Jökull - 01.12.1969, Qupperneq 54

Statistical studies of drift ice in particular
localities, correlated with the general annual
ice incidence might also be used to make the
ice forecasts more detailed and useful for the
public. Thus it should even be possible to pre-
dict the ice incidence in certain months for a
certain district. It is however only in the last
years that the ice records have become suf-
ficently complete to allow the collection of
the information necessary for such forecasts.
Fig. 5. Estimated (left) and actual (right) ice
incidence in Iceland. (Actual ice incidence in
1968/1969 was 4.7 months).
50 JÖKULL 19. ÁR
MONTHLY ICE FORECASTS
Even if a reasonable ice forecast for everv
month of the ice season may possibly be issu-
ed already at the end of November, it is de-
sirable to amend this forecast as soon as any
new information is gained.
Apart from aerial ice observations, the winds
to the north of Icelancl are probably among
the most important additional information
that is to be expected during the ice season.
It is an old experience in Iceland that the
west wind has some prognostic value for the
drift ice. A clergyman living in the island
Grimsey, Rev. M. Jónsson, writes in 1839:
“The clearest knowledge of the approach of
tlie sea ice is that in westerly storms, which
are generally more frequent and severe in early
winter than in other seasons, it drifts towards
the east and later arrives with onshore winds.’’
The method proposed here to take this effect
into account is the following:
We assume that. the monthly ice incidence
may be expressed by the equation:
i = l.l/(10aJ+t>E+c+1) (Eq. 2)
This is an almost analogous equation to the ,
expression of the annual ice incidence. The
notations are:
i = monthly ice incidence
a and b are constants
J = JM-temperature
E = pressure difference in the last month
between Cape Tobin and Galtarviti, in
mb.
c = coefficient depending on the season.
In this equation the minimum of i is 0, but
maximum can be 1.1. When the computed i
becomes more than one, we liowever put it
equal to 1.
Due to lack of data it is difficult to cleter-
mine the constants a and b and the variable
coefficient c. This lias therefore been done step-
wise. First of all we determine the approxi- v
mate proportion between a and b from the
linear regression formula:
i = aij + bjE + ci (Eq. 3)
where ai, bi and ci are constants. Using data
of J and E for the months February to June,