Chemical Modelling with CHASER and WRF/Chem in Japan

Abstract

In this paper results of studies with WRF/Chem and CHASER models are presented. The CHemical Atmospheric general circulation model for Study of Atmospheric Environment and Radiative forcing (CHASER) is a global chemical transport model (cf. Sudo et al. J Geophys Res 107(D21):4339, 2002a; Sudo et al. J Geophys Res 107(D21), 4586, 2002b; Takigawa et al., J Geophys Res 110, 2005). The gaseous and aerosol chemistry module is implemented in the CHASER in an on-line mode. CHASER is based on CCSR/FRCGC/NIES AGCM 5.7b, and the meteorology and radiation can be simulated in CHASER itself. The radiative feedback through the distribution of chemical species is taken into account. Daily forecasts have been available on internet since 1 January 2002. This forecasting system was developed for the use of daily flight planning for the Pacific Exploration of Asian Continental Emission (PEACE)-A (January 2002) and PEACE-B (April–May 2002) campaigns. A regional-scale chemical weather forecasting system based on WRF/Chem has been also developed. The lateral boundary for chemical species is taken from the 3-hourly output of CHASER. The modelled surface ozone was compared with the ground-based observations.

Appendix A: One-Way Nested Global Regional Model Based on CHASER and WRF/Chem

We are now also developing a one-way nested global-regional air quality forecasting (AQF) model system with full chemistry based on the CHASER (Sudo et al. 2002a) and WRF/Chem (Grell et al. 2005). Here, description and evaluation of model system are briefly given.

The global chemistry transport model (CTM) part is based on the CHASER model, which is based on CCSR/NIES/FRCGC atmospheric general circulation model (AGCM) version 5.7b. The basic features of the model have been already described in Sect.2. The regional CTM part is based on WRF/Chem (Grell et al. 2005). The databases used are the following:

• Anthropogenic emission data over Japan, except those from automobiles, are from the JCAP (Japan Clean Air Program) with 1×1km resolution (Kannari et al. 2007)

• Anthropogenic emissions from automobiles over Japan are from EAgrid2000 (East Asian Air Pollutant Emissions Grid Inventory) with 1×1km resolution (K. Murano, personal communication)

• Surface emissions over China and North and South Korea are from REAS (Regional Emission Inventory in Asia) with 0.5×0.5° resolution (Ohara et al. 2007)

• Surface emissions over Russia are from EDGAR (Emission Database for Global Atmospheric Research) with 1×1° resolution (Olivier et al. 1996)

Diurnal and seasonal variations in surface emissions are taken into account in the JCAP and EAgrid2000 data, and diurnal variations are also parameterized in emissions from REAS and EDGAR following averaged variations of JCAP. Weekly variation between workdays and holidays is also taken into account in the EAgrid2000 automobile emission data. Note here that emissions based on the statistics in 2000 are applied in the present study. Biogenic emissions are based on Guenther et al. (1993). The outer domain covers Japan with 15km horizontal resolution (152×52 grids for chemical species), and the inner domain covers the Kanto region with 5km resolution (111×111 grids for chemical species). The inner and outer domains in the regional CTM have 31 vertical layers (up to 100hPa). The two-way nesting calculation is applied in the regional CTM part.

The lateral boundary of chemical species in the regional CTM is taken from the global CTM. The output of the global CTM is linearly interpolated from the Gaussian latitude and longitude grid to a Lambert conformal conic projection for use in the regional CTM. The lateral boundary is updated every 3h and linearly interpolated for each time step. We did not include feedback from the regional CTM to the global CTM; that is, the one-way nesting calculation was done between the global and regional CTMs. The system is driven by meteorological data from NCEP for the global CTM part and from the mesoscale model (MSM) of the Japan Meteorological Agency (JMA) for the regional CTM part. A 15-h forecast has been produced four times daily at 00, 06, 12, and 18UTC with a lead time of 8–10h since July 2006, following a spin-up of 1 month for the global distribution of chemical species. The initial condition of the meteorological field for the regional CTM was taken from the MSM for each forecast, and the initial condition of chemical species was taken from the model output driven by the analysis meteorology.

To evaluate the model-calculated ozone, the surface ozone mixing ratio was compared to that observed at air quality monitoring stations. There are 251 stations observing surface ozone within the inner domain of the regional CTM as of August 2006. For the comparison of temporal variation, hourly averaged values of observed and modeled surface O₃ mixing ratios in August 2006 are shown in Fig.17.6. Observed ozone exceeded 100ppbv from 3 to 6 August at Hanyuu in Saitama Prefecture (36°10′28″N, 139°33′21″E, Fig.17.6a), which is downwind of the Tokyo metropolitan area. The maximum value in the observation was 162ppbv at 16UTC on 3 August. The model successfully reproduced the ozone maximum on 3 August. The maximum simulated value was 137ppbv in the model. The model also successfully captured the decrease from 3 to 7 August, but failed to show the rapid decrease on 8 August. Three typhoons (Maria, Somai, and Bopha) occurred during this period, and the difficulty of predicting the meteorological field may have led to the overestimation of ozone on 8 August. Both the model and observations indicate low levels of ozone from 14 to 17 August as typhoon 200610 (Wukong) approached Japan. The observed and modeled ozone exceeded 100ppbv on 11 and 13 August, and the model overestimated the ozone mixing ratio on 19 August. The modeled ozone mixing ratio was 135ppbv, whereas the observed ozone mixing ratio was 86ppbv. Daily variation in the ozone mixing ratio at nighttime was well reproduced by the model. The daily minimum of observed and modeled ozone exceeded 10ppbv on 12, 15, 27, and 28 August; except for these days, the ozone level was almost zero during nighttime. The comparison between the modeled and observed daily variation in surface ozone at Kodaira in Tokyo (35°43′42″N, 139°28′38″) is shown in Fig.17.6b.

Fig.17.6

Hourly observed (solid) and modelled (dashed) surface ozone mixing ratio in August 2006 at Hanyuu in Saitama prefecture (a) and Kodaira in Tokyo (b)

Maxima of observed and modeled surface ozone at Hanyuu appeared on 3 August, and the observed and modeled ozone mixing ratios at Kodaira were 140ppbv or higher on 5 and 6 August. The model tended to overestimate the daytime ozone maximum especially for cloudy days, and the discrepancy of daily maximum is larger in urban area compared to that in rural area. To evaluate the model performance, a set of statistical measures provided by the U.S. Environmental Protection Agency (US EPA 1991) was evaluated for stations in the inner domain of the model. The mean normalized bias error (MNBE), the mean normalized gross error (MNGE), and the unpaired peak prediction accuracy (UPA) were 7.1, 9.5, and 9.4%, respectively. These values are within the criteria range suggested by the U.S. EPA (MNBE< ±10–15%, MNGE< ±30–35%, and UPA< ±15–20%).