Tropospheric ozone is a major air pollutant and impacts human health and vegetation (HTAP, 2010; Yue and Unger, 2014). It is also an important greenhouse gas (IPCC, 2013). Tropospheric aerosols are also air pollutants and aggravate respiratory conditions (HTAP, 2010). Tropospheric aerosols also enhance radiative forcing in a negative (sulfuric acid particles) or positive (black carbon) way (IPCC, 2013), and they affect remote sensing, for example the measurement of greenhouse gases from space (Houweling et al., 2005; Uchino et al., 2012a). It is therefore very important to monitor tropospheric ozone and aerosols and to understand their temporal and spatial variations. On the one hand, aerosols transported from East Asia to western Japan have been observed by lidar, and their vertical distributions have been reported (Iwasaka et al., 1988: Murayama et al., 2001; Hara et al., 2009). On the other hand, ozone from Asia has been studied mainly via surface measurements (Akimoto et al., 1996; Yamaji et al., 2006). Continuous ozone vertical distributions measured by differential absorption lidar (DIAL) are very useful for studying transport processes and the origin of the ozone. To validate products of the Greenhouse gases Observing SATellite (GOSAT), we developed a two-wavelength (532 and 1064 nm) polarization lidar (hereafter referred to as Mie lidar) to observe vertical profiles of tropospheric and stratospheric aerosols and thin cirrus clouds at the National Institute for Environmental Studies (NIES), Tsukuba (36.05∘ N, 140.13∘ E), Japan, in 2009. In 2010 we also developed a DIAL to measure tropospheric ozone profiles (hereafter abbreviated as ozone DIAL). The ozone DIAL was installed in a container with the Mie lidar. In March 2011, we moved the lidar container to Saga (33.24∘ N, 130.29∘ E) in the Kyushu district of western Japan at a location 2.6 m above sea level. The ozone DIAL was modified in September 2012 (Uchino et al., 2014).

Mie lidar has been used to demonstrate the influence of high-altitude aerosols and cirrus clouds on the GOSAT product of the column-averaged dry air mole fraction of carbon dioxide (XCO2) retrieved from the Thermal And Near infrared Sensor for carbon Observation-Fourier Transform Spectrometer (TANSO-FTS) short-wavelength infrared (SWIR) spectral data onboard GOSAT. The XCO2 data were improved by taking the vertical profiles of aerosols and cirrus clouds measured by Mie lidar into account (Uchino et al., 2012a). The increases in stratospheric aerosols caused by the 2009 Sarychev eruption and the 2011 Nabro eruption were observed by Mie lidar (Uchino et al., 2012b).

Ozone DIAL has been used to validate the GOSAT ozone product retrieved from TANSO-FTS thermal infrared (TIR) spectral data (Ohyama et al., 2012), to observe ozone concentrations in the lower troposphere, and to compare the observed concentrations with those predicted by the Meteorological Research Institute Chemistry-Climate Model, version 2 (MRI-CCM2) (Deushi and Shibata, 2011). Use of Mie lidar and ozone DIAL will facilitate satellite product validation not only for GOSAT but also for upcoming satellites such as the TROPOspheric Monitoring Instrument (TROPOMI; Veefkind et al., 2012) and the Geostationary Environment Monitoring Spectrometer (GEMS; Bak et al., 2013). High-ozone episodes in the lower troposphere have been observed by lidar (Banta et al., 1998; Koutidis et al., 2002; Ancellet and Ravetta, 2005; Eisele and Trickl, 2005; Kuang et al., 2011). Those observational records were limited to at most 1 week. We made an 11-day continuous record during 20–31 March 2015.

In this paper we report an event during which high concentrations of ozone and aerosols were observed almost simultaneously below an altitude of 1.5 km over Saga on 22 March 2015 by Mie lidar and ozone DIAL. That event substantially impacted surface air quality. We also compared the observational results with those simulated by the models.

Mie lidar and ozone DIAL were installed in a container with dimensions of about 228 cm (width) by 683 cm (length) by 255 cm (height), as shown in Fig. 1. Mie lidar is a two-wavelength (532 and 1064 nm) polarization lidar based on a neodymium-doped yttrium–aluminum–garnet (Nd:YAG) laser; the characteristics are summarized in Table 1. The output energy at 532 and 1064 nm was 130 mJ, with a pulse repetition rate of 10 Hz. The diameter of the receiving telescope was 30.5 cm. The output signals from the photomultiplier tubes (PMT) and a silicon avalanche photodiode (APD) were processed by transient recorders with a 12 bit analog to digital converter and a photon counter.

The data analysis methods of Mie lidar and ozone DIAL have been described by Uchino et al. (2012b) and Uchino et al. (2014), respectively. We summarize the observation parameters obtained by Mie lidar. The backscattering ratio R is defined as R=(BR+BA)/BR, where BR and BA are the Rayleigh and Mie backscattering coefficients, respectively. Backscattering ratio profiles were derived by the inversion method (Fernald, 1984). The reference altitude was usually set between 9 and 12 km, where only molecular backscattering could be assumed in the absence of clouds. We assumed the lidar ratio LR (extinction-to-backscatter ratio) for aerosols to be 50 sr at 532 nm and 45 sr at 1064 nm based on the lidar ratios for Asian dust and pollutant aerosols summarized by Sakai et al. (2003), Anderson et al. (2003), and Cattrall et al. (2005). The following is a summary of their lidar ratios at 532 nm:

Asian dust 47 ± 18 sr (Sakai et al., 2003);

The total volume depolarization ratio D was defined as D=S/(P+S)×100(%), where P and S are the parallel and perpendicular components of the backscattered signals, respectively. The particle depolarization ratio Dp was obtained from the equation Dp=(D×R-Dm)/(R-1), where Dm is the atmospheric molecular depolarization ratio. We used a Dm value of 0.37 % for this lidar system; we calculated Dm from the spectral transmission data of the interference filter at 532 nm and the Rayleigh backscattering cross sections (Sakai et al., 2003). The value of Dp indicates whether the particles are spherical or non-spherical; large values indicate the presence of non-spherical particles. The backscatter-related Ångström exponent, Alp, the qualitative indicator of aerosol particle size, is defined by BA(λ)∝λ-Alp, where λ is the wavelength. Larger values of Alp indicate the predominance of smaller (i.e., submicrometer-sized) particles. The vertical resolution of these observational parameters was 150 m, and the time resolution was set to be 1 h for comparison with the Model of Aerosol Species in the Global Atmosphere (MASINGAR)-mk2 (Yukimoto et al., 2012). The lowest altitude of Mie lidar measurement was 225 m due to the imperfect overlap of the transmitter–receiver optical axes of the lidar system.

The ozone DIAL consisted of a Nd:YAG laser and a 2 m long Raman cell filled with CO2 gas that generated four Stokes lines from stimulated Raman scattering by CO2; the characteristics are summarized in Table 2. In this study, we used three Stokes lines (276, 287, and 299 nm). The output energies of these Stokes lines were about 8–9 mJ per pulse, with a pulse repetition rate of 10 Hz. The receiving telescope diameters were 10 cm for boundary layer ozone measurements and 49 cm for free-tropospheric-ozone measurements. The Mie lidar and ozone DIAL were synchronized by two pulse-delay generators.

The 276/287 nm and 287/299 nm wavelength pairs were used for ozone DIAL measurements in the altitude ranges 0.57–2.0 and 2.0–6.0 km, respectively. The effective vertical resolutions were 270 m for 0.57–2.0 km and 540 m for 2.0–6.0 km, respectively (Uchino et al., 2014). The time resolution was set to 1 h to facilitate comparison with MRI-CCM2. An aerosol correction was not made for ozone retrieval. Next, we report the continuous lidar observational results made at Saga from 20 March to 31 March 2015.

Figure 2a shows a time–altitude cross section of ozone volume mixing ratios observed by DIAL at Saga from 11:10 JST on 20 March to 14:33 JST on 31 March 2015. Lidar observations were not obtained from 15:56 JST on 27 March to 21:58 JST on 29 March 2015, mainly because conditions were rainy or cloudy. We made quality checks of the DIAL data. The gray regions in Fig. 2a correspond to areas where there were no observational data or the errors were larger than 10 %. The errors were computed from the lidar signal-to-noise ratios by use of Poisson statistics. Regions surrounded by a black rectangle are areas where the data were affected by aerosols and/or clouds with R larger than 2 at 299 nm. We calculated R assuming LR = 50 sr, without correcting for attenuation by ozone absorption. In the lowest row of Fig. 2a, we show hourly data of surface oxidant volume mixing ratios (Ox) at Takagimachi in Saga measured by the Saga Prefectural Environmental Research Center (https://www.pref.saga.lg.jp/web/at-contents/kankyo1/shisetsu/_40810/_41304/_67819.html). Takagimachi is located about 2.8 km northeast of the ozone DIAL site. Because the surface Ox was observed by a UV photometer, the contribution of other components such as peroxyacetyl nitrate (PAN) to oxidant concentrations was extremely low, and the oxidant volume mixing ratio was considered to be that of ozone.

Figure 2a indicates that the ozone volume mixing ratios measured by DIAL were usually about 50–70 ppbv during the study period. Comparatively high ozone concentrations, > 75 ppbv, were detected at altitudes of 0.57–3 and 0.57–2 km on 20–23 March and 30–31 March, respectively. Notably high ozone volume mixing ratios of 90–110 ppbv at altitudes of 0.57–1.5 km were observed from 03:00 to 20:00 JST on 22 March. These high ozone concentrations were also seen in the surface-photochemical-oxidants data, i.e., the Ox equaled 92–101 ppbv from 15:00 to 21:00 JST on 22 March, as shown in the lowest row in Fig. 2a. The maximum concentration of Ox was 101 ppbv at 16:00 JST. This maximum value was far above the environmental quality standard of 60 ppbv for hourly photochemical oxidants in Japan (https://www.env.go.jp/en/air/aq/aq.html).

Figure 3a, b, and c show time–altitude cross sections of the backscattering ratio (R), the total volume depolarization ratio (D), and the particle depolarization ratio (Dp), respectively, observed by Mie lidar at Saga from 09:24 JST on 20 March to 14:34 JST on 31 March 2015. Mie lidar data were not obtained from 15:56 JST on 27 March to 21:58 JST on 29 March 2015, mainly because conditions were rainy or cloudy. We made quality checks of Mie lidar data. Gray regions are areas where there were no observational data or the data were affected by clouds.

Aerosol layers with R in the range 2–4 almost always existed below an altitude of 2.5 km during 20–31 March 2015. An event of high aerosol loading with large values of R (> 8) was observed below altitudes of 1.5 km during 03:00–21:00 JST on 22 March, when the values of D were small (the mean ± standard deviation was 3.9 ± 2.1 %) compared with those before and after the event. The values of D were larger than 7.9 ± 2.1 % from 15:00 JST on 21 March through 15:00 JST on 23 March, except for 03:00–21:00 JST on 22 March. The main aerosol component during the event might have been submicrometer-sized spherical particles, because Dp was small (4 ± 2 %), and the wavelength exponent Alp was large (1.3 ± 0.3). In contrast, the main aerosol particles before and after the event may have been supermicrometer-sized, nonspherical mineral dust particles because Dp was comparatively large (13 ± 3 %) and Alp was 1.0 ± 0.2 (Sakai et al., 2003; Cattrall et al., 2005). When there were no clouds above, R at 1064 nm was estimated assuming Alp = 1.5 at the reference altitude, where very small amounts of aerosols were expected to be present, i.e., R = 1.06 ± 0.06 (D = 1.2 ± 0.5) at 532 nm in the altitude range 3–6 km. If the value of Alp was changed from 1.0 to 2.0 at the reference altitude, the uncertainty in Alp was estimated to be ±0.2. Alp was 0.3–2.0 in the 11-day Mie lidar record. The maximum errors of D and Dp were 0.1 and 2 % for R>2 at 532 nm.

During the same time period, high aerosol concentrations were also observed at the surface (Fig. 4). Hourly values of the mass concentrations of particulate matter with a diameter of 2.5 µm or less (PM2.5) at Takagimachi measured by the Saga Prefectural Environmental Research Center were 23 µg m-3 at 10:00 JST and increased up to a maximum value of 110 µg m-3 at 15:00 JST on 22 March; the concentrations were greater than 82 µg m-3 during 13:00–16:00 JST and decreased to 17 µg m-3 at 01:00 JST on 23 March. The daily mean value of PM2.5 was 50.6 µg m-3 for 24 h on 22 March at Takagimachi, larger than the environmental quality standard of 35 µg m-3 in Japan (https://www.env.go.jp/en/air/aq/aq.html).

Figure 7 shows the time–altitude cross sections of total aerosol extinction coefficients at 550 nm, and the ratios of dust extinction coefficients to total aerosol extinction coefficients simulated by MASINGAR-mk2 with potential temperatures over Saga for 20–31 March 2015. For the event on 22 March, the model predicted dust particles (about 60–100 %) in the altitude range 1–3 km, and it predicted sulfate (about 40–60 %) and dust (about 30–40 %) particles below 1 km. The numbers in parentheses indicate the ratio of each component's extinction coefficient to the total extinction coefficient. The dust particles descended to the surface in the afternoon (Fig. 7b). For the event on 30 March, MASINGAR mk-2 predicted dust particles (about 50–100 %) at altitudes of 1–6 km, and it predicted sulfate (about 50–80 %) and dust (about 0–20 %) particles below 1 km in the morning. Mie lidar data supported the model prediction because Dp was high (17 ± 6 %) at altitudes of 1–3 km and low (10 ± 3 %) below 1 km. For both events, small amounts of organic carbon, black carbon, and sea salt particles were predicted.

To identify the origin of the aerosols and related transport processes, three-dimensional backward trajectories of air parcels were calculated with the NOAA Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPRIT) model (Draxler and Hess, 1998; Stein et al., 2015). Air parcels were initially left at altitudes of 1500 m (Fig. 8a) and 500 m (Fig. 9a) over the lidar site at Saga. The trajectories were calculated for 3 days from 21:00 UTC on 21 (06:00 JST on 22) March 2015. Figures 8b and 9b show the time–altitude cross sections of dust and sulfate extinction coefficients simulated by MASINGAR mk-2 along the trajectory paths of Figs. 8a and 9a, respectively. Based on the results of the backward trajectories and the model simulations, the dust and sulfate particles on 22 March could have been transported within about 2 days from the Gobi Desert and the North China Plain (NCP), respectively, to the measurement site. The MASINGAR mk-2 simulation suggested that the dust particles emitted during 18:00–24:00 UTC on 19 March around 40∘ N and 105∘ E were responsible for the dust storm captured by the Mie lidar observation. The highest concentrations of SO2 and NO2 in the world were observed in the NCP during 2013–2015 by OMI onboard NASA's Aura satellite, based on Fig. 5 of Krotkov et al. (2016). These gases are important precursors of sulfate particles and ozone. Figure 10 shows horizontal maps of ozone volume mixing ratios at 925 hPa (an altitude of about 760 m) simulated by MRI-CCM2 at 21:00 JST on 19, 20, and 21 March and at 03:00 JST on 22 March 2015. These maps indicate that the high ozone concentrations could have been transported from the NCP to the Yellow Sea and then to Saga within about 2 days.

Because it was difficult to obtain observational data of surface ozone and sulfate particles over the NCP, including Beijing, on 19–20 March, we referred to the following papers with respect to those data. According to the ozonesonde measurements made by Wang et al. (2012), ozone concentrations ≥ 90 ppbv were observed over Beijing, China, in late March. Ma et al. (2016) reported a significant increase in surface ozone from 2003 to 2015 at Shangdianzi (40.65∘ N, 117.10∘ E), which is located about 100 km northeast of suburban Beijing, and the maximum daily average 8 h concentrations of ozone appear to have been > 100 ppbv in March 2015 based on Fig. 2 in their paper. High PM2.5 and submicron aerosol concentrations have been observed in Beijing (Zhang et al., 2013; Sun et al., 2015). Ozone and aerosol concentrations may therefore have been high in March 2015 over the NCP.

To elucidate the vertical transport processes of the aerosol and ozone in the lower troposphere over the measurement site, we show in Fig. 11 the time variations of the heights of the mixed layers from 2 hours after sunrise to 2 hours before sunset during 09:24 JST on 20 March through 14:34 JST on 31 March 2015. These altitudes were estimated from (1) the 1064 nm range-corrected backscatter signals, with a range resolution of 15 m and a time interval of 20 min, using the wavelet covariance transform method (Baars et al., 2008; Izumi et al., 2017), and (2) signals obtained from radiosonde data at Fukuoka and the JMA mesoscale analysis (MA) data over Saga using the parcel method (Holzworth, 1964). When the mixed layers developed in the afternoon, the heights of the mixed layers (1.5–2 km) estimated by Mie lidar were similar to those estimated by MA. Although the radiosonde data at 09:00 JST on 22 March found the height of the mixed layer to be 117 m (Stull, 1988), it was difficult for Mie lidar to detect the mixed layer because the lowest altitude of the Mie lidar measurement was 225 m.

The dust particles that originated from the Gobi Desert arrived at altitudes of 1–3 km over the lidar site at 06:00 JST on 22 March. When the mixed layer developed to 1.5–2 km at 11:00–15:00 JST on 22, the dust particles were assumed to be mixed into the boundary layer and then to reach the surface by entrainment, as simulated in Fig. 7b. This may explain the sharp increase in PM2.5 concentrations at the surface after 11:00 JST, as shown in Fig. 4. A similar phenomenon was observed over the northern Kyushu area during the dust event in late May–early June 2014 (Uno et al., 2016). A similar high-surface-ozone event was observed by eight ozonesonde measurements during 6–9 June 2003 over the Seoul metropolitan region (Oh et al., 2010).

By using ozone DIAL and a two-wavelength polarization (Mie) lidar, we made continuous measurements of ozone and aerosol concentrations over Saga during 20–31 March 2015. High ozone and high aerosol concentrations that occurred nearly simultaneously were observed in the altitude range 0.5–1.5 km from 03:00 to 20:00 JST on 22 March 2015. The ozone volume mixing ratio was larger than 100 ppbv. The aerosol extinction coefficient and AOD at 532 nm were larger than 0.5 km-1 and 1.5, respectively.

Backward trajectory analysis and the simulations by the MASINGAR mk-2 and MRI-CCM2 models indicated that mineral dust particles from the Gobi Desert and an air mass with high ozone and aerosol (mainly sulfate) concentrations that originated from the North China Plain could have been transported over the lidar site within about 2 days. Based on the lidar and surface measurement data and the simulation by the MASINGAR-mk2, there is a possibility that the air mass with high ozone and aerosol concentrations could have been transported from the lower troposphere to the surface by vertical mixing when the planetary boundary layer developed in the afternoon of 22 March 2015. The combination of ozone DIAL measurements with surface in situ ozone measurements is very useful for studying the process of descent of high ozone concentrations in the lower troposphere to the surface and the impacts on surface air quality. Such measurements of pollution plumes that descend from the free troposphere to the surface are highly recommended (HTAP, 2010).

MRI-CCM2 could approximately reproduce the high-ozone volume mixing ratios after some modifications of physical and chemical parameters. MASINGAR mk-2 successfully predicted high-aerosol-concentration events, but the predicted peak AOD was about one-third of the observed AOD. For further improvement of these models, it will be important to continue comparing these models with ozone DIAL, Mie lidar, and surface in situ ozone and particle measurements.

The reference and website for the data we used in the paper are provided in the text. The lidar and simulation data presented in the paper are available from the authors upon request (uchino.osamu@nies.go.jp).