10720
F. Prata et al.: Separation of ash and SO
2
be misleading because uncertainties in assumed parameters
may cause errors to cancel and lead to better results than oth-
erwise expected. Nevertheless, with the limited independent
observations available, accuracies in mass loadings appear to
be in the range of 20–50 %. Clarisse and Prata (2016) discuss
errors (precision and accuracy) in ash retrievals and suggest
areas where more research is needed. In this study, we focus
only on the accuracy in the retrievals for the Grímsvötn erup-
tion. Uncertainties that are identified due to cloudiness, lack
of thermal contrast (either ash that is either optically too thick
or optically too thin), radiometric errors, and estimates of
cloud-top and surface temperature are included in the error
budget. In the case of the ash retrievals for Grímsvötn, the
error estimates are within the expected range, giving an er-
ror of ±0.1 Tg or roughly 20–50 % of the estimated mass
of very fine ash. It is emphasized that this is not the total
mass emitted by the volcano, which is typically a few per-
cent of the total mass. It is however, the mass fraction that is
dispersed by the winds and the very fine ash that can cause
damage to aircraft jet engines. Individual mass loading er-
rors can be lower than 20 % and also much higher, depend-
ing mostly on contamination of the pixel by meteorological
cloud, but generally these are not validated because there
are no independent measurements of mass loading. IASI re-
trievals have a precision also in the range of 20–50 % but
their accuracy is unknown as no independent validation has
been done. IASI retrievals were biased high compared to the
SEVIRI and MODIS retrievals in this case and the cause is
not yet understood.
Retrieval methods are being continually improved and
there is an international effort (http://cimss.ssec.wisc.edu/
meetings/vol_ash15/) to intercompare retrieval schemes and
help reduce uncertainty. At the current time no firm con-
clusions have been made about retrieval accuracy as no ro-
bust validation has been made. Uncertainties can only be as-
sessed against independent observations and so far indepen-
dent measurements of mass loading as well as independent
measurements of atmospheric ash particle size distributions,
shapes, and composition are extremely sparse.
Tesche et al. (2012) and Ansmann et al. (2012) re-
port lidar measurements of ash mass concentrations in the
range of 100–340 µg m
−
3
. Moxnes et al. (2014) report val-
ues < 100 µg m
−
3
based on aircraft data and modelling.
These data, our data, and previous measurements from Ey-
jafjalljökull (lidar, and airborne and ground-based air qual-
ity) all provide adequate support for the assumptions used in
satellite-based infrared retrievals. The error estimates for the
Grímsvötn eruption used here are robust but should not be
extended to all ash retrievals or for any other eruption.
We estimate that the amount of ash transported towards
Europe between 22 and 25 May 2011 was 0.2–0.4 ± 0.1 Tg
(very fine ash). By comparison, Stohl et al. (2011) esti-
mated 8.3 ± 4.2 Tg of very fine ash to be emitted during
the Eyjafjallajökull eruption in April–May 2010, which is
an order of magnitude greater from an eruption that was
a factor of ∼ 2 smaller in total erupted mass than the
Grímsvötn eruption. Moxnes et al. (2014) estimate that a
total of 0.49 ± 0.1 Tg of very fine ash was emitted from
Grímsvötn, based on modelling results that utilized IASI re-
trievals not used in our study.
5
Possible column collapse and PDCs
The lower-level ash plume was beginning to form from 19:15
to 19:20 UTC and was fully developed by 20:00 UTC on
21 May. Ground-based observations (see Supplement pho-
tographs) show that the vent from Grímsvötn to Blágil in
the Laki area is about 60 km and the lower-level ash plume
reached there in about 1 h. The MODIS satellite data show
that the low-level ash layer (< 6 km high) was present off the
south coast of Iceland on the morning of 22 May and was also
clearly observed 24 h later (see Fig. 9). This layer appears to
be detached from the main eruption column. Photographic
evidence (see Fig. 1, panel b) shows a shallow ash cloud or
plume
4
at low level surrounding the main column (a “skirt”),
and another plume-like ash-rich layer higher up and at about
half the height of the column. These observations suggest the
possibility that the column may have undergone partial col-
lapse sometime during the evening of 21 May, causing an
outflow of ash, not dissimilar to the outflow often observed
from a collapsing thunderstorm. As large ash aggregates fall
through the column, enhanced by the presence of copious
amounts of water, for example see Telling et al. (2013b) for
a discussion of this process, ice would have formed on the
ash, increasing the size and fall speed and effectively remov-
ing particles from the column. These ice-coated ash aggre-
gates, sometimes termed volcanic hail would have fallen out
of the cloud very rapidly. The process of ash falling through
the column would have caused compression of the lower part
of the column and a mechanism for driving a gravity cur-
rent of ash outwards from the column. Such PDCs may have
supported plumes with ash rising from the regions immedi-
ately outside the vent area. The light southwest winds in the
lower troposphere favour propagation of the outflow towards
the west, as observed, but it is likely that the ash formed a
skirt surrounding the collapsing column. Column collapses
can also cause pyroclastic density currents, so that the two
mechanisms may not be seen as separate. A schematic of the
proposed processes is shown in Fig. 10.
The speculation that a partial collapse and/or generation of
an ash skirt (PDC without collapse) moving outwards from
the plume is supported by the photographs shown in Fig. 1
and the MODIS satellite image shown in Fig. 5. Jude-Eton
et al. (2012) showed that PDCs occurred during the 2004
Grímsvötn eruption (see the photographs in their Fig. 2a and
b). In these instances a column collapse is not required; PDC
4
We define an ash cloud as an identifiable structure wholly dis-
connected from the vent, whereas an ash plume has an identifiable
connection to the source vent.
Atmos. Chem. Phys., 17, 10709–10732, 2017
www.atmos-chem-phys.net/17/10709/2017/