A. Moberg et H. Alexandersson, HOMOGENIZATION OF SWEDISH TEMPERATURE DATA .2. HOMOGENIZED GRIDDED AIR-TEMPERATURE COMPARED WITH A SUBSET OF GLOBAL GRIDDED AIR-TEMPERATURESINCE 1861, International journal of climatology, 17(1), 1997, pp. 35-54
Homogeneity tests of long seasonal temperature series from Sweden, Den
mark, Finland, and Norway indicate that homogeneous series are rare an
d that an abrupt change of the relative mean level is a much more comm
on type of nonhomogeneity than a gradual change. Furthermore, negative
shifts were 20% more common than positive shifts. Homogenized tempera
ture anomaly series that were constructed for six 5 degrees latitude x
5 degrees longitude grid boxes indicate that the temporal pattern of
temperture changes has been similar in different parts of Sweden since
1861. The annual mean temperature over Sweden was found to have incre
ased by 0.68 degrees C from the period 1861-1890 to 1965-1994. The cor
responding changes for the seasons were: +0.18 degrees C (winter), +1.
40 (spring), +0.42 (summer) and +0.60 (autumn). A direct comparson sho
ws that non-homogeneities in the temperature series from individual gr
id boxes in a global data set can be as large as the total changes obs
erved. We estimate that a 95 per cent confidence interval for the erro
r, due to non-homogeneous long station records, in estimates of hemisp
heric temperature changes over land regions since the period 1861-1890
is +/-0.1 degrees C for the Northern Hemisphere and the globe and +/-
0.25 degrees C for the Southern Hemisphere. For a region consisting of
about five grid boxes, this error is +/-0.5 degrees C. The large non-
homogeneities in individual grid-box series in the global data set is
partly a consequence of the fact that homogeneous climate data are not
always easily available for the open research community. We urge that
efforts are made to improve this situation.