Systematic Errors for the International SGMIP (Stretched-Grid Model Intercomparison Project) Multi-Model Ensemble SimulationsPrimary Author: Fox-Rabinovitz, Michael
Additional Authors: Jean Cote Bernard Dugas, Michel Deque and John L. McGregor
Systematic Errors for the International SGMIP (Stretched-Grid Model Intercomparison Project) Multi-Model Ensemble Simulations
Michael Fox-Rabinovitz, University of Maryland at College Park, U.S.A.,
Jean Cote and Bernard Dugas, Environment Canada, Canada,
Michel Deque, Meteo-France, France, and
John L. McGregor, CSIRO, Australia
Variable-resolution GCMs using a global stretched grid (SG) with enhanced resolution over the region(s) of interest have proven to be an established approach to regional climate modeling providing an efficient regional down-scaling to mesoscales. This approach has been used since the early-mid 90s by the French, U.S., Canadian, Australian and other climate modeling groups along with, or as an alternative to, the current widely-used nested-grid approach. Stretched-grid GCMs are used for continuous/autonomous climate simulations as usual GCMs, with the only difference that variable-resolution grids are used instead of more traditional uniform grids. The important advantages of variable-resolution SG-GCMs are that they do not require any lateral boundary conditions/forcing and are free of the associated undesirable computational problems. As a result, SG-GCMs provide self-consistent interactions between global and regional scales of motion and their associated phenomena, while a high quality of global circulation is preserved, as in uniform grid GCMs. Climate simulation results obtained with the SG-GCMs have shown the maturity of the SG approach.
The international SGMIP-1 (Stretched-Grid Model Intercomparison Project, phase-1), using variable-resolution SG-GCMs developed at major centers/groups in Australia, Canada, France, and the U.S., has been conducted in 2002-2005. The next phase-2 of the project, SGMIP-2, has been conducted in 2005-2007. The results of the 12-year (1987-1998) climate simulations of SGMIP-1 and the 25-year (1979-2003) climate simulations of SGMIP-2 for a major part of North America as well as for the entire global domain, are available at the SGMIP web site: http://essic.umd.edu/~foxrab/sgmip.html, and are described by the authors in Fox-Rabinovitz et al. (2006). The SGMIP-2 high resolution multi-model ensemble simulations provided the possibility for a comprehensive analysis of systematic errors for regional and global products obtained with enhanced variable and uniform resolution.
The multi-model SGMIP regional climate simulations were conducted with enhanced 0.45 degree - 0.5 degree regional resolution for SG-GCMs, with the same or a similar number of global grid points as in a 1 degree x 1 degree global grid. The multi-model ensemble SGMIP global simulations are conducted with: (a) intermediate 1 degree x 1 degree resolution (with the same or a similar number of global grid points as in SGs), and (b) fine 0.5 degree x 0.5 degree resolution. The SGMIP SG-GCM simulations are analyzed in terms of studying the impact of high regional resolution on efficient downscaling to realistic mesoscales and regional climate variability. We focused mostly on studying the systematic errors for the multi-model ensemble results. The differences between the models have been also determined. The SGMIP multi-model ensemble results for the region compare well with reanalysis and observations, in terms of spatial and temporal diagnostics.
The major SGMIP results in terms of systematic errors are as follows:
1. Regional biases for time-averaged model products are mostly limited to about half (or less) of typical reanalysis or observational errors. Biases are larger, up to twice the reanalysis or observational errors (but only for the southern polar domain); note that our SGs have the North American area of interest. Overall, biases are within the uncertainties of the available reanalyses.
2. Both seasonal and interannual climate variability are well represented. Namely, annual cycles, seasonal differences, time series, and variances are close to those of observations or reanalyses.
3. Orographically induced precipitation and other products are well simulated at meso- and larger scales due to high-resolution regional forcing. The major positive regional impact from stretching is obtained from better resolved model dynamics and regionally-enhanced resolution of stationary lower-boundary forcing, i.e. orography and land-sea effects. In that sense, the improvements are obtained near small-scale terrain features and coastlines, and are reflected, for example, in the Appalachian and coastal precipitation.
Other SG-GCM results include the studies of systematic errors for: single model ensemble integrations (Fox-Rabinovitz et al. 2005), and the North American monsoon system (NAMS) for the multi-year time scales (Berbery and Fox-Rabinovitz 2003).
The future plans include: 1) conducting the phase-3 of the project, SGMIP-3, dedicated to simulation of future climate for the North American and other regions; and 2) contributing to the combined ensembles such as NARCCAP, including both nested and stretched grid models (using SGMIP-2 products for past climate and SGMIP-3 products for future climate).
The SGMIP results were presented to WMO/WGNE at its annual meetings in 2004, 2005, and 2006. The SGMIP products are available to national and international programs such as WMO/ WCRP/WGNE, CLIVAR, GEWEX, IPCC.
References
Fox-Rabinovitz, M. S., E. H. Berbery, L.L. Takacs, and R.C. Govindaraju, 2005: A multiyear ensemble simulation of the U.S. climate with a stretched-grid general circulation model?, Mon. Wea. Rev., 133, pp. 2505-2525.
Fox-Rabinovitz, M.S., J. Cote, M. Deque, B. Dugas, J. McGregor, 2006: Variable-Resolution GCMs: Stretched-Grid Model Intercomparison Project (SGMIP), J. Geophys. Res., 111, D16104, doi:10.1029/2005JD006520.
Berbery, E.H., and Fox-Rabinovitz, 2003: Multiscale diagnosis of the North American Monsoon System with a variable resolution GCM. J. Climate, 16, 1929-1947.
|
