Baseline radon detectors for shipboard use: Development and deployment in the First Aerosol Characterization Experiment (ACE 1)

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. D13, PAGES 16,743-16,751, JULY 20, 1998 Baseline radon detectors for shipboard use: Development and deployment in the First Aerosol Characteriation Experiment (ACE 1) S. Whittlestone and W. Zahorowski Australian Nuclear Science and Technology Organisation, Lucas Heights, New South Wales, Australia Abstract. A new design of two-filter radon detector has been developed for measurement of extremely low levels of radon in the harsh environments on board ships and remote islands. These were needed for the First Aerosol Characteriation (ACE 1) multiplatform experiment in the Southern Ocean. By employing an internal recirculation system and a wire mesh screen as the second filter it has been possible to reduce the power consumption by as much as a factor of 10 and the weight and cost by a factor of 2 compared to current designs of comparable sensitivity. A very high efficiency of 0.38 count 1 1 s- Bq- radon in the instrument has been achieved by counting while sampling. This is a key parameter because the larger this number, the smaller the volume and power consumption of the detector. Two air flow paths are used to separate the high flow rate needed to prevent loss of radon daughters to the walls of the detector and the lower flow rate needed to change the air sample in the instrument. As a result, the inlet air lines and delay chamber needed to remove thoron are compact. With a volume of 750 L the detectors used on board ships for ACE 1 had a sensitivity of about 0.2 counts -1 Bq - m -3 and a lower limit of detection of 40 mbq m -3 for a 1 hour count. An instrument with a volume of 10 m 3 and incorporating improvements made since ACE 1 could be expected to have sensitivity of 3.7 c s - Bq -1 m-xand a lower limit of detection of 2 mbq m -3. At 45 min the time resolution is twice as good as that of instruments using a low internal flow rate, but not as good as instruments with a moving filter, where the sampling period is precisely defined. Dual-flow loop radon detectors with screens have the virtues of simplicity and freedom from routine maintenance. This new technology extends the range of sites at which baseline radon measurements can be made to remote areas with little regular technical support and a harsh environment. 1. Introduction Radon and condensationucleus (CN) measurements were included in the International Global Atmospheric Chemistry (IGAC) Project's First Aerosol Characteriation Experiment (ACE 1) [Bates et al., this issue] to indicate the extent to which the air samples were influenced by recent passage over land. The concentrations of trace constituents in air which has been over land is likely to vary over distance scales of tens of kilometers. On the other hand, air which is free from radon has been away from land for a long time and can be considered to be "baseline" in the sense that it represents a large volume of the atmosphere. In a multiplatform experiment such as ACE 1, where the distance between platforms could be changed, it was valuable to know what distance could be considered close Paper number 98JD /98/98JD that the detector should be able to distinguish whether the radon levels in an air sample were above or below the baseline radon threshold. The selection of the threshold value of 100 mbq m -3 is discussed in detail by Whittlestone et al. [this issue]. It was accepted that there was a trade-off between sensitivity and practicality. The design aimed to achieve a precision of 10% at 100 mbq m -3 for a 1 hour count (at least 0.28 c s -1 Bq - m-3), with a detector volume as small as possible, but certainly less that 1 m 3, which was considered a practical upper limit. To be consistent with this precision at the threshold, the instrument should be able to hold its calibration within 10% during a 1 month voyage. Instruments available in 1993 with adequate sensitivity [Whittlestone, 1985] could not tolerate the movement of a ship. Other designs [Hutter et al., 1990] could not be implemented with sufficient sensitivity without exceeding a volume of 1 m. A development of an instrument described by Schery et al. [1980] had been used for detection of very low levels of thoron during Mauna Loa Observatory Photochemistry Experiment enough for air samples to be comparable. Radon measurements are widely used for similar types of air mass characteriation [Prospero and Carlson, 1970; Wilkness et al., 1973; Polian et al., 1986; Krit, 1990; Whittlestonet al., 1992]. Their application in ACE 1 is described by Whittlestonet al. [this (MLOPEX-II) [Whittlestonet al., 1996] and was considered a suitable basis for a radon detector for use at remote sites or on issue]. ships. This paper describes the development of the detectors At the planning stage of ACE 1 in 1993 there were no radon and their deployment at Macquarie Island (southwest of Tasinstruments suitable for deployment on ships. It was decided mania at 55.4øS, 159øE) on the Southern Surveyor and on the Copyright 1998 by the American Geophysical Union. 16,743 Discoverer. There is also discussion of the intercalibration of these instruments and the Cape Grim radon detector which was also used for ACE 1.

2 16,744 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS Signal to data logger In order to understand other developments it is necessary to expand on the simple description of the two-filter detector given above. A major challenge in the design of two-filter Photomultiplier inlet air detectors is to prevent the daughters from being plated out on 625 mesh wire the walls of the tank. Most two-filter detectors use a high flow ZnS screen (second filter)l! rate to ensure that the air passes from the inlet to outlet filters L delay in a time short compared to the mean plate-out time. In a chamber De to detector with a volume of 1 or 2 m 3 the plate-outime is a few remove minutes even when care is taken to avoid turbulence. Thus flow Exhaust k',, flow diffuser : internal flow: loop blower First Filter thoron Extemal flow loop blower Figure 1. Schematic diagram of dual-flow loop two-filter detector. 2. Review of the Two-Filter Radon Detector The new radon detectors for shipboard use belong to the class of instrument known as two-filter detectors. The name is derived from the mode of operation: an air sample is drawn through one filter which removes all radon and thoron decay products ("daughters"), then through a delay chamber in which some daughters are produced. Finally, the air passes through a second filter which retains the daughters. The daughters on the second filter have been produced in controlled conditions, so their number is proportional to the radon and thoron concentrations. There are many variations of the two-filter detector. The simplest design has an easily removable second filter. After a prescribed sampling period, the filter is removed and placed in an alpha particle detector, based typically on a inc sulphide scintillatot [Thomas and LeClare, 1970]. The instrument responds to both radon (222Rn, half-life 3.82 days) and thoron (222Rn, half-life 55.6 s). If thoron is present, it decays to 2 2pb which has a half-life of hours and causes an unwanted rates of about a cubic meter per minute are necessary. As a result, the second filter area and the power requirements for the air pump have to be large. The design of the Cape Grim detector [Whittlestone, 1985] is similar to that of Schery et al. [1980] in its use of continuous counting with its inherent long response time. However, in the Cape Grim detector a constant number of particles is maintained in the tank. The daughters become attached to the particles, which have a lifetime of many hours in a 9 m 3 tank. A modest 200 L min -1 flow rate is all that is necessary to obtain satisfactory results. This system, as with the other continuous designs described so far, is of limited use when thoron is present. However, compared to detectors using a high flow rate, elimination of thoron is much easier because a delay chamber of a reasonable sie can be installed in the inlet line. At 200 L min -1 a 1000 L delay chamber would be sufficient to remove thoron. A 9 m 3 detector would have to use a flow of 5 m 3 min-1 to reduce internal plate-out. A 25 m 3 delay chamber would be needed to achieve adequate reduction of thoron. Grumm et al. [1990] improved on existing designs by using a moving filter system and an alpha spectrometer to count the decays. Their detector was therefore able to detect radon and thoron simultaneously, which eliminated the need for a delay on the inlet to remove thoron. However, the sensitivity of this instrument, limited by the sie of solid state alpha particle detectors, is too low to be used in marine baseline atmospheres. None of the previous designs emphasied low power consumption, low tank pressure, ability to operate on a ship, or minimum maintenance. The specification for radon detectors for ACE 1 forced consideration of these factors. background count in an instrument intended to detect radon. It is possible to lower the thoron background count to an acceptable level by delaying air in the inlet by a few minutes. 3. Development of the Dual-Flow Since thoron has a half-life of less than a minute, it decays Loop Two-Filter Detector before entering the main delay chamber. This simplest form of two-filter detector can be automated by adding a filter changing mechanism. Hutter et al. [1990] have done this for an application where there was effectively no thoron in the air and continuous operation was essential. Their detector retains the well-defined time resolution of the simple The detector design for the ACE 1 instruments is shown schematically in Figure 1. It is a major redesign of the instrument of Schery et al. [1980], which will be referred to as design A. The external loop functions to change the air inside the detector, which is necessary on a timescale of about half an detector. Its sensitivity per unit delay chamber volume was hour to ensure that the overall time resolution of the instru counts- Bq-, which is a factor of 3 less than achieved ment is less than 45 min. This results in a manageable sie of on the Southern Surveyor. thoron delay and a modest sied inlet filter. For a 1000 L Another design for an automated two-filter detector [Schery detector a flow rate of about 50 L min -1 is satisfactory. The et al., 1980] sacrifices time resolution to gain substantially in sensitivity by installing an alpha particle detector inside the delay chamber (tank) so that it counts all the decays during high flow needed to prevent loss of daughters to the walls of the delay chamber is achieved by using a second blower inside the chamber to transport the daughters through the second sampling. The time resolution is not as good as the previous filter within about a minute of their formation. An air flow example because daughters sampled at one time decay later, diffuser is placed at the outlet of this blower to minimie resulting in effective smoothing with a time constant of about turbulence in the tank. 45 min. For the instruments used in ACE 1 this was the time The second major modification of design A was to use a wire for the instrument to reach 50% of its maximum value on screen in place of a fiber or membrane medium for the second application of a step change in radon concentration at the inlet. filter. This is possible because of the high diffusivity of radon

3 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS 16,745 Table 1. Design Parameters and Performance of Dual-Flow Loop Two-Filter Detectors Southern Macquarie Cape New Detector Description Prototype Surveyor Discoverer Island Grim Version Active volume of the tank, L ,500 9, Diameter or side, mm , Internal loop blower power, W External loop blower power, W , Screen average diameter, mm Length, mm Area, mm 2 2,435 7,600 9,800 9,800 24,300 Mesh, per inch Inner loop flow rate, L min Screen face velocity, cm s External loop flow rate, L min Screen or filter retention r, % Plate-out loss p, % Loss from decay in tank d, % Total collection efficiency e c, *% Screen alphactivity Bq - m -3 radon Estimated alpha counting efficiency e, % Calculated sensitivity, * counts - Bq - m Actual sensitivity, counts- Bq - m Actual/calculated sensitivity, % Sensitivity/volume, counts s- Bq- Limit of detection equal to concentration with 30% not tested error in 1 hour count, mbq m -3 Time to reach 50% of maximum count rate after step increase in radon concentration at the inlet, min The relevant parameters are also shown for the Cape Grim detector, whic has a single-flow loop and a particle generator. *Here, ec = (1 - p/100) (1 - d/100) r. This is the percentage of daughters generated in the active volume of the tank, which are collected on the screen. *Calculated sensitivity equal to (screen alph activity Bq - m -3 radon) e /100. daughters. Inside the tank there are no aerosols, so the daugh- 2. Most of the 3.05 min half-life 2 Spo atoms are collected ters remain as either atoms or small clusters of atoms no more than a few nanometers diameter [Raes et al., 1985]. They are known as "unattached" radon daughters because they are not on the second filter because of the short air residence time. In the Cape Grim design these atoms are lost by decay in the tank, which reduces the maximum possible efficiency by a attached to the larger aerosol present in most outdoor envifactor of 2. ronments. The high diffusivity of these unattache daughters means that a wire screen is able to remove them with high efficiency and very low flow impedance. Woven wire mesh has been used widely for detection of unattached radon daughters [George, 1972; Strong, 1988; Solomon and Ren, 1992], and the penetration and retention characteristics of the screens can be estimated using the theory of Cheng and Yeh [1980]. Whittlestone et al. [1994] examine the 3. This configuration confines the high-pressure components to just the holder for the second filter and the internal blower. In design A the entire tank had to withstand the pressure drop across the second filter. 4. Since the major power consumption in these detectors is by the air pump, use of screens makes it possible to dramatically reduce system power requirements. 5. The detector is less sensitive to daughters which may performance of screens in some detail and despite investigaenter the instrument from outside as a result of an ineffective tion of many relevant parameters, achieve only fair agreement first filter or leaks. This is because filter defects are more likely between experimental and calculated efficiency for the proto- to permit the entry of attache daughters. Unattache daughtype, a small, 32 L detector. Table 1 gives the design parame- ters are retained on even a very poor filter by diffusion. Once ters for this detector as well as for the detectors used in ACE inside the detector, the same principle applies to retention of 1. Here it can be seen that the actual sensitivity of the 32 L detector was 60% of the value calculated using equations derived in the appendix. It is considered that the major uncertainties in the calculation arise from the effect of nonuniform face velocity across the screen, the validity of the screen theory the daughters on the screen. Attached daughters from outside will pass through the screen and not contribute to the count. Although the authors recommend operating with positive pressure in the delay chamber, it is less critical in a system using a screen. An adjustable vent on the outlet of the tank (Figure 1) at high face velocities, the collection efficiency of alpha parti- is used to maintain an overpressure in the tank. In the deteccles on the inc sulphide scintillator, and the light collection tors used in ACE 1 the average overpressure was about 100 Pa. efficiency of the light guide from the scintillator to the photomultiplier. An inconvenience of using a screen is that it is necessary to count alpha particles emitted from both faces of the screen. There are several major advantages of using the dual-flow This complicates the design of the screen. This complicates the loop design with a screen in place of a fiber or membrane filter medium for the second filter: 1. As opposed to the Cape Grim design, it can be implemented with the robustness needed for shipboard use. design of the alpha particle detector-screen assembly. The alpha particle detector should be as small as possible, and the screen to scintillator distance should be minimied to obtain optimum efficiency and energy resolution. These requirements

4 16,746 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS Photomultiplier ZnS scintillator x Photomultiplier ii Photomultiplier I a. Macquarie Is and Discoverer air flow b. Southern Surveyor Photomultiplier screen ow, c. New Version 0 50mm I scale Figure 2. Light guide and screen designs of the radon detectors used in ACE 1. The performance of the instruments is given in Table 1. are antithetical to the need for a large enough screen to achieve high daughter collection and as large an area as possible to permit the air to pass with a minimum of resistance. Figure 2a illustrates the compromise reached for light guidescreen assemblies on the Discoverer and Macquarie Island. An air flow aperture just 20% smaller than the one in Figure 2a was unable to achieve an adequate flow with the blower available. Yet a larger aperture could be achieved only by increasing either the screen to scintillator spacing or the diameter of the scintillator. Already equal to 7 mm, the screen to scintillator spacing could not be increased without considerable loss of efficiency. An increase in screen diameter while retaining a 50 mm diameter photomultiplier would result in a reduction of light collection through the light guide. With the diameter shown the light guide attenuation was an average of a factor of 2. Although a higher attenuation may be acceptable, there is a law of diminishing returns applying to this procedure: a factor of 2 increase in aperture without increasing the screen to scintillator spacing requires a doubling of the screen diameter with a consequent quadrupling of the area from which the light guide must collect light. While some small improvement may i be possible, the design shown achieved a satisfactory 800 L/min for a power consumption of 22 watts. From Table 1 it is evident that the sensitivity per unit detector volume for the Discoverer and Macquarie Island detectors is low compared to the other dual-flow loop designs. This is because of poor light collection in the light guide. If the gain of the photomultiplier is increased to obtain better counting efficiency, the background count rate from impurities in the scintillation material increases to the point where the lower limit of detection becomes higher. For the Southern Surveyor a flat screen and two photomultipliers were used as shown in Figure 2b. Although the screen area was smaller, the net result was a 32% improvement in sensitivity per unit volume compared to the instrument on the Discoverer. Figure 2c shows the arrangement for a new version currently being tested. It uses a photomultiplier with 6 times the area and screen 2.5 times the area used in the instrument on the Discoverer (Table 1). The improvement in actual sensitivity is a factor of 2, which is close to the predicted improvement of a factor of 1.8. This improvement is worth the additional weight and fragility of the large photomultiplier and light guide. The new version has a sensitivity per unit volume a factor of 5 better than the moving filter design of Hutter et al. [1990]. Environmental factors affecting this type of detector are temperature, humidity, and pressure. These factors will affect the retention of the daughters on the screen by changing their diffusion characteristics. Provided the retention is high, this effect will be small for changes in conditions expected at any given site. For the maritime conditions of ACE 1 at sea level, variation of pressure, humidity, and ambient temperature were considered too small to cause serious changes in efficiency. One possible large influence could have been direct sunlight which could cause a temperature gradient which may result in turbulence leading to increased plate-out. Whittlestonet al. [1994] describe an experiment to check whether the 32 L detector was adversely affected by temperature gradients. With the detector axis horiontal a 1 kw radiant heating element was placed with its axis parallel to the air flow and displaced horiontally 0.3 m from the detector side. A 10% reduction of count rate was observed. From this it is concluded that tem- perature gradients encountered under more normal conditions are unlikely to significantly affect the performance of this detector. Although longer-term trials are desirable, this conclusion is confirmed by the stability of the calibrations of the instruments, which is discussed in the next section. Table 1 shows that the desired sensitivity (0.28 counts -1 Bq - m -3) was not quite achieved for the instruments used on the ships for ACE 1. The precision achieved at 100 mbq m -3 was closer to 14% rather than the intended 10%. It was only after more development that the new version, with an improved light guide and screen assembly, realied the potential of the dual-flow loop design. 4. Calibration The calibration procedures for the ACE instruments were established not only to ensure that the results were traceable to an internationally accepted standard, but also to check that the instruments would give the same results when placed side by side. Basic calibration required the determination of two pa- rameters, the sensitivity (response to changes in radon concentration) and instrumental background (the response where the ambient radon to be ero). Given that the design was intended

5 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS 16,747 to give a precision of 10% (1 standard deviation) at 100 mbq laboratory practice is to have the sources flushed continuously m -3, a calibration to within about 5% was considered acceptable. at about 10 ml min - for at least a week before use. For The sensitivity measurements were made in a similar way to that employed by Colle et al., [1995, 1996], who describe an international intercomparison of marine atmospheric radon changes in flow in the range 20 to 200 ml min- it was considered unlikely that there would be any detectable time lag in the source response beyond the necessary equilibration time 222 measurements. In this intercomparison the techniques em- for the air in the source holder. When more than 15 source ployed by the Australian Nuclear Science and Technology Organisation (ANSTO) laboratory achieved agreement within 2% over 15 trials with the standardied radon injections. The calibration for ACE 1 was performed by passing air through a dry radon source (model RN-25 by Pylon Electronics Corp. Canada) and calculating the radon concentration by dividing the stated radon output of the source by the flow, which was measured by a commercial gas meter with a stated accuracy of 2% (Toyo Co. Japan model ML-2500). The serial numbers of the three sources used for ACE 1 were A-261, A-327, and A-331. With radon production rates of about 3 Bq min - and volumes of air were passed through the source before sampling, no systemati change of radon output was observed with time in a series of trials with charcoal traps. In a typical trial the source was taken from its storage location where the flow was about 10 ml min - to the laboratory test rig which has a precision flow controller whose setting for these trials was 50 ml rain - After a delay of more than 30 rain (15 source volumes at this flow rate), four samples were taken at 40 rain intervals. The agreement between the traps was 3%, and there was no drift with time. Although the trials were intended to test the traps rather than the source, they demonstrated that the a typical air flow rate of 90 L min-, the radon concentration sampling protocol effectively removed problems associated was about 33 Bq m -3. Allowance had to be made for ambient with changes in source flow rate. radon concentrations which could rise to 15 Bq m -3 at Cape Grim, Tasmania [Gras and Whittlestone, 1992]. Fortunately, these periods were brief, and the problem was avoided by selecting calibration periods when the ambient radon concen- When Pylon sources were used for radon detector calibration, care was taken to ensure that the conditions were within the range validated by the standardiation procedures. The flow through the source itself was set to about 40 ml min- tration was below 3 Bq m -3, preferably in baseline conditions both in storage and when in use. For a calibration the output when levels were lower than 0.1 Bq m -3. The ambient radon concentration was estimated by linear interpolation between the concentration before and after the injection of radon from the source. Since the calibration procedure took 7 hours, there was a possibility of a change during the calibration. Depending on the equipment available, the meteorological conditions were monitored or radon concentrations measured by a second detector during the calibration. If it seemed likely that the ambient radon had varied by more than 1 Bq m -3, the calibration was rejected. The Pylon sources are subject to three types of error. First, they are not standards, and sources other than those used for ACE 1 have been found to yield a radon output up to 18% different from the stated output. Second, as reported by Colle et al. [1995], the output may depend on the flow rate through the source. Third, errors can result if there is insufficient time between a change of flow rate and use of the source. The first two of these possible errors were addressed by a rigorous calibration procedure. Traceability to a standard was achieved by calibrating A-261 against a standard liquid radium source supplied by the United States National Bureau of Standards (National Bureau of Standards, 1978). Air was passed through A-261 at a rate between 20 and 200 ml min- into a 4 L vessel. The flow was held steady for at least 12 hours before drawing a 90 ml sample from this volume into a scintillation cell. After standing for a period of 3 hours, a series of half hour counts were recorded, and the background count rate was subtracted. These counts were compared to the counts when the radon in the cell was obtained by transferring into it the entire radon content of the standard source after the source had been sealed for at least 30 days. A-261 was found to be within 5% of its stated output and independent of the flow rate. Tests for sensitivity to humidity did not reveal any difference between the output using air dried by silica gel and air at 80% humidity. Periodic spot tests gave no reason to suspect drifts in source output over either short or long timescales. The third error, from sampling too soon after a change of flow, was found to be a few percent 1 day after sampling if the source had been closed for a few weeks. Accordingly, the was injected into the detector input airstream immediately before the gas meter used to measure the total flow. Source A-327 was compared with A-261 by passing radon into the Cape Grim detector first from one and then the other source. A similar procedure was conducted using the Discoverer detector to standardise source A-331 against A-261. The agreement between the calibrated value and the stated output of each source was within the experimental accuracy, which was limited by flow accuracy and estimates of ambient radon concentration to 5%. The stability of the calibrations was best measured for the Cape Grim detector, which has automatic monthly calibrations in which radon at a known concentration from A-327 is in- jected into the detector for 4 hours. Over 1 year the standard deviation of the monthly calibration factors was 5.7%. The source was operated for 3 days and showed that the variability of the calibration factor over this period was 5.9%. Calibrations were carried out on the Discover over a 2 day period before ACE 1. The error on the mean calibration factor estimated from the variability of the measurements was 1.7%, somewhat better than the accuracy of the flow rate estimate (2%) and the accuracy of the estimate of ambient radon (5%). Diagnostic information recorded during the voyage confirmed that all flows and voltages were within ranges known not to affect the calibration by more than 2%. The accuracy of the calibrations was 7%, governed by the errors on the source (5%) and the background estimate (5%). The measurements were subject to this systematic error plus statistical errors from counting. As shown in Table 1, the counting error was 30% at a radon concentration of 40 mbq m -3. The instrument used on the Southern Surveyor was placed beside the Cape Grim detector prior to ACE 1. The agreement between the instruments was within the experimental error, indicating that the new design of detector was at least as stable as the Cape Grim detector. At Macquarie Island it was necessary to rely on the record of flows and voltages as an indication of stability. The test source with the instrument was neither stable nor standardied and so served only as a diagnostic tool in the event of major failure.

6 .. 16,748 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS Z 100 O m -,,;, O O o 5o n' 0 ß NOVEMBER 1995 APRIL HOUR OF DAY (universal time, UT) 0 UT is 10 am local time. There was no reason to doubt the stability of the detector during ACE 1. Evaluation of the instrumental background was indirect because it was not practicable to obtain a large volume of radon free air. Most of the background is the result of radioactivity in the scintillator material, and radon or thoron emitted by internal surfaces of the detector. To simplify the discussion and keep the values in perspective for instruments of different sensitivity, the values of background will be expressed in terms of the radon concentration which would be required to give the background count rate. At Cape Grim there were three facts which set limits on the background. An upper limit was set at 15 mbq m -3 because values of background plus ambient radon were sometimes as low as this. A lower limit of the background 1 mbq m -3 was set by the count rate when a new filter was installed and the flow stopped. An estimate of 10 _+ 5 mbq m -3 was made by comparison with radon daughter measurements. Although this estimate is adequate for air mass characteriation for ACE 1, and checking the performance of the new radon detectors (Cape Grim has the design described by Whittlestone [1985], the authors are attempting to improve it. However, this work is beyond the scope of this paper. The other instruments had lower limits of detection close to the lowest radon concentrations observed in baseline condi- tions, which were about 20 mbq m -3. The approach to evaluating their instrumental background was to assume that the lowest value observed in baseline conditions was 20 mbq m -3. Thus, if the average of the lowest 10 readings in baseline conditions was 60 mbq m -3, a background value of 40 mbq m -3 was assigned. Confirmation of this estimate was made for the two shipboard instruments by comparing the results when the vessels were close to Cape Grim. 5. Instrument Performance During ACE 1 Data recovery during the period was close to 100%. There were extended periods when the radon levels were at the lower limit of detection of the instruments, except at Cape Grim, loo 5O which has a more sensitive detector. The scatter that appears on the data in these periods can be attributed to counting statistics rather than fluctuations in radon level. The instruments on the Discover and at Cape Grim performed flawlessly, and calibrations before and after ACE 1 agreed within the 5% accuracy of the measurement. At Macquarie Island and on board the Southern Surveyor the data required some explanation and corrections Apparent Diurnal Variations of Radon Concentrations Macquarie Island Discussion of the diurnal variation in observations at Macquarie Island is included in this paper on instrumentation because it is not entirely certain that the variations are real. In this section we present tests which show that the variation is unlikely to be instrumental, as well as meteorological data which show that it is unlikely that there is a diurnal variation of the period of contact of the air samples with local soil. Figure 3. Composite daily variation of radon concentration Evidence that the variation is real comes from a comparison at Macquarie Island with error bars showing the standard error of the ACE 1 data with measurements taken in 1992 with a of the mean for each hour. The scales are offset for clarity. The completely different radon detector. Both types of detector, large error bars for 1992 are the result of lower instrumental when used in other locations, show no diurnal variations except precision. those which can reasonably be attributed to changes in ambient radon concentration. Figure 3 shows a composite day for November 1995, where the value for a given hour is the average of radon concentrations for that hour from every day in the month. Also shown is the composite day from April Periods of large, clearly continental radon concentrations have been excluded. To avoid bias, a complete day has been removed if any part of the day had to be excluded. These 2 months were deliberately chosen because their diurnal varia- tions had similar amplitudes and phases. The similarity proves that the diurnal variation observed in November 1995 was not anomalous. The same procedure was carried out for each month of 1995 and The respective diurnal variations averaged over each year are 37 _+ 6 mbq m -3 and 42 _+ 13 mbq m -3. The amplitude of the diurnal variation for each month, shown in Figure 4, varies between 10 and 60 mbq m -3. It is least in winter, which is reasonable for a diurnal effect driven by solar radiation at this latitude. The seasonal variation is more pronounced for 1995, which might be the result of the t?, 100 E E 0 I,-- 50 o o,< 0 ' ' I ' ' I ' ' I ' ' I ß 1995 / 1992, I I, I, I MONTH Figure 4. Diurnal variation of radon concentration at Macquarie Island for each month of 1992 and Data are missing for months 6 and 8 of 1992 and 11 and 12 of 1995.

7 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS 16,749 better sensitivity of the 1995 instrument. The ACE 1 period, November and December 1995, is not anomalous. A number of additional checks were undertaken to test whether the effect was instrumental. First, the high voltage supply to the photomultiplier was set to ero to check for electrical interference. The count rate was ero. A second test showed that there was no diurnal change in the ratio between the count rates for two pulse height discriminators set to different levels. This proved that the diurnal variation was not due to electrical interference, light leakage, beta particles on the filter or changes in photomultiplier efficiency. It was concluded that the variation was due to alpha particles, which means radon in the detector. There is no doubt that there is some radon generated inside the detector because trace quantities of radium are ubiquitous. But none of several other detectors similar to that on Macquafie Island shows a diurnal variation. It is also most unlikely that two different detectors on Macquarie Island should both have the same instrumental diurnal variation. If the diurnal variation is present in ambient air, there must be land upwind of the inlet. The inlet mast is 10 m high, and the land fetch before reaching the ocean is about 10 m of rounded stones, not a credible source of radon. However, a path at 280 ø from the station just touches a headland, and at 270 ø the land path is about 4 kin, long enough to yield about 50 mbq m -3 of radon from average soils. Figure 5 shows the vector average of wind speed and direction during ACE 1. There is a diurnal variation of about 4 m s- in wind speed and 10 ø direction. Although the average wind direction is never as low as 270 ø, there is a 20 ø standard deviation of the 10 rain wind directions for each hour. It is therefore possible that the diurnal shift of the average direction from 290 ø to 300 ø is enough to bring air over a longer land path at 0200 UT compared to 1400 UT. Against the case for a diurnal variation of radon concentrations from a local source is the observation that despite the diurnal variation of both composite wind direction and radon concentrations (Figures 3 and 5), hourly values of wind direction show no convincing correlation with concurrent radon measurements. Also, the island does not have the deep dry soils which are usually associated with radon fluxes high enough to produce the radon concentrations observed. This issue could only be resolved by direct measurement of radon flux. The authors believe that the diurnal variation is a true record of ambient radon at Macquarie Island but recommend further tests to prove that the effect is not instrumental. These would include measurement of the radon flux from the island near the station and a variety of tests of the instrument itself Partial Failure of the Radon Detector on the Southern Surveyor The detector on board the Southern Surveyor had two photomultipliers, of which one developed a fault which worsened over a period of 3 days from day 332 (November 28). The result was a decrease in detector efficiency by a factor of 2 over the 3 days, followed by intermittent bursts of electronic noise. It was fortunate that the initial period of change took place when the radon levels were at the lower limit of detection of the instrument. Even without the subsequent proof of instrument failure it was clear from the data record that the decline in count rate was instrumental because application of the back _ ß DIRECTION ß 290 '... '... '... t,,9,, HOUR OF DAY (UT) - 20 Figure 5. Composite daily variation of wind speed and direction formed from the vector average for each hour of day during ACE 1. ground level determined in the initial calibrations to the data at the end of the voyage led to negative radon levels. The data following the failure were corrected in two stages. First, the new instrumental background was estimated by matching the baseline periods to the response during prefailure baseline episodes. Next the calibration factor was multiplied by 2. This had no practical effect on data quality because the radon levels remained baseline for the remainder of the voyage. When the ship returned to port, the levels recorded were close to those observed prior to departure. This confirmed that the instrument was functioning, even if with reduced sensitivity, for the latter stage of the voyage. 6. Conclusions 10o- Wire mesh screens have been shown to be effective as the second filter in two-filter radon or thoron detectors. Use of a screen in place of a fiber or membrane filter medium is beneficial when cost, power consumption, and weight of the detector are of prime importance. In detectors of the new design used in ACE 1, it was possible to use a blower weighing 1 kg instead of the 10 kg unit needed had a filter been used. At the same time, there was a capital cost saving of a factor of 2 and a reduction of annual electricity consumption by a factor of 10 from 3000 kwh yr-1. Reduced pressure drop across the screen also reduces the pressure which the tank has to withstand, resulting in reduced construction costs. Thus, in many applications, the disadvantages of the screen would be strongly outweighed by the advantages. Full benefit of the screens in two-filter detectors is realied only when the flow through the first filter is separated from the flow through the screen. By matching the blower to the flow impedance in each flow path, it is possible to build a detector with very low power consumption. An added advantage of using a low flow through the sampling line for a radon detector is the reduction in volume of the delay needed to remove thoron. The dual flow loop radon detectors with screens have the virtues simplicity and freedom from routine maintenance. They can be built easily with detection limits as low as 1 mbq m -3. This technology extends the range of sites at which base- line radon measurements can be made to remote areas with

8 16,750 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS little regular technical backup and a harsh environment. For most environmental applications the 45 min response time of the instrument is a small disadvantage compared to the improvement in sensitivity by a factor of 5 for a given volume of estimate of geometrical efficiency. The count rate for radon concentration NRn is then C = e c(fa + Fc) detector when compared to the moving filter design. The new design was successful in ACE 1, with no major data Acknowledgments. B. Stenhouse of ANSTO assisted with design loss for the three new instruments. Data quality was excellent. and construction of the radon detectors. D. Hamilton of PMEL One surprising result, which needs independent confirmation, planned and executed construction of the instrument on the Discoverer and operated the instrument during ACE 1. Thanks are also due to the was the apparent diurnal variation of radon concentration at staff of the Cape Grim Baseline Atmospheric Pollution Station and I. Macquarie Island. Raymond at Macquarie Island for operating their respective radon detectors. B. Griffith of CSIRO Division of Fisheries organied operation and installation of the instrument on the Southern Surveyor. ACE 1 was supported by the U.S. National Science Foundation under grants ATM and ATM Appendix: Calculation of Two-Filter Detector Response Whittlestonet al. [1994] describe a computer code for calculating the response of a two-filter detector. This was used to derive the calculated values in Table 1. The physical principles are summaried here. The derivation closely follows the that of Evans [1969], with allowance for the special conditions applicable here. The calculation is in two parts, the first to obtain the concentration of radon daughters in air passing through the screen, and the second to obtain the concentration on the screen itself. The first part is different from that of Evans and will be explained in detail. Laminar flow is assumed, so a slice of air passes down the chamber, taking a transit time equal to the internal loop flow rate divided by the chamber volume. At any time the rate of change of the daughter concentration is in close analogy to Evans [1969, equation (15)] dna/dt = NRnitRn -- NA(&4 + itp) (m]) dn /dt = NAItA - NB(ItB + itv) dnc/dt = NBit - Nc(itc + itp) (A2) (A3) References Bates, T. S., B. J. Huebert, J. L. Gras, F. B. Griffiths, and P. A. Durkee, The International Global Atmospheric Chemistry (IGAC) Projects First Aerosol Characteriation Experiment (ACE 1): Overview, J. Geophys. Res., this issue. Busigin, A., A. W. van der Vooren, J. C. Babcock, and C. R. Phillips, The anture of unattached RaA particles, Health Phys., 40, , Cheng, Y. S., and H. C. Yeh, Theory of a screen-type diffusion battery, J. Aerosol Sci., 11, , Colle, R., et al., An international intercomparison of marine atmospheric radon 222 measurements in Bermuda, J. Geophys. Res., 100(D8), 16,617-16,638, Colle, R., et al., An international intercomparison of marine atmospheric radon 222 measurements in Bermuda, II, Results for the participating laboratories, J. Res. Natl. Inst. Stand. Technol., 101 (1), 21-46, Evans, R. D., Engineers' guide to the elementary behaviour of radon daughters, Health Phys., 17, , George, A. C., Measurements of the uncombined faction of radon daughters with wire screens, Health Phys., 23, , Gras, J. L., and S. Whittlestone, Radon and CN: Complementary tracers of polluted air masses at coastal and island sites, J. Radioanal. Nucl. Chem., 161, , Grumm, D., S. D. Schery, and S. Whittlestone, Two-filter monitor for low levels af 22øRn and 222Rn, paper presented at the 1990 International Symposium of Radon and Radon Reduction Technology, Environ. Prot. Agency, Atlanta, Ga., Feb , Hutter, A. R., A. C. George, M. L. Maiello, I. M. Fisenne, R. J. Larsen, where NA, NB, and N c are the concentrations of 2 8po, 2 4pb, and 214Bi, respectively, and ita, itb, and it c are the respective decay constants. The solutions to these are the same as Evans' [1969], with ita H. L. Beck, and F. C. Wilson, 222Rn progeny and 22øRn progeny as atmospheric tracers of air masses at the Mauna Loa Observatory, replaced by ita + it,p, etc. The difference is the plate-out time U.S. Dep. Energy Rep. EML-522, Natl. Telecommun. and Inf. Adconstant, which is assumed to be the same for each daughter. min., Springfield, Va., When short transit times are used, this assumption will not Krit, M. A., The China Clipper--Fast advective transport of radon cause significant errors from possible differences in the growth rich air from the Asian boundary layer to the upper troposphere near California, Tellus, Ser. B., 42, 46-61, rate of the different species [Busigin et al., 1981]. The concen- National Bureau of Standards, Radium-226 for radon analysis, Stand. tration at the filter is the concentration at a time equal to the Ref. Mater C, U.S. Dept. of Commer., Washington, D.C., May transit time through the tank The calculation of the alpha activities on the screen is also Polian, G., G. Lambert, B. Ardouin, and A. Jegou, Long-range transdone using the equations presented by Evans [1969]. Following port of continental radon in subantarctic and antarctic areas, Tellus, Ser. B, 38, , his equation (34), but accounting for screen retention by the Prospero, J. M., and T. N. Carlson, Radon-222 in the North Atlantic factor r s estimated using the theory of Cheng and Yeh [1980], trade winds: Its relationship to dust transport from Africa, Science, the rate of change for each species is the sum of the losses and 167, , gains: Raes, F., A. Jansens, and H. Vanmarke, A closer look at the behaviour of radioactive decay products in air, Sci. Total Environ., 45, , dfa/dt = QrdVA - FAItA (A4) Schery, S. D., D. H. Gaeddert, and M. H. Wilkening, Two-filter mondf /dt = QrdV + NAItA -- F itb (A5) itor for atmospheric 222Rn, Rev. Sci. Instrum., 51(3), , Solomon, S. B., and T. Ren, Counting efficiencies for alpha particles emitted from wire screens, Aerosol Sci. Technol., 17, 69-83, dfc/dt = QrdVc + F it - Fcitc (A6) Strong, J. C., The sie of attached and unattached radon daughters in room air, J. Aerosol. Sci., 19, , where F A, F, and F c are the activities on the screen of 2 8po, Thomas, J. W., and P. C. LeClare, A study of the two-filter method for 214pb, and 2 4Bi, respectively, and Q is the flow rate through radon-222, Health Phys., 18, , the screen. The counting efficiency e c for the screen was taken Whittlestone, S., A high sensitivity Rn detector incorporating a particle from the work of Solomon and Ren [1992] together with an detector, Health Phys., 49, , 1985.

9 WHITTLESTONE AND ZAHOROWSKI: BASELINE RADON DETECTORS 16,751 Whittlestone, S., E. Robinson, and S. Ryan, Radon at the Mauna Loa Observatory: Transport from distant continents, Atmos. Environ. Part A, 26(2), , Whittlestone, S., W. Zahorowski, and P. Wasiolek, High sensitivity two filter radon/thoron detectors deploying a wire or nylon screen as the second filter, Rep. E/718, Aust. Nucl. Sci. and Technol. Org., Lucas Heights, N. S. W., Australia, Whittlestone, S., S. D. Schery, and Y. Li, Thoron and radon fluxes from the island of Hawaii, J. Geophys. Res., 101(D9), 14,787-14,794, Whittlestone, S., J. L. Gras, and S. T. Siems, Surface air mass origins during ACE 1, J. Geophys. Res., this issue. Wilkness, P. E., R. A. Lamontagne, R. E. Larson, J. W. Swinnerton, C. R. Dickson, and T. Thompson, Atmospheric trace gases in the southern hemisphere, Nature Phys. Sci., 245, 46-47, S. Whittlestone and W. Zahorowski, ANSTO, Lucas Heights, New South Wales 2234, Australia. ( swh@atom.ansto.gov.au; wa@ ansto.gov.au) (Received June 13, 1997; revised January 19, 1998; accepted February 25, 1998.)