In 1990, a 10-year run was made (van den Dool, Saha and Toth 1991) with NMC's global spectral model (MRF) which had been used until then, almost exclusively, for NWP purposes. This climate experiment was run at T40L18, in conjunction with daily forecasts out to 90 days for the summer and fall of 1990 (van den Dool 1994). There have been many multi-year climate runs made after 1990, always at a resolution well below current NWP models. Often, these long runs are made for a very limited purpose, for instance, to study the impact of a change in a parameter in the boundary layer formulation or convection scheme. Many climate runs, have been made at NMC by the Coupled Model Project (Ji et al. 1994).
NMC's AMIP entry (Ebisuzaki and van den Dool 1993) was just one of these many runs. An important result of the NMC AMIP run was that the MRF model of vintage 1992, responded to anomalies in SST in the manner observed: i.e. the simulation of the Southern Oscillation Index and global angular momentum seemed quite realistic (after removing biases). The Model's mid-latitude and other `remote' connections, such as Indian rainfall, however, appear almost uncorrelated with reality.
Models have changed, and research efforts have, to a large extent, been organized in conjunction with the ReAnalysis Project at NMC (Kalnay and Jenne 1991). The ReAnalysis model is T62L28 in resolution and will be frozen for someyears to come. In this presentation, we will discuss an AMIP-type climate run with exactly the same T62L28 model and make comparisons to ReAnalysis data for the same years. `What is the difference in the modelled Earth's climate with and without atmospheric data-assimilation?' As of mid February 1995, 7 years (1985-91) have been completed. {It could well be, that more than 10 years will have been completed by May 1995.} We will present some of the classical monthly multi-year mean fields (e.g. standing waves in 500 mb height in January), where ReAnalysis can safely be assumed to be close to the truth and the difference of AMIP from ReAnalysis is thus the model error. When presenting some of the harder to observe fields such as the hydrological and variable mass components (precipitation, evaporation, precipitable water, global surface pressure, etc. see van den Dool et al. 1993), the difference between AMIP and ReAnalysis may point to the need to make improvements in both.