The coastal Gulf of Alaska (GOA) is subject to weather that is strongly impacted by its prominent coastal terrain. These effects must be included in specification of the atmospheric forcing of the ocean over the shelf of the GOA. It is possible to account for these effects, given only large-scale representations of the atmosphere such as available from the NCEP/NCAR Reanalysis, through a method known as downscaling. The downscaling used here is dynamical, and based on high-resolution numerical weather prediction (NWP) model simulations. Specifically, the Penn State/NCAR MM5 NWP model is run on a 15-km horizontal grid spacing using the NCEP/NCAR Reanalysis data set, which is on a 2.5 degree horizontal grid, as initial and boundary conditions. Mesoscale meteorological phenomena that are known to be common in the coastal GOA, e.g., barrier jets and gap flows, appear to be much better represented in the 15-km MM5 simulations than the original Reanalysis. These differences can be significant; a set of MM5 simulations for December 2000 and January 2001 showed average along-coast wind speeds in the northern GOA that averaged about 40% greater than their counterparts from the Reanalysis. Present work is using the MM5 driven by the Reanalysis to document how climate variability on large temporal and spatial scales is expressed in the coastal zone of the GOA.