Content area
Full Text
brief communicationstemperature data set, which provides
monthly temperature anomalies (with
respect to 196190) for 5[c5243][c450]5[c5243] grid boxes
based on land and ship reports2. A spatially
weighted average of the HadCRUT grid
boxes that have data from 1958 to 2000
gives a warming of 0.176 [c5243]C per decade,
which is significant at the 95% level. Doran et al. calculated spatially
smoothed trends using a technique that
allowed grid boxes (there are only 16 of
these south of 65o S) to have a radius of
influence of roughly 25% of the maximum
width of the image, which we believe is too
large, considering the highly localized
nature of the factors that influence the
climate at many of the stations. We suggest
that the interpolation has given too much
weight to grid boxes in data-sparse regions,
resulting in a misleading representation of
cooling over the continent, which is not
supported by in situ or remote-sensing data.Doran and colleagues Fig. 2 shows
annual and seasonal temperature trends for
19662000, with the largest cooling occurring in autumn over an area from the
southern Weddell Sea to the South Pole.
There is also pronounced cooling in spring
and in annual trends. However, their Fig. 2
does not show the marked warming on the
western side of the Antarctic Peninsula,
which is greatest during winter3. It is
unclear why the authors chose 19662000
for their analysis, as most of the temperature series in the HadCRUT data set begin
in the late 1950s.Although the spacing of stations around
the coast is reasonable in the eastern hemisphere, there are large gaps in the data for
the coast of west Antarctica, and only two
stations in the interior have long records:
those at the South Pole and Vostok (78.5[c5243] S,
106.9[c5243] E). The area in which Doran et al.
report the greatest cooling is devoid of
stations with long records, and HadCRUT
includes no data for this region.The warming on the western side of the
Antarctic Peninsula is...