japan and David Rothery of the Open University for the JERS-1 imagery,
and P. Mouginis-Mark for showing us the hardcopy SPOT scene. We particularly
wish to thank Pierre Vetsch of the Socie
Âte
ÂVolcanologique de Gene
Áve for information
on Erta `Ale’s recent behaviour; Seife Berhe, Solomon Kebede, Tadessa Mamo, Lema
Gelaget , Telahun, and Jon Rogers for enduring the Danakil; and Marta Bruno for
linguistic assistance.
App en di x. R a w Lan dsat T M b and 4 , 5, an d 7 d ata
Radiometric gains are 0´08148, 0´01083 and 0´006111 m WcmÕ2srÕ1mmÕ1DNÕ1
for bands 4, 5 and 7, resp ectively; biases are Õ0´1508, Õ0´03687 and
Õ0´01587 m W cmÕ2srÕ1mmÕ1for ban ds 4, 5 and 7, respectively. For each image,
DNxare given for the same 5Ö5 pixel grid in the order band 4, band 5, band 7.
Remote sensing of Erta `Ale volcano 1689
21 April 1984
41 36 36 32 30 57 71 105 54 33 43 69 255 255 56
40 38 36 33 28 62 200 255 68 35 118 118 255 255 132
40 38 36 33 28 59 209 255 187 20 46 116 255 255 255
40 40 40 36 30 59 87 213 119 47 90 90 255 255 146
38 39 38 38 41 54 54 51 64 60 38 50 52 64 73
24 June 1984
43 43 42 42 34 60 58 62 43 37 66 84 71 41 41
42 42 41 38 36 67 161 99 56 43 255 255 255 236 59
42 42 42 39 38 86 255 255 51 44 255 255 255 255 125
43 43 43 42 41 74 255 255 91 46 255 255 255 255 75
44 43 43 44 44 61 71 75 66 57 86 150 132 132 71
27 August 1984
63 58 50 46 35 88 88 94 82 60 80 117 151 97 68
62 53 44 43 36 90 111 248 173 64 117 255 255 236 110
60 53 48 43 37 90 223 255 255 83 121 255 255 255 255
59 56 52 47 43 88 140 255 255 127 145 255 255 255 49
59 56 52 47 43 81 87 105 117 82 91 158 255 247 113
30 Octo ber 1984
28 27 31 29 18 44 56 58 58 42 57 92 138 98 64
26 30 30 25 15 64 255 255 44 31 106 255 255 255 40
26 26 24 16 11 163 255 255 255 243 138 255 255 255 255
26 26 24 16 11 247 255 255 255 55 171 255 255 255 255
27 28 21 16 20 73 129 146 79 51 139 255 255 208 98
1 December 1984
25 25 30 24 13 37 37 55 78 31 78 74 164 129 49
24 25 22 16 11 91 255 255 164 46 139 255 255 255 130
25 38 20 12 12 255 255 255 255 121 255 255 255 255 255
25 27 20 15 17 255 255 255 255 55 255 255 255 255 255
27 26 26 29 32 128 128 255 255 46 194 255 255 255 143
2 Janu ary 1985
19 20 17 11 7 32 81 206 49 18 83 255 255 51 53
20 19 10 9 9 47 255 255 45 36 255 255 255 255 117
20 19 11 9 11 53 255 255 39 59 255 255 255 255 255
20 19 11 9 11 39 66 104 63 23 117 255 255 143 47
23 23 23 27 27 66 42 31 32 39 48 138 74 41 31
C. Oppenheimer and P. Francis1690
18 Janu ary 1985
21 22 25 19 9 44 73 62 62 27 70 224 255 111 71
21 19 15 10 9 183 255 255 19 19 147 255 255 255 0
21 19 14 10 10 255 255 255 255 72 214 255 255 255 255
22 23 20 16 20 165 255 255 36 36 148 255 255 255 255
24 22 23 26 29 46 69 121 53 53 78 157 255 255 92
7 March 1985
25 30 30 22 11 39 57 96 81 27 102 216 229 44 11
24 28 25 17 12 32 255 255 245 45 255 255 255 255 156
24 30 26 17 12 255 255 255 255 34 255 255 255 255 255
24 30 24 22 24 255 255 255 255 126 255 255 255 255 255
26 26 24 22 24 58 90 87 87 63 97 222 255 230 112
23 March 1985
31 34 30 30 20 61 146 141 141 56 118 255 255 190 38
31 34 30 30 20 255 255 255 106 106 255 255 255 255 255
30 35 27 27 16 83 255 255 205 32 255 255 255 255 255
31 36 22 22 16 99 255 255 69 48 80 255 255 255 184
29 30 32 32 33 45 51 53 49 52 70 122 107 72 72
5 Janu ary 1986
26 22 22 23 16 41 26 34 54 46 31 94 189 206 123
25 22 18 11 8 9 115 181 144 53 97 196 255 255 221
24 24 17 8 8 15 129 255 255 79 254 255 255 255 236
24 26 23 16 13 37 66 157 165 68 173 201 227 203 142
25 27 26 26 28 30 44 20 7 32 0 62 106 71 47
Refere nc es
d’ Abaddie, A.T., 1890, Ge
Âographie de l’E
Âthiopie: ce que j’ai entendu, faisant suite aÁ ce que
j’a i vu ( Paris: Mesn il ).
Barberi, F., Borsi, S., Ferr ara , G., Mar in elli, G.and Va re t, J., 1970, Relations between
tectonics and magmatology in the northern Danakil Depression (Ethiopia).
Phi losoph ical T rans actions of the R oyal S ociety of L on don, A 267, 293± 311.
Barberi, F., Chem ine
’e, J.L.and Va ret , J., 1973, Long-lived lava lakes of Erta Ale volcano.
Rev ue de G e
Âographie Physique et de Ge
Âologie Dynamique,15, 347± 352.
Barberi, F.and Varet , J., 1970, The Erta `Ale volcanic range. Bulletin Volc anologiq ue,
34, 848± 917.
CNR± CNRS Afar team, 1973, Geology of northern Afar (Ethiopia). Revue de Geographie
Physique et de Geologie Dynamique,15, 443± 490.
Cri sp, J.and Balo ga, S., 1990, A model for lava ¯ ows with two thermal components. Journal
of Geophysical Research,95, 1255± 1270.
Dain elli , G.and Mar in elli, O., 1906, Dell’Erta-ale, vulcano ritenuto attivo della Dancalia
settentrionale. Rivista Geogra® c a Italiana ,13, 261± 270.
Dain elli , G.and Mar in elli, O., 1907, Vulcani attivi della Dancalia. Ri vista Geogra® ca
Italiana,14, 16± 26.
Remote sensing of Erta `Ale volcano 1691
Dain elli , G.and Mar in elli, O., 1912, Risultati scienti® ci di un viaggio nella Colonia Eritrea.
Publicazioni del Istituto di Studi Superiori Pratici e di Perfezionamento di Firenzi.
Flynn , L.P., Mouginis-Mark, P.J. , Gradi e, J.C.and Luc ey, P.G., 1993, Radiative
temperature measurements at Kupaianaha lava lake, Kilauea volcano, Hawaii. Journal
of Geophysical Research,98, 6461± 6476.
Fran ci s, P.W., Oppe nhe ime r, C.and Ste venson, D., 1993, Endogenous growth of persistently
active volcanoes. Nature,366, 554± 557.
Gerl ac h, T.M., 1989, Degassing of carbon dioxide from basaltic magma at spreading centers:
I. Afar transitional basalts. Journal of Volcanology and Geothermal Research,39,
211± 219.
Har ris , A.J.L., Va ugh an, R.A.and Ro the ry, D.A., 1995, Volcano detection and
monitoring using AVHRR data: the Kra¯ a eruption, 1984, International Journal of
Rem ote S ensing,16, 1001± 1020.
Hilde bra ndt, J.M., 1875, Erlebnisse auf einer Reise von MassuÃa in das Gebiet der Afer und
nach Aden. Zeitschrift der G ese llschaft fur Erdkunde z u Berlin,10, 1± 38.
LeGuer n, F. , Car bon elle , J.and Taz ief f, H., 1979, Erta `Ale lava lake: heat and gas transfer
to the atmosphere. Journal of Volcanonology and Geothermal Research,6, 27± 48.
Matthews, S.J., Gard ewe g, M.C.and Sparks, R.S.J., 1997, The 1984 to 1996 cyclic
activity of Lascar volcano, northern Chile: cycles of dome growth, dome subsidence,
degassing and explosive eruptions. Bulletin of Volcanology, in press.
Mohr , P.A., 1962, The Ethiopian rift system. Bulletin of the Geophysical Observatory of Addis
Ababa,5 , 33± 62.
Munz ing er, W., 1869, Narrative of a journey through the Afar country. Journal of the Royal
Geographical Society,39, 188± 232.
Nesbitt, L.M., 1934, Desert and Forest: the exploration of Abyssinian Danakil (London:
Jonathan Cape) 450 pp. (Also published in 1935 by A. A. Knopf, New York, under
the title `Hell-hole of creation: the exploration of the Abyssinian Danakil’.)
Oppen hei mer , C., 1991, Lava ¯ ow cooling estimated from Landsat Thematic Mapper infrared
data: the Lonquimay eruption (Chile, 1989). Journal of Geophysical Research,96,
21865± 21878.
Oppen hei mer , C.and Franc is, P., 1997, Implications of longeval lava lakes for crustal growth
beneath Erta `Ale volcano, north Afar, Geology, submitted.
Oppen hei mer , C., Fra ncis, P.W., Rothery, D.A., Car lton, R.W.T.and Glaze, L.S.,
1993, Infrared image analysis of volcanic thermal features: Lascar volcano, Chile,
1984 ± 1992. Journal of Geophysical Research,98, 4269 ± 4286.
Roth ery, D.A., Borgia , A., Ca rlto n, R.W.and Oppe nhe ime r, C., 1992, The 1992 Etna
lava ¯ ow imaged by Landsat TM. Internati onal Journal of R emote Sensi ng,13,
2759 ± 2763.
Roth ery, D.A., Franci s, P.W.and Wo od, C.A., 1988, Volcano monitoring using short
wavelength infrared data from satellites. Journal of Geophysical Research,93,
7993 ± 8008.
Roth ery, D.A.and Oppenhei mer , C., 1994, Monitoring Mount Erebus by satellite remote
sensing. Antarctic R ese arc h Se ries,66, 51± 56.
Smithsonian Institution, 1992 a, Erta `Ale (Ethiopia). Bulle tin of the Gl obal Volc anism Network,
17, 5.
Smithsonian Institution, 1992 b, Erta `Ale (Ethiopia). Bulle tin of the Global Volcani sm Network,
17, 4.
Smithsonian Institution, 1992 c, Erta `Ale ( Ethiopia). Bul letin of the G lobal Volcanism Network,
17, 2.
Smithsonian Institution, 1995, Erta `Ale (Ethiopia). B ulletin of the Gl obal Volcanism Network,
20, 11.
Tanc re di, A.M., 1912, Un vulcano della Dancalia in eruzione. Rivista G eogra® ca Italiana ,
14, 490± 491.
Taz ief f, H., 1973, The Erta `Ale volcano. Rev ue de G e
Âographie Physique et de Ge
Âologie
Dynamique,15, 437± 441.
Taz ief f, H., 1994, Permanent lava lakes: observed facts and induced mechanisms. Journal of
Volcanology and Geothermal Research,63, 3± 11.
Tilli ng, R.I., 1987, Fluctuations in surface height of active lava lakes during 1972± 1974
Remote sensing of Erta `Ale volcano1692
Mauna Ulu eruption, Kilauea volcano, Hawaii. Journal of Geophysical Research,92,
13721± 13730.
Var et, J., 1971 a, Erta `Ale activity (Afar, Ethiopia). Bulletin Geophysical Observatory Addis
Ababa,13, 115± 119.
Var et, J., 1971 b, Sur l’activiteÂreÂcente de l’Erta `Ale (Dankalie, Ethiopie). Compte Rendue
Academie des Sciences de Paris,272, 1964 ± 1967.
Walk er, G.P.L., 1988, Three Hawaiian calderas: an origin through loading by shallow
intrusions? Journal of Geophysical Research,93, 14773± 14784.
White , R.S.and McKenz ie , D., 1989, Magmatism at rift zones: the generation of volcanic
continental margins and ¯ ood basalts. Journal of Geophysical Research,94, 7685 ± 7729.
