A32A-01 INVITED 10:20h
Regional Climate Modeling with Variable-Resolution Stretched-Grid GCMs
Variable resolution GCMs using a global stretched grid (SG) with enhanced regional resolution over one or multiple areas of interest represent a viable approach to regional climate modeling that has been established since the mid-1990's. The SG-approach provides an efficient regional downscaling to mesoscales and consistent interactions between global/large- and regional/meso- scales while preserving the high quality of global circulation. The SG-GCM ensembles have shown the potential in reducing the uncertainties of regional climate simulations. The international Stretched-Grid Model Intercomparison Project (SGMIP) has been initiated in 2002 for studying the global variable-resolution/stretched-grid approach to regional climate modeling. The phase-1 of the project, SGMIP-1, will be completed in 2005. The SG-GCMs participating in SGMIP-1 are the variable-resolution versions of the basic GCMs of the following four major meteorological centers/groups: (a) the Meteo-France, ARPEGE model, (b) the RPN/Canadian Meteorological Service, GEM model, (c) the Australian CSIRO C-CAM model, and (d) the U.S. NASA/GSFC GEOS model. The 12-year (1987-1998) AMIP-type SG-GCM simulations of the U.S. climate with enhanced, 0.5° regional resolution are analyzed in terms of studying: (a) an efficient/realistic downscaling to mesoscales; (b) interannual climate variability; (c) ENSO related and other anomalous regional climate events (floods, droughts, etc.) and major monsoonal circulations; (d) water and energy cycles. The analysis of SGMIP-1 multi-model ensemble integrations confirmed that a significant reduction of the uncertainty of regional climate simulations is achieved for the multi-model ensemble mean. The SGMIP-1 results are available on the web site (that is being updated): http://esic.umd.edu/~foxrab/sgmip.html SGMIP-1 laid a solid scientific foundation for conducting the new SGMIP-2 (phase-2 of SGMIP). SGMIP-2 will include multi-model ensemble simulation results for an extended period of over two decades. The SGMIP-2 effort will include the accompanying comparisons of enhanced uniform and variable resolution GCMs, with the emphasis on the multi-model ensemble mode. We are establishing a connection with the AMIP group on introducing the SGMIP-1 data and their analysis as a regional project and the SGMIP-2 data and their analysis as an enhanced uniform and variable resolution project, both with the emphasis on the U.S. climate. The SGMIP effort is beneficial to all the participants as well as to a broader regional (and eventually global) climate modeling community.
A32A-02 INVITED 10:35h
Downscaling With a Variable Resolution GCM
It has become rather popular in the last decade to use limited-area models as regional climate models (RCMs) to downscale from GCM simulations. This technique has proved rather successful for many locations. However, the technique can have difficulties for situations in which there is not an adequate progression of synoptic situations through the domain, such as may occur in the tropics A recent development is the application of variable-resolution GCMs as a downscaling technique. An obvious advantage is the avoidance of lateral boundary reflections, which can produce spurious vertical velocities and associated spurious rainfall near the boundaries of a limited-area model. Another advantage is that variable-resolution GCMs can avoid lateral boundary discrepancies which may arise if the GCM and RCM have dissimilar temperature or moisture biases. The model developed at CSIRO for these studies is the Conformal-Cubic Atmospheric Model (C-CAM), which obtains its variable resolution by means of the Schmidt transformation. For downscaling purposes, C-CAM has largely superseded the CSIRO limited-area model, DARLAM. The talk will commence with a description of the main features of the numerics and parameterizations of C-CAM. A large number of downscaling simulations have been performed with C-CAM, for a range of resolutions. For modest grid stretching, the model can be run in stand-alone mode. However, for strongly stretched applications it is desirable to include some form of nudging, at least in the coarsest regions, so as to ensure a reasonable climatology in the coarse region. In C-CAM, this nudging is provided just by the winds, either from a previous C-CAM simulation, from an atmosphere-ocean GCM, or from reanalyses, depending on the type of experiment being carried out. The wind nudging can either be "far-field" (only over the coarsest region) or "global" (at all grid points, but applied more weakly). The latter option is somewhat akin to the spectral nudging used by some of the limited-area modeling community to ensure their simulations produce a sequence of synoptic situations resembling the large-scale GCM or reanalysis. Both methods have been used at CSIRO for various applications. To provide an insight into the issues, a set of "big-brother" simulations has been performed, where a high-resolution global simulation is compared with a number of downscaled versions of itself. The remainder of the presentation will include examples from several C-CAM simulations. a) Climate change simulations for 1960-2100 with 60 km resolution over Australia, using SSTs and far-field forcing from the CSIRO Mk3 GCM for the A2 scenario. b) 40-year simulations over Australia, with both types of wind-nudging from NCEP reanalyses. As might be expected, the interannual variability is captured better by the simulation with "global" forcing. c) Several multi-decadal climate simulations over Tasmania at 14 km resolution. d) Simulations over Asia at 60 km resolution. e) Simulations of the transport of trace gases.
A32A-03 10:50h
Spectral Nudging to Eliminate the Effects of Domain Position and Geometry in Regional Climate Model Simulations
Previous regional climate simulations have exhibited great sensitivity to the size and position of the domain chosen for calculations. Using the Regional Atmospheric Modeling System (RAMS) we performed several experiments using a 7500 km x 5400 km grid over North America and 50 km resolution to investigate the physical and numerical mechanisms involved. The simulation period was June of 2000 and all experiments have NCAR/NCEP reanalysis as initial and boundary conditions and the same exact model settings. For each simulation we displaced the center of the grid in a different direction, always keeping the U.S. in the interior, out of the buffer zone. Circulation biases presented a large scale structure, organized by the Rocky Mountains, resulting from a systematic shifting of the synoptic wave trains that crossed the domain. The distortion of the large-scale circulation was produced by interaction of the modeled flow with the lateral boundaries of the nested domain and varied when the position of the grid was altered. This changed the large-scale environment among the different simulations and translated into diverse conditions for the development of the mesoscale processes that produce most of precipitation for the Great Plains in the summer season. As a consequence, precipitation results varied, sometimes greatly, among the experiments with the different grid positions. We performed additional experiments to assess the effect of the position of the domain on commonly studied sensitivities, such as those to soil moisture, and the response also showed variations among the simulations with different grid positions. To eliminate the dependence of results on the position of the domain we used a spectral nudging technique, where waves longer than 2,500 km above the boundary layer were relaxed to those from reanalysis. Moisture was not nudged at any level. This constrained the synoptic circulation to follow reanalysis while allowing the model to develop the small-scale dynamics responsible for the rainfall. We repeated the experiments with the different grid positions with spectral nudging and the sensitivity of precipitation to the position of the domain was virtually eliminated.
A32A-04 11:00h
Influence of domain size on RCM's internal variability
Pairs of simulations from the Canadian RCM (CRCM) using three different domain sizes are analyzed to investigate changes in internal variability with the increase of the integration domain. Members in the small ensembles differ only by a one-month lag in their initial conditions and all simulations cover a period of many years. As the domain increases, the weather systems stay longer in the RCM domain and the forcing from the lateral boundary conditions is then reduced. It was observed, in the larger domain, that the large-scale circulation differs sometimes between members of the ensemble and also from the one in the driving data. The internal variability in the larger domain reaches values comparable to the one in the driving GCM. Such a large internal variability requires an ensemble of simulations to cover the spectrum of possible solutions for a given set of lateral boundary conditions. Annual cycle of the root-mean-square-difference (RMSD) between pairs of simulations in the large domain shows that the winter RMSD is sometime higher than in summer. This result is different than the stronger RMSD values in summer showed in many studies (Giorgi and Bi, 2000, Christensen et al., 2001, Caya and Biner, 2004). Application of the spectral nudging technique to reduce internal variability for the largest domain is demonstrated in a last experiment. With this technique, the RCM large-scale circulation stays closer to the one of the driving data and this reduce the mismatch in large-scale circulation at the boundary outflow.
A32A-05 11:10h
Influence of Nesting Data and Domain on the Canadian Regional Climate Model's Hydrology.
Results from three 30-year runs (1961-1990) with the CRCM are used to investigate the effect of nesting data on the simulated climate over Quebec/Labrador (eastern Canada). The CRCM was run with a 45-km horizontal resolution, over a large domain covering most of North America (200x192 grid tiles), with 29 vertical levels and 15-minute time steps. Sea surface conditions (SST and sea-ice) were prescribed using monthly AMIP-II observations. In the various simulations, the CRCM was driven at its lateral boundaries by NCEP atmospheric objective reanalyses (NRA-1), ECMWF reanalyses (ERA-40 degraded to 2.5x2.5 degrees) and by CGCM2 output in current climate (Canadian GCM with CO$_{2}$ concentration following IS92a). Components of the model's hydrologic cycle are examined in detail for 12 rivers in northern Quebec and Labrador, totaling a combined drainage area of 435 000 km$^{2}$. Available streamflow observations are used to evaluate the model's climate when driven by reanalyses; the 30-year mean annual runoff is quite similar between simulations nested with NCEP and ECMWF reanalyses, and both have a negative bias when compared with observations. We also examine an additional 30-year CRCM run that was performed on a smaller domain (112x88 grid tiles), centered over Quebec, and nested with NCEP reanalyses. The model's mean annual runoff remains underestimated but the negative bias is reduced by about half on the smaller domain, mostly due to an increase in fall precipitation.
A32A-06 11:20h
Spatial scale separation of limited area models in the context of spectral nudging and isotropic discrete filtering
The concept of spatial scale separation is analyzed in terms of model evaluation and dynamical downscaling techniques. The main objective in downscaling is to obtain regional weather phenomena that are influenced by the local topography or other small scale features. The continental scale atmospheric state is assumed to be well resolved in global reanalyses data and thus it is to be retained by a regional model. For some meteorological conditions, like blocking weather situations, a regional model may simulate large scale features which do no agree with the large scale state of the global forcing data. The application of a spectral nudging technique forces the regional model to approximate the large scales to the reanalysis data for the whole integration area while the small scales are computed exclusively by the regional model. Large-scale control of regional climate modeling overcomes the fundamental problem of dealing with an ill-posed boundary value problem, but its usefulness depends on the specific application. The method of spectral nudging in regional climate modeling is discussed, and its utility in forcing the known large-scale state without suppressing regional and local variability is demonstrated. Also, spatial scale separation can be used for limited area model evaluation purposes. A two-dimensional discrete filter was developed which serves as a tool to classify meteorological fields according to their spatial dimensions by filtering certain wave number ranges. Thereby it performs an isotropic spatial scale separation of the limited area model fields. The filter algorithm is presented and examples demonstrate the scale separation of atmospheric fields on limited area grids which can be used for model evaluation, comparisons or process studies.
A32A-07 11:30h
A New 3-D Potential-Enstrophy-Conserving, Compressible, Nonhydrostatic Model for Weather and Climate Simulation on All Scales and Its Application to San Francisco
We present a unique atmospheric dynamics model useful from the microscale to the global scale and discuss its application to flow through San Francisco. The model, PECCAN (Potential Enstrophy Conserving Compressible Atmospheric Nonhydrostatic), solves the governing equations for a fully compressible 3-D nonhydrostatic atmosphere. The model is unique because it conserves the domain integrals of all of the following quantities simultaneously: (1) potential enstrophy in the limit of barotropic flow in 2-D, (2) total energy (kinetic plus potential plus internal energy) for frictionless flow in 3-D, (3) mass in 3-D, and (4) potential enthalpy (the product of specific heat, density, and potential temperature) for adiabatic flow in 3-D. A new feature of this model is that it generalizes the potential enstrophy and energy conserving numerical scheme of Arakawa and Lamb (1981) for the 2-D shallow water equations to discretize the nonlinear terms in the governing momentum equations of 3-D atmospheric flow. Various authors have demonstrated that, at least for 2-D shallow water flow, using a numerical scheme that maintains the various global conservation properties of the governing equations, especially that of potential enstrophy (PE), leads to the correct transfer of energy among scales in the model and eliminates spurious numerical sources of momentum and energy. Arakawa and Lamb (1981) have also shown that a PE-conserving numerical scheme for 2-D shallow water flow converges to the correct solution on a significantly coarser grid than does a non-PE-conserving scheme. PECCAN is written in general orthogonal curvilinear coordinates. Thus, the model simulates flows in Cartesian, cylindrical, spherical, and Lambert (polar stereographic and Mercator) coordinate systems, with stretched coordinates in any spatial direction. Stretched coordinates allow the model to cover most of the globe with a coarse horizontal grid while covering one or more regions of interest with a much finer grid. This approach eliminates the need for the use of limited-area boundary conditions. In the vertical, the model uses the altitude coordinate, although the numerical scheme can be generalized to the sigma coordinate. To advance the equations forward in time, the model uses any one of several explicit time integration schemes. These include the leapfrog, third and fourth order Runge-Kutta, second order Adams-Bashforth, and Heun schemes. To overcome the time step limitation due to sound waves, the model has the option of using a time-splitting technique in which the velocity divergence terms in the continuity and thermodynamic energy equations and the pressure gradient terms in the momentum equations are advanced using either a small time step explicitly or a large time step implicitly while the remaining meteorologically important terms are advanced using a long time step. Results discussed here include those from flows around buildings and through and above street canyons. In particular, we use a stretched grid that encompasses the globe at coarse resolution but focuses in on the city of San Francisco at high resolution ($<$ 20 m). Idealized flow tests and comparison with results from other authors will also be discussed. REFERENCES Arakawa, A. and Lamb, V. R. (1981). A potential exstrophy and energy conserving scheme for the shallow water equations. Mon. Wea. Rev., 109:18-36.
A32A-08 11:40h
Using Cloud-resolving Regional Model to Study Gravity Wave Generation by Hector Convection and Their Effect on Large-scale Circulation
Gravity waves generated by strong convection transport momentum and energy from troposphere into the middle atmosphere. This mechanism plays an important role in the dynamics of the middle atmosphere. However the related processes are yet not well quantified. The Darwin Area Wave Experiment (DAWEX) focused on observation of gravity wave propagation from the Hector convection at Tiwi Islands during October-December, 2001 using various instrumentation including 5 airglow images, radars, and boundary layer wind profiler. Here we present the results of a numerical case-study of Hector event observed on November 16, 2001 during DAWEX IOP2 aiming to better understand and quantify the convective generation of gravity waves, there further propagation in the middle atmosphere, interaction with the mean flow, and breaking above 100 km at the mesopause region. We employed a regional atmospheric modeling system (RAMS). RAMS is a comprehensive non-hydrostatic model with developed cloud microphysical parameterization and interactive (with cloud microphysics) radiative transport. The calculations were conducted in the horizontal domain of 400km x 500 km, that was extended vertically up to 130 km. Horizontal and vertical grid spacing were respectively 2km x 2km and from 30 to 1000m. The model was initialized from balloon soundings combined with the UKMO and URAP analysis profiles. The simulated storm has good timing and location. The intensity of the storm is in a good agreement with the observations by Australian Bureau of Meteorology Research Center C-band polarized radar. The latent heat release during the convection reached 4000 K/day. The vertical velocity perturbations caused by gravity waves propagated into the mesosphere reached 1 m/s at the altitude of 100 km. The wave phase speed and wave perturbations are compared with the airglow imager observations at Katherine and Wyndham sites that were located respectively about 350 and 450 km south of Tiwi Islans.
A32A-09 11:50h
Using Spectral (Bin) Microphysics in a Mesoscale Model to Investigate the Effect of Aerosols on Regional Precipitation.
A growing body of observational data suggests that the formation of precipitation is sensitive to aerosol concentration. For example, the effect of aerosols on precipitation has been noted downwind of urban centers along the California Coast and in Israel. The simulation of both orographic and convective precipitation along mountainous areas often involves both liquid and mixed precipitation processes that depend on the initial drop distribution. Spectral (bin) microphysics (SBM) embedded in a mesoscale model presents an ideal (if not the only) simulation platform to study the effect of aerosols on regional climate. The SBM is based on solving a system of equations for size distribution functions for water drops, three types of ice crystals (plates, columns, and dendrites) as well as snowflakes, graupel, and hail/frozen drops. A budget for aerosols is used to obtain the spectrum of condensation nuclei, which is used to obtain the initial drop spectrum. Primary and secondary ice nucleation are included in the model. Hydrometeors evolve through both diffusion and collisions. Breakup of large drops is also included. The SBM has been coupled with the three-dimensional mesoscale model, MM5, which allows SBM to simulate microphysics within a realistic, time-varying mesoscale environment. The new model has been used to simulate both convective and orographic precipitation. Results with MM5 SBM are compared to those obtained with MM5 using bulk parameterization. The MM5 SBM appears to better reproduce precipitation amounts, radar reflectivity, cloud structure, and cloud particle type than model simulations with bulk parameterization. We identify and discuss microphysical processes that are crucial to simulating the realistic development of both liquid and mixed precipitation. We explain why these processes cannot be well represented with bulk parameterization.
A32A-10 12:00h
A Comparative Analysis of MM5 and WRF High-Resolution Mesoscale Forecasts Over Variable Terrain
The growing need for accurate, lower mesoscale-limit, numerical weather forecasts provides the impetus for this comparative evaluation of the widely-used 5th generation National Center for Atmospheric Research/Penn State Mesoscale Model (MM5) and the Weather Forecasting and Research (WRF 2.0) Model. A series of simulations was conducted using a single domain, 500 km-squared forecast grids employing 1 km grid cell spacings centered on regions of relatively flat, mountainous, and complex (mixed) terrain and initialized using operational 12 km grid resolution National Center for Environmental Prediction (NCEP) Eta model data. Due to the problem size involved in performing these simulations, the Army High Performance Computing Research Center (AHPCRC) Cray X1 was utilized as the primary computational resource. Surface forecast validation data was provided by the Meteorological Assimilation Data Ingest System (MADIS). Statistical analyses are presented based on 10 meter surface winds, 2 meter temperature, 2 meter specific humidity, and cumulative precipitation data derived from model forecasts and observational records. Lastly, mesoscale model forecast skill implications are discussed in connection with microscale problems such as turbulence modeling using large eddy simulations and atmospheric signal propagation.
A32A-11 12:10h
Understanding the Value of High Resolution Regional Climate Modeling
Regional climate models derive benefits over global climate models from a more accurate representation of regional climate forcings, achieved through higher resolution orography, land-water contrasts and land surface characteristics. Regional forcings can produce statistically significant climate signals, particularly for processes forced directly by topography including orographic precipitation and monsoon circulations. Such high resolution climate scenarios are important for resource management and impact assessment. The Weather Research and Forecasting (WRF) model was designed specifically for high resolution applications, and provides an ideal tool for assessing the value of high resolution (1 - 10km grid-spacing) regional climate modeling. Simulations of the climate of western North America are performed using different model grid spacings. Model evaluation focuses on the statistics of cold-season precipitation, temperature and snowpack over the Pacific Northwest and warm-season temperature and convective precipitation over the Southwest associated with the North American Monsoon. The impact of physics parameterization sensitivity to model grid-spacing may overwhelm any benefits of high resolution simulation and is explored through sensitivity studies.