Jökull - 01.12.1969, Síða 101

is that now the regression line is diíferent,
generally indicating higher temperature for a
given ice incidence than in the period 1846—
1919. The reason is probably ntuch more
thorough observations of the ice, provided by
the Icelandic Meteorological Office which was
established in 1920. Tlie graph in Fig. 3 will
give at least 0.6° C higher temperature than
the graph in Fig. 2, if used to estimate the
temperature of the period 1591—1846. There
is however no reason to think that the ice
records before 1846 are ntore complete than in
the period 1846—1919. On the contrary, as
mentioned before, some increase of the appar-
ent incidence of ice in the period 1591—1846
is probably necessary in order to make that ob-
servation series comparable with the period
1846—1919. Having estimated the annual ice
incidence as explamed before the graph in
Fig. 2 is used to estimate the temperature
1591 — 1846. The result is given in Fig. 5,
showing by the dotted line the estimated
decadal running mean of the temperature. The
full line after ca. 1850 gives the observed temp-
erature.
CORRELATION OF TEMPERATURE
AND SEVERE YEARS
Having now estimated the temperature of
every decade from 1591 up to date, we can
compare it with the decadal number of severe
years in this period.
A scatter diagram indicated that there was
approximately a linear relation between the
decadal temperature and the expression
1/ VíL
where H is the decadal number of severe years,
as they have been defined here. In Fig. 4 we
see this correlation. The correlation coefficient
is 0.76, and it is significant to the 0.001 level.
The error of the estimated decadal tempera-
ture is 0.3 or less in some 84% of all cases. It
must however be emphasized that the un-
certainty of this estimate is greatest in mihl
periods, partly due to lack of data.
In Fig. 6 upper part we have used this rela-
tionship to estimate the temperature in the
centuries before 1591. The graph has though
been smoothed so as to give the running mean
Fig. 4. Correlation of temperature and
frequency of severe years (horizontal scale).
temperature of 30 years, one value plotted for
every decade, except for a period in the 15th
century when information is considered to be
too meager. This graph only gives the broad
lines of the history of temperature climate and
weeps out all the shorter fluctuations. For
comparison the temperature graph after 1600
AD is smoothed in the same way.
CORRELATION OF SEVERE YEARS
AND ICE INCIDENCE
In order to estimate the ice incidence in the
years before 1591 we use the correlation be-
tween severe years and ice incidence in tlie
period 1591 to 1940. The regression equation
turns out to be:
Ice incidence pr. decade
= — 1.4/ VlT-f 3.4
The correlation coefficient is 0.69, significant
to the 0.001 level. In Fig. 6 lower part we
have used this relationship to estimate the ice
incidence in the period before 1591. The
values have been smoothed so as to give the
30 year running mean, one value plotted for
every decade. As in the case of the tempera-
ture the values after 1591 have also been
smoothed in order to give 30 year running
rneans.
7
JÖKULL 19. ÁR 97