Perhaps the majority of regional climate modelers use one or another form of large scale (spectral) nudging, giving a variety of reasons for doing so. Foremost among those is the belief that the regional climate model (RCM) should maintain the driver model large scales, and that this is not going to happen to a satisfactory degree unless large scale nudging is used.

Yet, the experience of the Eta model at NCEP includes several indications that a nested model can improve on the large scales of its driver global model, provided a sufficiently large nested model domain is used. To this one can add a 9+9 member ensemble experiment of Fennessy and Altshuler of COLA, in which the Eta nested in a GCM dramatically improved on the simulation of the three-months 1993-1988 mean precipitation reflecting extraordinary flooding over the U.S. upper Mississippi basin. Finally, it is worth recalling the result of the NCEP North American Regional Reanalysis (NARR), in which the largest documented improvement of NARR over the driver global NCEP/NCAR Reanalysis was in January wind fits to rawinsondes at jet stream levels, a result clearly against the mainstream views of what a nested model's added value can be.

For a direct test of the issue ensemble experiments have been performed, presented in Veljovic et al. (Met. Zeitschrift 2010), in which the Eta was driven by 26 members of the ECMWF 32-day ensemble. Verified against ECMWF analyses, in two ways, one evaluating skill in placing 250 hPa jet stream winds, and the other inspecting the standard 250 hPa rms difference, they demonstrated that it is possible to improve on large scales of the driver ECMWF model more often than not so if large scale nudging is not used.

Additional experiments have been performed since, three of those so as to complete the 32-day experiments done with both the relaxation and the Eta lateral boundary condition (LBC) scheme to a total of six ensemble members. Note that the Eta scheme is using the global model data only along the single outermost line of grid points, and at the outflow boundary points is specifying one less variable than at the inflow points. Verification done using the two verification methods now strongly suggests an advantage of the Eta scheme.

More experiments are being evaluated as this is being written, including two 7-month forecasts driven by ECMWF seasonal forecasts, initialized 1 January 2010, and using both LBC schemes. Results will be discussed, with presumably more 7-month members added by the time of the conference.