Jökull - 01.12.2006, Qupperneq 73

Pleistocene rhyolitic volcanism at Torfajökull, Iceland
number of dated samples. (It is noteworthy that Fig-
ure 6 shows more scatter, for Nb and Sr, when analy-
ses from undated rhyolites are included.) Only when
further ages have been obtained will the robustness
of this near-linear trend be tested further. The third
trend, the pronounced drop in alkalinity after 83 ka,
has been followed by the dominance of subalkaline
rhyolite compositions during the Holocene. The ap-
parent permanence of this change is emphasised by
studies (McGarvie, 1984; Macdonald et al., 1990)
showing that the Holocene rhyolites, with one excep-
tion, the compositionally-zoned Hrafntinnusker erup-
tion, have compositions that are distinctly less evolved
than those of the ring-fracture rhyolites (McGarvie et
al., 1990). Furthermore, within the Holocene rhyo-
lites, for which there is reasonable stratigraphic con-
trol, only a slight trend of less-evolved compositions
being erupted with time is present (Macdonald et al.,
1990; McGarvie et al., 1990), which might indicate
that some process is buffering the compositions of
erupted Holocene magmas.
At least two viable models can explain the third
trend. The first is that the c. 16 km3 ring fracture
rhyolite eruption has depleted the magma system of
evolved melt, leaving only a subalkaline mush. For
the past c. 65 ka this mush has been periodically in-
truded by basaltic magmas, resulting both in the pro-
duction of small batches of rhyolitic melt, and (dur-
ing eruptive episodes) by extensive interactions be-
tween intruding basaltic magmas and rhyolitic melts.
Occasionally, melt generation and/or storage condi-
tions has favoured the differentiation of batches of
rhyolitic melt, leading (for example) to the production
of the Hrafntinnusker zoned eruption (McGarvie et
al., 1990). This could imply that episodes of basaltic
magma injection not only trigger eruptions, but may
contribute thermal energy that helps to generate fu-
ture batches of rhyolitic melt (see also Gunnarsson
et al., 1998). The second model ignores any direct
role for basaltic magma injection in rhyolite magma
production, and involves two steps, the first being
the production of batches of ’dominant volume’ rhy-
olitic magma (Smith 1979), and the second being
the differentiation of this starting composition. This
is broadly similar to models for other volcanic com-
plexes, such as Pantelleria (Mahood et al., 1990) and
Boina (Barberi et al., 1975). At Torfajökull, the sub-
alkaline compositions that typify Holocene eruptions
are assumed to represent the starting composition, and
although there is disagreement over how this melt is
produced (see Macdonald et al., 1990; Gunnarsson et
al., 1998), for present purposes this can be ignored.
Once the starting composition has been produced, it
can then differentiate towards more-evolved compo-
sitions such as comendites and pantellerites. Only at
one Holocene eruption (Hrafntinnusker, McGarvie et
al., 1990) has this occurred convincingly. Two possi-
bilities stem from this: firstly that the rate of produc-
tion of the starting composition is now less than it was
in the past (because there has not been a return to the
production of highly-evolved rhyolitic compositions
such as pantellerites); and secondly that rhyolite melt
replenishment is occurring at rates similar to those of
the past but repeated injections of basaltic magma at
c. 500-1000 year intervals (Larsen, 1984) are deny-
ing the accumulating magma of sufficient residence
time (and sufficient accumulated volumes) to differen-
tiate towards more evolved compositions. If the model
for the propagation of the Eastern Rift Zone to the
south-west is correct (e.g. Oskarsson et al., 1982,
1985) then this situation is unlikely to change because
episodic injections of mafic magmas into Torfajökull
(from various sources), will continue (Larsen, 1984;
McGarvie, 1985). Indeed, if an intensification of rift
zone mafic magma production accompanies this pos-
tulated rift propagation, this might well increase the
frequency of tholeiitic magma injections into Torfa-
jökull (McGarvie et al., 1990), though whether this
will increase or decrease the volume of rhyolitic melt
available for eruption is unclear at present.
In seeking a more viable explanation, it is note-
worthy that better-constrained examples of peralka-
line magma production and evolution such as those
at Olkaria and Naivasha in Kenya point to shorter
timescales of 103–104 rather than 105 years (Davies
and Macdonald, 1987; Heumann and Davies, 2002),
with even shorter timescales implicated at other sys-
tems (see Hawkesworth et al., 2004). However exam-
ples of such rapid production-differentiation need to
be reconciled with the broad general trend of older
JÖKULL No. 56 71