Uncertainties in modelling the spatial and temporal variations in aerosol concentrations

Aerosols play a key role in air quality (health aspects) and climate. In this thesis atmospheric chemistry transport models are used to study the uncertainties in aerosol modelling and to evaluate the effects of emission reduction scenarios on air quality. Uncertainties in: the emissions of gas and aerosol species in the inventories, in meteorological parameters, in the impact of orography on meteorology, all contribute to the uncertainties in gas and aerosol modelling and require high priority in order to estimate better the gas and aerosol concentrations for scientific research and policy making. The overall objectives of this thesis are to identify and quantify a few key uncertainties related to gas and aerosol regional and global scale modelling. These are: I. The impact of using two different emission inventories on gas and aerosol calculated concentrations. II. The role of the temporal and vertical distribution of emissions on gas and aerosol calculations. III. The impact of model resolution on aerosol calculations. IV. The impact of using two different meteorological driver models on gas and aerosol calculations. V. The impact of emission reduction scenarios on calculated air quality. VI. Strategies to evaluate model results with atmospheric measurements. In Chapter 1 an introduction to aerosols and aerosol modelling is given. In the following chapters, the subjects regarding the uncertainties in aerosol modelling are described. The first and the second subject of this thesis, the sensitivity in aerosol modelling to two different emission inventories, injection altitude and temporal variations of anthropogenic emissions, is described in chapter 2. We use the two-way nested global transport chemistry model TM5 focusing on Europe in June and December 2000. Two widely used emission inventories were available for this study. The EMEP inventory is a policy relevant European scale inventory which contains reported emissions by the member countries (http://www/emep.int), with more detailed information on the vertical, temporal and spatial distribution of the emissions. The AEROCOM inventory (Dentener et al., 2006) is a compilation of global scale aerosol and precursor emissions for the year 2000, and was used in the IPCC AR5 assessment report. The simulations of gas, aerosol concentrations and aerosol optical depth (AOD) with the two emission inventories are compared with EMEP gas and aerosol surface based measurements, AERONET sun photometer retrievals and MODIS satellite data. We evaluated the impact of the EMEP and AEROCOM emission inventories on aerosol concentrations and aerosol optical depth (AOD) in Europe for June and December 2000. There are substantial differences between annual emissions included in the two inventories. It also appears that differences are found in the vertical distribution of the SO2 and NOx emissions. Despite these differences, for most aerosol species and aerosol precursor gases TM5 simulates the spatial and temporal distribution over Europe relatively well when compared to observations. Spatial correlations, based on monthly mean surface concentrations, are often quite high (> 0.7) and many EMEP measurement stations show high temporal correlation with the simulations using EMEP and AEROCOM. However, from the comparison with surface observations, we conclude that the AEROCOM inventory overestimates the emissions of aerosol precursor gases SO2 and NOx and NH3 emissions, especially in June. Furthermore, a lack in seasonal varia tion in the AEROCOM inventory and uncertainty in the vertical distribution of emissions (SO2 and NOx), contributes to disagreement of model and surface observations. For NH3 it seems that the inclusion of recent abatement measures in the EMEP inventory indeed leads to a better agreement with measured concentrations. The large differences in surface concentrations between the simulations are not equally reflected in corresponding differences in computed column aerosol and AOD. Model AOD computations using the AEROCOM and EMEP emission inventories reveal good agreement with surface based AERONET sun photometer observations and AOD retrieved from MODIS. Spatial patterns over Europe of AOD differ due to the varying contributions of mineral dust and inorganic aerosol, as observed by satellite and confirmed by model simulations. An evaluation of the impact on aerosol of the temporal distribution (daily, weekly and seasonal) of emissions reveals that the concentrations of most aerosol components are not strongly influenced by introduction of a high temporal resolution of emissions. The exception is aerosol nitrate and its precursor gases NOx, and NH3. However, seasonal temporal variation of the emissions do play an important role for all gas and aerosol calculations, and need to be included to accurately calculate aerosol concentrations and it’s influence on air quality and climate change. The third subject of this thesis is to study the model resolution dependency on aerosol calculations, see chapter 3. For this work the mesoscale Transport of Atmospheric Pollutants Model (TAPOM) is used. Firstly, we evaluated the TAPOM performance in calculated aerosol surface concentrations and aerosol optical depth (AOD) values for the greater Milan area in Italy during June 2001. Secondly, we used the model to study scale issues in aerosol modelling at three horizontal resolutions (5x5 km, 10x10 km and 20x20 km) through calculations of AOD and other aerosol properties for the same area and period. Model calculations of sulphate aerosol concentrations on a 5x5 km horizontal resolution show a reasonable agreement with measurements, i.e. within a factor 1.5 of the measurements. A comparison of aerosol optical depth calculated with the model and surface based sun photometer measurements revealed some discrepancies, which can be roughly divided in two clusters: 1. On clear, dry sky days there is a relatively good correlation between model and sun photometer AOD and satellite AOD. 2. A second group of model calculated AOD and measurements appear to be uncorrelated. The discrepancies of model results and measurements are for some days related to the underestimation of PM emissions and the lack of natural dust at the boundaries. Another reason could be related to the presence of cirrus clouds appearing as AOD in the measurements. Model calculations at 5x5 and 10x10 km horizontal resolution show a good internal agreement, whereas a model version using 20x20 km resolution loses some details regarding the spatial distribution of AOD. Daily mean and maximum coinciding model 5x5 km AOD values with MODIS and MISR AOD agree better with the observations than the computed AOD using the 20x20 km resolution. The largest differences in aerosol calculated concentrations between the different resolutions are associated with high RH (relative humidity) conditions leading to large amounts of computed aerosol water and higher AOD values. A further scale related issue is the non-linearity of aerosol formation at especially high relative humidity. These uncertainties are expected to be even more uncertain in large scale models that use fairly simplified parameterization relating relative humidity to cloud formation. Given the uncertainties in cloud screening of the satellite products, the reliability of these satellite products is often questionable and it has turned out to be difficult to make comparisons to model results at the mesoscale (like in this study). This problem is even larger for global scale models, with resolutions higher than 100x100km and less detailed emission inventories, where local and regional information are not taken into account. This study also showed the importance of accurate model boundary conditions for realistic gas, aerosol and AOD computations in regional/urban scale air quality modelling. In chapter 4, the fourth subject of thesis is described, i.e. studying the impact of two different meteorological models (MM5 and WRF) on aerosol and O3 calculations with the chemistry transport model CHIMERE. The area and period of interest is the Po valley region (Italy) for January and June 2005. The meteorological data sets used for the study were created in the frame of the Po valley air quality Model Inter-comparison (POMI) exercise, which is coordinated by the Institute of Environment and Sustainability, JRC, Ispra, Italy (http://aqm.jrc.it/POMI/). First we evaluate for January, June and the whole year the calculated meteorological parameters by MM5 and WRF (temperature, wind speed, wind direction, relative humidity and precipitation) with observations. The analysis shows that the overall performance of both models is similar, however some small differences are still noticeable. On a yearly basis, the temperatures are mainly underestimated (less by WRF) when compared to observations and the values of relative humidity are in general overestimated (less by MM5). WRF output follows better the hourly pattern of relative humidity. We had only two stations available with wind data. The wind speed is well reproduced for Ispra monitoring site (especially by WRF) but for Mantova is largely overestimated by both models (less by MM5). Both models do not reproduce well the wind direction. The rainfall is in general overestimated, however the MM5 output shows lower rainfall values. Secondly, we analyze the impact of using MM5 and WRF on calculated PM10 and O3 concentrations. In general the model underestimates the observed PM10 concentrations with both the MM5 and WRF meteorology, from now on denoted as CHIMERE/MM5, CHIMERE/WRF respectively. The PM10 concentrations for January are a factor 1.6 higher for CHIMERE/MM5 than CHIMERE/WRF. The concentrations of gases and aerosols at ground level strongly depend on the profile of the planetary boundary layer. Therefore we examined the PBL profiles and the latent and sensible heat fluxes which are responsible for the PBL formation. The difference and the larger underestimation in PM10 concentrations by CHIMERE/WRF are related to the differences in heat fluxes and the resulting PBL heights calculated by WRF. In general the PBL height by WRF meteorology is a factor 2.8 higher at noon in January than calculated by MM5. This could result in a better vertical mixing of the aerosols than CHIMERE/MM5, causing lower aerosol concentrations at the surface. Changing the Noah LSM scheme in our WRF pre-processing for the 5-layer soil temperature model, leads to an increase of the calculated PM10 concentrations of 30% for January 2005 when compared to the simulation using Noah LSM. This study also showed that the difference in microphysics scheme has an impact on the profile of cloud liquid water (CLW) calculated by the meteorological driver and therefore on the production and removal of SO4 = aerosol. For June the differences in PM10 concentrations between the model simulations using MM5 and WRF are small. Analyzing the heat fluxes and the diurnal behaviour of the PBL height we observe small differences between the two meteorological models. In the previous chapters, three major uncertainties which contribute to modelled aerosol calculations are described. It is crucial to understand how important these sources of uncertainties contribute to the model outcome, especially when the model results are used for policy relevant studies. In the last chapter of this thesis (chapter 5) we performed a case study to evaluate the impact of emission reduction scenarios on PM2.5 and O3 calculated concentrations for the Po valley area, by using the chemistry transport model CHIMERE. Comparing calculated surface aerosol concentrations by the model for the standard case (no emission reductions) with observations for January and June 2005 shows that the monthly mean PM10 concentrations are in general underestimated by a factor 1.4. To evaluate the impact of emission reduction scenarios we reduced the emissions of NOx, PM2.5, SO2, VOC or NH3 by 50% separately for each component and for different source sectors, together with 5 additional scenarios. The 50% reduction of the emissions corresponds to the application of the current legislation for PM2.5 for Italy which should be met in 2010 in respect to the emissions of the base year 2000 (EURODELTA exercise: http://aqm.jrc.it/eurodelta/, which is carried out in the frame of the Clean Air For Europe programme, CAFE). The most effective scenarios to abate PM2.5 concentration are based on the nonindustrial combustion plants, i.e. domestic heating (SNAP2) and road traffic (SNAP7) sectors, for which the NOx and PM2.5 emissions are reduced by 50%. These scenarios reduce the monthly calculated PM2.5 concentrations for the Po valley area for January on average by 1-6µg/m3. However, these emission reduction scenarios for domestic heating and traffic do not have an effect on lowering the number of days for which the planned European limit value (Directive 2008/50/EC) of 25µg/m3 is exceeded. Only by combining the emission reductions of NOx, PM25, SO2, VOC and NH3 for domestic heating, road traffic and agriculture (SNAP10) results in a larger reduction of PM2.5 calculated concentrations over the larger area in the Lombardy region (~20%), which corresponds with the findings of a similar study performed by APAT (the Italian National Agency for environmental protection and technical services). In the APAT study more than one species was reduced for different activity sectors. Our study also showed that a more effective pollutant reduction (emissions) per tonne of pollutant reduced (concentrations) for the greater Milan area is obtained by reducing the primary PM2.5 emissions for road traffic. The most effective scenario on (secondary) PM2.5 decrease for which precursor emissions are reduced is achieved by reducing SO2 emissions by 50% for road traffic. Our study showed that during summer, the largest reductions in O3 concentrations are achieved for SNAP 7 emission reductions, when volatile organic compounds (VOCs) are reduced by 50%. It appears that around 40% of the PM2.5 concentration for Milan is caused by the emissions from surroundings of the Lombardy region as well as by longrange transport from elsewhere (as reflected as model boundary conditions). Therefore effective abatement of air pollution does need the consideration of the transboundary aspects of air pollution. Our study showed that differences may occur between measurement (ground based and space born) data sets for the same variable and time period. This complicates the use of observations for model evaluation purposes. The main conclusions of this thesis can be summarized as: • The amounts and the temporal and vertical distributions of the emissions are important for gas and aerosol calculations, especially when calculations are compared to surface measurements. • Going to a finer model resolution results in better agreement between calculated aerosol and AOD values and observations. • The PBL height and vertical mixing of the lower troposphere determine to a large extend the surface gas and aerosol concentrations. These boundary layer characteristics are strongly determined by the sensible and latent heat fluxes at the surface. • Local emission reductions result in a local improvement in air quality.Longrange transport is however also very important for reducing local pollution. • Surface and space born measurements of gas phase and aerosol components are crucial for evaluating model results. However inconsistencies in measurement data sets make it complicated for reliable comparisons with model simulations.