Keywords: air pollution, parallel platforms, meteorological modelling.

During the last decade a considerable effort has been made on improving the computer efficiency of sophisticated air quality modelling systems. The objective of this contribution is to show the results of testing the air quality modelling system, MM5-CMAQ [1] in different platforms and with different configurations. The MM5-CMAQ modelling system is a very sophisticated air quality modelling system which is used to simulate the dynamics of the pollution concentrations in the atmosphere by using the Navier-Stokes fluid dynamics partial differential equation system. MM5 is a non-hydrostatic meteorological model developed by Penn State and NCAR (US) [1] and CMAQ [2] is the Community Multiscale Air Quality Modelling System developed by EPA (US). These models require a substantial amount of input information such as hourly and pollutant emission data, meteorological and chemical initial and boundary conditions, land use data and topographic information. This is a huge amount of information. For the emission data, we have used the EMIMO model [3] to provide the amount of pollution emitted per grid cell, per hour and per activity. The MM5 and CMAQ models are fully parallelized for OpenMP and MPI environments and different tests have been done. Due to the Courant-Levy condition, the time step for the meteorological module and for the CMAQ module depends on the spatial grid resolution and the maximum wind speed in the hree-dimensional domain so that the time step changes every time step and the value depends on the grid resolution. In this contribution we will show results of the MM5-CMAQ modelling system applied over the whole of Europe with a meteorological domain defined as 95x90 horizontal layers (50 km spatial resolution) and 23 vertical layers (up to 100 mb) and the CMAQ modelling system over a domain of 88x83 with 12 vertical layers. In this contribution we have used the MM5 meteorological mesoscale model developed by Pennsylvania State University (USA) and NCAR (National Centre for Atmospheric Research, USA) [1]. The CMAQ model is the Community Multiscale Air Quality Modelling System developed by EPA (USA) [2] and EMIMO is the Emission Model developed by San José et al. [3]. MM5 is a well recognized non-hydrostatic mesoscale meteorological models which uses global meteorological data produced by global models such as GFS model (NCEP, USA) to produce high resolution detailed three dimensional fields of wind, temperature and humidity which are used in our case as input for the photochemical dispersion model CMAQ. In addition to MM5 output data, EMIMO model produces for the specific required spatial resolution, hourly emission data for different inorganic pollutants such as particulate matter, sulphur dioxide, nitrogen oxides, carbon monoxide and total volatile organic compounds VOC's. The VOC's are divided according to SMOKE (Sparse Matrix Operator Kernel Emissions. In this particular case, the EMIMO model is applied by using the annual EMEP official emissions from 2003 for the whole of Europe. The full system is called OPANA V3 since the V2 included adaptations of the MEMO model (REMEST) and on-line implementation of the SMVGEAR implicit technique with the CBM-IV chemical carbon bond mechanism.

We have presented a system called OPANA (Operational Atmospheric Numerical pollution model for urban and regional Areas) which has been improved during the last ten years. OPANA has included different versions of second generations of air quality models such as MEMO and CHEMA (adapted), nowadays it includes MM5, CMAQ and EMIMO as part of the full system (including a version of MIMO and CAMO, cellular automata model, called MICROSYS). The system is on service producing on a daily basis, air quality forecasts for several cities and regions such as Madrid city (Spain), Las Palmas de Gran Canaria (Canary Islands, Spain) and Leicester City (Great Britain). During the last two years several experiments have been performed on different parallel platforms, all of them cluster oriented (distributed memory). During the last year, several experiments have been performed on dual core machines (shared memory) and mixed systems (shared and distributed memory). In general, CMAQ ran faster as more processors were used, as long as the data was written locally. The higher the spatial resolution, the smaller the time step and the CPU increases exponentially. The results show that the dual core AMD 4200 and 4800 obtain an excellent performance compared to one processor and PIV 3.4 Ghz with a faster clock. More experiments are expected in the CESVIMA supercomputer center with 400 nodes during the next months.

1

Grell, G., J. Dudhia, and D. Stauffer 1994: A Description of the Fifty Generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note, TN-398+STR, 117 pp.

2

Byun, D.W., J. Young, G. Gipson, J. Godowitch, F. Binkowsky, S. Roselle, B. Benjey, J. Pleim, J.K.S. Ching, J. Novak, C. Coats, T. Odman, A. Hanna, K. Alapaty, R. Mathur, J. McHenry, U. Shankar, S. Fine, A. Xiu, and C. Lang, 1998. Description of the Models-3 Community Multiscale Air Quality (CMAQ) model. Proceedings of the American Meteorological Society 78th Annual Meeting Phoenix, AZ, Jan. 11-16, 1998. pp. 264-268.

3

San José R., Peña J.I., Pérez J.L. and González R.M., EMIMO: an emission model. 292-298. ISBN: 3-540-00840-3 Springer-Verlag, 2003.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)