Meteorological data is one of key input parameters when modelling long-range transport and deposition of atmospheric pollutants. Most of the meteorological parameters required for modelling of pollutant atmospheric transport are not available from the routine synoptic or aerological observations. Moreover, observation stations are randomly distributed over surface, whereas the modelling needs data on a regular grid. Therefore, it is necessary to use a pre-processing system, which can prepare gridded consistent meteorological parameters with certain temporal resolution. The methodology of meteorological preprocessing has been worked out over many years at MSC-E. Originally it was applied at a regional scale (EMEP domain), later it was adapted to a global scale.

Below the main aspects of meteorological data preparation at the MSC-E are briefly described. More detailed information about meteorological data processing and evaluation at MSC-E is available in technical reports.


Meteorological drivers

Atmospheric transport models of MSC-E utilize off-line meteorological information. This means that meteorological data are not generated in the process of calculations, but periodically supplied into the model as input data. Therefore, meteorological data are prepared in advance and stored in the same model grid as used in the transport models. Therefore, meteorological drivers are used to generate sets of meteorological parameters consistent in space and time. At present time three following drivers are used: WRF, MM5, and GEM.

The Fifth-Generation Mesoscale Model (MM5) is the model designed to simulate or predict mesoscale atmospheric circulation. MM5 was worked out at Pennsylvania State University and National Center for Atmospheric Research (NCAR). This driver was adapted for simulation of meteorological parameters over the EMEP domain.

The Global Environmental Multiscale Model (GEM) is an integrated forecasting and data assimilation system developed in the Recherche en Provision Numerique (RPN), Meteorological Research Branch (MRB), and the Canadian Meteorological Centre (CMC). This model is used for generation of meteorological data over the global scale.

The Weather Research & Forecasting Model (WRF) is a next-generation mesoscale numerical weather prediction system designed to serve both operational forecasting and atmospheric research needs. It has been developed in a collaborative partnership, among the National Center for Atmospheric Research (NCAR), the National Oceanic and Atmospheric Administration (the National Centers for Environmental Prediction (NCEP)) and several other agencies. This model is capable of simulating meteorological fields both at regional and local scales. Gradual transition from MM5 and GEM to WRF is ongoing.


Input data

ECMWF operational analysis data data are currently used as input information for meteorological drivers listed above. These data are used as initial conditions and are assimilated during simulation cycle.


Spatial scales and resolutions. Map projections

MSC-E perform numerical modeling of air pollutants atmospheric transport in different spatial scales (global, regional, local) and map projections (latitude/longitude and polar stereographic). All the meteorological drivers currently used are multi-scale. Only the WRF model can work both in latitude/longitude and polar stereographic projections. The potentialities of meteorological drivers related to spatial scales/grid projections are illustrated in Fig.1.

Global-scale modeling is performed with spatial resolutions from 5ox5o to 1ox1o (latitude/longitude grid). Regional/local scale modeling is carried out with resolutions from 50x50 km2 to 5x5 km2 (polar stereographic projection). Nesting procedure is used to provide consistent meteorological data sets at different scales. The example of multi-scale calculation results is shown in Fig. 2.


Fig. 1. Potentialities of meteorological drivers
related to spatial scales/grid projections


Global domain, 1ox1o   EMEP domain, 50x50 km   Czech domain, 10x10 km



Fig.2. Calculated spatial distributions of monthly averaged air temperature at 2m in February 2008 in different scales


Calculation cycle

Fig. 3. Calculation cycle scheme

The process of preparation of meteorological data for the long period of time (one or several years) is organized as the sequence of short period meteorological model runs. Each of these runs consists of the two parts: spin-up and forecast (Fig. 3). The lengths of these periods can be different (it depends on model. To filter out the high-frequency oscillations each short period model run begins with the digital filtering. The modeling results of the forecast period are stored and further used as the meteorological input for the EMEP chemical transport models.


Meteorological parameters

List of the parameters involved in transport of HMs and POPs and their usage in modeling is given in Table 1.


Table 1. Meteorological parameters used in modelling of atmospheric transport of HMs and POPs are presented in the table below
Parameter Notation Dimension Usage
 Surface pressure ps 2D  Air density, atmospheric transport
 Components of wind velocity U,V 3D  Atmospheric transport
 Air temperature Ta 3D  Air density, atmospheric chemistry, dry deposition
 Water vapour mixing ratio qv 3D  Air density, dry deposition
 Liquid water mixing ratio qw 3D  Atmospheric chemistry, in-cloud scavenging
 Ice mixing ratio qi 3D  In-cloud scavenging
 Stratigorm precipitation Rs 3D  Wet removal
 Convective precipitation Rc 3D  Wet removal
 Eddy diffusion coefficient Kz 3D  Vertical eddy diffusion
 Monin-Obukhov lengh* L 2D  Stability, dry deposition
 Surface temperature Ts 2D  Natural emission and re-emission
 Snow cover height Hs 2D  Natural emission and re-emission


Evaluation of pre-processed meteorological data

In order to perform evaluation of the output of meteorological models simulation results are compared with measurement data. Two kinds of measurements are used in the evaluation: surface measurements and upper-air observations.

To compare calculated precipitation fields with measurements surface observations of daily precipitation amount from the GCOS Surface Network (GSN) are employed.

Near-surface and upper-air observations from the NOAA/ESRL Radiosonde Database (RAOB) with 12-hour resolution are used for the comparison of computed and observed data. The following meteorological variables are involved in the comparison: air temperature, air humidity, geopotential height, and wind speed components U and V.

To characterize the level of agreement between the calculated and measured values of meteorological variables the following statistical indicators are used: BIAS, root mean square error (RMSE), and correlation coefficient (Rcorr).

The examples of the comparison of modeling and measurement data are presented in Fig. 4-5. Temporal variations of air temperature at two particular RAOB sites are given in Fig.4. Scatter plot of annual mean near-surface calculated and measured temperatures is shown in Fig 5.


Fig. 4. Comparison of near-surface air temperature calculated by WRF and GEM models with measurements of the meteorological sites 72520 (USA) and 2836 (Finland) for January and July 2001


Fig. 5. Scatter plots of measured annual mean
surface air temperature at 2m height obtained
from RAOB database versus calculated by GEM
modes for 2001

Two main types of averaged statistics are calculated:

  • spatial indicators averaged on time moments (RMSEspace, Rcorr-space);
  • temporal indicators averaged on measurement sites (RMSEtime, Rcorr-time).

The examples of vertical profiles of site-averaged statistical indicators are given in Fig. 6.

In addition to the comparison with measurements, the spatial distribution of selected meteorological parameters calculated are compared with the data of two meteorological re-analysis products: ECMWF (ERA-40) and NCEP-DOE. Precipitation amount is also compared with the dataset based on satellite measurements prepared by the GPCP project. For example, the spatial distributions of annual precipitation amount simulated by the GEM model, obtained from ECMWF ERA-40 and NCEP/DOE re-analyses, and according to GPCP product for 2001 are presented in Fig. 7.




Fig. 6. Comparison of air temperature for January 2001 obtained by WRF and GEM simulations with 12-hour measurements from RAOB database. Site-averaged vertical profiles of statistical indicators


GEM model
gridded precipitation of GPCP
ERA-40 reanalysis
NCEP/DOE reanalysis
Fig. 7. Spatial distribution of annual precipitation amount for 2001 calculated by GEM model, obtained from re-analyses ERA-40 and NCEP/DOE, and according to GPCP product based on satellite data