Partner und Internationale Organisationen
(Englisch)
|
AT, BE, BG, CH, CY, DE, DK, EE, ES, FI, FR, GR, HU, IT, LT, NL, NO, PL, PT, RO, SE, TR, UK
|
Abstract
(Englisch)
|
Numerous air quality simulations with the 3-dimensional Fifth generation Mesoscale Model (MM5) off-line coupled to the Comprehensive Air Quality Model (CAMx) at PSI showed that the parameters of the interface that links the two models might significantly affect the concentration of pollutants. One of the key processes is the vertical turbulent diffusion implemented in the interface. The current schemes are not capable of simulating vertical transport correctly. The main objective of the COST proposal at the time of its submission to the State Secretariate for Education and Research (SER) was the improvement of those schemes. A qualified PhD student with a meteorological background was required to perform the main scientific work. Due to difficulties in the recruitment process it was only in October 2007 when Iakovos Barmpadimos, a master student of the University of Reading, took up employment at PSI. In August 2008 he became one of the Swiss coordinator of COST 728 (together with Balazs Szintai, MeteoSwiss). The first part of the project focuses on the simulation and analysis of 3 episodes for Switzerland where extensive experimental data are available: January / February 2006, June 2006 and January 2007. Meteorology and air quality were simulated for 3 nested domains (Europe, Central Europe and Switzerland with 27, 9 and 3 km resolution, respectively). Concentrations of numerous chemical species were estimated, mainly those relevant for ozone formation and for the production of organic and inorganic aerosols. Carbon monoxide, which is a slowly reacting chemical compound, was used as a suitable proxy of the effect of meteorology on air pollution. The results of MM5 and CAMx were compared with measurements from the NABEL stations, which measure both meteorological variables and concentrations of common air pollutants, and with additional measurements of particulate matter taken at specific sites. The winter episode 2006 was characterized by two persistent inversion episodes ('Hochnebel') leading to very high aerosol concentrations over the Swiss Plateau and in southern Switzerland, and by 2 periods with larger wind velocities. It became evident that meteorology and air quality can be reproduced fairly well for advective conditions. For stagnant situations, however, MM5 and its post-processor MM5CAMx are not capable of simulating the atmospheric profile and the surface concentration of pollutants correctly. Carbon monoxide during low wind periods was underestimated by a factor of 2 to 3. Predictions of aerosol concentrations were also too low. However, the relative chemical composition of particulates is reproduced fairly well by CAMx. In summer 2006 the performance of the models regarding ozone and aerosol concentrations during high wind periods are similar as in winter. At high-pressure conditions, however, peak ozone and aerosol concentrations are substantially underestimated. This underperformance is supposed to be mainly due to deficiencies either of meteorological initial / boundary data (COSMO from MeteoSwiss), of MM5 and its parameter settings or of the MM5CAMx interface. Since there is no further development of MM5 we decided to move to its successor, the Weather Reasearch & Forecast (WRF) Model driven by ECMWF global data. MM5 was running on the linux cluster at PSI, whereas WRF model runs are carried out on Cray XT5 at the Swiss National Supercomputing Centre (CSCS, Manno) since March 2010. Preliminary results with standard parameter settings show a slight improvement of the meteorological fields compared to the MM5 driven by COSMO data. In the second part of the project the analysis of specific scenarios was extended along with other COST partners to a mesoscale meteorological model evaluation and intercomparison study, which was carried out for two two-month periods: August / September 2003 and January / February 2004. The domain of interest was northern Italy, the focus being on the Po Basin. Aim of the study was to evaluate model performance and to explain the discrepancies between different models and between models and observations in an area with challenging topography such as the Po Basin. Considering the selected time periods, during August 2003 a major heat wave dominated over Europe and caused, among others, very high ozone concentrations. January and February 2004 also included periods with very high levels of air pollution (in particular PM10) in Po Basin. Meteorology and air quality output were provided by five modelling systems. PSI ran MM5 and WRF offline coupled with CAMx. The Technical University of Madrid provided output from MM5 offline coupled with the Community Multiscale Air Quality (CMAQ) model and the online coupled meteorology-chemistry model WRF/Chem. The University of Brescia provided output of the CESCAM model. The four models will be hereafter referred to as MM5CAMx, WRFCAMx, MM5CMAQ, WRFChem and CESCAM, respectively. All models ran in 5km horizontal resolution at the domain covering Po Valley with the same emission data provided by the Italian Regional Environmental Agencies (ARPA). The model output is evaluated againist the ARPA network of 55 meteorological stations and 51 air quality stations providing quality checked data. In August-September 2003 MM5CAMx, WRFCAMx, MM5CMAQ and WRFChem overestimate wind speed by 0.8, 0.9, 0.3 and 0.6 m/s, respectively, on average over all lowland stations. The average observed wind speed is only 1.8 m/s. The average temperature, which is 22.6 ºC, is underestimated by 1.7, 2.1, 0.3 and 1.0 ºC, respectively. In January and February 2004 more challenging weather conditions led to poorer model performance in terms of wind speed. The model overestimation is 1.4, 1.2, 0.9 and 1.8 m/s respectively. Temperature predictions improved for most models in winter, the bias being -0.4, -0.8, +0.6 and +1.4 ºC respectively and the observed average temperature being 2.8 ºC. Apart from the surface data, the vertical profiles of wind speed, potential temperature and relative humidity of MM5CAMx were examined and compared with soundings. Model agreement with the observations was found to be inadequate in both modeling periods, at all heights and at both modelling scenarios. The above statistics and further examination of the wind speed, dew point and temperature time series (not shown) reveal that MM5CAMx, WRFCAMx and WRFChem models have difficulties in representing the aforementioned challenging weather conditions. This is reflected in the output of the chemistry models. As seen in ozone timeseries for all types of stations, MM5CAMx, WRFCAMx and WRFChem do not follow adequately the daily cycle or the monthly variation of the ozone maxima in August and September 2003. Conversely, MM5CMAQ which has better weather predictions also has satisfactory ozone predictions. Although MM5CAMx and MM5CMAQ use the same meteorology model, their meteorology outputs differ considerably. The possible causes of this feature are being investigated and they include the model initial and boundary conditions, the selection of boundary layer scheme and the cloud physics.
|
