SENSITIVITY ANALYSIS OF INFLUENCING FACTORS ON PM 2.5 NITRATE SIMULATION

Transcription

1 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, SENSITIVITY ANALYSIS OF INFLUENCING FACTORS ON PM. NITRATE SIMULATION Hikari Shimadera, *, Hiroshi Hayami, Satoru Chatani, Yu Morino, Yasuaki Mori, Tazuko Morikawa, Kazuyo Yamaji, Toshimasa Ohara Central Research Institute of Electric Power Industry, Abiko, Japan Toyota Central R&D Labs., Inc., Nagakute, Japan National Institute for Environmental Studies,, Japan Japan Weather Association, Tokyo, Japan Japan Automobile Research Institute,, Japan Japan Agency for MarineEarth Science and Technology, Yokohama, Japan. INTRODUCTION Fine particulate matter (PM. ) has been of increasing concern because of its adverse effects on human health. The Ministry of the Environment of Japan introduced an air quality standard for PM. in 9. Although PM. concentrations have decreased in recent years in Japan, the standard for PM. is not attained in most urban areas in Japan. Although air quality models (AQMs) with reasonable accuracy are essential tools to develop and evaluate measures to attain the standard, current air quality models cannot adequately simulate atmospheric mass concentrations of PM. and its components in the Kanto region of Japan (Morino et al., ). This study focused on secondary inorganic aerosols among major PM. components. Air quality simulations were conducted in the Kanto region in winter and summer by using the Community Multiscale Air Quality modeling system (CMAQ) (Byun and Schere, ) with several configurations. Because large discrepancies occurred between observed and simulated NO concentrations as described in section, sensitivity analyses were conducted in order to investigate factors influencing the model performance for NO simulation.. METHODOLOGY Fig. shows modeling domains and locations of observation sites for PM. and its components. The horizontal domains consist of three domains from domain (D), covering a wide area of East Asia, to domain (D), covering most of the Kanto region. The horizontal resolutions and the number *Corresponding author: Hikari Shimadera, Central Research Institute of Electric Power Industry, Abiko, Abiko, Chiba 79, Japan; simadera@criepi.denken.or.jp of grid cells are, and km, and 9, and for D, domain (D) and D, respectively. The vertical layers consist of sigmapressure coordinated layers from the surface to hpa with the middle height of the first layer being approximately m. Meteorological fields were produced using the Weather Research and Forecasting model (WRF) (Skamarock and Klemp, ) version.. driven by the final analysis data and the highresolution global sea surface temperature analysis data of the National Centers for Environmental Prediction. WRF was configured with the Kain Fritsch cumulus parameterization in D and D, the Asymmetrical Convective Model version (ACM) for the planetary boundary layer parameterization, the WRF singlemoment class microphysics scheme, the PleimXiu land surface model, the rapid radiative transfer model for the longwave radiation and the Dudhia scheme for the shortwave radiation. Grid nudging was applied to wind, temperature and humidity in D and D Emission data were produced according to the method described by Chatani et al. (). Anthropogenic emissions in D were derived from the INTEXB Asian emission inventory for SO, NO X, CO, PM and volatile organic compounds (VOC), and from the regional emission inventory in Asia (REAS) for. Anthropogenic emissions in D and D were estimated with the Japan AutoOil Program (JATOP) vehicle emission estimate model and the GeoreferenceBased Emission Activity Modeling System (GBEAMS). Ship emissions were derived from emission inventories developed by the National Maritime Research Institute (NMRI) and the Ocean Policy Research Foundation (OPRF). Biogenic VOC emissions were estimated with the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version.. Table summarizes five kinds of CMAQ configurations (MM). Initial and boundary

2 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, Table CMAQ configurations. M M M M M AQM CMAQ v.7. CMAQ v.7. CMAQ v. CMAQ v.7. CMAQ v. Domain D, D, D D D D, D, D D Horizontal advection Yamartino Yamartino Yamartino PPM Yamartino Vertical advection Yamartino Yamartino Yamartino PPM WRF Horizontal diffusion Multiscale Multiscale Multiscale Multiscale Multiscale Vertical diffusion ACM ACM ACM ACM ACM Photolysis rate Lookup table Online Lookup table Online Online Gas phase SAPRC99 (EBI) SAPRC99 (EBI) SAPRC99 (ROS) SAPRC99 (EBI) SAPRC99 (EBI) Aerosol phase AERO AERO AERO AERO AERO Cloud phase ACM ACM RADM ACM ACM concentrations for D were obtained from the Model for Ozone and Related Chemical Tracers version (MOZART). CMAQ simulations were conducted for periods from November to December, and July to,, with the first seven days being initial spinup periods. M is the baseline simulation case in this study. Results of M were used for boundary concentrations for M, M and M, and are compared with the observation data in the next section.. MODEL EVALUATION Fig. shows observed and Msimulated mean concentrations of PM. and its components including organic aerosol (OA), elemental carbon (EC), SO, NO and NH, and gaseous and at the observation sites in winter. The Msimulated mean concentrations of PM., EC except at, and SO approximately agreed with the observations. However, the model clearly underestimated OA concentrations at the observation sites, and overestimated NO, and concentrations at. Fig. shows observed and Msimulated mean concentrations of PM. and its components, and gaseous and at the observation sites in summer. The Msimulated mean concentrations of PM., EC, SO and concentrations approximately agreed with the observations. However, the model clearly underestimated OA concentrations, and clearly overestimated NO and concentrations. The overestimation of NO concentration is likely attributed to overestimation of NH NO production. The agreement of mean PM. concentrations between the observation and simulation in this study does not indicate a good model performance because it is due to compensation of the overestimations and underestimations of PM. components. Results of sensitivity analyses to D Elevation (m) 7 D D Fig.. Modeling domains and locations of observation sites for PM. and its components.

3 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, Obs. M PM. PM. OA PM. EC PM. SO PM. NH. (μg m ).. Fig. Comparisons of observed and Msimulated mean concentrations in winter. PM. 9 PM. OA. PM. EC Obs. PM. SO M PM. NH (μg m ).. Fig. Comparisons of observed and Msimulated mean concentrations in summer. investigate influencing factors on the overestimation of NO are presented in the next section.. SENSITIVITY ANALYSES. Different CMAQ Configurations For simplification, results of sensitivity analyses were compared by percentage differences of mean concentrations from each baseline cases in land areas with an elevation less than m in D (target area) during November to December, and July to,. Fig. shows comparisons of results of MM. Because of common meteorological, emission and Difference from M NO N O M M M M PM. NH Fig. Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for MM. NO N O PM. NH boundary concentration data, temporal variation patterns of simulated concentrations at the observation sites were very similar to each other (not shown). Although the results of M and M, which are CMAQ version.7. (the same version as M), were relatively similar to those of M, it was indicated that online calculation of photolysis rate slightly reduced NO in Kanto in summer. The results of M were the most different from the others due to various different configurations. M, which is CMAQ version. in which minimum eddy diffusivity reduced from. to. m s, predicted higher groundlevel concentrations under stable conditions during nighttime.. Temperature and Relative Humidity NH NO is a semi volatile compound can exist in gaseous and particulate phase mainly depending on temperature and relative humidity. Sensitivity analyses of temperature and relative humidity were conducted in D by M. In sensitivity cases, temperature or relative humidity used in aerosol module was uniformly changed by ± K (T, T) or by ±% (RH, RH). Fig. shows results of the sensitivity analyses. In higher temperature and lower relative humidity, NH NO easily shifts to the gas phase. Note that temperature and relative humidity affect not only gas/aerosol partitioning in aerosol module but other processes such as heterogeneous reaction

4 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, Difference from M Fig. Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for sensitivity analyses of temperature and relative humidity. The results indicate even errors of temperature and humidity that sometimes occur can cause over or underestimation of NO.. NO X Emission Sensitivity analysis of NO X emission was conducted in D by M. In sensitivity cases, NO X emission was uniformly changed by from % to Difference from M Fig. Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for sensitivity analysis of NO X emission. Difference from M NO NO NO N O N O T T RH RH N O PM. NH PM. NH PM. NH Fig. 7 Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for sensitivity analysis of seasonal variation of emission. NO E E E E NO NO N O N O N O PM. NH PM. NH PM. NH % (E, E, E, E). Fig. shows results of the sensitivity analysis. Although the sensitivity of NO X emission to NO concentration was large, the sensitivity to NO concentration was relatively small. Because uncertainty in total NO X emission in Japan is probably smaller than %, improvement of NO X emission is unlikely to be a key factor for improvement of NO simulation.. Seasonal variation of Emission Sensitivity analysis of emission was conducted in D to D by M. For sensitivity runs, a seasonal variation of emission was estimated according to processes for N O emission estimate in Japan (National Institute for Environmental Studies, ). Using the seasonal variation, total emission in D increased by % in winter and decreased by % in summer. Fig. 7 shows results of the sensitivity analysis. The change of emission directly affected concentration, and substantially affected formation of NH NO. In Japan, observation data of concentration are not sufficiently available for model evaluation. Therefore uncertainty in emission estimate is probably large and improvement of emission may be one of the key factors for improvement of NO simulation.. and Dry Depositions Gaseous and are efficiently removed from the atmosphere through dry deposition. While mean dry deposition velocities of and in the target area in daytime ranged from to cm s in M, Neuman et al. () estimated much higher dry deposition velocities of in daytime ( to cm s ). Sensitivity analysis of and dry deposition velocities was conducted in D by M. Difference from M NO N O Vd PM. NH Fig. Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for sensitivity analysis of and dry deposition velocities. NO Vd. N O PM. NH

5 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, In sensitivity cases, the dry deposition velocities were uniformly multiplied by (Vd) and. (Vd.). Fig. shows results of the sensitivity analysis. Higher dry deposition velocities substantially decrease concentrations of and, and consequently NH NO. Because of difficulty in evaluation of dry deposition process, uncertainty in simulated deposition velocity is probably large. Therefore, and dry deposition velocities can be one of the key factors for improvement of NO simulation.. N O Heterogeneous Reaction Sensitivity analysis of N O heterogeneous reaction was conducted in D by M. In sensitivity cases, the heterogeneous reaction probability (Γ NO ) was set to (no reaction; Γ) and. (upper limit; Γ.). In addition, sensitivity runs using parameterization methods of the probability used in AERO (Γae) and AERO (Γae), which are older aerosol modules of CMAQ, were conducted. Fig. 9 shows results of the sensitivity analysis. NO concentrations in Γ case were lower by about % in winter and % in summer than those in M. These rates indicate the Difference from M Fig. 9 Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for sensitivity analysis of N O heterogeneous reaction. Difference from M NO NO N O N O Γ. Γ Γae Γae PM. NH PM. NH Fig. Percentage differences of mean concentrations from M in the target area in winter (left) and summer (right) for ModMulti case. NO NO N O N O PM. NH PM. NH contribution of NO production through the heterogeneous reaction in M. Therefore, uncertainty in the heterogeneous reaction is not a dominant cause of the overestimation of NO concentration. However, modification of the Γ NO parameterization method can slightly mitigate the overestimation..7 Modification of Multiple Factors Multiple factors that can mitigate the overestimation of NO concentration were simultaneously applied to M (ModMulti). ModMulti case is different from M in using online calculation of photolysis rate, the modified seasonal variation of emission, fivefold dry deposition velocities of and, and the Γ NO parameterization method of AERO. Fig. shows results of the sensitivity analysis. The results showed substantial decrease of and NO concentrations, particularly in summer. Figs. and show observed, M and ModMultisimulated mean concentrations of PM. and its ionic components, and gaseous and at the observation sites in winter and summer, respectively. In winter, the difference between M and ModMultisimulated concentrations was relatively small because the effect of higher dry deposition velocities compensated the effect of larger emission. In summer, concentrations of ModMulti case were substantially lower than those of M mainly because of the combined effect of higher dry deposition velocities and smaller emission in summer. Changes in and concentrations affected formation of NH NO. PM. concentrations of ModMulti case were lower than those of M because of lower concentrations of ionic components, particularly NO. Consequently, NO concentrations of ModMulti case were closer to the observations than those of M. Using fivefold dry deposition velocities may be unrealistic and has no theoretical support. However, the results indicate that such drastic modification may be required to improve the model performance for NO concentration.. SUMMARY Air quality simulations using CMAQ were conducted in the Kanto region of Japan in winter and summer. Although the model approximately reproduced EC and SO concentrations, it clearly underestimated OA and clearly overestimated NO and concentrations.

6 Presented at the th Annual CMAS Conference, Chapel Hill, NC, October 7, Obs. M ModMulti PM. PM. SO PM. NH. (μg m ).. Fig. Comparisons of observed, M and ModMultisimulated mean concentrations in winter. Obs. M ModMulti PM. PM. SO PM. NH (μg m ) Fig. Comparisons of observed, M and ModMultisimulated mean concentrations in summer. Sensitivity analyses were conducted in order to investigate factors influencing the model performance for NO simulation. The investigated factors include several CMAQ configurations, temperature and relative humidity, NO X emission, emission, and dry depositions, and N O heterogeneous reaction. The analyses showed considerable sensitivities of emission and and dry depositions to NO concentration. Because there still remain large uncertainties in estimates of emission and and dry depositions, these may be key factors to improve the model performance for NO simulation.. ACKNOWLEDGEMENTS This research was supported by the Environment Research and Technology Development Fund (C) of the Ministry of the Environment, Japan. We are grateful to NMRI and OPRF for providing the ship emission inventories. 7. REFERENCES Byun, D. and K. L. Schere. : Review of the governing equations, computational algorithms, and other components of the models community multiscale air quality (CMAQ) modeling system. Appl. Mech. Rev., 9, 7, Chatani, S., T. Morikawa, S. Nakatsuka, S. Matsunaga, and H. Minoura, : Development of a framework for a highresolution, threedimensional regional air quality simulation and its application to predicting future air quality over Japan. Atmos. Environ.,, 9. Morino, Y., S. Chatani, H. Hayami, K. Sasaki, Y. Mori, T. Morikawa, T. Ohara, S. Hasegawa, and S. Kobayashi, : Intercomparison of chemical transport models and evaluation of model performance for O and PM. prediction Case study in the Kanto area in summer 7. J. Japan Soc. Atmos. Environ.,, (in Japanese). National Institute for Environmental Studies, : National Greenhouse Gas Inventory Report of JAPAN. Neuman, J. A., D. D. Parrish, T. B. Ryerson, C. A. Brock, C. Wiedinmyer, G. J. Frost, J. S. Holloway, and F. C. Fehsenfeld. : Nitric acid loss rates measured in power plant plumes. J. Geophys. Res., 9, D. Skamarock, W. C. and J. B. Klemp. : A timesplit nonhydrostatic atmospheric model for weather research and forecasting applications. J. Comput. Phys., 7,.