SUMMARY Binned and unbinned drifter velocities in key flows were comparable with historical estimates. A qualitative overview of the drifter and numerical trajectories showed the higher resolution model to be much more realistic in terms of its coverage of the basin and its ability to simulate the pathways revealed by the drifters. The coverage of the domain by the 0.28^o numerical trajectories tracked coherent flows and displayed notable regions where there were few or no trajectories. Comparisons of the mean binned drifter velocity fields and the MKE fields with the model fields showed a number of unrealistic features in the 0.28^o field: the large anticyclone at 37^oN near Cape Hatteras, the displacement of the Northwest Corner too far eastward across the basin, and a zonal flow extending across the basin around 25^o to 28^oN. These features are not present in the 0.1^o model and the flow is seen to be more realistic than in the coarser resolution model. A statistical James test was used to determine if the model mean velocities were significantly different to the drifter means. Locations where the means differed from the observations in each model were mainly in the regions qualitatively identified to be erroneous in the models. The variance in the 0.1^o model was much more realistic than that of the 0.28^o model over the entire domain. The representation of the variability was especially improved in the GS and the AC in the 0.1^o model over that found in the coarser model. Observed time and length scales have characteristic values of 2-4 days and 20-50 km, respectively. Time scales were too long in the 0.28^o run, while in the 0.1^o simulation they were not statistically different to the observed values in the zonal direction. A couple of outlying high ratios in the 0.1^o meridional direction biased the statistics, which otherwise are close to one over most of the domain. The related length scales are too short in the 0.28^o model, while the distributions of the drifter and 0.1^o scales are very similar, with the exception of higher occurrences of meridional scales in the 30-40 km range. These results show that the 0.1^o, 40-level POP much more faithfully reproduces the observed circulation. It can be concluded that if POP is to be used as the ocean component of a global coupled prediction system it will need to have a horizontal resolution of at least 0.1^o and 40-levels. A simulation at this resolution is currently underway on the IBM SP3 at the Navy Oceanographic Center. It should be noted that the wind forcing and domain of the two models differ. Also one model does not have a mixed layer. Comparisons of statistical quantities calculated from this 0.1^o simulation and the earlier 0.1^o Smith et al. (2000) run indicate that the effects arising from the use of the different wind products are unlikely to be significant. Finally, we have not tried to determine the individual effects of increased horizontal and vertical resolution. To separate out the responses to all these factors, a suite of high-resolution model runs would be needed, requiring a large allocation of computational resources. Ongoing improvements in super-computing technologies are likely to make such suites of runs common place in the near future.		Abstract Introduction Model Descriptions Surface Drifters Numerical Trajectories Eulerian Statistics Mean Velocity Vectors MKE Ellipses EKE James test Lagrangian Statistics Diffusivities Time Scales Length Scales Summary Appendix A Appendix B Acronyms References Acknowledgements Forum