Jim Overland
A comment sometimes made for sea ice and wave forecasting models is that if the forecast was good, it was the result of the model and if the forecast was poor, it was the result of the atmospheric forcing. As PIPs3 moves toward higher spatial resolution, the full plastic nature of the ice equations will be resolved for the first time, which places a more stringent requirement on the accuracy of the atmospheric numerical forecasts. The early experience of high-resolution atmospheric forecast models was that they tended to get the right details in the wrong places, rather than bland forecasts everywhere with a coarse resolution model. This may happen with PIPs3.
Comments were made in this section by Tim Hogan, Steve Burk, Axel Schweiger and Jim Overland. The two issues discussed were resolution, i.e. matching the scales of atmospheric process to ice process, and verification/validation.
Resolution
Results from LEADEX showed that the spectrum of deformation for sea ice has energy at
higher frequencies than the deformation of atmospheric forcing. Atmospheric scales tend to be
large (500 km) and fast (10 m/s) compared to ice motions (10 cm/s). Even when the scales of
deformation overlapped, such as in an atmospheric front (50 km), the front moves through so
rapidly that there is not enough time for the ice to make major adjustments. SIMI showed that the
ice responds on all scales to large post-frontal air masses of large fetch (500 km) and long duration
(3 days). Thus, based on air-ice coupling, the space scale of an ice model (10 km) does not need
to be matched by the space scale of the atmospheric model. An accurate forecast from an
atmospheric model on a 50 km scale should be adequate for forcing the ice model. Steve Burk pointed out, however, that to accurately model the development of storms on the 500 km scale, one needs to have a grid spacing of 30 km to adequately resolve the internal physics
in the atmospheric model.
Verification/Validation
Tim Hogan gave an overview of the NOGAPS (Navy Operational Global Atmospheric
Prediction System). The current 4.0 version is spectral (T159) with 0.75 degree resolution (~90
km) and has 24 layers. It has a sophisticated model physics and model/data assimilation cycle.
The plan for the year 2002-2003 is a resolution of 50-60 km and perhaps 30 km by 2005.
NOGAPS will likely provide the forcing for PIPs3. Another model, COAMPS, is a regional
nested model. While it would be difficult to cover the Arctic with COAMPS, it would be available
for sub-basin models.
While the issue for PIPs3 is to consider additional algorithms, the issue for current
generation atmospheric forcing models is that they generally contain the right physical processes,
but differ in the parameterization and implementation of these processes and how these processes
balance to provide the forecasts. For example, one model may have better cloud distributions and
another may have better cloud radiative properties, but they both give a similar cloud radiative
forcing. Boundary layer temperatures are a balance of large tendencies from radiative, advective
and turbulent processes; small changes in parameterization can give large changes in results. For
example, the NOGAPS appears to have a 0.6%C/day warm bias below 800 mb in the Arctic.
Models can systematically differ in the intensity and tracks of surface low pressure systems.
Thus, the first recommendation is that the NOGAPS model be evaluated in the Arctic against in-situ data and other operational models. This includes both the wind stress fields and the radiative forcing. It is important to understand the systematic bias of this particular model.
The second recommendation is that the PIPs3 model be tested with NOGAPS data. Thus, the systematic biases of the atmospheric forcing are included in evaluation of the PIPs3 model system. For example, the air/ice drag coefficients are probably model dependent based on the model's simulation of the actual Arctic Boundary Layer. The surface radiation budget will be a function of the model's cloud parameterization.