Jökull - 01.01.2013, Side 99

Mass balance of Mýrdalsjökull ice cap
The underestimated elevation translates to an av-
erage error of approx. 0.65◦C in surface temperature
at the ice cap’s peak. In an attempt to account for this
error in surface elevation and account for the temper-
ature threshold between solid and liquid precipitation
lying slightly above 0◦C on the plateau (Ólafsson and
Haraldsdóttir, 2003), only simulated winter precipita-
tion at temperatures below 2◦C is considered. The
choice of this critical value is the author’s best esti-
mate in light of the available data and previous knowl-
edge on the subject. If the temperature bias resulting
from the underestimated elevation is taken into ac-
count, the value used here is close to that used in some
mass balance models (1◦C, Jóhannesson et al., 1995).
It will furthermore exclude the largest rain events in
the early autumn when rain may be expected to seep
through the snowpack and hence be lost from the win-
ter layer.
Figure 5 shows the mean simulated precipitation
for the same five winters (Sept.–April), as covered by
the mass balance measurements, from Sept. 2006 un-
til April 2011. The greatest simulated precipitation
during winter is on average above 7 m at the southeast
edge of the plateau. The mean simulated precipita-
tion compares well with the average measured winter
mass balance (accumulation) at the sites during 2006–
2011, but errors are considerably greater for individ-
ual years (not shown). The observed precipitation is
furthermore successfully reproduced at the two low-
land stations, Vík in Mýrdalur and Vatnsskarðshólar,
and the dataset has previously been verified at sev-
eral locations in the region immediately east of the ice
cap (Nygaard et al., 2013). Large parts of the sys-
tematic underestimate (4.5%) at M1 (located in a de-
pression in the ice cap) and the overestimate (10%)
at M3 (located at the top of the ice cap) may be ex-
plained by snow drifting, which will respectively en-
hance the winter balance at M1 through deposition,
and remove a part of the snow falling at M3 through
scouring. Errors may also be associated with the at-
mospheric model, in particular the parameterization
of atmospheric moisture as well as too coarse model
resolution for resolving the orography of the ice cap.
On an annual basis, errors related to individual precip-
itation events can be large as found by comparison of
the observed and simulated precipitation at Vík and
Vatnsskarðshólar (not shown). These errors tend to
cancel out when summed over several years, but they
are often a result of poor atmospheric analysis forcing
the atmospheric model. However, the temporal res-
olution of the mass balance measurements does not
allow for an analysis of individual events at the ice
cap plateau.
The simulated mean winter snowfall (directly
based on the simulation of solid precipitation and not
a critical temperature value) underestimates the win-
ter mass balance, while the mean total winter precipi-
tation (snow as well as rain) overestimates the balance
by a smaller amount, approx. 13%, 17% and 30% at
sites M1, M2 and M3 respectively (not shown). When
compared to the measured winter balance and the high
density of the cores, depicted in Figure 2, the differ-
ent errors in simulated snowfall and total precipitation
(Figure 5) may to a large extent be explained by the
high frequency of rainfall on the ice cap. That is: 1)
Simulated snowfall on the glacier will not include rain
that refreezes within the snow layer, thus it underes-
timates the winter balance. 2) The total precipitation
will include rainfall that refreezes within the winter
layer and contributes to its high density, as well as
early autumn rainfall that seeps through the layer and
is lost, thus overestimating the winter mass balance.
There are indications of a slight trend towards
smaller amounts of precipitation and snowfall at the
survey sites during 2000–2011. The possible decrease
in snowfall is the result of relatively more frequent
rainfall on the glacier in a warming climate. This is
supported by the trend being most prominent at sites
M1 and M2, but least evident at M3 which is at the
highest elevation on the ice cap. However, these re-
sults are not statistically significant, and neither is the
small increase in observed and simulated precipitation
at both Vatnsskarðshólar and Vík. Further analysis is
therefore not performed, but these results imply a pos-
sible change in the relation between precipitation at
the lowland and mountain locations, due to changes
in the structure of the atmospheric flow impinging on
the mountains.
JÖKULL No. 63, 2013 99