Jökull - 31.12.2001, Qupperneq 15

Geomorphology of Ódáðahraun
Several major sedimentary bodies were identified
in the study area: (1) the sandur in front of Dyngju-
jökull (Figure 3; J & K); (2) three parallel sand
tongues on Frambruni lava west of Dyngjufjöll ytri
(Figure 3; N); (3) the long stretch of sand running
from Suðurárbotnar towards the NE across Fremriná-
mar and almost reaching the road no. 1 (Figure 3; L);
and (4) the jökulhlaup deposits along Jökulsá á Fjöll-
um (Figure 3; P and R). The elongated form of the
aeolian sedimentary deposits suggests a transport pro-
cess towards the north-east. Identifying the source of
these sediments would be very useful, enabling man-
agers to try to bind the sediment at the source (i.e. stop
erosion) rather than stopping the transport process at
a later stage and thereby creating unstable sediment
sinks.
Although our land cover classification succeeded
in revealing some characteristics of the land degrada-
tion process, the ultimate trigger of this process re-
mains a mystery. In general, the spatial distribution of
the barren, severely eroded landscape and the volcanic
rift zone coincide well. One can therefore ask to what
extent the severe land degradation in Ódáðahraun re-
gion is the result of endogenous processes, namely
volcanic destruction caused by tectonic activity, lava
ílows, tephra outfalls, and particularly jökulhlaups.
Lava flow fields cover a third of the investigated Land-
sat TM scene area, and the most recent flows erupted
less than two decades ago. Is it possible that erupting
lava flows would trigger erosion by destroying veg-
etated land cover? The recently formed (1970s-80s)
Krafla lavas provide no evidence of such a triggering
mechanism. Instead, rough lava surfaces probably act
as sediment traps and sinks, slowing the pace of north-
eastward sediment transport. In the case of Frambruni
lava, for example, sand can move forward only by te-
diously filling in the hollows and depressions of the
lava field (e.g. in the form of the three sand stretches
across the lava flow). In addition, normal faults and
fissures are natural sediment traps, and are often filled
with aeolian sand. Smooth pahoehoe type lava sur-
faces, however, may act as important pathways for ae-
olian sand transport.
In addition to aeolian sedimentary deposits, there
are vast flood plain deposits along the Jökulsá á Fjöll-
um valley (cf. P and R in Figure 3), illustrating the
magnitude of past jökulhlaup events. Our results sug-
gest that flood events proceeded towards the north-
west, perhaps extending from the saddle between Sel-
landafjall and Bláfjall to the depression currently oc-
cupied by Mývatn. With the available data, however,
we can only speculate on these matters. Nonetheless,
jökulhlaups may have played a significant role in the
initiation of the degradation process, as flood waters
destroy vegetation by both slashing it and burying it
with sediment when flow competence decreases.
Several details of the degradation process remain
still unexplained e.g., sediment provenance, the role
of climatic change in vegetation decline and the po-
tential effect of anthropogenic factors such as defor-
estation and grazing by horses and sheep. A final
key question with regard to land degradation is the
undisturbed state of this region and the age of the ero-
sion process, as it has yet to be discovered whether
the whole area has ever been fully vegetated and cov-
ered with soil. More detailed dating methods and sed-
imentological studies combined with palaeoecologi-
cal reconstructions in the future may throw light on
this matter.
Evaluating the classification
The field checks revealed a common problem in re-
mote sensing: how to define an appropriate nomen-
clature for the classification? Image classification
schemes often cluster different land cover types, as
observed in the field, into a single class due to similar
reflectance. On the other hand, these schemes can also
generate several classes despite apparent homogeneity
of the site in the field. In real life, furthermore, there
is no predetermined nomenclature for describing geo-
morphological features; nature appears as a complex
continuum without sharp boundaries between mem-
bers. Therefore, we found it unfeasible to subject the
classification to a quantitative, statistical evaluation.
The current mapping result should rather be evalu-
ated in the appropriate context and in relation to its
prospective applications.
The image pixel values (digital numbers), and
consequently the results of our classification, are pri-
marily controlled by the colour of the lava and sed-
iment, their textural properties (roughness), and the
JÖKULLNo. 51 13