Understanding Temporal Variations of Atmospheric Radon-222 around Japan Using Model Simulations

Abstract

Temporal variations of atmospheric radon-222 (²²²Rn) observed at four Japan Meteorological Agency stations in Japan by the Meteorological Research Institute were analyzed using an online Global Spectral Atmosphere Model-Transport Model (GSAM-TM). Monthly and diurnal variations and a series of synoptic high-²²²Rn events were extracted from 5 years to 12 years of ²²²Rn observations during 2007–2019. Observed seasonal patterns of winter maxima and summer minima, driven mainly by monsoons, were well reproduced by the GSAM-TM based on existing ²²²Rn emission inventories, but their absolute values were generally underestimated, indicating that our understanding of ²²²Rn emission processes in East Asia is lacking. The high-resolution model (∼ 60 km mesh) demonstrated that observed consecutive high-²²²Rn peaks at several-hour timescales were caused by two ²²²Rn streams from different regions and were not well resolved by the low-resolution model (∼ 200 km mesh). GSAM-TM simulations indicate that such cold-front-driven events are sometimes accompanied by complicated three-dimensional atmospheric structures such as stratospheric intrusion over the front, significantly affecting distributions of atmospheric components. A new calculation approach using hourly ²²²Rn values normalized to daily means was used to analyze the diurnal ²²²Rn cycle, allowing diurnal cycles in winter to be extracted from ²²²Rn data that are highly variable due to sporadic continental ²²²Rn outflows, which tend to obscure the diurnal variations. Normalized diurnal cycles of ²²²Rn in winter are consistent between observations and model simulations, and seem to be driven mainly by diurnal variations of planetary boundary layer height (PBLH). These results indicate that ²²²Rn in the near-surface atmosphere, transported from remote source regions, could vary diurnally by up to 10 % of the daily mean mainly owing to local PBLH variations, even without significant local ²²²Rn emissions.

1. Introduction

Atmospheric concentrations of greenhouse gases (GHGs) have been monitored continuously by the Japan Meteorological Agency (JMA) since the early 1990s under the Global Atmosphere Watch program of the World Meteorological Organization (WMO/GAW) (Watanabe et al. 2000). Atmospheric GHG mole fractions vary over a range of timescales due to atmospheric transport and chemistry as well as their sources and sinks at the surface. To determine synoptic-scale transport of GHGs and air pollutants from land regions observationally, the monitoring of atmospheric radon-222 (²²²Rn) has been undertaken at JMA stations since the 2000s. Because ²²²Rn is a radioactive noble gas emitted almost exclusively from land at an approximately known emission rate and decays with a known half-life (3.82 d), it is an excellent tracer of atmospheric transport of gases released from land, aiding our understanding of the atmospheric variability of GHGs with land-based sources and sinks (e.g., Biraud et al. 2000). Since the beginning of the 21st century, East Asia has become the predominant emitter of GHGs worldwide because of rapid economic growth, particularly in China (Friedlingstein et al. 2020; Saunois et al. 2020; Tian et al. 2020), increasing the importance of ²²²Rn observation in the region as an indicator for GHGs land emission signals. There have been several studies of ²²²Rn observations in East Asia (Zahorowski et al. 2005; Moriizumi et al. 2008; Wada et al. 2012, 2013), all of which have indicated that continental ²²²Rn emissions and monsoon-driven transport seasonality play major roles in causing atmospheric ²²²Rn variations in the region.

Radon-222 is also utilized widely in validating atmospheric transport models through model–observation comparisons. In several studies, a semihomogeneous annual mean ²²²Rn emission distribution (Jacob et al. 1997) has been used in atmospheric ²²²Rn simulations (e.g., Taguchi et al. 2011; Belikov et al. 2013), generally reproducing atmospheric ²²²Rn variations globally. However, ²²²Rn emissions are highly affected by the spatiotemporal variations of its source, radium (²²⁶Ra), in soil and soil water content, as considered in some ²²²Rn emission estimations (Conen and Robertson 2002; Hirao et al. 2010a). Model simulations based on thus estimated emissions still differ from observations (e.g., Zhang et al. 2021), with the understanding of ²²²Rn emission processes requiring improvement, although emission data are sometimes modified by inverse analyses as needed (Hirao et al. 2010b).

The combination of ²²²Rn observations and model simulations can be useful in elucidating various-scale atmospheric processes common to all atmospheric components. Most studies reported on synoptic to monthly variations using global models (e.g., Wada et al. 2013; Zhang et al. 2021) or on diurnal to synoptic variations using regional models (e.g., Chevillard et al. 2002; Chambers et al. 2019). Ishijima et al. (2018) analyzed synoptic ²²²Rn peaks observed and simulated by multiple-resolution versions of an atmospheric general circulation model (AGCM), determining the scale of atmospheric processes required to produce high-²²²Rn peaks and also revealed that the distribution of atmospheric components around cold fronts is generally explained by equivalent potential temperature. There have been no studies to date involving the analysis of diurnal, synoptic, and monthly variations of atmospheric ²²²Rn concentrations based on decadal observation data using a single global model. In this study, temporal variations of atmospheric ²²²Rn concentrations observed at four stations in Japan, including a new remote station, were analyzed on diurnal, synoptic, and monthly timescales with the aid of an AGCM-based transport model. We also attempted a new approach to better extract diurnal cycles from highly variable ²²²Rn data due to sporadic continental ²²²Rn outflow events.

2. Observations

2.1 ²²²Rn observation stations

Atmospheric ²²²Rn concentrations were observed at four JMA stations: three monitoring atmospheric GHGs (WMO/GAW)—Ryori (RYO; 39°N, 142°E), Minamitorishima (MNM; 24°N, 154°E), and Yonagunijima (YON; 24°N, 123°E)—and a non-GAW weather observation station—Chichijima (CCJ; 27°N, 142°E)—with the latter three being on islands with the same names. Locations are shown in Fig. 1 and Table 1. ²²²Rn observation data from RYO, MNM, and YON have been used in previous studies (Wada et al. 2010, 2012), but that from CCJ is used for the first time in this study.

Fig. 1.

(a) Distributions of constant ²²²Rn emissions around Japan (Jacob et al. 1997), and (b, c) of monthly emissions (Hirao et al. 2010a) in January and July. Climatological wind-field of the T319 model first layer and mean sea-level pressure for 2007–2019 are shown by vectors and contours, respectively, together with locations of the ²²²Rn observation stations used in this study (white triangles).

Station RYO is on a hilly cape on the Pacific coast in the northeast of the main island of Japan. Westerly winds predominate throughout the year, with an annual mean wind speed of ∼ 4 m s⁻¹. Yonagunijima island is the westernmost territory of Japan, and is 350 km and 100 km east of China and Taiwan, respectively. The station is ∼ 500 m inland in the northeast of the island, and it is surrounded by grassland and sugarcane fields. Prevailing winds are from the NNE–NE or S. Minamitorishima island is the southeastern-most territory of Japan in the Pacific Ocean, ∼ 2000 km SE of Tokyo. It is a very small, isolated island formed by a coral reef. The station is near the east coast of the island, which has an easterly prevailing wind and is an ideal background air monitoring station. Chichijima island is ∼ 1000 km SSE of Tokyo, in the Pacific Ocean. The station is on top of a mountain near the NW coast of the island, exposed almost exclusively to maritime air. Air intakes are installed at ∼ 20 m above ground level (a.g.l.) on top of towers at RYO, MNM, and YON, and at ∼ 15 m a.g.l. at CCJ. The stations have different characteristics, including near-perfect remote stations (MNM and CCJ), a continental out-flow regional station (YON), and a local station with the strong local ²²²Rn source (RYO).

Radon-222 observation data are available on the website of The World Data Centre for Greenhouse Gases (WDCGG, https://gaw.kishou.go.jp/).

2.2 Method

A high-precision ²²²Rn measurement system, developed by the National Institute of Advanced Industrial Science and Technology and the Meteorological Research Institute (MRI) of Japan (Wada et al. 2010), was used to continuously monitor atmospheric ²²²Rn concentrations. The system employs the electrostatic collection method, which collects positively charged progenies of ²²²Rn in sample air on a PIN photodiode with a high negative voltage applied, and detects alpha particles emitted by progeny nuclides. Data obtained every 10 min were averaged for 1 hour at 1 hour interval. The detection limit is 0.16 Bq m⁻³ for the hourly mean data. Analyzers are calibrated by cross-checking with an analyzer calibrated at Nagoya University, Japan (Iida et al. 1991, 1996).

3. Model

3.1 Model description

We used a newly developed online Global Spectral Atmosphere Model–Transport Model (GSAM-TM). It is a spectrum model, employing a reduced Gaussian grid (Juang 2004) to simulate physical quantities in nearly homogeneous grid sizes globally to reduce computation cost. Tracer transport is driven by the JMA operational global numerical weather prediction model (December 2012 version) (Japan Meteorological Agency 2013), which is also used to produce the JRA-55 Reanalysis (Kobayashi et al. 2015). The model has a semi-implicit semi-Lagrangian framework with improved vertical mass conservation and computational cost through the separate adoption of horizontal (Gravel and Staniforth 1994) and vertical (Yoshimura and Matsumura 2003) advection schemes. Only parts of sub-grid-scale tracer vertical mixing and transport are imported from the Meteorological Research Institute Chemistry–Climate Model, version 2 (MRI–CCM2) (Deushi and Shibata 2011), including the prognostic Arakawa–Schubert cumulus parameterization scheme (Arakawa and Schubert 1974) and the level-2 turbulence closure scheme (Mellor and Yamada 1974) to simulate deep cumulus convection and vertical turbulent mixing, respectively.

We used two different horizontal resolution versions, TL95 (1.875° × 1.875°) and TL319 (0.56° × 0.56°), for the maximum wavenumber of triangular truncation to investigate the horizontal resolution dependency of reproducibility for observed atmospheric ²²²Rn concentrations (referred to hereafter as TL95 and TL319). The model has 60 layers of sigma-pressure hybrid vertical coordinates from the surface to the 0.1 hPa level (Simmons and Burridge 1981). The layer thickness almost linearly increases with altitude, being roughly 30–300 m below 2 km altitude and 300–800 m for 2–10 km attitudes. The horizontal wind fields are nudged toward those of the JRA-55 Reanalysis (Kobayashi et al. 2015) using a simple Newtonian relaxation technique (Tanaka et al. 2003). In this study, the nudging uses a relatively short relaxation time of 0.1 d for all model layers, considering that the base model of GSAM-TM (especially the TL319 version) is almost the same as that of JRA-55.

3.2 ²²²Rn emission data

Two different types of ²²²Rn emission data were used to elucidate the impact of emission differences on simulated atmospheric ²²²Rn concentrations: (1) emission of 1 atom cm⁻² s⁻¹ at 60°S–60°N land, 0.005 atoms cm⁻² s⁻¹ at 50°S–60°N ocean, 70–60°S, and 60–70°N, and zero from other areas (Jacob et al. 1997) (Fig. 1a); (2) monthly emission based on spatiotemporal variations of soil ²²²Rn emissions with soil water content, texture, and temperature, as well as the ²²⁶Ra content of soil particles (Hirao et al. 2010a) (Figs. 1b, c). Hereinafter, these emissions are referred to as “constant” and “monthly” emissions, respectively. To make the constant emission map, the original landsea mask distribution of the GSAM-TM was used. To make the monthly emission map, the original 1° × 1° resolution data by Hirao et al. (2010a) were regridded for TL95 and TL319 resolutions, precisely maintaining the global total emission for each month. In model, there is neither land grid nor its ²²²Rn emission around the two small remote island stations (MNM and CCJ), but oceanic emissions exist there. It should be noted for YON that the TL95 model has no land grid at the location of the main island of Taiwan, close to YON, and thereby no ²²²Rn emission from Taiwan in the constant emission map for TL95 (Fig. S1a), whereas that for TL319 and monthly emission maps for both resolutions have ²²²Rn emissions from Taiwan (Figs. S1b, S2a, b). The location of RYO, being on the Pacific coast in reality, is between land and ocean grids in all cases of emission maps and model resolutions (Figs. S1, S2).

3.3 ²²²Rn simulation procedure

We simulated atmospheric ²²²Rn concentrations for the four stations for the period January 2005 to December 2019 with two different horizontal resolutions, two different emission datasets, zero ²²²Rn initial concentrations, and with the first 2 years as spin-up. For comparison with hourly observation data from each station, model results were sampled by three-dimensional linear interpolation of hourly model outputs for eight grids surrounding the station, based on longitude, latitude, and geopotential height.

4. Results

4.1 Hourly ²²²Rn observations

Hourly mean atmospheric ²²²Rn concentrations observed at the four JMA stations are plotted in Fig. 2, together with the data acquisition rate (ratio of the number of hourly observations to total hours in a month), means, medians, and upper and lower quartiles of data for each month. Atmospheric ²²²Rn concentrations display clear seasonal cycles with maxima in winter and minima in summer at all stations, with no significant long-term trend. Anomalously high concentrations occur irregularly due to natural causes in all seasons at all stations. The causes are almost exclusively continental transport of ²²²Rn, as discussed in Section 4.3, while more local conditions such as planetary boundary layer height (PBLH), land-sea breeze, and local emissions are also likely to affect atmospheric ²²²Rn at RYO, as discussed in Section 4.4. As discussed for the MNM, YON, and RYO stations by Wada et al. (2012, 2013) and Ishijima et al. (2018), seasonal patterns are generally explained by seasonal exchanges of continental and maritime air masses over the western North Pacific, consistent with seasonal changes in wind fields (Figs. 1b, c). Japan is in the East Asian Monsoon region, so it is affected by ²²²Rn outflow from the Eurasian continent in winter, and in summer it is covered by maritime air containing less ²²²Rn primarily due to the Ogasawara high, which is usually centered off the southeastern coast of the main island of Japan in summer. Synoptic high-²²²Rn events occur at the stations more frequently in winter when low-pressure systems develop over the Pacific (Fig. 1b), forming cold fronts with rapid ²²²Rn transport from the continent behind the frontal line. Such fronts usually migrate eastward, passing over the stations, which then record higher ²²²Rn concentrations. This typical pattern has been described previously by Ishijima et al. (2018), based on composite analyses of high ²²²Rn events at remote stations. Although ²²²Rn observations at station CCJ are reported here for the first time, meteorological processes driving seasonal and synoptic variations there are similar to those at station MNM, as deduced from their locations, windfield seasonality (Fig. 1), and observed concentrations (Fig. 2). The magnitudes of ²²²Rn concentration at each station roughly reflect the ²²²Rn transport path distance from continental source regions, considering the short ²²²Rn half-life of only ∼ 3.8 days. This is discussed further in the following section.

Fig. 2.

Hourly means (black dots) of observed ²²²Rn concentrations, and their monthly means (blue circles), medians (red line), upper and lower quartiles (tops and bottoms of error bars, respectively), and data acquisition rates (gray bars) at the four ²²²Rn observation stations. The data accusation rate is the ratio of the number of hourly observations to total hours in a month. X-ticks correspond to 00:00 JST January 1.

Data acquisition rates were generally > 90 % for the entire observation period, with those at station YON being slightly lower because the station is unmanned. Data may also be missing because of periodic maintenance and typhoon weather etc.

4.2 Seasonal variations of observed and simulated ²²²Rn concentrations

Figure 3 reports means, medians, and upper and lower quartiles of ²²²Rn concentrations observed and simulated monthly at each station over the whole observation period (Table 1). Correlation coefficient and standard deviation between monthly mean ²²²Rn concentrations of each model simulation and observation are also shown. Winter maxima and summer minima commonly occur at MNM, CCJ, and YON for both observations and model simulations. Results for MNM and CCJ are similar due to their similar locations and surrounding environments as remote island stations in the Pacific, far from continental sources. However, the ²²²Rn concentrations at MNM are almost half those observed at CCJ, reflecting the greater distance from the continental sources. Assuming that the ²²²Rn transport path from the continental source region to CCJ and MNM in winter (December to March) is roughly linear, the elapsed time and speed of ²²²Rn transport from CCJ to MNM were estimated as 3.5 days and 4.0 m s⁻¹, respectively, based on mean December–February observations and a CCJ–MNM distance of ∼ 1200 km. The model results gave speeds of 4.1–4.7 m s⁻¹. These speeds seem to be consistent with mean wind speeds (Fig. 1b). Such estimations could provide an index for transport timescales from CCJ to MNM, despite them having different source regions. At RYO, strong local emissions obscure monsoon-driven atmospheric ²²²Rn seasonality.

Fig. 3.

Monthly climatological means, medians, and upper and lower quartiles of atmospheric ²²²Rn concentration observed at the four stations and simulated by the GSAM-TM model. Only model results corresponding to observation data were used to calculate monthly values. Correlation coefficient and standard deviation (in mBq m⁻³ in parentheses) between monthly mean ²²²Rn concentrations of each simulation and observation are also indicated in the same color as the mark in the legend.

The models well simulated the observed monthly patterns at MNM, CCJ, and YON, as shown by the high correlation coefficients of more than 0.9, although ²²²Rn levels were slightly underestimated by models, especially at CCJ and YON. At MNM and CCJ, there is little difference between model horizontal resolutions and ²²²Rn emission types, indicating that ²²²Rn variations are controlled predominantly by synoptic-scale atmospheric transport. Monthly emission simulations match observations slightly more accurately, with higher values for all seasons, particularly at YON. Especially, the lower-resolution model with monthly emission showed the closest values to the observation. It is due to the relationship between station location and land emission area in model. Both constant and monthly emission data gave very low oceanic emissions (0.10 mBq m⁻² s⁻¹ and 0.14 mBq m⁻² s⁻¹, respectively) in grids surrounding YON station as Yonagunijima island is very small (< 30 km²), as accurately reproduced by the high-resolution model (TL319). However, this was not the case for the low-resolution model (TL95) with monthly emission because grids to the west of YON include eastern Taiwan, which has high land emission (Fig. S2a), causing higher ²²²Rn values at YON in TL95 than in TL319. In contrast, constant emission simulation by TL319 showed higher values than that by TL95 throughout the year, because there is no land grid at the location of Taiwan in TL95, as mentioned in Section 3.2. Such situations occur for atmospheric component simulations depending on station location and emission distribution. If high-resolution emission data are realistic, a higher-resolution model should better reproduce atmospheric observations at a station such as YON. Nevertheless, if the emission data are not realistic, even if they are of high resolution, a lower-resolution model sometimes better reproduces observations with the unrealistic emission distribution being smoothed out, as is likely at YON. At RYO, the difference in atmospheric ²²²Rn simulations between emission types is pronounced. Moreover, correlation coefficients for constant emission simulations at RYO show the model resolution dependency, with the lowest correlation coefficient for TL95 (0.6) against high correlation for TL319 (0.85). This is a case that larger emission distribution difference between resolutions for constant emission than that for monthly emission (Figs. S1, S2) is reflected in simulated ²²²Rn, coupled with the coastal location. Monthly emission simulations significantly underestimate observations, whereas constant emission simulations better reproduce observed ²²²Rn levels. These results may simply reflect terrestrial emissions from Japan affecting RYO, considering the prevailing wind direction from the west. Monthly emissions in northern Japan are much lower than Eurasian continental emissions (Figs. 1b, c) and land emissions of constant emission (Fig. 1a). Emissions in northern Japan may therefore be significantly underestimated in monthly emission data. Inverse modeling by Hirao et al. (2010b) indicates that emissions estimated by Hirao et al. (2010a) may be underestimated by ∼ 30 %, although no station in the northern main island was used. Continental emissions of monthly emission data seem to be more realistic than those of constant emission data, considering monthly emission results were better at YON, which is most affected by continental emissions among the four stations. The underestimation by models at MNM, CCJ, and YON may be due to the underestimation of emission data for wider East Asian regions (Hirao et al. 2010a), with simulations based on the same constant emission also underestimating observations in China and Korea (Zhang et al. 2021). Throughout comparisons between observations and model simulations at four JMA stations in this section, monthly emission data seems to be slightly better than constant emission data, as also indicated by score values in Fig. 3, although results at RYO are not so simple as mentioned above.

4.3 Synoptic high-²²²Rn events

High-²²²Rn events are sometimes observed almost simultaneously at two or three of the four JMA stations, especially in cold seasons, due to frontal activity excited by lows. Figure 4 shows an example of such simultaneous events at YON, CCJ, and MNM in 2017. High-²²²Rn events occurred during January 20–22, January 30–February 1, and February 2–3 in 2017 at YON, CCJ, and MNM, with the second and third events also being observed on January 30–31 and February 2 at YON, respectively. For all cases, high- ²²²Rn events were first observed at YON, then at CCJ, and finally at MNM. In each event, a cold front holding high ²²²Rn concentrations behind the frontal line moved eastward over the stations, with high ²²²Rn levels being observed in the station order. Model simulations generally reproduce temporal high-²²²Rn patterns with synoptic timescales of 2–5 days from trough to trough. However, the reproducibility of shorter peaks (< 12 h) sometimes highly depends on the horizontal resolution of the model. The clearest such case was observed on January 21, 2017, at CCJ. The first and second peaks were observed at 01:00 JST and 10:00 JST, respectively. The TL319 model with monthly emissions also simulated two similar peaks, but 1 hour later at 02:00 JST and 11:00 JST, and with constant emission, it showed a very small first peak. The TL95 model simulated only one peak on that day. The cases on January 30–31 and February 1–2 at YON (Fig. 4a) also demonstrate that only TL319 can resolve the observed multiple peaks.

Fig. 4.

Radon-222 concentrations observed at (a) YON, (b) CCJ, and (c) MNM for the period January 18 to February 4, 2017, and simulation results. Observation data were sometimes missing, and full hourly model outputs are plotted.

The horizontal distribution and latitude–pressure cross-section of atmospheric ²²²Rn concentrations at the longitude of CCJ on January 21, 2017 (07:00 JST), are shown in Fig. 5, corresponding to a trough between the two high peaks (Fig. 4b), as simulated by TL95 and TL319 with monthly emissions. Contours of equivalent potential temperature and potential vorticity are also shown as indices for distributions of atmospheric components around a cold front and stratospheric air, respectively. Potential vorticity was calculated from 6-hourly JRA-55 Reanalysis data, spatiotemporally interpolated (Sawa et al. 2008). TL95 shows no trough in ²²²Rn concentrations around CCJ (Fig. 5a), but TL319 shows an arcuate shape of low ²²²Rn trough over CCJ (Fig. 5c). To the north and south of CCJ, there are two elongated ²²²Rn streamlines originating from northern China/Korea and southern China/Taiwan, respectively. The southern and northern branches (Fig. 5c) were observed as the first and second peaks, respectively, on January 21 at CCJ (Fig. 4b). The significant difference in TL319 results at CCJ on January 21 is attributable to the emission in southern China in January being much greater than in July and than the constant emission (Fig. 1). Model resolutions finer than 60 km (corresponding to TL319) are needed to reproduce the general shape of observed high-²²²Rn peaks at remote stations (Ishijima et al. 2018), but here the sharp edge of the observed peak could not be reproduced, even at high resolution, as also indicated in Fig. 4, especially at CCJ. Latitude–pressure cross-sections clearly show that the southern branch involves ²²²Rn accumulated behind the frontal surface at around 23°N (Fig. 5d), and the northern branch at around 29°N is visible only in the TL319 result. Radon-222 was seemingly uplifted from near-surface layers to around the 500 hPa level (Figs. 5b, d), although it was likely to be uplifted over the continent and then transported eastward, based on our model simulation results. A low ²²²Rn trend from ∼ 450 hPa down to ∼ 750 hPa over CCJ might have been caused by stratospheric air intrusion, as shown by the potential vorticity isoline of 1 PVU (PVU = 10⁶ m² s⁻¹ K kg⁻¹). It seemed that ²²²Rn, uplifted on the China Mainland, was eastward transported by strong westerlies in the middle troposphere, then the high ²²²Rn region for the 800–500 hPa was divided by the stratospheric intrusion (Fig. S3). Upward transports looked mainly occurring ahead of the frontal line, as indicated by negative vertical velocities, in the warm sector with less ²²²Rn. Thus, it is unlikely that ²²²Rn around the 500 hPa level originated from the nearsurface maritime air. Reid and Vaughan (2004) observed a tropopause fold behind a cold front together with active convection over the frontal surface, based on aircraft observation. The GSAM-TM simulations thus indicate that complicated three-dimensional atmospheric structures may sometimes be produced over cold fronts and related lows, significantly affecting spatiotemporal variations of atmospheric constituents.

Fig. 5.

Horizontal distribution at the lowest layer of the model and latitude-pressure cross-sections of ²²²Rn concentrations at 07:00 JST, January 21, 2017, when a high-²²²Rn event was observed at CCJ (solid white circle), as simulated by the TL95 and TL319 models with the monthly emissions. Black and white contours represent equivalent potential temperature (K) and potential vorticity (PVU = 10⁶ m² s⁻¹ K kg⁻¹), respectively. Potential vorticity was calculated from 6-hourly JRA-55 Reanalysis data, spatiotemporally interpolated.

4.4 Diurnal cycle of ²²²Rn

Mean diurnal cycles of observed and simulated ²²²Rn and simulated PBLH [calculated following the method of Seidel et al. (2012)] at hourly interval (00:00–23:00 JST) for each month at the four stations are shown in Figs. 6 and 7. Diurnal cycles were calculated in two ways, either as the simple mean of hourly data for each hour of each month over the whole observation period, or as hourly data divided by the daily mean and subtracted by one, averaged for each hour of each month. The latter is referred to as δ here, the value normalized to the daily mean. For each day, after calculating the daily mean from 24 hourly data in that day, hourly data are divided by the daily mean, and subtracted by one. δ is the mean of thus obtained values for each hour of each month. In both methods, if a day had missing hourly data, even for 1 hour, that day was not included in the diurnal cycle calculations. Δ also represents the simple mean of hourly data for each hour of each month, but is subtracted by the monthly mean, to make the diurnal amplitude seasonality more distinct.

Fig. 6.

Monthly mean diurnal ²²²Rn cycles, observed and simulated, and simulated PBLH cycles at RYO (a, b) and YON (c, d). δ is the value normalized by daily mean, and Δ is the deviation from the monthly mean.

Fig. 7.

Same as Fig. 6 but for MNM (a, b) and CCJ (c, d).

Most previous studies have reported that the diurnal cycle of atmospheric ²²²Rn concentration is associated mainly with varying PBLH and local or quasi-local ²²²Rn emissions (e.g., Moriizumi et al. 2008; Chambers et al. 2019), while remote stations such as MNM or CCJ have been regarded as not having a diurnal cycle, with no local emissions. The atmospheric ²²²Rn concentration largely changes day by day mainly owing to synoptic transport of ²²²Rn from continental source emissions (Fig. 4). The lack of local sources and the existence of sporadic synoptic high-²²²Rn events obscure diurnal cycles at such remote stations. Normalization of hourly data to the daily mean may at least remove the synoptic variation signals. Conversely, synoptic transport of ²²²Rn from continental sources is needed to cause ²²²Rn diurnal cycles at such a remote station with no significant local ²²²Rn emissions, otherwise there is not enough ²²²Rn supply there. Thus supplied ²²²Rn in the near-surface atmosphere is affected by the local PBLH variations over the station and could be diurnally varied. We therefore first attempted the δ-calculation approach to extract diurnal signals at remote stations in this study.

The ²²²Rn has clear diurnal cycles at RYO (Figs. 6a, b), with both absolute and δ values varying in phase, almost opposite that of PBLH. The similar variations of absolute and δ values indicate that the diurnal cycles are caused mainly by local sources and diurnal variation of PBLH, with local sources including ²²²Rn emissions from upwind areas nearby on the main island of Japan, with westerly winds predominating throughout the year. Moriizumi et al. (2008) reported that Cape Ochiishi station on Hokkaido island displays diurnal ²²²Rn variations only during November–February, when seasonal northwesterly winds supply ²²²Rn from local areas upwind. The GSAM-TM also well simulates diurnal cycles with morning maxima and afternoon minima. The observed amplitude of ²²²Rn diurnal variations is larger in summer than in winter, despite the monthly mean being lower in summer, whereas the models do not show such seasonality of diurnal amplitude. This observed trend has also been reported in East Asia and Europe (Zahorowski et al. 2005; Moriizumi et al. 2008; Zimnoch et al. 2014). The mechanisms for this have not yet been established, but the results have been simply explained by surface heating and active vertical mixing owing to longer daytime insolation in summer. However, this is inconsistent with the monthly mean PBLH maxima in winter around Japan (Figs. 6a, c, 7a, c). In monsoon regions, during the wet season (summer), surface solar heating energy is used mainly to evaporate water, whereas in the dry season (winter), greater surface sensible heat flux is available to drive vertical dry mixing (Chan and Wood 2013), consistent with our simulated PBLH results (Figs. 6a, c, 7a, c). Land and sea breezes should be also considered as possible causes affecting ²²²Rn diurnal cycles at RYO, which is located on a mountain facing the Pacific Ocean. Although the prevailing wind direction is from the west in all seasons, the wind changes from a land breeze at nighttime to a sea breeze in daytime during April–September (Fig. S4). This could lead to higher and lower ²²²Rn concentrations in nighttime and daytime, respectively, as does the PBLH variation, with the observed diurnal amplitude being relatively large during these months. Furthermore, in winter strengthened westerlies brings more continental ²²²Rn, attenuating diurnal amplitudes, as discussed below. Thus, larger fractions of continental ²²²Rn could enhance the diurnal amplitude reduction in winter. The model also reproduces this observed trend of larger diurnal amplitude in summer but with smaller seasonal changes than those observed, most likely due to the lower model resolution being unable to reproduce diurnal changes in local wind direction (Fig. S4). To summarize, diurnal variations of PBLH, land and sea breezes, and ²²²Rn from local and remote sources are likely to be important in explaining the seasonal variation of the diurnal amplitude of ²²²Rn at RYO. The observed diurnal amplitudes in winter, which are lower than those simulated, may also be due to the seasonality of local ²²²Rn emissions. Lower temperatures could reduce ²²²Rn diffusion from soil to the atmosphere, due to freezing of soil water and/or temperature decrease in soil. The location of the station in complex terrain may also enhance the effects of local meteorology. However, these other factors are beyond the scope of this study.

At station YON, only observations display large diurnal variations with morning maxima and afternoon minima in summer, but models show negligibly small or opposite phase variations (Figs. 6c, d), because model grids around YON are oceanic with very low ²²²Rn emission (Figs. 1c, S1, S2). Observations, therefore, capture local ²²²Rn emission signals, as the station is surrounded by grassland and agricultural fields. A lower PBLH in summer fosters the accumulation of locally emitted ²²²Rn during night to morning. Remotely transported ²²²Rn, enhanced in winter, is related to the diurnal amplitude seasonality for observation and simulations, similarly to that at RYO. Trends in δ values in winter are similar at stations MNM, CCJ, and YON, although land and sea breezes are not considered for these stations because they are on small islands where there is little diurnal change in wind direction (Fig. S4).

At MNM and CCJ, diurnal cycles of absolute ²²²Rn concentrations are negligible or unclear (Figs. 7a, c). However, δ values from winter to spring (November–May) indicate clear diurnal cycles with notable day-time minima, with this temporal pattern being similar for both observations and models (Figs. 7b, d). The minima timings are similar to those of PBLH maxima, suggesting that PBLH variation drives ²²²Rn diurnal cycles. At the MNM and CCJ island stations, local ²²²Rn sources are not significant, and in model emissions around the stations are also negligibly small because there are no land grid-points around the stations. To evaluate the influence of oceanic emissions, we performed a test simulation using the TL95 model and modified emission maps, whose oceanic emissions were set to zero. Considering δ diurnal cycles in cold seasons at MNM and CCJ are well reproduced at both model resolutions (Figs. 7b, d), only TL95 simulation is expected to be adequate to obtain quantitative information for that. Consequently, the oceanic emissions seem to have no effect on the diurnal cycles, showing almost the same results as the original emission case (Fig. S6), although absolute ²²²Rn levels are slightly reduced especially for monthly emissions with larger oceanic emissions (Section 4.2). These results are consistent with the suggestion that variations of ²²²Rn concentrations in the near-surface atmosphere over the stations, which have been transported from the continental sources, are controlled by diurnal variations of the local PBLH. Therefore, clear diurnal cycles of δ, for both observations and models, occur only in months with high ²²²Rn concentrations due to active continental outflow of ²²²Rn (Figs. 1, 3, 4). In order to better understand what kind of condition contributes to the ²²²Rn diurnal cycles represented by δ, we performed an additional analysis for January, February, March, November, and December (Fig. S7 and its explanation in the supplement). It is found as a consequence that δ values highlight ²²²Rn diurnal cycles under conditions of lower ²²²Rn level and/or PBLH at remote stations. It means that variability and phase of δ diurnal cycle tend to be larger and closer to that of the monthly mean diurnal cycle, respectively, when the ²²²Rn level and/or PBLH are relatively low. They also suggest the possibility that PBLH is high when the ²²²Rn level is high due to the continental transport, and vice versa. It seems to be reasonable that PBLH generally increases with increasing wind speed and decreases with increasing atmospheric stability. In summer, the diurnal cycle of ²²²Rn seems not to be well extracted at remote stations even by the δ approach, because the signal-to-noise ratio is too low.

It is difficult to detect diurnal cycles with absolute values because δ values indicate very small diurnal amplitudes (generally < 10 % of daily means) in November–March. Similar diurnal variations of δ values occur at YON during November–December when the contribution of continental-origin ²²²Rn is more significant than that of local-origin ²²²Rn. Small δ amplitudes in those months at RYO are likely due to the same effect, although the contribution of local ²²²Rn is more significant there, even in winter.

Using δ values rather than absolute concentrations seems to successfully derive diurnal cycles from ²²²Rn concentrations when the station is less and more affected by local-origin ²²²Rn and episodic continental ²²²Rn transport, respectively. Especially, δ highlights ²²²Rn diurnal cycles when ²²²Rn level and/or PBLH are relatively low. At stations with significant diurnal cycles (e.g., RYO), both absolute and δ values similarly represent true diurnal cycles. The combined use of both absolute and δ values to represent observed and simulated ²²²Rn diurnal cycles has been helpful in elucidating the relative contributions of local and continental-transport ²²²Rn. Monthly mean diurnal amplitudes of δ at remote stations are actually very small, being mostly below 10 % of the daily mean ²²²Rn. Nevertheless, the amplitude and variation phase are quite similar for observation and model simulations. These results are notable, because small diurnal cycles are almost ignored when using only the absolute values (Figs. 7a, c). δ value seems to well function to validate the model performance of simulating the diurnal behavior of atmospheric components in the near-surface atmosphere, even when the atmospheric component shows large synoptic variability and/or very small diurnal variation. Diurnal cycles are important especially for carbon dioxide (CO₂), because relative contributions of photosynthesis and respiration of land ecosystems vary between daytime and nighttime. For inverse model estimation of CO₂ emissions, model performance for reproducing diurnal cycles significantly affected by PBLH reproducibility seriously affects CO₂ emission estimates (e.g., Stephens et al. 2007). In this study, we discussed only temporal variations of ²²²Rn diurnal cycles, with a newly introduced δ calculation method, and a three-dimensional approach would be needed to further improve our understanding of atmospheric component behaviors around the observation stations.

5. Summary

Atmospheric ²²²Rn concentrations observed at four JMA stations in Japan were analyzed for climatological mean monthly and diurnal variations and synoptic high ²²²Rn events, compared with GSAM-TM simulations.

Observed ²²²Rn concentrations show clear seasonal variations at all stations, with no significant long-term trends. Monthly ²²²Rn values at the remote stations MNM, CCJ, and YON commonly reach maxima and minima in winter and summer, respectively, mainly due to the seasonality of atmospheric transport. These stations in the East Asian Monsoon region are affected by ²²²Rn outflow from Eurasian continent sources in winter, and by maritime air containing less ²²²Rn in summer, primarily due to the Ogasawara high. The atmospheric transport model GSAM-TM well reproduces monthly variations but tends to underestimate ²²²Rn concentrations, mainly due to the emission inventories used. At RYO, the strong local-emission influence masks monsoon-driven atmospheric ²²²Rn seasonality. The model–observation comparison at RYO indicates that ²²²Rn emission on the main island of Japan still has not been well understood and therefore could largely depend on the emission estimations. Model resolution dependency is dominantly found at RYO due to the difference of the constant emission distributions around the coastal station between the model resolutions. Model–observation comparisons for monthly mean ²²²Rn concentrations in this study imply the monthly varying emission data shows slightly better performance than the constant emission data around Japan.

A case study of synoptic high ²²²Rn events caused by continental outflow indicates that the TL319 model (∼ 60 km mesh) successfully simulates several-hour-timescale peaks, whereas the TL95 model (∼ 200 km mesh) cannot, as reported by Ishijima et al. (2018). Simulated ²²²Rn latitude–pressure cross-sections at the time of an event indicate the possibility that complicated three-dimensional atmospheric structures accompanied by stratospheric intrusion could be formed around a cold front by compositive causes, significantly affecting atmospheric component distributions.

Monthly mean ²²²Rn diurnal cycles were analyzed. At RYO, clear diurnal cycles are observed and well simulated, primarily due to PBLH variations combined with a local-emission influence throughout the year. Diurnal amplitudes are enhanced by land and sea breezes in summer but reduced by remotely sourced ²²²Rn in winter. At MNM and CCJ, diurnal cycles are not clear due to the lack of local emission and sporadic synoptic events by continental outflows in winter, with simply averaged absolute ²²²Rn values being used. However, by averaging hourly ²²²Rn concentrations normalized by the daily mean, the resulting δ values at MNM, CCJ, and YON show relatively clear diurnal cycles with daytime minima in winter at up to 10 % of the daily mean in both observations and simulations. The normalization procedure may successfully highlight diurnal cycles based on highly variable raw ²²²Rn data affected by continental outflows. Monthly mean diurnal cycles of δ in winter at remote stations seem to mainly reflect those under the condition of low ²²²Rn level and/or PBLH, reasonably suggesting that PBLH is high when ²²²Rn level is high due to the continental transport, and vice versa. However, even the δ approach cannot extract the diurnal cycle signal in summer when the signal-to-noise ratio is too low. Diurnal cycles extracted using the δ calculation, for both observations and simulations, are driven mainly by diurnal variations of the PBLH. Results demonstrate the possibility that diurnal variations of ²²²Rn transported from remote source regions are controlled mainly by local PBLH variations, even without significant local ²²²Rn emissions.

Supplements

Figure S1 shows distributions of the constant ²²²Rn emission around Japan for TL95 and TL319 resolutions and locations of ²²²Rn observation stations. Figure S2 is the same as Fig. S1 but for the monthly ²²²Rn emissions. Figure S3 shows horizontal distribution at 600 hPa and longitude-pressure cross-sections of ²²²Rn concentrations at 31.7°N, 29.5°N, and 27.2°N at 07:00 JST, January 19, 20, and 21, 2017, when high-²²²Rn part was seen over CCJ in model simulations (Fig. 5), as simulated by the TL319 model with the monthly emissions. Figure S4 shows monthly mean diurnal cycles of wind direction and speed, observed and simulated. Figure S5 is the same as Fig. 6, but results of TL319 simulations at RYO and YON are replaced by those of TL95 simulations with modified emission maps, whose oceanic emissions are set to zero. Figure S6 is the same as Fig. S5 but for MNM and CCJ. Figure S7 shows the normalized standard deviation of mean δ diurnal cycle for selected ²²²Rn data (see explanation for Fig. S7) and its correlation coefficient with the mean δ diurnal cycles for all data, in January, February, March, November, and December. Table S1 shows the correlation coefficient and standard deviation between hourly means of model simulations and observation.

Acknowledgments

We thank all members of the JMA for taking care of ²²²Rn measurement systems at the observation stations. This study was financially supported by the JSPS KAKENHI (19H01975) and the Global Environmental Research Coordination System from the Ministry of the Environment of Japan (MOE1454 and MOE1953).

References

•  Arakawa,  A., and  W. H.  Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment, Part I. J. Atmos. Sci., 31, 674-701.
•  Belikov,  D. A.,  S.  Maksyutov,  M.  Krol,  A.  Fraser,  M.  Rigby,  H.  Bian,  A.  Agusti-Panareda,  D.  Bergmann,  P.  Bousquet,  P.  Cameron-Smith,  M. P.  Chipperfield,  A.  Fortems-Cheiney,  E.  Gloor,  K.  Haynes,  P.  Hess,  S.  Houweling,  S. R.  Kawa,  R. M.  Law,  Z.  Loh,  L.  Meng,  P. I.  Palmer,  P. K.  Patra,  R. G.  Prinn,  R.  Saito, and  C.  Wilson, 2013: Off-line algorithm for calculation of vertical tracer transport in the troposphere due to deep convection. Atmos. Chem. Phys., 13, 1093-1114.
•  Biraud,  S.,  P.  Ciais,  M.  Ramonet,  P.  Simmonds,  V.  Kazan,  P.  Monfray,  S.  O'Doherty,  T. G.  Spain, and  S. G.  Jennings, 2000: European greenhouse gas emissions estimated from continuous atmospheric measurements and radon 222 at Mace Head, Ireland. J. Geophys. Res., 105, 1351-1366.
•  Chambers,  S. D.,  E.-A.  Guérette,  K.  Monk,  A. D.  Griffiths,  Y.  Zhang,  H.  Duc,  M.  Cope,  K. M.  Emmerson,  L. T.  Chang,  J. D.  Silver,  S.  Utembe,  J.  Crawford,  A. G.  Williams, and  M.  Keywood, 2019: Skill-testing chemical transport models across contrasting atmospheric mixing states using Radon-222. Atmosphere, 10, doi: 10.3390/atmos10010025.
•  Chan,  K. M., and  R.  Wood, 2013: The seasonal cycle of planetary boundary layer depth determined using COSMIC radio occultation data. J. Geophys. Res.: Atmos., 118, 12422-12434.
•  Chevillard,  A.,  P.  Ciais,  U.  Karstens,  M.  Heimann,  M.  Schmidt,  I.  Levin,  D.  Jacob,  R.  Podzun,  V.  Kazan,  H.  Sartorius, and  E.  Weingartner, 2002: Transport of ²²²Rn using the regional model REMO: A detailed comparison with measurements over Europe. Tellus B, 54, 850-871.
•  Conen,  F., and  L. B.  Robertson, 2002: Latitudinal distribution of radon-222 flux from continents. Tellus B, 54, 127-133.
•  Deushi,  M., and  K.  Shibata, 2011: Development of a Meteorological Research Institute Chemistry-Climate Model version 2 for the study of tropospheric and stratospheric chemistry. Pap. Meteor. Geophys., 62, 1-46.
•  Friedlingstein,  P.,  M.  O'sullivan,  M. W.  Jones,  R. M.  Andrew,  J.  Hauck,  A.  Olsen,  G. P.  Peters,  W.  Peters,  J.  Pongratz,  S.  Sitch,  C.  Le Quéré,  J. G.  Canadell,  P.  Ciais,  R. B.  Jackson,  S.  Alin,  L. E. O. C.  Aragão,  A.  Arneth,  V.  Arora,  N. R.  Bates,  M.  Becker,  A.  Benoit-Cattin,  H. C.  Bittig,  L.  Bopp,  S.  Bultan,  N.  Chandra,  F.  Chevallier,  L. P.  Chini,  W.  Evans,  L.  Florentie,  P. M.  Forster,  T.  Gasser,  M.  Gehlen,  D.  Gilfillan,  T.  Gkritzalis,  L.  Gregor,  N.  Gruber,  I.  Harris,  K.  Hartung,  V.  Haverd,  R. A.  Houghton,  T.  Ilyina,  A. K.  Jain,  E.  Joetzjer,  K.  Kadono,  E.  Kato,  V.  Kitidis,  J. I.  Korsbakken,  P.  Landschützer,  N.  Lefèvre,  A.  Lenton,  S.  Lienert,  Z.  Liu,  D.  Lombardozzi,  G.  Marland,  N.  Metzl,  D. R.  Munro,  J. E. M. S.  Nabel,  S.  Nakaoka,  Y.  Niwa,  K.  O'Brien,  T.  Ono,  P. I.  Palmer,  D.  Pierrot,  B.  Poulter,  L.  Resplandy,  E.  Robertson,  C.  Rödenbeck,  J.  Schwinger,  R.  Séférian,  I.  Skjelvan,  A. J. P.  Smith,  A. J.  Sutton,  T.  Tanhua,  P. P.  Tans,  H.  Tian,  B.  Tilbrook,  G.  van der Werf,  N.  Vuichard,  A. P.  Walker,  R.  Wanninkhof,  A. J.  Watson,  D.  Willis,  A. J.  Wiltshire,  W.  Yuan,  X.  Yue, and  S.  Zaehle, 2020: Global carbon budget 2020. Earth Syst. Sci. Data, 12, 3269-3340.
•  Gravel,  S., and  A.  Staniforth, 1994: A mass-conserving semi-Lagrangian scheme for the shallow-water equations. Mon. Wea. Rev., 122, 243-248.
•  Hirao,  S.,  H.  Yamazawa, and  J.  Moriizumi, 2010a: Estimation of the global ²²²Rn flux density from the Earth's surface. J. Health Phys. Japan, 45, 161-171.
•  Hirao,  S.,  H.  Yamazawa, and  J.  Moriizumi, 2010b: Inverse modeling of Asian ²²²Rn flux using surface air ²²²Rn concentration. J. Environ. Radioact., 101, 974-984.
•  Iida,  T.,  Y.  Ikebe, and  K.  Tojo, 1991: An electrostatic radon monitor for measurements of environmental radon. Res. Lett. Atmos. Electr., 11, 55-59.
•  Iida,  T.,  Y.  Ikebe,  K.  Suzuki,  K.  Ueno,  Z.  Wang, and  Y.  Jin, 1996: Continuous measurements of outdoor radon concentrations at various locations in East Asia. Environ. Int., 22, 139-147.
•  Ishijima,  K.,  M.  Takigawa,  Y.  Yamashita,  H.  Yashiro,  C.  Kodama,  M.  Satoh,  K.  Tsuboi,  H.  Matsueda,  Y.  Niwa, and  S.  Hirao, 2018: Analysis of high radon-222 concentration events using multi-horizontal-resolution NICAM simulations. SOLA, 14, 111-115.
•  Jacob,  D. J.,  M. J.  Prather,  P. J.  Rasch,  R.-L.  Shia,  Y. J.  Balkanski,  S. R.  Beagley,  D. J.  Bergmann,  W. T.  Blackshear,  M.  Brown,  M.  Chiba,  M. P.  Chipperfield,  J.  de Grandpré,  J. E.  Dignon,  J.  Feichter,  C.  Genthon,  W. L.  Grose,  P. S.  Kasibhatla,  I.  Köhler,  M. A.  Kritz,  K.  Law,  J. E. Penner,  M.  Ramonet,  C. E.  Reeves,  D. A.  Rotman,  D. Z.  Stockwell,  P. F. J.  Van Velthoven,  G.  Verver,  O.  Wild,  H.  Yang, and  P.  Zimmermann, 1997: Evaluation and intercomparison of global atmospheric transport models using ²²²Rn and other short-lived tracers. J. Geophys. Res., 102, 5953-5970.
• Japan Meteorogical Agency, 2013: Outline of the operational numerical weather prediction at the Japan Meteorological Agency. Appendix to WMO technical progress report on the Global Data-processing and Forecasting System (GDPFS) and Numerical Weather Prediction (NWP) Research. Japan Meteorological Agency, 43-61. [Available at https://www.jma.go.jp/jma/jma-eng/jma-center/nwp/outline2013-nwp/index.htm.]
•  Juang,  H.-M. H., 2004: A reduced spectral transform for the NCEP seasonal forecast global spectral atmospheric model. Mon. Wea. Rev., 132, 1019-1035.
•  Kobayashi,  S.,  Y.  Ota,  Y.  Harada,  A.  Ebita,  M.  Moriya,  H.  Onoda,  K.  Onogi,  H.  Kamahori,  C.  Kobayashi,  H.  Endo,  K.  Miyaoka, and  K.  Takahashi, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5-48.
•  Mellor,  G. L., and  T.  Yamada, 1974: A hierarchy of turbulence closure models for planetary boundary layers. J. Atmos. Sci., 31, 1791-1806.
•  Moriizumi,  J.,  T.  Ohkura,  S.  Hirao,  Y.  Nono,  H.  Yamazawa,  Y.-S.  Kim,  Q.  Guo,  H.  Mukai,  Y.  Tohjima, and  T.  Iida, 2008: Continuous observation of atmospheric ²²²Rn concentrations for analytic basis of atmospheric transport in East Asia. J. Nucl. Sci. Technol., 45, 173-179.
•  Reid,  H. J., and  G.  Vaughan, 2004: Convective mixing in a tropopause fold. Quart. J. Roy. Meteor. Soc., 130, 1195-1212.
•  Saunois,  M.,  A. R.  Stavert,  B.  Poulter,  P.  Bousquet,  J. G.  Canadell,  R. B.  Jackson,  P. A.  Raymond,  E. J.  Dlugokencky,  S.  Houweling,  P. K.  Patra,  P.  Ciais,  V. K.  Arora,  D.  Bastviken,  P.  Bergamaschi,  D. R.  Blake,  G.  Brailsford,  L.  Bruhwiler,  K. M.  Carlson,  M.  Carrol,  S.  Castaldi,  N.  Chandra,  C.  Crevoisier,  P. M.  Crill,  K.  Covey,  C. L.  Curry,  G.  Etiope,  C.  Frankenberg,  N.  Gedney,  M. I.  Hegglin,  L.  Höglund-Isaksson,  G.  Hugelius,  M.  Ishizawa,  A.  Ito,  G.  Janssens-Maenhout,  K. M.  Jensen,  F.  Joos,  T.  Kleinen,  P. B.  Krummel,  R. L.  Langenfelds,  G. G.  Laruelle,  L.  Liu,  T.  Machida,  S.  Maksyutov,  K. C.  McDonald,  J.  McNorton,  P. A.  Miller,  J. R.  Melton,  I.  Morino,  J.  Müller,  F.  Murguia-Flores,  V.  Naik,  Y.  Niwa,  S.  Noce,  S.  O'Doherty,  R. J.  Parker,  C.  Peng,  S.  Peng,  G. P.  Peters,  C.  Prigent,  R.  Prinn,  M.  Ramonet,  P.  Regnier,  W. J.  Riley,  J. A.  Rosentreter,  A.  Segers,  I. J.  Simpson,  H.  Shi,  S. J.  Smith,  L. P.  Steele,  B. F.  Thornton,  H.  Tian,  Y.  Tohjima,  F. N.  Tubiello,  A.  Tsuruta,  N.  Viovy,  A.  Voulgarakis,  T. S.  Weber,  M.  van Weele,  G. R.  van der Werf,  R. F.  Weiss,  D.  Worthy,  D.  Wunch,  Y.  Yin,  Y.  Yoshida,  W.  Zhang,  Z.  Zhang,  Y.  Zhao,  B.  Zheng,  Q.  Zhu,  Q.  Zhu, and  Q.  Zhuang, 2020: The global methane budget 2000–2017. Earth Syst. Sci. Data, 12, 1561-1623.
•  Sawa,  Y.,  T.  Machida, and  H.  Matsueda, 2008: Seasonal variations of CO₂ near the tropopause observed by commercial aircraft. J. Geophys. Res., 113, D23301, doi:10.1029/2008JD010568.
•  Seidel,  D. J.,  Y.  Zhang,  A.  Beljaars,  J.-C.  Golaz,  A. R.  Jacobson, and  B.  Medeiros, 2012: Climatology of the planetary boundary layer over the continental United States and Europe. J. Geophys. Res., 117, D17106, doi:10.1029/2012JD018143.
•  Simmons,  A. J., and  D. M.  Burridge, 1981: An energy and angular-momentum conserving vertical finite-difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758-766.
•  Stephens,  B. B.,  K. R.  Gurney,  P. P.  Tans,  C.  Sweeney,  W.  Peters,  L.  Bruhwiler,  P.  Ciais,  M.  Ramonet,  P.  Bousquet,  T.  Nakazawa,  S.  Aoki,  T.  Machida,  G.  Inoue,  N.  Vinnichenko,  J.  Lloyd,  A.  Jordan,  M.  Heimann,  O.  Shibistova,  R. L.  Langenfelds,  L. P.  Steele,  R. J.  Francey, and  A. S.  Denning, 2007: Weak northern and strong tropical land carbon uptake from vertical profiles of atmospheric CO₂. Science, 316, 1732-1735.
•  Taguchi,  S.,  R. M.  Law,  C.  Rödenbeck,  P. K.  Patra,  S.  Maksyutov,  W.  Zahorowski,  H.  Sartorius, and  I.  Levin, 2011: TransCom continuous experiment: Comparison of ²²²Rn transport at hourly time scales at three stations in Germany. Atmos. Chem. Phys., 11, 10071-10084.
•  Tanaka,  T. Y.,  K.  Orito,  T. T.  Sekiyama,  K.  Shibata,  M.  Chiba, and  H.  Tanaka, 2003: MASINGAR, a global tropospheric aerosol chemical transport model coupled with MRI/JMA98 GCM: Model description. Pap. Meteor. Geophys., 53, 119-138.
•  Tian,  H.,  R.  Xu,  J. G.  Canadell,  R. L.  Thompson,  W.  Winiwarter,  P.  Suntharalingam,  E. A.  Davidson,  P.  Ciais,  R. B.  Jackson,  G.  Janssens-Maenhout,  M. J.  Prather,  P.  Regnier,  N.  Pan,  S.  Pan,  G. P.  Peters,  H.  Shi,  F. N.  Tubiello,  S.  Zaehle,  F.  Zhou,  A.  Arneth,  G.  Battaglia,  S.  Berthet,  L.  Bopp,  A. F.  Bouwman,  E. T.  Buitenhuis,  J.  Chang,  M. P.  Chipperfield,  S. R. S.  Dangal,  E.  Dlugokencky,  J. W.  Elkins,  B. D.  Eyre,  B.  Fu,  B.  Hall,  A.  Ito,  F.  Joos,  P. B.  Krummel,  A.  Landolfi,  G. G.  Laruelle,  R.  Lauerwald,  W.  Li,  S.  Lienert,  T.  Maavara,  M.  MacLeod,  D. B.  Millet,  S.  Olin,  P. K.  Patra,  R. G.  Prinn,  P. A.  Raymond,  D. J.  Ruiz,  G. R.  van der Werf,  N.  Vuichard,  J.  Wang,  R. F.  Weiss,  K. C.  Wells,  C.  Wilson,  J.  Yang, and  Y.  Yao, 2020: A comprehensive quantification of global nitrous oxide sources and sinks. Nature, 586, 248-256.
•  Wada,  A.,  S.  Murayama,  H.  Kondo,  H.  Matsueda,  Y.  Sawa, and  K.  Tsuboi, 2010: Development of a compact and sensitive electrostatic Radon-222 measuring system for use in atmospheric observation. J. Meteor. Soc. Japan, 88, 123-134.
•  Wada,  A.,  H.  Matsueda,  S.  Murayama,  S.  Taguchi,  A.  Kamada,  M.  Nosaka,  K.  Tsuboi, and  Y.  Sawa, 2012: Evaluation of anthropogenic emissions of carbon monoxide in East Asia derived from the observations of atmospheric radon-222 over the western North Pacific. Atmos. Chem. Phys., 12, 12119-12132.
•  Wada,  A.,  H.  Matsueda,  S.  Murayama,  S.  Taguchi,  S.  Hirao,  H.  Yamazawa,  J.  Moriizumi,  K.  Tsuboi,  Y.  Niwa, and  Y.  Sawa, 2013: Quantification of emission estimates of CO₂, CH₄ and CO for East Asia derived from atmospheric radon-222 measurements over the western North Pacific. Tellus B, 65, 18037, doi:10.3402/tellusb.v65i0.18037.
•  Watanabe,  F.,  O.  Uchino,  Y.  Joo,  M.  Aono,  K.  Higashijima,  Y.  Hirano,  K.  Tsuboi, and  K.  Suda, 2000: Interannual variation of growth rate of atmospheric carbon dioxide concentration observed at the JMA's three monitoring stations: Large increase in concentration of atmospheric carbon dioxide in 1998. J. Meteor. Soc. Japan, 78, 673-682.
•  Yoshimura,  H., and  T.  Matsumura, 2003: A Semi-Lagrangian scheme conservative in the vertical direction. CAS/JSC WGNE Res. Activ. Atmos. Oceanic Modell., 33, 19-20. [Available at https://www.wcrp-climate.org/WGNE/blue_book.html.]
•  Zahorowski,  W.,  S.  Chambers,  T.  Wang,  C.-H.  Kang,  I.  Uno,  S.  Poon,  S.-N.  Oh,  S.  Werczynski,  J.  Kim, and  A.  Henderson-Sellers, 2005: Radon-222 in boundary layer and free tropospheric continental outflow events at three ACE-Asia sites. Tellus B, 57, 124-140.
•  Zhang,  B.,  H.  Liu,  J. H.  Crawford,  G.  Chen,  T. D.  Fairlie,  S.  Chambers,  C.-H.  Kang,  A. G.  Williams,  K.  Zhang,  D. B.  Considine,  M. P.  Sulprizio, and  R. M.  Yantosca, 2021: Simulation of radon-222 with the GEOS-Chem global model: Emissions, seasonality, and convective transport. Atmos. Chem. Phys., 21, 1861-1887.
•  Zimnoch,  M.,  P.  Wach,  L.  Chmura,  Z.  Gorczyca,  K.  Rozanski,  J.  Godlowska,  J.  Mazur,  K.  Kozak, and  A.  Jeričević, 2014: Factors controlling temporal variability of near-ground atmospheric ²²²Rn concentration over central Europe. Atmos. Chem. Phys., 14, 9567-9581.