Geoscience Reference
In-Depth Information
fluxes and variables (e.g., precipitation, evaporation, radiation and SAT). The fore-
casted atmospheric state will differ from the true atmospheric state because of the
inability to perfectly specify the initial atmospheric conditions, as well as short-
comings in the model physics, resolution, and parameterizations. In NWP, there
is an ongoing process of restoring the model back toward the true atmospheric
state through assimilation of observations. In this process, the forecasts, adjusted
by data assimilation (the blended fields generally termed “analyses”), represent
the new initial state to generate the next forecast (similar techniques are being
applied to sea ice models, see
Section 9.4
). Assimilation data for NWP models
are primarily tropospheric observations of temperatures, pressure height, winds,
and humidity obtained from radiosonde profiles and satellite retrievals. Surface
variables such as precipitation, evaporation, radiation fluxes, and SAT are typically
not assimilated and are simply model forecasts. Surface conditions, such as sea ice
cover, sea surface temperatures, snow cover, and vegetation cover are generally
prescribed from observations.
Analyses and forecasts from operational numerical weather predictions systems
are not ideal for climate studies (e.g., to study year-to-year variability in precipi-
tation or temperature). The reason is that operational systems are constantly being
refined so as to get better forecasts. As a result, if one were to look at a time series
of outputs from an operational NWP system, one would encounter pseudo-climate
signals (jumps) attributed to the frequent changes in these systems. Atmospheric
reanalyses are special forms of NWP that are intended to get around this problem
by using unchanging (static) versions of the atmospheric model and assimilation
system.
The first reanalysis attempted was the NCEP/NCAR effort (Kalnay et al.,
1996
;
Kistler et al.,
2001
). A number of the igures in this topic are based on NCEP/
NCAR data. The NCEP/NCAR reanalysis has involved recovery and assembly of
numerous atmospheric data sets, which are quality-controlled and then assimilated
with the static assimilation and forecast system. It is fair to say that the NCEP/
NCAR reanalysis revolutionized climate research. The NCEP/NCAR reanalysis
effort is still ongoing, and now provides more than fifty years (continually updated)
of global atmospheric analyses and surface fields.
Although a number of more modern reanalysis are now in existence, based
on better models and data assimilation approaches, they have all built on lessons
learned from the NCEP/NCAR effort. One important lesson is that even with a
static model and assimilation system, pseudo-climate signals will still be present
owing to changes in the amount and quality of assimilation data. Users of reanal-
ysis data must always be aware of this problem, especially for time series and
trend analyses. Prior to 1958, the frequency of radiosonde reports in the Arctic is
very low. Radiosonde coverage increased after 1958, and again in the early 1970s.
Satellite data (temperature and humidity information) began to be incorporated in
the 1970s. Starting in 1979, drifting buoy data from the IABP began to provide reg-
ular reports of surface pressure over the Arctic Ocean, helping to constrain the field
of atmospheric mass.
