Over the past 5-10 years, the Environmental Modeling Center (EMC) of NCEP has embraced and benefited from the conviction that the advancement of global and regional prediction models is accelerated and improved by the joint development of virtually the same models for both weather and seasonal-to-interannual (SI) climate prediction. The basis for this conviction is scientific and technological. Scientifically, it is readily acknowledged that climate is the integrated manifestation of daily and weekly weather events. In turn, the character of those weather events is modulated by such longer-term factors as quasi-persistent planetary-scale anomalies of order weeks, the seasonal cycle and slowly changing seasonal anomalies of ocean and land states (temperature, sea ice, soil moisture, snowpack and vegetation). Finally, the latter seasonal anomalies are a cumulative response to weather events. Technologically, the rapid advance in computer power has resulted in global or regional models for SI climate prediction being executed today at the spatial resolutions of weather prediction models of 5-7 years ago.

This presentation presents pathways of model development in EMC that reflect commitment to the weather-climate connection. The foremost example is the implementation in NCEP's present Climate Forecast System (CFS) for SI climate prediction the same global coupled atmosphere/land model as used in a recent version of NCEP's medium-range Global Forecast System (GFS). Moreover, the same atmospheric vertical resolution of 64 levels (L64) is used presently in both the GFS and CFS. Aside from the shared atmosphere/land model (albeit executed at T256 horizontal resolution in the GFS and T62 resolution in the CFS), the GFS uses analyzed SST over ocean, while the CFS utilizes predicted SST via its fully coupled ocean model (MOM-3 from GFDL). For the global atmosphere/land model of its next-generation CFS, targeted for T126/L64 resolution, EMC is now testing the June 2005 version of its operational GFS global atmosphere/land model. Thus upgrades to EMC's global atmosphere/land model are now tested on three time scales: 1) the medium-range of 1-2 weeks, 2) the seasonal range of 1-12 months (with coupled ocean component), and 3) in free-running multi-year "CMIP" mode (with coupled ocean component) of one or more decades (up to 50 years or more in some tests).

A second example of the commitment to the weather-climate connection in EMC model development is the EMC goal of utilizing unified physical parameterizations in the GFS, CFS, and NCEP's mesoscale NWP system known as the North American Model (NAM). The GFS version now being tested in the next-generation CFS includes the same land surface model (known as the Noah Land Surface Model, or Noah LSM) as operationally executed in the GFS. Additionally, EMC implemented a virtually identical version of the Noah LSM in the May 2005 implementation in NCEP's operational mesoscale North American Model (NAM). This milestone of a unified treatment of the land surface component (via the Noah LSM) in NCEP's mesoscale, medium-range, and SI climate-range systems is the culmination of a nearly 10-year commitment to the following goal explicitly undertaken in the mid 1990's by EMC and the GCIP and GAPP sub-programs of NOAA's Climate Program: to develop a land surface component for SI climate prediction by developing, testing and advancing a land surface model first at the mesoscale scale (including the diurnal cycle), followed by testing, refining and implementing at the global medium-range scale, and culminating in now testing and (soon) implementing in the CFS at the SI-climate scale. The initial (mid-1990's) land surface development and testing at the mesoscale was at resolutions of order 50-km then typical of NWP mesoscale models -- not far removed from the roughly 100-km resolution now characteristic EMC testing of CFS upgrades.

Aside from the unification of the land surface component in NCEP mesoscale, medium-range and SI-range prediction models, EMC is pursuing the unification of other components of its physical parameterizations in its models at these three forecast ranges, including shortwave and longwave radiation, cloud microphysics, planetary boundary layer and convection. The unification of the convection scheme is expected to be the most difficult because of the high sensitivity of convection schemes to spatial scale and the different relative importance of tropical maritime convection and mid-latitude continental convection in the GFS/CFS versus the NAM.

While pursuing the aforementioned strategy of unified physics in NCEP models, EMC is also committed to model diversity, but in the setting of ensemble prediction. EMC has operationally implemented ensemble prediction at all three forecast ranges (short-range, medium-range, SI-range). While the individual ensemble members in NCEP ensemble systems thus far are mostly members with different atmospheric initial conditions, EMC is accelerating its efforts to add members representing different physical parameterizations (and different parameter values in its physical parameterizations), as well as entirely different models. Thus while the mainline "deterministic control member" of the ensemble systems at NCEP may share unified physics, other members of the ensemble will apply different physical parameterizations and different initial conditions for land states and ocean states.

As a third example of the commitment to the weather-climate connection in EMC model development is the extensive testing of its three mesoscale models as Regional Climate Models (RCMs) for SI climate prediction. Presently EMC executes three mesoscale models: WRF-NMM, Eta model, and the Regional Spectral Model (RSM). While the WRF-NMM model is NCEP's highest resolution deterministic model in its NAM system, all three mesoscale models are executed operationally in NCEP's Short-range Ensemble Forecast System (SREF). But more importantly for this presentation, we will illustrate recent extensive hindcasts tests by EMC of the Eta and RSM mesoscale models executed in fully predictive mode as RCMs for SI-range seasonal climate forecasts, for both summer and winter seasons. Here the Eta and RSM RCMs are driven by lateral boundary conditions and SST fields predicted by the parent global CFS. The seasonal predictive skill of the RCMs is then compared with that of the parent CFS, especially with respect to precipitation, near-surface air temperature, orographic signatures, and low-level jets.