Geoscience Reference
In-Depth Information
observed data. Data assimilation can also fill in gaps in observations and provide bet-
ter estimates of parameters that are difficult to observed directly (e.g., ice thickness).
There are a number of assimilation methods, ranging from the trivial such as “direct
replacement” of modeled with observed data (either of ice velocity or concentration)
to more sophisticated approaches such as optimal interpolation or Kalman filtering.
These account for the error characteristics of both the model and data, as well as
the spatial distribution of the data, to minimize error in a statistical sense. Another
approach is a variational scheme, where a “cost function” that measures the mismatch
between model and data is minimized under the constraint that appropriate model
quantities (e.g., mass, momentum, energy) are conserved (see Ghil and Malanotte-
Rizzoli,
1991
). Data assimilation methods were first developed for atmospheric
models to improve weather forecasts (Charney, Halem, and Jastrow,
1969
). A good
example of the direct replacement method applied to sea ice velocities is the study of
Maslanik and H. Maybee (
1994
) Examples of more sophisticated schemes to assim-
ilate ice velocity include Meier, Maslanik, and Fowler (
2000
), Meier and Maslanik
(
2003
) and J. Zhang et al. (
2003
). Assimilation of observed ice concentrations has
been addressed by Thomas and Rothrock (
1989
,
1993
) and Thomas et al. (
1996
).
Perhaps the most notable effort as of this writing is the Pan-Arctic Ice Ocean
Modeling and Assimilation System (PIOMAS), developed by the University of
Washington Seattle Polar Science Center (Zhang and Rothrock,
2003
; Schweiger
et al.,
2011
). PIOMAS provides estimates of Arctic sea ice volume from 1979
onward with regular updates. It consists of a multicategory thickness and enthalpy
distribution sea ice model coupled with the Parallel Ocean Program developed at
the Los Alamos National Laboratory. Satellite-derived sea ice concentration data are
assimilated into the model to improve ice thickness estimates. Sea surface temper-
ature data are assimilated in ice-free areas. Atmospheric forcing for of the model,
including 10 m winds, near surface air temperature, precipitation, specific humidity,
evaporation, sea level pressure, downward longwave radiation, and cloud cover (to
compute downward solar radiation) are obtained from the NCEP/NCAR reanalysis.
PIOMAS covers the region north of 48
o
N latitudes and is driven at its boundaries
with input from a global ocean model (Schweiger et al.,
2011
). The system is under-
going constant refinement. Results from PIOMAS show a downward trend in sea ice
thickness from 1979 to present consistent with the observed downward trend in sea
ice extent.
Figure 9.11
shows the mean annual cycle in ice volume from PIOMAS
for the 1979-2011 period and for selected individual years. For the 1979-2011 aver-
age, volume ranges from 28,700 km
3
in April to 12,300 km
3
in September. The low
volume in recent years compared to the long-term average is clear. The estimated ice
volume of 3,400 km
3
for September 2012 is the lowest in the PIOMAS record.
9.5
Global Climate Models
GCMs represent one of our most valuable tools to both better understand changes
in climate that have occurred in the past, changes over the instrumental record,
and those that can be expected in the future. The analysis of global climate models
