Identifying tropospheric baseline air masses at Mauna Loa Observatory between 2004 and 2010 using Radon-222 and back trajectories


  • Scott D. Chambers,

    Corresponding author
    1. Australian Nuclear Science and Technology Organization, Institute for Environmental Research, Kirrawee, NSW, Australia
    • Corresponding author: S. Chambers, Australian Nuclear Science and Technology Organisation, Institute for Environmental Research, Locked Bag 2001, Kirrawee DC, NSW 2232, Australia. (

    Search for more papers by this author
  • Wlodek Zahorowski,

    1. Australian Nuclear Science and Technology Organization, Institute for Environmental Research, Kirrawee, NSW, Australia
    Search for more papers by this author
  • Alastair G. Williams,

    1. Australian Nuclear Science and Technology Organization, Institute for Environmental Research, Kirrawee, NSW, Australia
    Search for more papers by this author
  • Jagoda Crawford,

    1. Australian Nuclear Science and Technology Organization, Institute for Environmental Research, Kirrawee, NSW, Australia
    Search for more papers by this author
  • Alan D. Griffiths

    1. Australian Nuclear Science and Technology Organization, Institute for Environmental Research, Kirrawee, NSW, Australia
    Search for more papers by this author


[1] We use 7 years of hourly radon observations at Mauna Loa Observatory (MLO), together with 10-day back trajectories, to identify baseline air masses at the station. The amplitude of the annual MLO radon cycle, based on monthly means, was 98 mBq m–3 (39 –137 mBq m–3), with maximum values in February (90th percentile 330 mBq m–3) and minimum values in August (10th percentile 8.1 mBq m–3). The composite diurnal radon cycle (amplitude 49 mBq m–3) is discussed with reference to the influences of local flow features affecting the site, and a 3-hour diurnal sampling window (0730–1030 HST) is proposed for observing the least terrestrially influenced tropospheric air masses. A set of 763 baseline events is selected, using the proposed sampling window together with trajectory information, and presented along with measured radon concentrations as a supplement. This data set represents a resource for the selection of baseline events at MLO for use with a range of trace species. A reduced set of 196 “deep baseline” events occurring in the July–September window is also presented and discussed. The distribution (10th/50th/90th percentile) of radon in deep-baseline events (8.7/29.2/66.1 mBq m–3) was considerably lower than that for the overall set of 763 baseline events (12.3/40.8/104.1 mBq m–3). Results from a simple budget calculation, using sonde-derived mixing depths and literature-based estimates of oceanic radon flux and radon concentrations in the marine boundary layer, indicate that the main source of residual radon in the lower troposphere under baseline conditions at MLO is downward mixing from aged terrestrial air masses in the upper troposphere.

1 Introduction

[2] Due to the rapid industrialization and population growth throughout central and southern East Asia [Kim et al., 2011; Zhang et al., 2009; Ohara et al., 2007; Zhang et al., 2004], the frequency and magnitude of anthropogenic pollution incursions into the atmosphere over the Pacific Basin are increasing [e.g., Liu et al., 2003; Martin et al., 2003; Board et al., 1999; Dibb et al., 1996; Balkanski et al., 1992, Merrill et al., 1989]. The magnitude of such events is of interest from both an environmental standpoint (regarding rates of nutrient/toxin deposition to surface waters of the Pacific and cross-boundary transport of atmospheric pollutants as far as west coast U.S.: He et al. [2012]; Dunlea et al. [2009]; Pochanart et al. [2004]; Guttikunda et al. [2001]; Streets et al. [2000]; Holmes and Zoller [1996]) and a regulatory one (regarding regional emission estimates and targets).

[3] The first step toward quantifying pollution events over the Pacific is defining reference “baseline” concentrations. Reliable identification of baseline periods is also important in the long-term monitoring of trends in climatically active gases [Perry et al., 1999; Pochanart et al., 1999; Thoning et al., 1989].

[4] Radon-222 (radon) is a relatively short-lived (3.82-day half-life), naturally occurring, radioactive gas whose oceanic flux is, on average, 2–3 orders of magnitude smaller than the corresponding terrestrial value [Schery and Huang, 2004; Lambert et al., 1982; Turekian et al., 1977; Wilkening and Clements, 1975]. Being a noble, poorly soluble gas that does not accumulate in the atmosphere, it is an ideal tracer of recent (<2–3 weeks) air-mass contact with ice-free terrestrial sources. Of the reported methods that have been used to investigate the frequency, intensity, and spatial extent of pollution incursions to remote regions of the atmosphere, the method based on radon is unique in that it is an unambiguous indicator of potential land-based pollution sources, irrespective of whether the corresponding fetch region is dominated by natural or anthropogenic sources of pollutants. Furthermore, radon emissions are more uniformly distributed over terrestrial regions than any anthropogenic species. For studies in which a generic indicator of terrestrial influence is required, radon-based techniques are therefore preferable to “tagged source” or “marked source” techniques [e.g., Huebert et al., 2001; Houwelling et al., 2000].

[5] The aim of this study was to use long-term radon concentration measurements and back trajectories to identify “baseline” air masses arriving at the Mauna Loa Observatory (MLO) and to characterize their seasonal variability. This will be accomplished using a radon-derived 3-hour diurnal sampling window, and increasingly stringent selection criteria based upon back trajectory analysis, and radon concentration distributions.

[6] All times reported in this study refer to Hawaiian Standard Time (HST = UTC-10 hours) and Northern Hemisphere seasons (summer: June–August; autumn: September–November; winter: December–February; spring: March–May). Radon concentrations, reported in mBq m–3, have not been standard temperature and pressure corrected.

2 Site and Methods

2.1 Atmospheric Radon Over the Northern Pacific Basin

[7] The vertical structure of the atmosphere overlying the northern Pacific Basin comprises a mixed layer (ML), typically 400–1000 m above sea level (asl), a cloud-containing layer (CCL; 0.2–2.5 km deep, when present), a trade-wind inversion (TWI; 100–300 m deep), the lower troposphere (LT; extending from the TWI to 3–5 km asl; characterized by seasonally varying trade-wind influenced flow), the upper troposphere (UT; ≥5 km asl; characterized by westerly jet-stream–influenced flow), the tropopause (8–10 km asl in winter; 15–17 km asl in summer), and the stratosphere above.

[8] As a result of the seasonal migration of the North Pacific anticyclone, the mid-North Pacific LT switches between easterly dominated trade-wind flow in May through September to westerly dominated flow for the remainder of the year [Hahn et al., 1992]. In mid-summer, air masses within the northeast trade-wind regime can spend weeks over the North Pacific without making landfall [Lee et al., 1994; Harris et al., 1992]. The UT of this region, however, is influenced by the westerly flow of the subtropical jet stream throughout the year.

[9] Over continental regions, radon concentrations usually decrease very rapidly with altitude, often dropping by more than an order of magnitude across the daytime terrestrial boundary layer [Williams et al., 2011; Liu et al., 1984]. Over the eastern North Pacific, however, a very different radon profile exists. In the summer of 1994, Kritz et al. [1998] conducted 11 vertical radon profiles near the Californian coastline between the surface and 12 km asl. An envelope representing the spread of data retrieved during these profiles [Kritz et al., 1998, Table] is shown in Figure 1. The highest concentrations (0.7–1.5 Bq m–3), found in the ML/CCL, are attributable to synoptic-scale advection from the adjacent continental source combined with a small oceanic radon flux. Above the TWI, radon in the LT decreases rapidly to minimum values between 2 and 3.5 km. Between 4 and 6 km, however, the spread of concentrations increases once again to a secondary maximum in the UT. The contrasting levels of terrestrial influence on air within the UT and LT over the mid-North Pacific in summer reflects the switch in dominant influence from light northeasterly trade-wind flow below 5 km and westerly jet-stream flow above 5 km [e.g., Gupta et al., 2004; Hahn et al., 1992; Balkanski et al., 1992].

Figure 1.

Envelope of radon concentrations measured during 11 vertical profiles flown over the northeast Pacific between 3 June and 16 August 1994 adapted from Kritz et al. [1998].

2.2 MLO

[10] The MLO lies on the upper northern flank of the Mauna Loa volcano (4170 m asl). Situated in the mid-North Pacific (nearest land fetch ≈6300 km to the west or ≈4000 km to the east), the location of MLO (19.54°N, 155.58°W) is well removed from significant continental pollution sources. For appropriately selected conditions, the high elevation of the station (3397 m asl), which is well above the typical TWI (2–2.5 km), provides direct access to air masses between 3.3 and 4.2 km asl [Perry et al., 1999; Ryan, 1997; Lee et al., 1994; Harris et al., 1992].

2.3 MLO Radon Observations

[11] A new 1500 L dual flow loop, two-filter radon detector was installed at MLO in December 2003. For details regarding the two-filter detection method in general, and the specific implementation employed at MLO, the reader is referred to Chambers et al. [2011], Whittlestone and Zahorowski [1998], Newstein et al. [1971], and Thomas and Leclare [1970].

[12] The MLO radon detector provides continuous, hourly observations, has a response time of 45 minutes (defined here as the time to reach 50% of the maximum count rate after a step increase in radon concentration), and, at the time of commissioning, had a lower limit of detection (defined here as the radon concentration at which the counting error is 30%) of 30 mBq m–3.

[13] Sampling was conducted at 38 m agl from an adjacent 40 m tower using an inlet line that terminates in a “goose-neck” style inlet to minimize the ingestion of precipitation. All detector output was recorded at 30-minute intervals. Final concentrations (and corresponding meteorological data) were aggregated to hourly values in postprocessing. The adopted time stamping convention is such that, for instance, the 1200 HST radon concentration is derived from the sum of half-hourly counts taken from 11:30 to 12:30.

[14] Instrumental background checks were routinely performed on the detector every 3 months, and calibrations were performed monthly using a 19.5 ± 4% kBq Radium-226 source (Pylon Electronics;, traceable to National Institute of Science and Technology standards.

[15] The analyses presented in sections 3 and 4 are based on a total of 7 years of observations (January 2004 to December 2010). Accounting for all calibration events, instrumental background checks, and maintenance periods or failures, the average data recovery rate for this period was >94%.

2.4 Air-Mass Back Trajectories

[16] For regions with complex local meteorology, or when the longer-term air-mass fetch history can be highly variable, employing a combination of radon observations and air-mass back trajectories provides a far superior means of interpreting fetch behavior and pollutant concentrations than interpretations based on local meteorological measurements alone. All back trajectories for this study were calculated using the PC version of the National Oceanic and Atmospheric Administration's HYSPLIT v4.0 model [Draxler and Hess, 1998]. HYSPLIT is an offline Lagrangian transport model that was forced with meteorological fields taken from National Centers for Environmental Prediction Final Operational Global Analysis data (original data are available from

2.5 Identifying Baseline Events at MLO

[17] The fact that, by virtue of its elevation, the MLO site samples directly from the LT sets it apart from many other World Meteorological Organization Global Atmosphere Watch Baseline stations, which tend to sample air that has spent the majority of its recent history in the atmospheric boundary layer (ABL). The Cape Grim Baseline Air Pollution Station (CGBAPS) in Tasmania, for example, samples ABL air that has been in direct contact with a range of surfaces on synoptic timescales, including the Tasmanian landmass, the Australian mainland, and the vast expanse of the Southern Ocean. Consequently, within the CGBAPS record, a clear distinction exists between monthly mean radon concentrations and corresponding 10th percentile values, with the latter predominantly representing the oceanic ABL (Figure 2a). In such cases, baseline air is identified with the extremely “clean” oceanic airstream, and baseline events can be isolated all year round using radon measurements alone, by specifying a constant threshold concentration (e.g., 100 mBq m–3; dashed line in Figure 2a) below which terrestrial influence (and, by extension, anthropogenic pollution) is considered to be minimized.

Figure 2.

Monthly mean radon concentrations and corresponding 10th percentile values at (a) Cape Grim and (b) MLO. The Cape Grim radon baseline threshold of 100 mBq m–3 is indicated with a dashed line.

[18] In contrast, mean monthly radon concentrations at MLO tend to be much lower than at CGBAPS, and there is a large variability in monthly 10th percentile radon concentrations that overlaps significantly with the range of mean monthly values (Figure 2b). These characteristics are a consequence of the fact that MLO is a mid-ocean site that samples predominantly from the LT, with remote fetch variations occurring mainly on seasonal (rather than synoptic) timescales. This precludes the specification of a constant baseline threshold radon concentration. Furthermore, while it would theoretically be possible to use the month-by-month 10th percentile radon concentration as a variable baseline threshold, it would not be possible to distinguish between air masses that had low radon as a result of the air being aged and well mixed in the LT and cases when the radon concentration was low due to the influence of a stratospheric injection event (e.g., tropopause folding as a result of fronts or other severe weather events).

[19] For the reasons outlined above, the use of radon measurements alone to identify baseline events at MLO is problematic, and we instead choose to employ a combination of back trajectories and radon concentrations for this purpose. Since it is the intention of MLO observations to represent tropospheric conditions, for the purposes of this study we define baseline events to mean tropospheric air masses, specifically, that have been removed from significant terrestrial pollution sources for at least 2 weeks. Such air masses are considered to be well mixed in the LT and representative of the seasonally varying background tropospheric conditions in the mid-North Pacific. The selection process employed seeks to minimize both the influences of stratospheric air intrusions as well as all local island effects.

3 Results

[20] Prior to establishing radon- and trajectory-based criteria for baseline (least terrestrially perturbed) tropospheric events observed at MLO, and because this is the first publication of results from the improved MLO radon detector, we first discuss the seasonal and diurnal cycles of radon concentration over the 2004–2010 observation period.

3.1 Seasonal Radon Cycle at MLO

[21] The mean annual radon concentration at MLO between 2004 and 2010, based on all hourly data, was 102 mBq m–3, with individual years ranging from 84 to 115 mBq m–3 (Table 1). In contrast, the mean annual radon concentration between 1997 and 2003, based on all hourly measurements from the previous detector, was 156 mBq m–3, with individual years ranging from 135 to 182 mBq m–3. While a detailed comparison of the two detectors is beyond the scope of this study, tests have indicated that the two data sets can provide a similar monthly mean distributions in the 2200–0800 HST diurnal window if a constant scaling factor (of approximately 0.76) is applied to the 1997–2003 data. However, the 1997–2003 data set is not suitable for defining radon concentrations in the 0730–1030 HST diurnal window proposed in section 3.2 of this article.

Table 1. Yearly Mean and Median Radon Concentrations at MLO for All Hourly Data, Data Within the 2000–0800 HST Window, and Data Within the 0730–1030 HST (Nonwinter) or 0830–1130 HST (Winter) 3-Hour Window
 All Hourly Data2000–0800 HST Window3-Hour Window
YearMean RnMedian RnMean RnMedian RnMean RnMedian Rn

[22] The seasonal cycle of radon at MLO, based on the sampling window proposed in this study (see section 3.2), was characterized by a late-winter maximum (February 90th percentile: 330 mBq m–3) and late-summer minimum (July 10th percentile: 8.1 mBq m–3). The amplitude of the seasonal radon cycle at MLO, based on mean monthly values, was 98 mBq m–3; a summary of monthly concentration statistics for the observation period is provided in Table 2. The observed seasonal variability (Figures 3 and 4) was characterized by low values throughout summer and high values throughout late winter and early spring, consistent with earlier reported studies [e.g., Zahorowski et al., 2005; Balkanski et al., 1992].

Table 2. Monthly Mean Radon Concentrations and Corresponding Distribution (10th–90th Percentiles) Based on Observations at MLO in the 3-Hour Diurnal Window Between 2004 and 2010 Inclusive
Month10th25th50th75th90thMean RnSDN%
  1. SD, standard deviation.

Figure 3.

Seasonal variability of radon concentration at MLO as daily mean values of observations within the 0730–1030 HST diurnal sampling window for the period 2004–2010.

Figure 4.

Monthly distributions of hourly radon concentration at MLO for the period 2004–2010 based on data within the 0730–1030 HST sampling window.

[23] Much of the observed variability can be attributed to seasonal migration of the Intertropical Convergence Zone and subtropical jet stream [Gregory et al., 1997; Talbot et al., 1997; Balkanski et al., 1992; Kritz et al., 1990; Harris and Kahl, 1990; Merril et al., 1989; Darzi and Winchester, 1982]. During the winter-spring transition, the subtropical jet stream is positioned over central or southern Asia and transports dust and/or anthropogenic pollutants (lifted by deep convection or frontal activity) across the North Pacific at a latitude observed by MLO [e.g., Andreae et al., 1988; Anderson and Larson, 1974]. This is evident in higher radon concentrations and variability in these months, as shown in Figure 3. In contrast, the low radon concentrations and variability characteristic of summer months at MLO are mainly attributable to LT air masses that are virtually stagnant within the Pacific, which, in some cases, have not made landfall for weeks [e.g., Fuelberg et al., 1999; Browell et al., 1996; Lee et al., 1994]. For these cases, conditions in the lower atmosphere of the mid-North Pacific are approaching equilibrium with respect to rates of vertical radon exchange. In late autumn to early winter, it is common for air masses originating from North America to travel as far as Hawaii into the Pacific Basin [Jaffe et al., 1997; Balkanski et al., 1992]. However, as evident from Figure 3, these events affect radon concentrations much less than the jet-stream transport in late winter to early spring.

[24] As evident from the 10th percentile radon values in Table 2 and Figure 4, as well as numerous other investigations [e.g., Jaffe et al., 1997, CO; Holmes and Zoller, 1996, mineral aerosols; Harris and Kahl, 1990, trajectories; Savoie and Prospero, 1989, n.s.s. SO442–], July to September corresponds to the main period of least terrestrial influence on air masses in the region.

3.2 Diurnal Radon Cycle

[25] The location of MLO, near the summit on the northern flank of a large volcano, adds considerable complexity to sampling efforts [Ryan, 1997; Hahn et al., 1992; Mendonca, 1969]. The locally observed wind patterns are a combination of 1) tropospheric flow around a large obstacle and 2) anabatic/katabatic flow up and down the flanks of the volcano. This makes the validity of data interpretations reliant upon an understanding of the mountain wind field [Ryan, 1997].

[26] Since the slope of the terrain in the vicinity of the station is ~8° (as is typical of shield volcanos), anabatic (upslope) winds during the day and katabatic (downslope) winds during the night have a substantial impact on observations at MLO over the diurnal cycle. Upslope winds generally prevail during the day, and observations are strongly influenced by the local island surface [Ryan, 1997; Bodhaine, 1996; Ridley and Robinson, 1992; Mendonca, 1969]. Downslope winds prevail at night, however, and observations are usually considered to be representative of the lower troposphere above the station [Perry et al., 1999; Ryan, 1997; Lee et al., 1994; Harris et al., 1992]. While evidence points to some air masses arriving at MLO from heights in the troposphere between 4.8 and 5 km, on only 8% of the time were they from above 4.2 km [Ryan, 1997].

[27] Due to the diurnal switching of flow regimes at the station, it is a well-established practice to define sampling windows within the diurnal cycle to better target air masses of interest. Various meteorological and tracer techniques have been used to define sampling windows to isolate periods of consistent downslope drainage flow, which is usually considered to be the best mechanism for bringing tropospheric air to the site with minimum terrestrial influences (e.g., 0000–0800 HST: Bodhaine [1996]; Dentener et al. [1999]; 2000–0600 HST: Mendonca [1969]; 2200–0800 HST: Perry et al. [1999]; Hahn et al. [1992]; 2200–1000 HST: Ridley and Robinson [1992]). In contrast to other tracer species, radon is not affected by airborne or waterborne pollutant sources, or secondary chemical processes, and is therefore an unambiguous indicator of terrestrial influence on an air mass. As such, it is the most reliable tool for determining an appropriate diurnal sampling window that minimizes local island effects on MLO observations, an important step in the identification of baseline events.

[28] Diurnal composite radon concentrations at MLO based on the 2004–2010 data set are presented in Figure 5. We note that, due to the superior performance of the present MLO radon detector, the details of this diurnal curve differ from the 2001 results of Zahorowski et al. [2005] using the previous detector system. The diurnal cycle of radon observed at MLO is closely linked to the diurnal evolution of the local boundary layer (BL), which, as discussed by Mendonca [1969], is markedly different to the typical diurnal BL evolution over land.

Figure 5.

Diurnal variability of radon concentration at MLO. (a) Distributions (10th/50th/90th percentiles) of hourly radon concentrations for all years/seasons. (b) Median hourly radon concentrations in summer and winter.

[29] At a flat, inland site, near-surface radon concentrations typically increase through the night as local emissions are trapped within the overlying stable layer, reaching their maximum around sunrise when stability is strongest and near-surface gradients are large. Concentrations then reduce rapidly throughout the morning as the BL grows, entraining less radon-rich air from the overlying residual layer, and increasing the volume within which near-surface radon, and new radon emissions, are mixed. Minimum concentrations are usually observed in the early afternoon when the BL depth is at its maximum (1–2 km). Generally speaking, the radon surface source term remains fairly constant throughout the entire diurnal cycle.

[30] In contrast, at MLO, radon concentrations are relatively stable throughout much of the night (Figure 5). From around 2100–2200 HST until sunrise, the station is under the influence of a katabatic drainage flow [Mendonca, 1969; Bodhaine, 1996; Hahn et al., 1992] that is ~55 m deep on average [Mendonca, 1969]. Under these flow conditions, the incoming tropospheric air stream (typically low in radon) picks up a small additional amount of radon from the 6–7 km of mountainside fetch that it traverses on the way down from the volcano's caldera, 770 m above MLO. After sunrise (~0600 HST; Figure 5), the katabatic flow slows and eventually stops. During this period, the regional-scale horizontal tropospheric flow is frequently oriented east-west [Bodhaine, 1996; Hahn et al., 1992] and there has been sufficient heating to break down the capping inversion of the local nocturnal boundary layer, but the anabatic flow regime has not yet become established [Ryan, 1997]. Low-radon tropospheric air continues to arrive at the MLO site and is now brought directly to the surface by the entrainment process as the daytime mixed layer begins to grow. This results in a mid-morning minimum radon concentration, indicating the “cleanest” conditions prior to the onset of anabatic winds (~1100 HST; Figure 5). Ridley and Robinson [1992] also reported the onset of anabatic winds at MLO to occur around 1100 HST, while Mendonca [1969] suggests they start an hour earlier.

[31] The mixed layer continues to grow for several hours after the onset of anabatic winds, to reach a maximum height of around 600 m in the mid-afternoon [Mendonca, 1969]. Importantly, however, under anabatic flow conditions, the source of air is the marine boundary layer (not the lower troposphere) and the radon fetch region becomes the lower mountain slopes and low-lying island regions, which are more soil rich than the upper mountain slopes. Not only is the radon source strength much higher in these regions, but the land fetch is almost an order of magnitude greater (50–60 km). The combination of longer fetch and higher radon flux throughout the anabatic flow regime result in maximum radon concentrations being observed at MLO in the mid- to late-afternoon, despite mixing depth being comparatively deep at this time. Anabatic wind conditions are the most likely to contain anthropogenic emissions from nearby townships or natural emissions from the surrounding tropical vegetation. Assuming that tropospheric and marine BL air have similarly low initial radon concentrations, and accounting for the change in BL depth, the enhancement of radon concentration under anabatic flow is almost 20 times greater than that observed under katabatic flow.

[32] After sunset (~1800 HST; Figure 5), the anabatic flow slows, and eventually stops, resulting in a gradual reduction in radon concentrations until the katabatic flow is reestablished around 2100–2200 HST.

[33] The observation of a consistent and relatively low radon concentration during the main katabatic flow period (e.g., ~2000–0600 HST) has previously been interpreted as being indicative of a nocturnal sampling window that is characteristic of the well-mixed lower troposphere [e.g., Zahorowski et al., 2005; Dentener et al., 1999]. However, evidence presented here from the improved radon detector of a further drop in radon concentration to a mid-morning minimum can only be interpreted as sampling air with an even smaller terrestrial influence (i.e., more direct from the lower troposphere).

[34] For sufficiently strong tropospheric flow, during the lull period in radiation winds (~0800–1000 HST), tropospheric air may arrive directly at the MLO mast without significant land contact. Whether it is via direct contact with the tropospheric airstream, or mixing of tropospheric air downward by entrainment in the vertically growing morning BL, there can be no doubt that a minimum in the diurnal radon signal must correspond to a minimum terrestrial influence in the sampled air because of the unique properties of radon.

[35] It is of interest to consider the possible effects of contributions from local radon sources during downslope flow conditions in the 2200–0600 HST window. The observatory is located 6–7 km north of the volcano's summit. Under stably stratified nocturnal conditions, which give rise to katabatic flows, a nocturnal drainage flow develops down the flanks of the volcano. It has been estimated that this drainage flow has an average depth of h = 55 m and that the entire 40 m sampling mast at MLO lies within the resultant nocturnal boundary layer (NBL) of the volcano flank on approximately 94% of evenings [Mendonca, 1969]. As tropospheric air masses travel from the volcano summit to MLO, they pick up small amounts of radon along the d = 6–7 km fetch. Using a representative downslope wind speed of V = 6 ms–1 [Mendonca, 1969] and a radon flux density of F = 1.3 mBq m–2 s–1 for thin, dry Hawaiian soils [Whittlestone et al., 1996] in lieu of measurements over the ~30-year-old lava flow on the upper flanks of the volcano (i.e., an upper estimate), the elevation of radon concentrations in the NBL above tropospheric background levels (C–Cb) can be estimated as:

display math

This compares well with the magnitude of the drop in concentrations observed when the downslope wind abates after sunrise (Figure 5a).

[36] Assuming that advection of lower tropospheric air past the tower, and/or entrainment of tropospheric air within the growing BL, are the predominant contribution to the observed radon signal during the morning transition period, we propose that the 0730–1030 HST window (i.e., constituting the 0800, 0900, and 1000 HST samples, Figure 5a) provides the best opportunity for tropospheric air masses to reach the MLO measurement mast with minimal terrestrial influence. In support of this finding, condensation nuclei (CN) measurements at MLO by Bodhaine [1996] in summer, and others by Walega et al. [1992], also indicated a minimum at around 1000 HST, with an associated shift in wind direction, and sharp increase in CN values thereafter (see also section 3.5).

[37] To further test the suitability of the above proposed 3-hour sampling window for investigating baseline conditions at MLO year round, we separated the 7 years of observations by season and compared diurnal cycles for the seasonal extremes (Figure 5b). For winter, 0830–1130 HST was the most suitable sampling window, but for all other seasons 0730–1030 HST remained the most suitable sampling window.

[38] Ultimately, the choice of sampling window at MLO will be investigation specific. The 3-hour (0730–1030 HST) window excludes anabatic, and reduces katabatic effects, and is therefore most relevant for measurements of trace species that may be affected by local influences downslope of the station (e.g., anthropogenic sources). For measurements that may be more sensitive to aerosols or gaseous emissions from either the caldera (summit) or the intervening lava field than they are to emissions from lower lying parts of the island, then the 0830–1130 HST window (which further reduces katabatic effects, but includes some anabatic influence) may be more appropriate. However, both of these suggested 3-hour windows are very restrictive and may exclude too many data points for some studies. In such cases, we would recommend an extended sampling window that includes the nocturnal downslope flow, but avoids the daytime onset of anabatic winds (say, 2200–1000 HST). For sufficiently large data sets, however, if all traces of local terrestrial/volcanic influence need to be minimized, the 0730–1030 HST window is the most effective means to characterize least terrestrially effected lower tropospheric air masses over the mid-North Pacific from MLO data sets.

3.3 Identification of MLO Baseline Periods

[39] For characterization of baseline air masses at MLO, we have cataloged the least terrestrially perturbed events over the period 2004–2010 using a combination of back-trajectory analysis, radon concentration distributions, and radon-derived diurnal sampling window. Progressive analysis of the radon concentration distributions was used to provide an indication of the efficacy of each of the trajectory selection steps. Throughout this process, several potential selection techniques were ruled out on the basis of the resulting radon distributions.

[40] In all, there are five steps in the selection process outlined below: steps 1 and 2 minimize remote continental influence (e.g., from Asia and North America), step 3 minimizes the terrestrial influence of other islands in the Hawaiian chain, step 4 minimizes local terrestrial influence (from the island of Hawaii) resulting from anabatic/katabatic flow, and step 5 targets the slowest moving central North Pacific air masses representative of persistent anticyclonic conditions (i.e., excluding strong jet-stream influence and frontal passage/tropopause folding events).

[41] Initially, we prepared a summary database containing a 10-day MLO back trajectory for every hour of the 7-year observational period, into which we incorporated the matching hourly radon time series (resulting in 57,774 valid events of a possible 61,368).

  • Step 1.From the summary database, we rejected all trajectories that passed over land within the 10-day hind-cast period, as defined by the North Pacific Basin boundary shown in Figure 6 (leaving 23,359 events).
  • Step 2.From the step 1 subset, we further rejected events whose trajectories traveled for any length of time below 1.5 times the marine boundary layer depth, as calculated by HYSPLIT, to avoid contaminated outflows from distant continents trapped within the marine ABL (leaving 18,767 events).
  • Step 3.To minimize possible influences on observed radon concentrations from other islands in the Hawaiian chain, we then defined a 6° × 6° grid cell around MLO (see Figure 6, red dotted line) and rejected any of the step 2 trajectories that traveled below 3500 m asl within this region at any time except the last 2 hours of transit (leaving 8963 events).
  • Step 4.A diurnal composite plot of radon concentration based on the remaining 8963 events in the step 3 database indicated a mid-morning minimum at the same time as previously shown in Figure 5a. Consequently, we next applied our 0730–1030 HST sampling window to minimize local island radon influences (due to anabatic/katabatic flows), which reduced the set of possible baseline events to 1621.
  • Step 5.When the central North Pacific atmosphere is almost stagnant, as often occurs in summer, MLO air masses typically stay within 1000–2000 km of Hawaii for the entire 10-day hind-cast period. Such near-stagnant air masses are most likely free from strong jet-stream influences or frontal passage mixing events (e.g., tropopause folding). To isolate these slowest moving air masses, we grouped the remaining 1621 events in 3-month blocks (January–March, April–June, July–September, October–December) based on monthly 10th percentile radon concentrations (Table 2) and performed a cluster analysis on each group's trajectories (based on five cluster groups). Clusters within each 3-month group whose mean path traveled farthest from MLO always exhibited the broadest distribution of radon concentrations. We therefore retained only three (of five) clusters per 3-month group, corresponding to the slowest moving air masses. The remaining 763 events were used to characterize baseline tropospheric conditions at MLO.
Figure 6.

Representations of the assumed Pacific Basin boundary for air masses (black dotted line) and imposed grid cell around the sampling location (red dotted line).

[42] Distributions of hourly radon concentrations corresponding to the events retained at each step of the above selection process are presented in Figure 7a. Although each step of the process results in a reduction in median radon concentration of the associated group of events, steps 1 (identifying Pacific trajectories to remove obvious distant land contact) and 4 (applying the 3-hour sampling window to remove direct local land influence by anabatic/katabatic flows) had the largest effect on reducing the terrestrial signal of observed air masses.

Figure 7.

(a) Radon distributions (10th/50th/90th percentiles) for each stage of the data-selection process and (b) progressive reduction in number of valid events.

[43] The distribution (10th/50th/90th percentile) of radon concentrations for the final 763 events was 12.3/40.8/104.1 mBq m–3. Radon distributions, by 3-month group, of these events, along with the number of events on which each distribution is based, are shown in Figure 8. Based on median values, there is still a small seasonal variability in baseline radon concentrations at MLO (amplitude 35 mBq m–3), which highlights how challenging it can be to find central North Pacific air masses in the nonsummer months that show little signs of recent terrestrial influence, despite the large distance from the nearest continents.

Figure 8.

(a) Three-monthly distribution of hourly radon concentrations at MLO for the period 2004–2010, based on identified baseline tropospheric events within the 0730–1030 HST diurnal sampling window (c.f. Figure 3b) and (b) total number of baseline events arranged by 3-month group.

[44] The complete set of 763 hourly events and their corresponding radon concentrations is available as a supplementary database to this article [see Supplementary_DB.xlsx]. This database is intended for use in a range of MLO baseline studies in which radon criteria can be introduced to set additional requirements regarding residual terrestrial influences on selected air masses.

[45] Since no air mass identified within this set has had direct contact with land for more than 10 days, the remaining radon variability will most likely reflect the degree of exposure/interaction of the air mass to terrestrial sources at the last point of land contact; a similar variability might be expected in pollutants of terrestrial origin observed under baseline conditions (see section 3.5).

3.4 Least Terrestrially Perturbed Air in the North Pacific

[46] Based on 90th percentile values, a winter-spring maximum and summer minimum are still evident in Figure 8a, indicating a vestigial continental influence (i.e., within the past 2–3 weeks) on the identified baseline air masses.

[47] With a view to facilitating interannual investigations of the most pristine (or deep baseline) conditions at MLO, we now focus attention specifically on the period July–September of each year, when overall terrestrial influence is minimized. Of the 763 baseline events identified between 2004 and 2010, 196 occur in the July–September window and are cataloged in Table 3 as deep-baseline events. The distribution (10th/50th/90th percentile) of radon concentrations for deep-baseline events was 8.7/29.2/66.1 mBq m–3, respectively. Our median estimate of 29.2 mBq m–3 for deep-baseline air events in the mid-northern Pacific agrees well with other estimates in the literature for “clean” North Pacific air (8–37 mBq m–3, Balkanski et al. [1992]; 17–37 mBq m–3, Lambert et al. [1982]; 25–39 mBq m–3, Moore et al. [1977]).

Table 3. Deep-Baseline Events
YearJulian DayHourRadonσYearJulian DayHourRadonσYearJulian DayHourRadonσ
  1. Least terrestrially affected summertime air masses arriving at MLO in the period 2004–2010 identified by day, time, and radon concentration (mBq m–3). Error estimates have been provided for each radon concentration by expressing the standard deviation (σ) of the hourly count rate as a concentration. Times indicate midpoints of 1-hour integration periods. See text for details.


3.5 Evaluation and Application of the Radon Baseline Data Set

[48] To evaluate our proposed baseline selection method, we performed preliminary analyses on corresponding MLO CH4, CO2, and O3 data sets ( It should be noted that, for the purpose of these examples, only simple averaging was performed on the trace-gas results for Figures 9c and 9d. Since no consideration was given to seasonal trends, absolute concentrations shown may not be truly representative.

Figure 9.

(a and b) Diurnal composite CH4 and CO2 at MLO from 2004–2010 and (c and d) comparison of mean CH4 and O3 in all baseline MLO air masses with corresponding concentrations in the 1st and 4th quartile radon baseline air masses. Red lines in (b) placed as guides to the eye following the nocturnal and near-noon trends in CO2 concentration.

[49] In support of our choice of diurnal sampling window, the 7-year diurnal composite of hourly deviations from the daily mean MLO CH4 measurements (Figure 9a) were found to exhibit a distinctly lower mean concentration within the 0730–1030 HST window than the more traditionally adopted 2200–0800 HST or 2000–0600 HST diurnal sampling windows, implying a small CH4 source from either the caldera or intervening lava field. Similarly, the corresponding diurnal composite CO2 record (hourly deviations from the daily mean; Figure 9b) shows a break in slope over the 0730–1030 HST diurnal window between periods of more uniformly varying concentration under strictly katabatic or anabatic flow (Figure 9b, red lines). Since the mid-volcano flanks are not well vegetated, this slight reduction in CO2, compared to nocturnal values, most likely reflects the removal of a small CO2 source—either from the caldera or along the 6–7 km of upper volcano fetch—when the katabatic flow is “switched off,” rather than CO2 drawdown during the initial stages of anabatic wind development.

[50] We now demonstrate the value of our radon-derived baseline events using hourly MLO measurements of O3 and CH4. Prior to entering the Pacific Basin, identified “baseline” events experience different degrees of interaction with terrestrially based pollution sources. A record of each baseline event's “potential for being polluted” is thus contained in the vestigial radon concentration, enabling the information in Table 3, as well as the supplementary database, to be used to better interpret the variability of trace-gas measurements under baseline conditions. The filled black circles in Figures 9c and 9d represent monthly mean concentrations of CH4 and O3, respectively, under baseline conditions. We then ranked the monthly baseline data by radon concentration and took events corresponding to the 1st quartile radon concentrations to represent baseline events that had experienced minimal terrestrial influence (a low “pollution potential”). Likewise, we took events corresponding to the 4th quartile radon concentrations to represent baseline events with the highest pollution potential. The open circles and open squares in Figures 9c and 9d, respectively, represent CH4 and O3 concentrations of baseline air masses with the lowest and highest pollution potential. Clearly, for periods of the year when tropospheric continental outflow events dominate (i.e., February–April), there can be a marked variability in the pollutant concentrations within even baseline air masses. Most importantly, though, radon observations provide a means of characterizing these influences.

4 Radon Sources for the Least Terrestrially Perturbed Events

[51] Although very small, radon concentrations of air masses associated with the set of deep-baseline events identified in section 3.4 were still not negligible. Given the range of vertical radon distributions observed through the troposphere over the North Pacific, discussed in the introduction (see Figure 1), it is of interest to determine whether the main source of residual radon in these baseline air masses is the ocean surface below (via the marine boundary layer) or distant land masses to the west (via the upper tropospheric jet stream), since this will also reflect the dominant source of other atmospheric contaminants observed under deep-baseline conditions at MLO.

[52] Several estimates of the oceanic radon flux density in the North Pacific are available in the literature, varying from Fsfce = 0.07 to 0.0925 mBq m–2 s–1 [Schery and Huang, 2004; Guedalia et al., 1974; Peng et al., 1974]. Radon concentrations in the marine boundary layer (MBL) over the mid-North Pacific are typically CMBL = 30–50 mBq m–3 [Zahorowski et al., 2005; Balkanski et al., 1992; Lambert et al., 1982; Moore et al., 1977]. Although summertime MBL depths (hMBL) can range from 400 m to 3 km asl, including coupled cloud layers [Wood and Bretherton, 2004; Kaneyasu et al., 2000; Gregory et al., 1997; Wu et al., 1997; Lee et al., 1994; Johnson et al., 1993; Hahn et al., 1992; Fuelberg et al., 1999; Greenberg et al., 1990; Andreae et al., 1988; Kritz, 1983], summertime values calculated for our 7-year study period from nearby atmospheric soundings at Lihue on the Hawaiian island of Kauai [IGRA, 2011] were typically around hMBL = 900 m asl. We therefore take Fsfce = 0.08 mBq m–2 s–1, CMBL = 40 mBq m–3, and hMBL = 900 m. Assuming equilibrium conditions, an estimate for the MBL radon venting flux (Fvent) through the trade-wind inversion can then be obtained from a simple budget calculation (see also Figure 10a):

display math

where λ = 2.0982 × 10–6 s–1 is the radioactive decay constant for radon. This negligible venting flux is consistent with a strongly defined trade-wind inversion inhibiting mixing from the MBL into the lower troposphere [Kritz, 1983], as expected for slow-moving air masses within an anticyclonic weather system.

Figure 10.

(a) Schematic of the Pacific Basin atmosphere to which the budget estimate applies and (b) comparison of relative flux magnitudes across the interfaces identified in (a) for deep-baseline tropospheric events for a range of Fsfce and CMBL values.

[53] Taking the transition between the upper and lower troposphere to be at around hLT = 4 km (Figure 1), and using our median estimate of CLT = 29.2 mBq m–3 for the least terrestrially perturbed air in the mid-North Pacific LT (section 3.4), we can furthermore estimate the radon flux between the upper and lower troposphere as:

display math

[54] To gauge the robustness of these calculations, given the considerable uncertainty in oceanic radon flux term and MBL radon concentration, we also applied the budget above to low (Fsfce = 0.07 mBq m–2 s–1; CMBL = 30 mBq m–3), high (Fsfce = 0.0925 mBq m–2 s–1; CMBL = 50 mBq m–3), and crossed (Fsfce = 0.0925 mBq m–2 s–1; CMBL = 30 mBq m–3) parameter range cases (see Figure 10b). For all cases, downward transport of radon from the UT was the dominant influence on deep-baseline LT radon concentrations.

[55] These estimates suggest that the predominant source maintaining residual radon concentrations in the lower troposphere during the deep-baseline events identified in our study is a small downward transport of air from aged terrestrial air masses in the upper troposphere. This air was presumably transported more than 6000 km eastward across the North Pacific in the subtropical jet stream.

5 Summary and Conclusions

[56] We have described the 1500 L dual flow loop two-filter radon detector that was installed at the MLO in December 2003 and summarized the first 7 years (2004–2010) of continuous hourly observations. Yearly mean radon concentrations at MLO over the observation period varied from 84 to 115 mBq m–3. The seasonal cycle of radon at MLO was characterized by a late-winter maximum (Febuary 90th percentile: 330 mBq m–3) and late-summer minimum (August 10th percentile: 8.1 mBq m–3), with an amplitude (based on mean monthly values) of 98 mBq m–3 (39–137 mBq m–3).

[57] The annually averaged diurnal cycle of radon was characterized by a late-afternoon maximum, mid-morning minimum (with an amplitude of 49 mBq m–3), and relatively stable values between the hours of 2200–0600 HST. A pronounced dip in radon concentration (30 mBq m–3 below the stable nocturnal value) was observed mid-morning, corresponding to a local lull in anabatic/katabatic wind on the flanks of the Mauna Loa volcano. Within this dip, a 3-hour window comprising hourly samples collected between 0730 and 1030 HST, was chosen as the most appropriate temporal window for sampling the least terrestrially affected (baseline) tropospheric air masses.

[58] A comprehensive table of baseline events was compiled, using the new 3-hour sampling window, together with 10-day HYSPLIT back-trajectory information, over the entire seasonal cycle for the 7-year period and is catalogued along with corresponding radon concentrations and percentiles in a supplementary database. This database represents a comprehensive resource for the selection of baseline events at MLO using combined trajectory and radon-based criteria. Furthermore, a short list of 196 deep-baseline events in July–September was presented and discussed. The distribution (10th/50th/90th percentile) of radon concentrations in deep-baseline events was 8.7/29.2/66.1 mBq m–3, which is in close agreement with existing estimates of clean air in the Pacific Basin.

[59] Finally, we used estimates of MBL depth, oceanic radon flux, and radon concentrations in the MBL, together with a simple budget calculation, to estimate the source of the residual radon in the deep-baseline lower tropospheric air masses. Our conclusion was that, despite a nearest westerly land fetch of over 6000 km, most residual radon present observed in the lower troposphere at MLO during deep-baseline events is likely to have been mixed downward from aged terrestrial air masses in the upper troposphere. Similar processes presumably affect the concentrations of other trace species monitored at MLO under deep-baseline conditions.


[60] The authors would very much like to acknowledge the ongoing help and support of National Oceanic and Atmospheric Administration, Earth System Research Laboratory, Global Monitoring Division, specifically Ed Dlugokencky, Kirk Thoning, and Samuel Oltmans for access to hourly CH4, CO2, and O3 records, and staff at the Mauna Loa Observatory, particularly Paul Fukumura. We are also very grateful for the technical support of ANSTO staff Ot Sisoutham and Sylvester Werczynski in preparing the hardware and software of the MLO radon detector.