At present, more than 60 million people are added to cities each year worldwide, and approximately 325 cities have a population of more than one million, compared to 270 such cities in 1990 [1]. In recent decades, air quality has deteriorated remarkably in the large cities of developing countries. The export of air pollutants from urban to regional and global environments is of major concern because of wide-ranging potential consequences for human health and for cultivated and natural ecosystems, visibility degradation, weather modification, radiative forcing, and changes in tropospheric oxidation (self-cleaning) capacity. Rapidly increasing urbanization will be a major environmental driving force in the 21st century, affecting air quality on all scales (e.g., local, regional, and global) [1]. Large numbers of people are likely to be exposed to unhealthy ozone (O₃) concentrations when ground-level ozone accumulates in urban metropolitan areas under certain weather conditions [2], and quantitative atmospheric dispersion models will be of great help to provide effective decision support systems for planners and administrators.

The main pollutants emitted into the atmosphere in urban areas are sulfur oxides (SOx), nitrogen oxides (NOx), carbon monoxide (CO), volatile organic compounds (VOCs), metal oxides, and particulate matter (PM/aerosols), mostly consisting of black carbon, sulfates, nitrates, and organic matter. The oxidization of CO, VOCs, and NOx produces ozone in the planetary boundary layer (PBL), which has an important impact on human health in urban areas. The recent economic developments of China, Korea, and Japan are very rapid due to the rapid growth of heavy industrial operations. The rapidly growing urbanization in these countries will cause wide-ranging potential consequences in terms of environmental problems. These areas have been suffering from severe air pollution problems for the past decade, such as high particulate matter (PM) concentrations and poor visibility [3,4,5,6]. As industrial activity and the number of automobiles increases, the emission of volatile organic carbons (VOCs) and NOx (NO + NO₂) will also significantly increase. Both VOCs and NOx play critical roles in ozone formation in the troposphere [7,8], and the variations in their concentrations and the ensuing effects on the ozone production rate can be characterized as either NOx-sensitive or VOC-sensitive [7,9,10,11,12]. Thus, obtaining a better understanding of the relationship between ozone precursors (VOCs, NOx) and ozone formation in east Asia is one of the critical pre-required pieces of information necessary to develop effective ozone control strategies [13,14,15].

In Japan, VOC, NOx, and SOx were categorized as hazardous air pollutants (HAPs) in 1996 and are on the list of Substances Requiring Priority Action published by the Central Environmental Council of Japan [16]. The Central Environmental Council published the second report on the “Future direction of measures against hazardous air pollutants” in October 1996. The second report also proposed that voluntary actions to reduce emissions, as well as investigations of hazards, atmospheric concentrations and pollution sources of those substances, should be promoted. Although industrial emissions of VOCs, NOx, and SO_X in Japan have been decreasing in recent years, primarily due to voluntary reductions from industrial sources, the risks of exposure to these pollutants have remained largely unknown. In this study, a rapid change of ozone and its precursors were simulated during the spring of 2008. The simulation results contribute to a better understanding of ozone variability in coastal areas of the Sea of Japan [17]. A three-dimensional chemical/dynamical regional model (WRF/Chem V-3.3) was applied to analyze the causes of the rapid changes in ozone during this period.

Ozone is not emitted directly into the air; instead, it forms in the atmosphere as a result of a series of complex chemical reactions between oxides of nitrogen (NOx) and hydrocarbons, which together are precursors of ozone. Ozone precursors have both anthropogenic (man-made) and biogenic (natural) origins. Motor vehicle exhaust, industrial emissions, gasoline vapors, and chemical solvents are some of the major sources of NOx and hydrocarbons. Many species of trees and other plants emit hydrocarbons, and fertilized soils release NOx [18,19].

In the presence of ultraviolet sunlight (hν), oxygen (O₂) and nitrogen dioxide (NO₂) react in the atmosphere to form ozone and nitric oxide (NO) [18,19]. The resultant ozone reacts with NO to form nitrogen dioxide. A steady state is attained through these reactions:

NO₂ + hν → NO + O

NO + NO₂ + H₂O → 2HONO

HONO + hν → NO + OH

O + O₂ → O₃

CO + 2O₂ → CO₂ + O₃

O₃ + NO → NO₂ + O₂

Even in the absence of anthropogenic emissions, these reactions normally result in a natural background ozone concentration of 25 to 45 ppb [20,21]. Ozone cannot accumulate further unless volatile organic compounds (VOCs), which include hydrocarbons, are present to consume or convert NO back to NO₂ [19,22,23], as in the following:

VOC + NO → NO₂ + other products

This equation is a simplified version of many complex chemical reactions. Because NO is consumed by this process, it is then no longer available to react with ozone. When additional VOCs are added to the atmosphere, a greater proportion of the NO is oxidized to NO₂, resulting in a greater amount of ozone formation. Anthropogenic sources of NO lead to higher levels of NO₂ in the atmosphere, and this NO₂ is then available for photolysis. The formation and increase in ozone concentrations occur over a period of a few hours. Shortly after sunrise, NO and VOCs react in the presence of sunlight to form ozone. Throughout the morning, ozone concentrations increase while NO and VOCs are depleted. Eventually, either the lack of sunlight, NO, or VOCs limits the production of ozone. This diurnal cycle varies greatly depending on site location, emission sources, and weather conditions. Precursor emissions of NO and VOCs are necessary for ozone to form in the troposphere. Understanding the nature of when and where ozone precursors originate may help forecasters factor in day-to-day changes in emissions. For example, if region’s emissions are dominated by mobile sources, then emissions, and thus the ozone that forms from them, may depend on day-to-day commuting patterns. The primary sources of NOx production are combustion processes, including industrial and electric generation processes, and mobile sources such as automobiles [24]. Mobile sources also account for a large portion of VOC emissions. Industries such as the chemical industry or others that use solvents also account for a large portion of VOC emissions. Biogenic VOC emissions from forested and vegetation areas may impact urban ozone formation in some parts of the country. Biogenic NOx emission levels are typically much lower than anthropogenic NOx emission levels [24].

In this study, a preliminary evaluation of the model performance was performed by comparing results to the online observations obtained from different monitoring stations over the past few years (2000–2008) in Japan. The spatial and temporal dynamics were studied using simulations performed with the Weather Research and Forecast (WRF) model [25] coupled with the Chemistry model (WRF/Chem) [26]. Due to the geographical complexity of the study area, an important part of the study was also to evaluate the model’s ability to represent the measured meteorological conditions and surface ozone levels.

The development and occurrence of photochemical pollution episodes have been studied using air quality models (AQM) as they incorporate the contributing atmospheric physical and chemical processes [27,28,29,30,31,32]. The fully coupled weather chemistry model, WRF/Chem, is the next generation model currently used by many researchers for air quality studies [33,34,35,36,37] over different regions. Misenis [35] used WRF/Chem to study the air quality of the Houston–Galveston area and reported that the model planetary boundary layer (PBL) and land surface model (LSM) schemes affected the simulation of chemical species. Their study indicated that the Yonsei University non-local diffusion PBL scheme has given better results for meteorological variables than the Mellor-Yamada-Janjic PBL scheme for the ozone predictions. Jiang [36] studied a continuous photochemical pollution episode in Hong Kong using WRF/Chem to examine the meteorological processes contributing to the formation of high ozone and reported that the northerly air stream associated with high temperatures, stable boundary layer and clear sky conditions favored the high ozone formation in Hong Kong city. De Foy [37] evaluated the WRF model for the complex wind flows in the Mexico City basin area with data from field campaigns using statistical techniques, cluster analysis of flow trajectories and concentration measurements. Their study of ozone showed the influence of the local and regional scale circulations and their modulation by the synoptic–scale flow patterns to govern the short-range transport in the Mexico City region. Tie [38] studied the performance of the WRF/Chem for the simulation of ozone and its precursors in the Mexico City region using in situ aircraft measurements of chemical species. They reported that the model was able to capture the timing and location of the ozone concentrations, their association with city plumes and that the model underestimated the ozone mixing ratios by about 0–25%. Zhang [39] studied the air quality over Mexico City using WRF/Chem and reported that the model performs much better during daytime than nighttime for both chemical species and meteorological variables and different combinations of the available PBL and land surface schemes did not reduce the errors.

Chatani and Sudo [40] observed and analyzed the seasonal variation of ozone concentration in the Japan domain by using the WRF/Chem, CHASER model with monitoring station data during 1996–2005. They tried to evaluate the performance of the WRF/Chem, CHASER model with monitoring data during this period. Masanori [41] evaluated vertical ozone profiles simulated by WRF/Chem using Lidar observation over the Kanto region of Japan on 27–29 July and 16–21 August 2005. Kurokawa [42] showed the influence of meteorological variability on interannual variations of the springtime boundary layer ozone over Japan between 1981–2005. Kurokawa [43] depicted an analysis of episodic pollution of photochemical ozone during 8–9 May 2007 over Japan using the nesting RAMS/CMAQ modeling system. Hayasaki [44] described Episodic pollution of photochemical ozone during 8–9 May 2007 over Japan. Saikawa [45] analyzed the impact of China’s vehicle emissions on regional air quality in 2000 and 2020. It is evident that the early research work on ozone simulation by WRF/Chem in the Japan domain did not show any time series ozone concentration with observation data to find out an ozone episode. However, the use of WRF/Chem to predict the ozone concentration in Japan domain is very rare.

In this study an attempt has been made to examine the evolution of surface ozone and other precursor emissions like NOx over the Coastal Areas of the Sea of Japan using WRF/Chem, an online chemistry model. The performance of WRF/Chem in the simulation of ozone concentration distribution in the region that have occurred during the spring (middle of March) condition and the results from this study could provide useful information about ozone formative meteorological processes to air quality regulatory agencies and health administrators. The spring climate in the study region is characterized with strong land–ocean thermal gradients and the resulting mesoscale sea breeze circulation. The concentration distribution of ozone in 14–16 March 2008 during spring with sufficient observations was selected to assess the model performance.

The Weather Research and Forecasting (WRF) model is a mesoscale numerical weather prediction system designed to serve both operational forecasting and atmospheric research needs. The development of the WRF model has been a collaborative partnership, principally among the National Center for Atmospheric Research (NCAR), the National Oceanic and Atmospheric Administration (NOAA), the National Center for Environmental Prediction (NCEP), the Forecast Systems Laboratory (FSL), the Air Force Weather Agency (AFWA), the Naval Research Laboratory, Oklahoma University, and the Federal Aviation Administration (FAA). The WRF model is a fully compressible and Euler nonhydrostatic model. It calculates winds (u, v, and w), the perturbation potential temperature, the perturbation geopotential, and the perturbation surface pressure of dry air. It can also optionally output other variables, including turbulent kinetic energy, the water vapor mixing ratio, the rain/snow mixing ratio, and the cloud water/ice mixing ratio. The model physics include bulk schemes, mixed-phase physics of cloud-resolving modeling, and multi-layer land surface models ranging from a simple thermal model to full vegetation and soil moisture models. The full vegetation and soil moisture land surface models can include snow cover and sea ice, turbulent kinetic energy predictions or non-local K schemes for planetary boundary layer calculations, long wave and shortwave schemes with multiple spectral bands, and a simple shortwave model. On the other hand, the WRF Preprocessing System (WPS) is used to prepare a domain (region of the earth) for WRF. The WPS consists of three independent programs: geogrid, ungrib, and metgrid. The purpose of geogrid is to define the simulation domains and interpolate various terrestrial data sets to the model grids. The ungrib program reads GRIB files, degribs the data, and writes the data in a simple format, called the intermediate format. The metgrid program horizontally interpolates the intermediate-format meteorological data that are extracted by the ungrib program into the simulation domains defined by the geogrid program. A detailed description of the WRF and WPS model can be found on the WRF website [26].

In addition to a dynamical calculation, a chemical model (WRF/Chem) is fully (on-line) coupled with the WRF model. A detailed description of WRF/Chem is given by Grell et al. [26]. The version of the model (version 3.3) used in the present study includes simultaneous calculations of dynamical parameters (e.g., wind, temperature, boundary layer, and clouds), transport (advective, convective, and diffusive), dry deposition [46], gas phase chemistry, radiation and photolysis rates [47,48], and surface emissions, including online calculations of biogenic emissions. Ozone chemistry is represented in the model by a modified Regional Acid Deposition Model version 2 (RADM2) gas-phase chemical mechanism [49], which includes 158 reactions among 36 species, in conjunction with the Secondary Organic Aerosol Model (MADE/SORGAM) of aqueous reactions [50].

Annual and monthly pollutant concentrations observed at monitoring stations operated by Japanese local governments and the Japanese Meteorological Agency [84,85] (JMA) are available on the website [86]. This site contains data that has been collected since 1970, which are valuable for evaluating the long-term or short-term trends of pollutant concentrations in Japan. Based on this database, the Japanese government has annually reported surface ozone levels in Japan, with the long-term trend showing that these ozone levels have increased over time [87]. The first observation data set of 1,045 monitoring stations that continued to monitor photochemical oxidants between 1996 and 2008 were used in this study to investigate the pattern of ozone concentrations in March and to compare it with the simulation results. Most of the monitoring stations are located in populated coastal areas in Japan, as shown in the left map of Figure 2, because these stations are operated primarily to ensure adherence to environmental quality standards (EQSs). This database has been compiled from surface ozone concentrations recorded as daytime and nighttime averages and from the maximum concentrations of photochemical oxidants. Daytime corresponds to a twelve-hour period from 06:00 a.m. to 06:00 p.m. and time corresponds to a twelve-hour period from 06:00 p.m. to 06:00 a.m. of Japan Standard Time. Photochemical oxidants include other trace oxidants such as H₂O₂ and peroxyacyl nitrates (PANs), but instruments that detect only ozone have been officially approved by the Japanese government because differences in concentrations of photochemical oxidants and ozone are expected to be small [40]. Therefore, differences in concentrations between ozone and photochemical oxidants can be ignored for monthly variations in observed surface ozone concentrations. Additionally, the second observation data set of the Acid Deposition Monitoring Network in East Asia (EANET) was prepared to compare with the WRF/Chem simulation results, whereas the EANET operates monitoring stations located primarily in remote areas. Observed concentrations of surface ozone at 10 EANET monitoring stations in Japan were used to validate the performance of the simulation in Japanese background areas from 2000–2008. The locations of EANET monitoring stations are shown in the middle map of Figure 3. They are scattered in remote areas throughout Japan.

The concentrations of surface ozone at EANET monitoring stations in Japan show seasonal variations, with low values in the summer (August) and higher values in the spring (Middle of March). The average ozone concentrations in March at different EANET monitoring stations were found to be approximately 37 to 55 ppb, compared to approximately 33.5 ppb in populated Japanese areas. Figure 3 shows monthly variations of observed surface ozone concentrations, averaged over 2000–2008 at each of 10 EANET monitoring stations. The red line in Figure 3 shows monthly variations of observed concentrations of surface ozone above populated Japanese areas averaged over 1996–2008.

The two observation data sets have shown that there have been air pollution episodes in the middle of March at most of the monitoring stations, in which the concentrations of ozone, nitrogen dioxide (NOx), volatile organic compounds (VOCs) and particulate matter (PM) are considerably high, hence causing lower visibility [86]. In this paper, a serious and continuous photochemical pollution episode in Japan during the anti-cyclone of 14–16 March 2008 was investigated using the WRF/Chem. Chatani and Sudo [40] observed and analyzed the ozone and other trace gas concentrations, as well as the corresponding weather conditions in this heavy-pollution episode. Our aim was to further understand the physical and chemical mechanisms of this episode by use of the new generation air quality model WRF/Chem. In Section 6.2, the simulation results were carefully compared with observation data and analyzed with the focus on local chemical production and regional transport.

The past several year standard ozone average of two different data sets of different monitoring stations (Figure 3) shows an average monthly variation of observed ozone concentrations. From analysis of air pollution records of the last few years, the monthly average concentration of ozone is higher in March in most of the EANET monitoring stations and the monitoring stations in the populated area. Figure 3 was prepared by the data set of monthly average ozone concentrations from 2000 to 2008 of 10 EANET monitoring stations and of the data set of monthly average concentration from 1996 to 2008 for the 1,045 monitoring stations in the populated area. Due to the trends of higher ozone concentration in March in Japan [40], an attempt has been taken to examine the evolution of surface ozone from 14 to 16 March 2008 by WRF/Chem simulation. The highest ozone concentration occurred from 14 to 16 March 2008, lasting for 2–3 days. These monitoring stations are Rishiri, Tappi, Happo, and Banryuko and are representative of the coastal region; Ijirako and Yusuhara are the representative of the inland area; Sado, Oki, Ogasawara, Hedomisaki are located in the island. Prior to the episode, the surface ozone concentrations were low on 13 March (about 50–60 ppb) and exceeded to 60 ppb at the monitoring stations Rishiri, Tappi and Oki on March 14 and 15 and reached 72 ppb on 14 March at Tappi. The peak higher value of ozone concentration was identified 72 ppb on March 14 at Tappi, which could be called an episode (Figure 4). Figure 4 shows time series (3 hours interval) zone concentrations from 00:00 UTC on 1 March to 00:00 UTC on 31 March of 2008. The black circle on Figure 4 shows the peak ozone concentration (72 ppb). Although these periods do not characterize a very severe ozone episode, they are considered important as the hourly average ozone values at 2 EANET monitoring stations exceeded 60 ppb, a threshold which can seriously affect people suffering from respiratory deficiencies [88,89] and therefore indicate moderate severe ozone pollution for human health. However, the hourly average ozone concentration at all EANET monitoring stations exceeded 50 ppb, which exceeded the air quality standard of Japan for hourly average ozone concentrations [90].

By using the new generation of the regional air quality model WRF/Chem V3.3, a two-day heavy-pollution episode in March 2008 was investigated. Rapid ozone changes from 00:00 UTC on 14 March to 00:00 UTC on 16 March were observed throughout the entire domain of the coastal area of the Sea of Japan. During this 2-day period, the daytime and nighttime concentrations of ozone changed from 50 to 30 ppb throughout the entire domain. To understand the processes controlling this rapid change in ozone concentrations, a three-dimensional regional chemical/dynamical model (WRF/Chem) was applied. The results showed that the calculated trends of ozone concentrations and ozone precursors are consistent with the measured trends. The calculated magnitudes and diurnal variations of ozone were slightly underestimated relative to the measured values in most of the rural areas, while they were overestimated in populated urban areas, suggesting that industrial emissions or emissions from biomass burning in populated urban areas are significantly underestimated in this study and that their estimates need to be improved. The model’s relatively coarse horizontal grid spacing in this study may have led to the overestimation in the O₃ calculation. Therefore, the resolution of the WRF/Chem model in this study may not be high enough to predict the ozone concentrations over populated Japanese areas. Additionally, because the model is capable of calculating rapid changes in ozone, it is suitable for studying the causes of the changes in ozone. The analysis of the model results suggests that weather conditions play important roles in controlling the surface ozone in the coastal areas of the Sea of Japan. Under the winter weather system, the high ozone concentrations, which were chemically formed in the inland areas, were rapidly transported to the downwind region of the coastal areas, resulting in low ozone concentrations in the urbanized region of East Asia. This study illustrates that the WRF/Chem model is a useful tool for studying the high variability of ozone concentrations. The study of ozone variability has important implications for the prediction of ozone concentrations and for control strategies of high ozone events in the coastal areas of the Sea of Japan.