Journal Archive

In Japan, a national standard has been established for acid gases (SO₂ and NO₂) and their concentrations have been monitored by a national monitoring network that consists of approximately 1500 sites throughout Japan (Ministry of Environment, Japan, 2017). Conversely, because NH₃ is not regarded as an air pollutant, as mentioned above, monitoring networks and official observation methods for NH₃ have not yet been established in Japan. In this context, for this study, observations of the NH₃ concentration have been conducted at multiple sites in the Tokyo metropolitan area since April 2015. This manuscript introduces the observational results from April 2015 to March 2017 at the five sites shown in Fig. 1. The sites were located in urban (Shinjuku), rural (Hino and Komae), remote (Tsukui), and agricultural areas (Hiratsuka).

Fig. 1. 
Model domains and locations of the five monitoring sites for atmospheric NH₃ to be compared with the calculated concentrations (1: Shinjuku [urban], 2: Komae [rural], 3: Hino [rural], 4: Tsukui [remote], and 5: Hiratsuka [agricultural]).

In this study, a passive sampler, manufactured by Ogawa Shokai Co., Ltd., was adopted as the observational method for the NH₃ concentration. The Ogawa passive sampler has been used for NH₃ sampling inside and outside of Japan (e.g., Matsumoto et al., 2010; Roadman et al., 2003). The sampling was carried out every two weeks basically. Captured NH₃ on the sampling filter was extracted into pure water (5 mL), and the amount was detected using the Flow Injection Analysis method. The concentration was derived by deducting a blank value on unused filter from the collected NH₃.

In this study, a modeling analysis was performed using CMAQ (Community Multiscale Air Quality; Byun and Ching, 1999) version 4.7.1. The selected chemical reaction scheme was the same as that of Sakurai et al. (2015). The meteorological data in the three-dimensional space was calculated using WRF (Weather Research Forecast model; Skamorock et al., 2008) version 3.7.1. The global objective analysis data (FNL) of the National Centers for Environmental Prediction (NCEP) were used in the WRF simulation as the initial and boundary conditions. In addition, RGT_SST of NCEP was used for the sea surface temperature. The simulation period was from March 2015 to March 2016. As shown in Fig. 1, the modeling domain of WRF had a nesting system with grid resolutions of 45 km (East Asia as Domain 1), 15 km (Japan as Domain 2), and 5 km (the Kanto region as Domain 3). The domain sizes were 3,825×4,950 km² for Domain 1, 1,095×1,200 km² for Domain 2, and 335×335 km² for Domain 3 with the domain center at 36° N and 140° E. The vertical layers consisted of 30 sigma-pressure layers from the surface to 100 hPa with the top height of the lowest layer being approximately 22 m.

Air quality simulations based on CMAQ were conducted only in Domains 2 and 3 due to the limited information regarding the continental emission inventories during the recent years. Inland anthropogenic emissions in Domains 2 and 3 were derived from EAGrid-Japan 2010 (Fukui et al., 2014). NO_x emissions from vehicles were modified by reducing them by approximately 75% according to the reduction rate of the annual averaged concentration of NO_x observed at all Motor Vehicle Exhaust Monitoring Stations in Japan from 2010 (416 sites) to 2015 (413 sites) (Ministry of Environment, Japan, 2017). Ship emissions were derived from the emission inventories developed by the Ocean Policy Research Foundation (2013). In addition, GEIA (Global Emission InitiAtive database) and GFED Ver. 3.1 (van der Werf et al., 2010) were applied for vegetable origin VOCs and biomass burning origins, respectively. The volcanic origin SO₂ emissions were also taken into account in the same way as in Sakurai et al. (2015). The boundary concentration of air pollutants in Domain 2 was derived from MOZART-4 (Model for Ozone and Related chemical Tracers version 4) (Emons et al., 2010).

Fig. 2 introduces the seasonal variation in the NH₃ concentrations observed from April 2015 to March 2017 at the sites indicated in Fig. 1. Because the observations indicated that the level and seasonal variation in the observed concentrations at the Hino (rural), Komae (rural), and Tsukui (remote) sites were nearly the same, the observations at the Komae (rural) and Tsukui (remote) sites were terminated in April 2016.

Fig. 2. 
Seasonal variation in the NH₃ concentrations observed from April 2015 to April 2017 at the sites indicated in Fig. 1.

It appears that the observed concentrations increased toward the warm seasons, and relatively higher concentrations were observed at the Shinjuku (urban) and Hiratsuka (agricultural) sites throughout the study period. Fig. 3 shows the horizontal distribution of the annual NH₃ emission used in the simulation for Domain 3 (5-km grid resolution). The annual emission amounts of NH₃ in each grid were Shinjuku: 4.92 ton km^-2 year^-1, Komae: 4.52 ton km^-2 year^-1, Hino: 3.24 ton km^-2 year^-1, Tsukui: 1.36 ton km^-2 year^-1, and Hiratsuka: 4.60 ton km^-2 year^-1. Fukui et al. (2014) estimated that the annual emission amount of NH₃ in Japan was 404,393 ton year^-1 in 2010 and that the agricultural and human sources contributed 66% and 18% of the total amount, respectively. The large emission amounts at the Shinjuku (urban) and Hiratsuka (agricultural) sites originate from human and agricultural emission sources, respectively. Therefore, for these two sites, the relationship between the emission amount and the observed concentrations was consistent. Conversely, even though the emission amount at the Komae (rural) site, which originated primarily from human sources, was nearly as large as that at the Shinjuku (urban) and Hiratsuka (agricultural) sites, a relatively lower concentration of NH₃ was observed there. The Komae (rural) site is located in a suburb of Tokyo, and the human activities there are substantially different from those near the Shinjuku (urban) site. Therefore, the inconsistency between the emission amount and the concentration at the Komae (rural) site suggests that the estimated emissions around the Tokyo suburb might be overestimated compared to the actual situation.

Fig. 3. 
Horizontal distribution of the annual NH₃ emissions (ton km^-2 year^-1) used in the simulation for Domain 3 (5-km grid resolution).

Seasonal averaged concentrations over the two years at the Shinjuku (urban), Hino (remote), and Hiratsuka (agricultural) sites are summarized in Table 1. The table also shows the proportion of NH₃ emissions for each season compared to the annual amount. In general, the emission strength of NH₃ based on volatilization increases as the temperature rises. Accordingly, the emission ratio reached its maximum in the summer season (from July to September) at all sites. Even though the seasonal averaged concentrations at the Shinjuku (urban) and Hino (rural) sites reached their maximums in the summer season, the concentration at the Hiratsuka (agricultural) site became high not only in the summer but also in the winter (from January to March) and the autumn (from October to December). The inconsistency between the emission and the observed concentration at the Hiratsuka (agricultural) site suggests that there might be uncertainties in the seasonal variation in the NH₃ emitted from agricultural sources.

To examine the model performance for atmospheric ammonia, comparisons between the observed and simulated NH₃ were conducted at the five sites, as shown in Fig. 4, during the period from April 2015 to March 2016. In addition, Table 2 shows statistics of mean value of simulated concentration (mean Sim.), mean value of observed concentration (mean Obs.), normalized mean bias (NMB) and correlation coefficient (r) between the simulation and observation during the period from April 2015 to March 2016. NMB indicates a level of model overestimation (plus value) and underestimation (minus value), and the calculated NMB in this study were ranged from -56% (at Hiratsuka in the cold season) to 170% (at Hino in the warm season).

Fig. 4. 
Comparison between the observed and simulated NH₃ concentrations during the period from April 2015 to March 2016.

It was found that the model overestimated the observed NH₃ concentration in the warm season and underestimated the concentration in the cold season. In particular, the values of NMB in the warm season at Hino (rural) and Komae (rural) sites reached 170% and 70%, respectively. Regarding the simulation result for the Komae (rural) site, the observations and simulation suggest that NH₃ emissions from human sources around the Tokyo suburb might be larger than in reality, especially in the warm season.

As for the Hino (rural) site, there is a consistency in the relationship between the emissions and the observed concentrations, as mentioned in the previous section. However, the simulated concentrations at the Hino (rural) site in the warm season became unexpectedly higher despite the smaller emission amount of NH₃ in its grid. Fig. 5 indicates the ensemble mean of the diurnal variation in the simulated NH₃ in the sampling period from July 21 to August 3, 2015, when the largest overestimations were simulated at all five sites. The simulation was configured so that the emission strength of NH₃ increased in the daytime according to the temperature rise and human activities. Accordingly, the increase of NH₃ concentration was simulated in the late morning at the Shinjuku (urban), Komae (rural), and Tsukui (remote) sites. The simulated concentration at Shinjuku (urban) increased earlier than those at Komae (rural) and Tsukui (remote), and it seems to be derived from the NH₃ emission from vehicles during the morning rush hours.

Fig. 5. 
Ensemble mean of the diurnal variation in the simulated NH₃ in the sampling period from July 21 to August 3, 2015, when the largest overestimations were simulated at each site.

Conversely, the daily maximum concentration of NH₃ was simulated in the early morning at the Hino (rural) and Hiratsuka (agricultural) sites. The simulated high concentration in the early morning was inconsistent with the diurnal variation in the NH₃ emissions. To evaluate why high concentrations appeared in the early morning during the summer season at these two sites, ensemble mean of the spatial distributions of the NH₃ concentration at 5 AM (Japan Standard Time) in the sampling period from July 21 to August 3, 2015, simulated by WRF/CMAQ, are illustrated in Fig. 6. In addition, Fig. 7 indicates a comparison between the observed and simulated wind direction in the same period at the nearest AMeDAS (Automated Meteorological Data Acquisition System) station to the Hino (rural) site, which is located in Hachioji city and is approximately 5 km west-northwest of the Hino site. The simulated wind direction at Hachioji showed reasonable agreement with the observations and it was found that a land breeze (north wind) generally prevailed in the middle of the night.

Fig. 6. 
Ensemble mean of the spatial distributions of the NH₃ concentration at 5 AM (Japan Standard Time) in the sampling period from July 21 to August 3, 2015.

Fig. 7. 
Comparison between the observed (Obs) and simulated (Sim) wind direction from July 21 to August 3, 2015 (JST) at the nearest AMeDAS station to the Hino (rural) site, which is located in Hachioji city and is approximately 5 km west-northwest from the Hino site.

Sakurai et al. (2003) also reported that higher concentrations of NH₃ were observed at night in Tokyo in the summer of 2002. As shown in Fig. 3, a large emission area of NH₃ exists in the northern part of the Kanto region originating from agricultural sources. Because a land breeze (north wind) generally prevails at night in the Kanto region, Sakurai et al. (2003) concluded from the simulation analysis that the higher concentration of NH₃ observed in Tokyo at night was caused by the transportation of NH₃ emitted in the northern part of the Kanto region. Accordingly, Fig. 6 shows that the NH₃ emitted in the northern part of the Kanto region was transported south by the land breeze and that the concentrations at Hino (rural) and Hiratsuka (agricultural) sites were affected by the transported NH₃. In addition, vertical turbulence is generally reduced due to the cooling of the land surface at night. This likely leads to a low planetary boundary layer as well as a reduction in the vertical mixing. Therefore, it is also suggested that the higher concentration of NH₃ in the early morning at the Hino (rural) and Hiratsuka (agricultural) sites occurred due to the reduced dilution of the local NH₃ emission and the transported NH₃ in the simulated lower planetary boundary layer in the early morning.

Hayami (2013) reported the seasonal variation in the NH₃ emission strength from agricultural sources, which was modified considering the fertilization time in Japan. In its estimation, the emission amounts in the warm and cold seasons were decreased and increased, respectively. According to the model overestimation at the Hino (rural) and Hiratsuka sites in the summer as shown in Fig. 4, it is suggested that the agricultural NH₃ emission in the northern part of the Kanto region might be larger than the actual emission in the summer season. Moreover, the simulated concentration underestimated the observations in the cold season (October 2015 to March 2016) at the Hiratsuka (agricultural) site (NMB=- 56%). As a result, it is suggested that there might not actually be a large difference in the agricultural NH₃ emissions between seasons. The discussion here supports the seasonal variation in the NH₃ emission estimated by Hayami (2013).