Geoscience Reference
In-Depth Information
sector of the Arctic Ocean is quite cloudy for all months. Nevertheless, even mean
monthly cloudiness is rather poorly quantified, perhaps only within 5-10 percent.
As reviewed by Curry et al. (
1996
), there are several reasons for this unsatisfying
state of affairs. Although cloud fraction (the part of the celestial dome covered by
cloud) is a standard variable of synoptic weather reports, spatial coverage of obser-
vations is spotty and the measurement itself is somewhat subjective. The most
comprehensive surface-based data set for the central Arctic Ocean is still from the
Russian NP program. In addition, during the long polar night, cloud observations
are difficult, especially under moonless skies. An analysis by C. Hahn, S.Warren,
and J. London (
1995
) of Arctic Ocean cloud amounts observed in winter under
moonlit and moonless conditions suggests an underestimate of total cloudiness
by about 5 percent. There are a number of satellite-derived gridded data sets. For
example, gridded satellite-derived estimates of cloud parameters (cloud amount,
cloud top height, optical thickness, and so on) with global coverage are provided
through the International Satellite Cloud Climatology Project (ISCCP) (Rossow
and Schiffer,
1991
). Gridded polar-specific cloud products from 1982 onward are
provided as part of the AVHRR Polar Pathfinder effort (Maslanik et al.,
1997
).
Cloud detection in both data sets is based on a combination of visible and infrared
wavelength imagery. For example, the ISCCP-C1 (daily) and C2 (monthly) data
sets for the polar regions are based on AVHRR data from channels 1 (0.58-0.68
µm, 1 µm = 10
−6
m) and channels 4 (10.3 to 11.3 µm). The Arctic Polar Pathfinder
(APP-x) effort includes cloud properties along with surface radiation fluxes and
other variables on a daily basis (Key and Intrieri,
2000
). Cloud fraction is also
available from the MODIS sensor.
There are fundamental problems with satellite retrievals in high latitudes. In the
visible wavelengths, clouds and the snow/ice surfaces have essentially the same
reflectance, making cloud discrimination very difficult. Infrared measurements are
limited by the fact that temperature differences between clouds and the surface are
usually small. Because of the low-level temperature inversion, cloud tops may be
warmer than the surface.
A. Schweiger and J. Key (
1992
) show a wide discrepancy between the annual
cycle of total cloud amount in the Arctic reported by S. Warren and his colleagues
(
1988
) from surface stations and from ISCCP-C2 data, although except for the cen-
tral Arctic Ocean the basic spatial patterns are in general agreement. L. Wilson, J.
Curry, and S. Ackerman (
1993
) suggested that some of the discrepancies in win-
ter might be attributable to the frequent occurrence of ice crystals (diamond dust)
in the atmosphere, which might be identified as low-level cirrus by the ISCCP
algorithms. The ISCCP-C products have been superseded by the ISSCP-D series,
which includes improvements to mitigate the problems found over polar surfaces.
However, it has also been shown (Schweiger et al.,
1999
) show that the ISSCP-D1
product still has significant problems and overestimates cloudiness in winter and
summer compared with the NP observations. Moreover, the annual cycle of cloud-
iness is reversed from that observed at surface stations. In a more recent study, R.
Eastman and S. Warren (
2010b
) find that interannual variability in cloud amounts
