Jökull - 31.12.2001, Qupperneq 7

Geomorphology of Odáðahraun
leading to meltwater accumulation within the Gríms-
vötn caldera lake, causing regular glacial bursts onto
Skeiðarársandur in the south. Catastrophic floods
in the north, however, are distinctively less frequent
than towards Skeiðarársandur in the south (Tómasson,
1973; Björnssonand Einarsson, 1991.
Geomorphological evidence of past floods can
, be seen throughout the surroundings of Vatnajök-
ull, primarily along Jökulsá á Fjöllum and the Ás-
byrgi canyon. Elíasson (1977) reports three prehis-
toric flood events in the Jökulsá river canyon: 4600
BP, 3000 BP and 2000 BP. Written records kept by
local farmers in the area indicate that at least eight
jökulhlaups have taken place between 1477 and 1934
(Björnsson and Einarsson, 1991). The most recent
major jökulhlaup in the north occurred in 1729 (ísaks-
son, 1985). The gauging records of Jökulsá á Fjöllum
operated by the National Energy Authority indicate
16 minor jökulhlaup events in the period 1976-2002
(National Energy Authority 2001).
The latest eruption under Vatnajökull in Decem-
ber 1998 was not associated with flooding. The erup-
tion took place inside the Grímsvötn caldera and did
not melt enough ice to produce a jökulhlaup. An ear-
lier fissure eruption in November 1996 in Gjálp, north
of Grímsvötn, produced large amounts of meltwater
and resulted in a jökulhlaup towards the south. Guð-
mundsson et al. (1997) pointed out that an eruption
slightly further north would have resulted in a jök-
ulhlaup on the northern margin of the Vatnajökull.
Traces of meltwater from two small cauldrons at the
southeastern caldera rim of Bárðarbunga, were also
detected in Jökulsá á Fjöllum (Kristmannsdóttir et al.,
1999). It is widely acknowledged that the area is en-
tering a period of renewed volcanic activity (Larsen et
al., 1998), and this potential threat calls for better un-
derstanding of geoecological conditions in the north.
Aeolian processes
Our study region is almost entirely devoid of vege-
tation or well-developed soil cover, though eroding
steep-edged remnants of loess-type soil, rofabörð, can
be found in sheltered spots. Arnalds (1992b) has mon-
itored rofabörð erosion in the field, and a nationwide
project by the Agricultural Research Institute (RALA)
has classified erosion into five categories based on the
severity of the process (Arnalds et al., 1997).
Aeolian sand is abundant throughout the study
area, filling cracks and depressions on the lava fields,
and blanketing the sandur between Askja and Dyngju-
jökull. This sand only rarely forms distinctive dune
formations, probably due to the highly irregular and
rough terrain, and to inadequate sediment input rela-
tive to the weak trapping efficiency of the scant veg-
etation. Thick sand beds do occasionally form in
places favourable for sediment deposition, however,
and these may evolve into proper dunes, often cov-
ered with sparse lyme grass (Elymus arenarius) and
fescue species (Festuca sp.). At the northern mar-
gin of the current study region, advancing fronts of
thick wind-blown sediment bury vegetation, threaten-
ing the surrounding pasturelands and human commu-
nities. The extensive dune fields near the Nýjahraun
lava flow are partly vegetated, and the southern stoss
faces of dunes are often grass covered and gently slop-
ing, while the northem slip faces are typically steep,
barren, and wind scoured (Figure 2).
MATERIALS AND METHODS
A floating Landsat TM quarter scene, acquired on
1992-07-14 and covering an area of ca. 135x70km
was used for constructing a land cover map for the
study area. Of the total of seven bands available
from the TM sensor, we used the following four bands
in the image analysis: 2(green, 520-600nm), 4(near
infra-red, 760-900 nm), 5 (middle infra-red, 1550-
1750nm), and 7 (middle infra-red, 2080-2350 nm).
The inspection of individual TM bands revealed sub-
stantial striping in the visible wavelength bands 1, 2
and 3. As they are also highly correlated, only one
visible band (TM 2; green) was included in the clust-
ering. Infra-red bands were included in the analy-
sis due to their suitability for vegetation discrimina-
tion, soil moisture determination and mineral and rock
type identification. At the time, the digital elevation
model was still under preparation and hence, no to-
pographic normalisation was performed for the data.
As land cover characteristics in this remote area are
poorly understood, we were unable to conduct a su-
pervised classification of the image. Instead, we per-
JÖKULL No. 51 5