NUMERICAL INVESTIGATION WITH A PHYSICALLY-BASED REGIONAL INTERPOLATORFOR OFF-LINE DOWNSCALING OF GCMS - FIZR

A novel approach for regional climate modeling based on an off-line do wnscaling of GCM simulations is described and illustrated with a one-m onth simulation example. The model is physically based and it requires outputs from a previous GCM integration. The methodology is based upo n the premise that much of ''small-scale'' variability (i.e., for spat ial scales below current GCM resolution) is often the result of surfac e forcings rather than small-scale dynamical effects. Following on thi s consideration, the present work seeks to address the question of reg ional climate diagnostics by combining precomputed GCM atmospheric lar ge-scale transports of momentum, heat, and moisture, called ''the dyna mics,'' with recomputed GCM subgrid-scale parameterized effect, called ''the physics,'' including an additional mesoscale forcing term that is parameterized in terms of large-scale flow resolved by GCM coupled with fine-scale geophysical surface fields. This combination is integr ated in a prognostic mode on a high-resolution grid over a chosen limi ted area of the earth. This is an original one-way nesting technique a nd it offers major advantages over simpler techniques used to interpol ate GCM outputs down to finer scales. While the dynamics, inferred fro m a GCM and solely projected on the high-resolution grid, does not int erfere with smaller scales, the proposed model does simulate the nonli near vertical interactions. The model is nicknamed FIZR: FIZ serves to remind that the model is physically based, and R stands for regional. To validate the FIZR approach the authors have performed a test on a 0.50 resolution grid over the west coast of North America to downscale January conditions simulated by the Canadian Climate Centre second-ge neration general circulation model (GCMII). Due to the coarse spatial resolution of GCMII, the West Coast January simulated precipitation pa ttern suffers from a lack of mesoscale details. To circumvent this res olution problem, we have conducted an experiment with FIZR in which la rge-scale dynamics is inferred from GCMII outputs and interpolated on the grid, the entire GCMII physics package is recalculated on a high-r esolution grid using geophysical fields of vegetation types, soil char acteristics, background albedos, and SSTs. A mesoscale forcing term is further added to parameterize orographic upslope and downslope in ter ms of large-scale flow resolved by GCMII coupled with high-resolution topography. Preliminary results indicate that FIZR can add a number of realistic mesoscale features to GCMII simulation, especially in preci pitation and hydrological surface fields. The mesoscale hydrologic fea tures are consistent with the thermal features simulated by the downsc aling FIZR technique. Although the scope of this study is restricted t o a single January scenario, it does provide evidence that FIZR is a p romising technique for downscaling of GCM.