Winter 2001

We will present a fully IMPLICIT baroclinic primitive equation model representing the circulation on the time scales greater than the inertial time scales and on spatial scales exceeding the Rossby radius. We do beleave that this model is capable of the baroclinic variational data assimilation for a year-long data sets.

Since the model is intended for data assimilation studies, the forward model was developed simultaneously with the tangent linear and adjoint code. The non-linear forward model utilizes the tangent linear and adjoint code for an efficient solving of the implicit scheme equations on each time step (BiCG technique).

We will discuss the numerical scheme of the model and some preliminary results of data assimilation and simulation of the circulation in the Tsushima Strait and Western Bank regions. The model is controlled via the initial distributions of temperature and salinity, boundary conditions for temperature, salinity and sea surface height on the open boundaries, and sea-surface fluxes.

In the present study, the accuracy of selected two-stream approximations is quantified using a global GCM-generated dataset (nearly 50 000 atmospheric columns with clouds, aerosols, and absorbing gases), by comparing their results to delta-16-stream reference multiple-scattering calculations. Out of the two-stream methods tested, the Practical improved flux method (PIFM) provided the best results, the delta-Eddington approximation being almost as good. For PIFM, the rms errors in top-of-the-atmosphere and surface net fluxes and in total column solar absorption were only 1.57, 2.18 and 2.11 Wm-2, respectively. However, all two-stream approximations tested had a negative bias of about 1-2 Wm-2 in atmospheric absorption.

In addition to the overall error statistics, the following points will be discussed: (1) how the errors depend on solar elevation?; (2) how large are the errors in the treatment of individual factors such as water clouds, ice clouds, and aerosols?; (3) is it possible to improve the accuracy of two-stream calculations? The ansver to point (3) is in fact positive, in particular, as what comes to the results at low solar elevations.

The simulated winter NAO variability compares well with the observations. Cross-spectral analysis reveals that the dipole characteristics emerge on time scales longer than 15 days, wheres they are absent on synoptic time scales. On intraseasonal time scales the NAO shows basically a red spectrum. There is little statistical evidence for the presence of significant low-frequency intraseasonal oscillations of the NAO.

The strength of the winter-averaged NAO has a marked effect on the statistical properties of intraseasonal NAO variability. During high NAO winters low-frequency intraseasonal variability of the NAO is less (more) pronounced compared to low NAO winters. It is shown that this asymmetry leaves its imprint on the near-surface temperature response to a forcing by the NAO, particularly over Europe. Finally, the results are discussed in terms of NAO-related changes in medium-range, extended-range, and seasonal atmospheric predictability.