- Email: [email protected]
Weather system scale variation in radon-222 concentration of indoor air
The Science of the Total Environment 284 Ž2002. 157᎐166
Weather system scale variation in radon-222 concentration of indoor air Jane E. Rowe1, Mike Kelly U , Laura E. Price Department of En¨ ironmental Science, Uni¨ ersity of Lancaster, Lancaster LA1 4YQ, UK Received 24 January 2001; accepted 10 May2001
Abstract Radon-222 concentration measurements using the charcoal adsorption technique were made continuously over periods of up to 2 years in three houses, two rural and one urban. The measurement records for all houses show a variation in radon concentration on a seasonal scale on which is superimposed shorter periods of relatively large magnitude fluctuations on a scale longer than the typically 4᎐5-day measurement duration. Regression analysis using meteorological data from a weather station remote from the houses shows that much of the variation in both monthly and 3-day mean radon concentration can be explained by regional scale external temperature variation. Wind speed, rainfall and barometric pressure apparently influence the 3-day values to a much smaller degree. Differences in the radon levels between the houses reflect the different geological radon potential of the two areas, with notably higher levels over limestone than glacial till and sandstone. Other differences within and between houses reflect house construction and occupancy factors. 䊚 2002 Elsevier Science B.V. All rights reserved. Keywords: Indoor radon; Concentration variation; Meteorological effects
1. Introduction Radon in the air of domestic buildings derives mainly from the adjacent subsoil gas by advective transport, driven by a pressure difference across the housersubsoil boundary ŽNazaroff et al., U
Corresponding author. Tel.: q44-1524-593937; fax: q441524-593985. E-mail address: [email protected] ŽM. Kelly.. 1 Present address: Environment Agency, Cameron House, White Cross Industrial Estate, Lancaster, LA1 4 XQ, UK.
1988.. The effect of changes in meteorological conditions on this system has been seen in a number of studies based on continuous recording of indoor radon levels and meteorological variables at individual houses. These have shown that indoor radon levels fluctuate widely on a variety of time scales, with periods from diurnal to annual Že.g. Nazaroff et al., 1985; Cliff et al., 1987; Furrer et al., 1991; Wilson et al., 1991; Porstendorfer et al., 1994; Hubbard et al., 1996.. ¨ The aim of this investigation was to determine the degree of variability in radon levels on a time
0048-9697r02r$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 8 - 9 6 9 7 Ž 0 1 . 0 0 8 7 6 - 2
158
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
scale corresponding to the time resolution of the charcoal adsorption passive monitoring system, i.e. ; 5 days, as this method is used for reconnaissance and property valuation purposes in a number of countries, although not recommended for radon dosimetry. An additional aim was to determine the extent to which such variability could be explained by regional weather conditions, as characterised by readily available data from a standard meteorological station. For this purpose, indoor radon concentration was monitored by the charcoal adsorption method in three occupied houses over periods of up to 2 years. Two of the houses were located in a bedrock area of high radon potential and the third in a low potential area ŽRowe et al., 1993..
2. Method The 222 Rn activity concentration in air Žhenceforth referred to as radon concentration . was passively sampled using activated charcoal in canisters based on those of George Ž1984. Žsupplied by Canberra Industries .. The deployment period was normally 4᎐5 days, but exceptionally 3᎐8 days, to suit occupant’s requirements. Radon-222 adsorbed by the charcoal was quantified by counting the ␥ photon emission of its progeny 214 Pb, after leaving the sealed canister for a suitable period for establishment of radioactive equilibrium. The efficiency calibration of the HPGe detectors was determined by spiking a ‘baked out’ charcoal canister with a certified mixed nuclide standard. The calculated 222 Rn activity in the canister
was converted into a radon concentration in air using the radon adsorption efficiency factors for charcoal of George Ž1984., which take into account the reduction of radon adsorption by adsorbed water by gravimetrically determining the charcoal moisture content. Since these efficiency factors were for a charcoal canister of different geometry, a further conversion factor was determined by exposing 10 charcoal canisters to a known 222 Rn concentration in air Žusing the facilities of the UK National Radiological Protection Board.. The conversion factor was taken as the mean ratio of the predicted against the observed values. The propagated errors, from the conversion factor, the counting efficiency and the sample count, give a 2 error of approximately 30% in the measured radon concentration. Details of the monitored houses’ construction and of the local geological conditions are given in Table 1. They are located in north-west England, UK, one in the city of Lancaster and two in the village of Arnside and were chosen for their different combinations of local geology and house construction and, especially, the willingness of the occupants to change the charcoal canisters. Charcoal canisters were sited in a main room on each of the floors of the houses and changed in the early morning after a suitable exposure period. Although it was not practicable to obtain detailed information about the occupant’s habits which might affect room ventilation, the use of fires or periods when the houses were unoccupied were noted for houses A and B. The standard meteorological data came from the Lancaster University weather station, located approximately 5 km from Lancaster and 21 km from Arnside, i.e. daily values of maximum and
Table 1 House details House Žlocation.
Bedrockrsediment
Type
External walls
Ground floor
A ŽArnside.
Carboniferous limestone Carboniferous limestone Glacial tillrCarboniferous sandstone
Detached 2-storey
Rendered brick with cavity Limestone with rubble filled cavity Brick with cavity
Suspended floor above bedrock Suspended floor above bedrock Suspended floor above cellar
B ŽArnside. C ŽLancaster .
Semi-detached 2-storey Terraced 2storey q cellar
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
minimum temperature, barometric pressure, rainfall and mean wind speed. To assist in data analysis, the radon series for all three properties were converted by interpolation into a regularly sampled series with a 3-day interval. Mean values for each meteorological variable for concurrent 3-day periods were also calculated. A simple linear regression model was used to investigate the relationships between radon concentration and the recorded meteorological variables, with the general form: A m s a1 q a2 X q a3 Y q a4 Z
159
floor rooms. A strong seasonal effect is evident in these data, with a clear annual cycle of low summer and high winter levels. Superimposed on the seasonal cycle, are irregular fluctuations of an intermediate periodicity, e.g. with a mean peak to peak period of 18 days at house A Ž1991.. These can be of substantial magnitude, with deviations from annual mean radon levels almost as large as those shown by the seasonal cycle, e.g. up to 128% deviation from the mean compared to 140% deviation for the seasonal cycle Ž1991 ground floor, house A..
Ž1.
where X, Y and Z are meteorological variables.
3. Results 3.1. Radon concentration time-series The radon concentration time-series for the ground floor rooms of houses A᎐C are shown in Figs. 1᎐3. Similar results were obtained for first
3.2. Annual mean radon concentration and long period ¨ ariation
The annual and monthly mean indoor radon concentrations calculated from the measurements are shown in Table 2. Notably, the annual mean values for individual houses are similar for successive years ŽA and C.. The first floors have lower values than the ground floors for A and B, but are similar at C. However, the cellar at C has
Fig. 1. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse A, ground floor, 1991..
160
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
Fig. 2. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse B, ground floor, 1991..
substantially higher values than for the floors above, as might be expected. The weighted mean annual concentrations takes into account different room occupancies, using the factors obtained by the UK National
Radiological Protection Board, i.e. 45% for ground floor living rooms and 55% for first floor bed rooms ŽWrixon et al., 1988.. The weighted values for the two houses on limestone ŽA and B. are similar Ž430᎐437 and 416 Bq my3 . and con-
Fig. 3. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse C, ground floor, 1991..
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166 Table 2 Monthly mean
222
Rn concentrations ŽBq my3 .
House A 1991 Ground January 768 February 734 March 688 April 427 May 355 June 309 July 108 August 195 September 260 October 580 November 550 December 600 Annual mean 467 Weighted mean 430 a
161
1992
House B
House C
1991
1991
1992a
First
Ground
First
Ground
First
Cellar
Ground
First
Cellar
Ground
First
751 725 534 342 274 280 53 93 173 476 493 564 399
807 565 591 584 208 142 114 179 494 702 584 745 479
818 452 446 445 139 76 53 85 298 665 493 726 402
991 814 682 460 562 705 307 376 442 768 744 828 631
ND 332 302 223 175 238 53 82 154 298 363 402 240
95 118 97 123 63 61 119 113 67 74 92 99 97
30 32 24 20 13 13 12 15 16 25 28 39 23
32 36 22 19 16 12 9 12 15 23 25 34 23
105 100 96 106 144 108 105 53 135 37 63 80 96
41 38 38 38 32 20 23 20 30 26 27 31 31
30 26 19 23 20 13 16 9 23 26 23 29 22
437
416
23
26
Data substantially incomplete.
siderably higher than for the house on sandstone ŽC, 23᎐26 Bq my3 .. The monthly mean radon concentrations Ž A m . reveal the seasonal variation mentioned above and, consequently, show a high degree of negative correlation with measures of temperature. Table 3 shows the regression parameter values Žstandard errors in parentheses . and the goodness of fit given by the coefficient of determination Ž R 2 . for the Eq. Ž1. regression model with only one variable considered, i.e. the monthly mean value
of the daily maximum temperature in degrees Celsius. This gave a slightly higher degree of correlation than either the monthly mean daily minimum or median temperatures. Whilst the daily mean temperature might be expected to be the best measure, for a system with a linear response, the data were not available to calculate this. Points of note in the table are the high degree of explanation of the monthly radon variation, 87᎐94% at A and a somewhat lower 72᎐87% at B
Table 3 Regression of indoor radon concentration against monthly mean daily maximum temperature w R 2 and parameter values with standard errors in parentheses, Eq. Ž1.x House
Floor
a1
a2
R2 Ž%.
House A 1991
Ground First Ground First Ground First Cellar Ground First
957.9 Ž58.5. 929.3 Ž44.6. 1066.5 Ž72.5. 1015.3 Ž80.3. 1070.2 Ž84.1. 509.8 Ž43.8. 101.8 Ž18.1. 40.0 Ž3.7. 41.3 Ž2.6.
y40.9 Ž4.5. y44.2 Ž3.4. y48.3 Ž5.5. y51.0 Ž6.1. y35.6 Ž6.5. y21.4 Ž3.2. y0.7 Ž1.4. y1.5 Ž0.3. y1.7 Ž0.2.
89.3 94.3 88.3 87.4 75.3 82.8 0.2 72.3 87.0
House A 1992 House B 1991 House C 1991
162
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
and C, compared to a total lack of association for the cellar at C. Parameter a1 values reflect the bedrock radon potential, with high values for the limestone bedrock area ŽA and B. and low for the glacial tillrsandstone ŽC.. The relatively low value for house B first floor indicates greater degree of isolation from the bedrock source than occurs at house A, due to construction or habit factors. The contrast between the two areas in parameter a2 values is interesting, with the higher values for A and B indicating a more direct coupling to the source, for whatever reason. 3.3. Intermediate period radon concentration ¨ ariation
In order to test the hypothesis that the intermediate period variation, at least in part, is a function of regional scale weather conditions, the mean radon concentrations for 3-day intervals in the first year Ž1991. were regressed against the 3-day mean meteorological data series. Tests showed that the meteorological data and the ground floor radon concentration data for all three houses were normally distributed and, hence, suitable for regression analysis. For the first floor, data for houses A and C departed only slightly from a normal distribution, whilst those for house B were normally distributed. The fit of the linear regression model wEq. Ž1.x between radon and various combinations of meteorological variables for both floors of all three houses is given by the R 2 values in Table 4. It is
clear from these modelling results that temperature is the most important single variable that may be used to describe the radon series, here shown by daily maximum temperature. The percentage fit is highest at A and ranges overall from 52 to 74%. This is a reduction in the degree of association with temperature compared to the monthly data set but, since the fit remains relatively high despite a 10-fold increase in the number of data, it indicates a degree of control by short period temperature fluctuations in addition to the seasonal influence. For the best case Žhouse A., for temperature alone, the regression parameters with standard errors in parentheses are estimated as a1 s 964.6 Ž32.7. and a2 s y40.56 Ž2.5.. Mean wind speed Žkm hy1 . was also noted to improve the fit when used as a second regressor alongside temperature. Rainfall and pressure produced some improvements in model fit for some, but not all houses. However, it is important to note the danger of models becoming over-parameterised, especially where additional regressors only serve to provide minor increases in model fit. In general, a simple model combining maximum temperature and wind speed, with or without rainfall, provides not only a common model structure for all three properties, but also furnishes a reasonably good fit for each location whilst remaining parametrically efficient. Table 5 gives the R 2 and regression parameter values for the Eq. Ž1. model with X as mean daily maximum temperature, Y as total rainfall in mm, and Z as mean wind speed in km hy1 . Overall, the models
Table 4 Correlation coefficient Ž%. for regression between 3-day mean values of radon concentration and meteorological variables House A 1991 Ground Max temp Wind speed Rainfall Pressure Max tempq wind Max tempq rain Max tempq pressure Max tempq windq rain Max tempq windq pressure q rain
74.7 2.8 0.6 5.4 82.1 80.3 77.5 83.3 83.6
House B 1991 First 74.0 6.4 1.4 10.5 87.3 81.0 81.0 87.9 87.9
Ground 52.0 7.6 1.3 3.6 63.4 56.8 53.1 63.4 67.8
House C 1991 First 68.8 1.5 0.2 0.8 76.1 72.4 70.2 76.1 77.2
Ground 53.4 7.1 8.6 3.7 62.4 63.1 55.8 63.6 66.8
First 64.0 5.2 8.4 1.8 71.0 73.0 64.9 74.2 78.5
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
163
Table 5 Regression of radon concentration Ž1991 3-day means. against mean daily temperature Ž a1 ., total rainfall Ž a2 . and mean wind speed Ž a3 . w R 2 and parameter values with standard errors in parentheses, Eq. Ž1.x House House A House B House C
Ground First Ground First Ground First
a1
a2
a3
a4
R2 Ž%.
1179.0 Ž41.5. 1296.6 Ž41.1. 1320.0 Ž79.1. 668.5 Ž37.1. 48.7 Ž2.5. 47.3 Ž2.5.
y42.7 Ž1.9. y48.9 Ž1.9. y36.2 Ž3.8. y25.0 Ž1.8. y1.5 Ž0.1. y1.3 Ž0.1.
y11.4 Ž2.7. y19.7 Ž2.6. y19.0 Ž4.4. y8.0 Ž1.8. y0.6 Ž0.2. y0.6 Ž0.2.
y12.0 Ž4.6. y10.1 Ž4.6. 0 0 0 0
83.3 87.9 63.4 76.1 62.4 71.0
explain 62᎐88% of the radon variation, with house A again giving the best fit. For the ground floor of each house in 1991, the radon concentrations predicted by the models described by Table 4 are compared with the measured values in Figs. 1᎐3. Some of the unexplained variation will be due to unusual occupant living conditions, such as the noted unoccupied periods at A. For house C, more sections of radon data lie outside the estimated model but this is likely to be due to the large amount of missing data, which makes the model inefficient. The radon concentration was also recorded during 1992 in house A and much less completely
from C. For the ground floor of A, the modelled data closely fitted the measured concentrations, with an R 2 value of 84% ŽFig. 4.. The estimated model parameters are and a1 s 1281 Ž51.7., a2 s y48.8 Ž2.40., a3 s y1.1 Ž5.61. and a4 s y14.8 Ž2.96.. However, although the same regressors are used in the 1991 and 1992 models, namely maximum temperature, wind speed and rain, the parameters multiplying the rainfall and wind terms are notably different. 3.4. Prediction of radon concentration time-series In order to validate the regression models esti-
Fig. 4. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse A, ground floor, 1992.. For comparison, the radon concentration predicted by the 1991 model is also shown.
164
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
mated from the 1991 data, the best fit model for house A in 1991 was used to predict the 1992 radon concentration from the 1992 meteorological data. The predicted concentrations are compared with the measured values in Fig. 4 Žnote this also includes the directly modelled values as described above.. The general trend of the predicted fit is good, with an R 2 of 80.7%, which compares well with the 84% fit produced by direct modelling of the measured 1992 radon concentration.
4. Discussion and conclusions The difference in radon concentrations between the three houses is due primarily to the influence of bedrock geology on soil gas levels. Thus, the weighted mean annual radon concentrations for the two houses on limestone Ž416᎐437 Bq my3 . are considerably higher than the corresponding average value for the UK of 20 Bq my3 ŽKendall et al., 1994., whereas the values for the house on glacial till and sandstone are close to the average Ž23᎐26 Bq my3 .. These values are in keeping with what is known about the relative radon potential of the rock types of the Arnside and Lancaster areas. The local Carboniferous limestone has been assigned to a ‘high’ radon potential category by Appleton and Ball Ž1995., due to its high fissure permeability, especially in the near surface zone of karstic weathering. This category agrees with the 3᎐10% incidence of houses with radon levels above the 200 Bq my3 action level determined by the National Radiological Protection Board surveys ŽLomas et al., 1996.. Whilst the local Carboniferous sandstone has a ‘moderate’ classification, the cover of glacial till reduces this radon potential to the ‘low’ category ŽAppleton and Ball, 1995., corresponding to the - 1% incidence above the action level seen in the NRPB surveys ŽMiles et al., 1996.. The seasonal cycle seen in monthly radon concentrations is similar to that obtained from UK national surveys by combining shorter period data ŽWrixon et al., 1988.. The shorter period variation in radon levels characteristic of all houses can also be seen in long records obtained from else-
where by instrument based measurements Že.g. Hubbard et al., 1996.. In addition, such records have higher frequency variations which have been filtered out in our case by the passive sampling method used. It has been shown that much of the long and intermediate period variation in the charcoal canister measurements of radon concentration can be related to the regional climate, as described by the meteorological data from a remote weather station. For both the monthly and 3-day data, most of the variation is explained by temperature, using mean daily maximum temperature as its measure. This accounts for 72᎐94% and 52᎐75% of the variation for monthly and 3-day variation, respectively. The explanation for this strong negative relationship is that external temperature is inversely related to the outdoor᎐indoor temperature difference, when the indoor temperature is kept constant. Such constancy, at a relatively high temperature, is the aim of occupants, being obtained by control of heating and ventilation. The underlying radon transport mechanism will be the stack effect, which occurs when a higher indoor than outdoor temperature generates a pressure difference which then draws in small volumes of soil gas with high radon concentrations ŽNazaroff et al., 1988.. However, Hubbard et al. Ž1996. demonstrated that, although this was the normal condition for Swedish houses, some showed the opposite effect, with reduced indoor radon level as temperature difference increased, due to increased ventilation by outdoor air with low radon concentrations. Other meteorological variables, which appeared to have a contributory, but lesser effect on radon concentration, are wind speed, rainfall and barometric pressure. These have also been previously identified as influencing radon transport from the ground into houses. Wind pressure on building surfaces is widely assumed to have a positive effect on radon entry from the soil by generating an over-pressure on the surface of the building shell ŽNazaroff et al., 1988. or by developing an under-pressure by the Bernoulli effect of wind blowing over open chimneys ŽKendall et al., 1994.. Conversely, Riley et al. Ž1996. suggest that soil gas radon levels could be diluted by wind
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
induced inflow of air. Rainfall can also have an influence on the radon concentration, by causing relative differences in permeability and soil air pressures below and beyond the house, as can barometric pressure where the ground is differentially saturated or frozen ŽNazaroff et al., 1988.. The temporal scale of the periodicity in the radon levels, and in the meteorological variables with which the radon levels are correlated, indicates a relationship with weather systems of a large spatial scale. In this region of north-west England, the weather systems are dominated by the passage of the North Atlantic depressions, which give major variations in air mass temperature together with frontal precipitation, elevated wind speeds and reduced pressure. In such a climate, it is obvious that single charcoal adsorption measurements cannot be used to estimate mean radon concentration for periods much longer than the measurement interval. The variability in the degree of association of radon concentration with weather system variables shown between rooms and between houses will be due to the additional effects of house construction and occupancy factors. Within houses A and C, the close similarity of radon concentration between floors suggests a high degree of air exchange between the two floors. In contrast, for house B, the upper room appears to be relatively isolated from the lower one, giving radon levels typically half those of the ground floor. Although the regression models identified have provided good approximations to the radon concentration time-series measured in the three houses, only linear, constant-parameter models have been considered. Initial work using the recursive regression technique ŽYoung, 1999. for the radon series and meteorological variables for house A has shown that the parameters of the regression model exhibit significant time variability throughout the year. This could indicate that either there is an additional regressor influencing the measured radon concentration, which has not yet been identified, or that one of the meteorological regressors is in fact influencing the radon concentration in a non-linear way. Therefore, future work should focus on identifying the nature
165
and character of the time varying parameter sequence, which when measured or described mathematically can be incorporated in the model structure, resulting in a much improved model fit.
Acknowledgements We would like to thank the occupants of the houses for their patience and help with the radon sampling. References Appleton JD, Ball TK. Radon and background radioactivity from natural sources: characteristics, extent and relevance to planning and development in Great Britain. Technical Report. 1995; WPr95r2, British Geological Survey. 1995:93 pp. Cliff KD, Wrixon AD, Green BMR, Miles JCH. Concentrations of radon in dwellings in the United Kingdom. In: Hopke PK, editor. Radon and its decay products. Washington, DC: American Chemical Society, 1987:536᎐559. Furrer D, Crameri R, Burkart W. Dynamics of Rn transport from the cellar to the living area in an unheated house. Health Phys 1991;60:393᎐398. George AC. Passive, integrated measurement of indoor radon using activated carbon. Health Phys 1984;46:867᎐872. Hubbard LM, Mellander H, Swedjemark GA. Studies in temporal variations of radon in Swedish single-family homes. Environ Int 1996;22:S715᎐S722. Kendall GM, Miles JCH, Cliff KD, Green BMR, Muirhead CR, Dixon DW, Lomas PR, Goodridge SM. Exposure to radon in UK dwellings. Report NRPB-R272. Chilton: National Radiological Protection Board, 1994 43 pp. Lomas PR, Green BMR, Miles JCH. Radon atlas of England. Report NRPB-R290. Chilton: National Radiological Protection Board, 1996 16 pp. Miles JCH, Green BMR, Lomas PR. Radon affected areas: England. Documents of the NRPB 1996;7Ž2.:2᎐9. Nazaroff WW, Fuestal H, Nero AV, Revzan KL, Grimsrud DT, Essling MA, Toohey RE. Radon transport into a detached one-storey house with a basement. Atmos Environ 1985;19:33᎐46. Nazaroff WW, Moed BA, Sextro RG. Soil as a source of indoor radon: Generation, migration and entry. In: Nazaroff WW, Nero AV, editors. Radon and its decay products in indoor air. Wiley-Interscience, 1988:57᎐112. Porstendorfer J, Butterweck G, Reineking A. Daily variation ¨ of the radon concentration indoors and outdoors and influence of meteorological parameters. Health Phys 1994; 67:283᎐287. Riley WJ, Gadgil AJ, Bonnefous YC, Nazaroff WW. The effect of steady winds on radon-222 entry from soil into houses. Atmos Environ 1996;30:1167᎐1176.
166
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
Rowe JE, Kelly M, Hewitt CN. The occurrence of high indoor radon levels in Carboniferous bedrock areas of NW England. Radiat Prot Dosim 1993;26:201᎐205. Wilson DL, Gammage RB, Dudney CS, Saultz RJ. Summertime elevation of 222 Rn levels in Huntsville Alabama. Health Phys 1991;60:189᎐197. Wrixon AD, Green BMR, Miles JCH, Cliff KD, Francis EA,
Driscoll CMH, James AC, O’Riordan MC. Natural radiation exposure in UK dwellings. Report NRPB-R190. Chilton: National Radiological Protection Board, 1988 188 pp. Young PC. Nonstationary time series analysis and forecasting. Prog. Environ. Sci. 1999;1Ž1.:3᎐48.
Weather system scale variation in radon-222 concentration of indoor air Jane E. Rowe1, Mike Kelly U , Laura E. Price Department of En¨ ironmental Science, Uni¨ ersity of Lancaster, Lancaster LA1 4YQ, UK Received 24 January 2001; accepted 10 May2001
Abstract Radon-222 concentration measurements using the charcoal adsorption technique were made continuously over periods of up to 2 years in three houses, two rural and one urban. The measurement records for all houses show a variation in radon concentration on a seasonal scale on which is superimposed shorter periods of relatively large magnitude fluctuations on a scale longer than the typically 4᎐5-day measurement duration. Regression analysis using meteorological data from a weather station remote from the houses shows that much of the variation in both monthly and 3-day mean radon concentration can be explained by regional scale external temperature variation. Wind speed, rainfall and barometric pressure apparently influence the 3-day values to a much smaller degree. Differences in the radon levels between the houses reflect the different geological radon potential of the two areas, with notably higher levels over limestone than glacial till and sandstone. Other differences within and between houses reflect house construction and occupancy factors. 䊚 2002 Elsevier Science B.V. All rights reserved. Keywords: Indoor radon; Concentration variation; Meteorological effects
1. Introduction Radon in the air of domestic buildings derives mainly from the adjacent subsoil gas by advective transport, driven by a pressure difference across the housersubsoil boundary ŽNazaroff et al., U
Corresponding author. Tel.: q44-1524-593937; fax: q441524-593985. E-mail address: [email protected] ŽM. Kelly.. 1 Present address: Environment Agency, Cameron House, White Cross Industrial Estate, Lancaster, LA1 4 XQ, UK.
1988.. The effect of changes in meteorological conditions on this system has been seen in a number of studies based on continuous recording of indoor radon levels and meteorological variables at individual houses. These have shown that indoor radon levels fluctuate widely on a variety of time scales, with periods from diurnal to annual Že.g. Nazaroff et al., 1985; Cliff et al., 1987; Furrer et al., 1991; Wilson et al., 1991; Porstendorfer et al., 1994; Hubbard et al., 1996.. ¨ The aim of this investigation was to determine the degree of variability in radon levels on a time
0048-9697r02r$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 8 - 9 6 9 7 Ž 0 1 . 0 0 8 7 6 - 2
158
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
scale corresponding to the time resolution of the charcoal adsorption passive monitoring system, i.e. ; 5 days, as this method is used for reconnaissance and property valuation purposes in a number of countries, although not recommended for radon dosimetry. An additional aim was to determine the extent to which such variability could be explained by regional weather conditions, as characterised by readily available data from a standard meteorological station. For this purpose, indoor radon concentration was monitored by the charcoal adsorption method in three occupied houses over periods of up to 2 years. Two of the houses were located in a bedrock area of high radon potential and the third in a low potential area ŽRowe et al., 1993..
2. Method The 222 Rn activity concentration in air Žhenceforth referred to as radon concentration . was passively sampled using activated charcoal in canisters based on those of George Ž1984. Žsupplied by Canberra Industries .. The deployment period was normally 4᎐5 days, but exceptionally 3᎐8 days, to suit occupant’s requirements. Radon-222 adsorbed by the charcoal was quantified by counting the ␥ photon emission of its progeny 214 Pb, after leaving the sealed canister for a suitable period for establishment of radioactive equilibrium. The efficiency calibration of the HPGe detectors was determined by spiking a ‘baked out’ charcoal canister with a certified mixed nuclide standard. The calculated 222 Rn activity in the canister
was converted into a radon concentration in air using the radon adsorption efficiency factors for charcoal of George Ž1984., which take into account the reduction of radon adsorption by adsorbed water by gravimetrically determining the charcoal moisture content. Since these efficiency factors were for a charcoal canister of different geometry, a further conversion factor was determined by exposing 10 charcoal canisters to a known 222 Rn concentration in air Žusing the facilities of the UK National Radiological Protection Board.. The conversion factor was taken as the mean ratio of the predicted against the observed values. The propagated errors, from the conversion factor, the counting efficiency and the sample count, give a 2 error of approximately 30% in the measured radon concentration. Details of the monitored houses’ construction and of the local geological conditions are given in Table 1. They are located in north-west England, UK, one in the city of Lancaster and two in the village of Arnside and were chosen for their different combinations of local geology and house construction and, especially, the willingness of the occupants to change the charcoal canisters. Charcoal canisters were sited in a main room on each of the floors of the houses and changed in the early morning after a suitable exposure period. Although it was not practicable to obtain detailed information about the occupant’s habits which might affect room ventilation, the use of fires or periods when the houses were unoccupied were noted for houses A and B. The standard meteorological data came from the Lancaster University weather station, located approximately 5 km from Lancaster and 21 km from Arnside, i.e. daily values of maximum and
Table 1 House details House Žlocation.
Bedrockrsediment
Type
External walls
Ground floor
A ŽArnside.
Carboniferous limestone Carboniferous limestone Glacial tillrCarboniferous sandstone
Detached 2-storey
Rendered brick with cavity Limestone with rubble filled cavity Brick with cavity
Suspended floor above bedrock Suspended floor above bedrock Suspended floor above cellar
B ŽArnside. C ŽLancaster .
Semi-detached 2-storey Terraced 2storey q cellar
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
minimum temperature, barometric pressure, rainfall and mean wind speed. To assist in data analysis, the radon series for all three properties were converted by interpolation into a regularly sampled series with a 3-day interval. Mean values for each meteorological variable for concurrent 3-day periods were also calculated. A simple linear regression model was used to investigate the relationships between radon concentration and the recorded meteorological variables, with the general form: A m s a1 q a2 X q a3 Y q a4 Z
159
floor rooms. A strong seasonal effect is evident in these data, with a clear annual cycle of low summer and high winter levels. Superimposed on the seasonal cycle, are irregular fluctuations of an intermediate periodicity, e.g. with a mean peak to peak period of 18 days at house A Ž1991.. These can be of substantial magnitude, with deviations from annual mean radon levels almost as large as those shown by the seasonal cycle, e.g. up to 128% deviation from the mean compared to 140% deviation for the seasonal cycle Ž1991 ground floor, house A..
Ž1.
where X, Y and Z are meteorological variables.
3. Results 3.1. Radon concentration time-series The radon concentration time-series for the ground floor rooms of houses A᎐C are shown in Figs. 1᎐3. Similar results were obtained for first
3.2. Annual mean radon concentration and long period ¨ ariation
The annual and monthly mean indoor radon concentrations calculated from the measurements are shown in Table 2. Notably, the annual mean values for individual houses are similar for successive years ŽA and C.. The first floors have lower values than the ground floors for A and B, but are similar at C. However, the cellar at C has
Fig. 1. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse A, ground floor, 1991..
160
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
Fig. 2. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse B, ground floor, 1991..
substantially higher values than for the floors above, as might be expected. The weighted mean annual concentrations takes into account different room occupancies, using the factors obtained by the UK National
Radiological Protection Board, i.e. 45% for ground floor living rooms and 55% for first floor bed rooms ŽWrixon et al., 1988.. The weighted values for the two houses on limestone ŽA and B. are similar Ž430᎐437 and 416 Bq my3 . and con-
Fig. 3. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse C, ground floor, 1991..
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166 Table 2 Monthly mean
222
Rn concentrations ŽBq my3 .
House A 1991 Ground January 768 February 734 March 688 April 427 May 355 June 309 July 108 August 195 September 260 October 580 November 550 December 600 Annual mean 467 Weighted mean 430 a
161
1992
House B
House C
1991
1991
1992a
First
Ground
First
Ground
First
Cellar
Ground
First
Cellar
Ground
First
751 725 534 342 274 280 53 93 173 476 493 564 399
807 565 591 584 208 142 114 179 494 702 584 745 479
818 452 446 445 139 76 53 85 298 665 493 726 402
991 814 682 460 562 705 307 376 442 768 744 828 631
ND 332 302 223 175 238 53 82 154 298 363 402 240
95 118 97 123 63 61 119 113 67 74 92 99 97
30 32 24 20 13 13 12 15 16 25 28 39 23
32 36 22 19 16 12 9 12 15 23 25 34 23
105 100 96 106 144 108 105 53 135 37 63 80 96
41 38 38 38 32 20 23 20 30 26 27 31 31
30 26 19 23 20 13 16 9 23 26 23 29 22
437
416
23
26
Data substantially incomplete.
siderably higher than for the house on sandstone ŽC, 23᎐26 Bq my3 .. The monthly mean radon concentrations Ž A m . reveal the seasonal variation mentioned above and, consequently, show a high degree of negative correlation with measures of temperature. Table 3 shows the regression parameter values Žstandard errors in parentheses . and the goodness of fit given by the coefficient of determination Ž R 2 . for the Eq. Ž1. regression model with only one variable considered, i.e. the monthly mean value
of the daily maximum temperature in degrees Celsius. This gave a slightly higher degree of correlation than either the monthly mean daily minimum or median temperatures. Whilst the daily mean temperature might be expected to be the best measure, for a system with a linear response, the data were not available to calculate this. Points of note in the table are the high degree of explanation of the monthly radon variation, 87᎐94% at A and a somewhat lower 72᎐87% at B
Table 3 Regression of indoor radon concentration against monthly mean daily maximum temperature w R 2 and parameter values with standard errors in parentheses, Eq. Ž1.x House
Floor
a1
a2
R2 Ž%.
House A 1991
Ground First Ground First Ground First Cellar Ground First
957.9 Ž58.5. 929.3 Ž44.6. 1066.5 Ž72.5. 1015.3 Ž80.3. 1070.2 Ž84.1. 509.8 Ž43.8. 101.8 Ž18.1. 40.0 Ž3.7. 41.3 Ž2.6.
y40.9 Ž4.5. y44.2 Ž3.4. y48.3 Ž5.5. y51.0 Ž6.1. y35.6 Ž6.5. y21.4 Ž3.2. y0.7 Ž1.4. y1.5 Ž0.3. y1.7 Ž0.2.
89.3 94.3 88.3 87.4 75.3 82.8 0.2 72.3 87.0
House A 1992 House B 1991 House C 1991
162
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
and C, compared to a total lack of association for the cellar at C. Parameter a1 values reflect the bedrock radon potential, with high values for the limestone bedrock area ŽA and B. and low for the glacial tillrsandstone ŽC.. The relatively low value for house B first floor indicates greater degree of isolation from the bedrock source than occurs at house A, due to construction or habit factors. The contrast between the two areas in parameter a2 values is interesting, with the higher values for A and B indicating a more direct coupling to the source, for whatever reason. 3.3. Intermediate period radon concentration ¨ ariation
In order to test the hypothesis that the intermediate period variation, at least in part, is a function of regional scale weather conditions, the mean radon concentrations for 3-day intervals in the first year Ž1991. were regressed against the 3-day mean meteorological data series. Tests showed that the meteorological data and the ground floor radon concentration data for all three houses were normally distributed and, hence, suitable for regression analysis. For the first floor, data for houses A and C departed only slightly from a normal distribution, whilst those for house B were normally distributed. The fit of the linear regression model wEq. Ž1.x between radon and various combinations of meteorological variables for both floors of all three houses is given by the R 2 values in Table 4. It is
clear from these modelling results that temperature is the most important single variable that may be used to describe the radon series, here shown by daily maximum temperature. The percentage fit is highest at A and ranges overall from 52 to 74%. This is a reduction in the degree of association with temperature compared to the monthly data set but, since the fit remains relatively high despite a 10-fold increase in the number of data, it indicates a degree of control by short period temperature fluctuations in addition to the seasonal influence. For the best case Žhouse A., for temperature alone, the regression parameters with standard errors in parentheses are estimated as a1 s 964.6 Ž32.7. and a2 s y40.56 Ž2.5.. Mean wind speed Žkm hy1 . was also noted to improve the fit when used as a second regressor alongside temperature. Rainfall and pressure produced some improvements in model fit for some, but not all houses. However, it is important to note the danger of models becoming over-parameterised, especially where additional regressors only serve to provide minor increases in model fit. In general, a simple model combining maximum temperature and wind speed, with or without rainfall, provides not only a common model structure for all three properties, but also furnishes a reasonably good fit for each location whilst remaining parametrically efficient. Table 5 gives the R 2 and regression parameter values for the Eq. Ž1. model with X as mean daily maximum temperature, Y as total rainfall in mm, and Z as mean wind speed in km hy1 . Overall, the models
Table 4 Correlation coefficient Ž%. for regression between 3-day mean values of radon concentration and meteorological variables House A 1991 Ground Max temp Wind speed Rainfall Pressure Max tempq wind Max tempq rain Max tempq pressure Max tempq windq rain Max tempq windq pressure q rain
74.7 2.8 0.6 5.4 82.1 80.3 77.5 83.3 83.6
House B 1991 First 74.0 6.4 1.4 10.5 87.3 81.0 81.0 87.9 87.9
Ground 52.0 7.6 1.3 3.6 63.4 56.8 53.1 63.4 67.8
House C 1991 First 68.8 1.5 0.2 0.8 76.1 72.4 70.2 76.1 77.2
Ground 53.4 7.1 8.6 3.7 62.4 63.1 55.8 63.6 66.8
First 64.0 5.2 8.4 1.8 71.0 73.0 64.9 74.2 78.5
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
163
Table 5 Regression of radon concentration Ž1991 3-day means. against mean daily temperature Ž a1 ., total rainfall Ž a2 . and mean wind speed Ž a3 . w R 2 and parameter values with standard errors in parentheses, Eq. Ž1.x House House A House B House C
Ground First Ground First Ground First
a1
a2
a3
a4
R2 Ž%.
1179.0 Ž41.5. 1296.6 Ž41.1. 1320.0 Ž79.1. 668.5 Ž37.1. 48.7 Ž2.5. 47.3 Ž2.5.
y42.7 Ž1.9. y48.9 Ž1.9. y36.2 Ž3.8. y25.0 Ž1.8. y1.5 Ž0.1. y1.3 Ž0.1.
y11.4 Ž2.7. y19.7 Ž2.6. y19.0 Ž4.4. y8.0 Ž1.8. y0.6 Ž0.2. y0.6 Ž0.2.
y12.0 Ž4.6. y10.1 Ž4.6. 0 0 0 0
83.3 87.9 63.4 76.1 62.4 71.0
explain 62᎐88% of the radon variation, with house A again giving the best fit. For the ground floor of each house in 1991, the radon concentrations predicted by the models described by Table 4 are compared with the measured values in Figs. 1᎐3. Some of the unexplained variation will be due to unusual occupant living conditions, such as the noted unoccupied periods at A. For house C, more sections of radon data lie outside the estimated model but this is likely to be due to the large amount of missing data, which makes the model inefficient. The radon concentration was also recorded during 1992 in house A and much less completely
from C. For the ground floor of A, the modelled data closely fitted the measured concentrations, with an R 2 value of 84% ŽFig. 4.. The estimated model parameters are and a1 s 1281 Ž51.7., a2 s y48.8 Ž2.40., a3 s y1.1 Ž5.61. and a4 s y14.8 Ž2.96.. However, although the same regressors are used in the 1991 and 1992 models, namely maximum temperature, wind speed and rain, the parameters multiplying the rainfall and wind terms are notably different. 3.4. Prediction of radon concentration time-series In order to validate the regression models esti-
Fig. 4. Measured and modelled radon concentration, based on the temperature, wind and rainfall regression model, with model standard error bounds Žhouse A, ground floor, 1992.. For comparison, the radon concentration predicted by the 1991 model is also shown.
164
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
mated from the 1991 data, the best fit model for house A in 1991 was used to predict the 1992 radon concentration from the 1992 meteorological data. The predicted concentrations are compared with the measured values in Fig. 4 Žnote this also includes the directly modelled values as described above.. The general trend of the predicted fit is good, with an R 2 of 80.7%, which compares well with the 84% fit produced by direct modelling of the measured 1992 radon concentration.
4. Discussion and conclusions The difference in radon concentrations between the three houses is due primarily to the influence of bedrock geology on soil gas levels. Thus, the weighted mean annual radon concentrations for the two houses on limestone Ž416᎐437 Bq my3 . are considerably higher than the corresponding average value for the UK of 20 Bq my3 ŽKendall et al., 1994., whereas the values for the house on glacial till and sandstone are close to the average Ž23᎐26 Bq my3 .. These values are in keeping with what is known about the relative radon potential of the rock types of the Arnside and Lancaster areas. The local Carboniferous limestone has been assigned to a ‘high’ radon potential category by Appleton and Ball Ž1995., due to its high fissure permeability, especially in the near surface zone of karstic weathering. This category agrees with the 3᎐10% incidence of houses with radon levels above the 200 Bq my3 action level determined by the National Radiological Protection Board surveys ŽLomas et al., 1996.. Whilst the local Carboniferous sandstone has a ‘moderate’ classification, the cover of glacial till reduces this radon potential to the ‘low’ category ŽAppleton and Ball, 1995., corresponding to the - 1% incidence above the action level seen in the NRPB surveys ŽMiles et al., 1996.. The seasonal cycle seen in monthly radon concentrations is similar to that obtained from UK national surveys by combining shorter period data ŽWrixon et al., 1988.. The shorter period variation in radon levels characteristic of all houses can also be seen in long records obtained from else-
where by instrument based measurements Že.g. Hubbard et al., 1996.. In addition, such records have higher frequency variations which have been filtered out in our case by the passive sampling method used. It has been shown that much of the long and intermediate period variation in the charcoal canister measurements of radon concentration can be related to the regional climate, as described by the meteorological data from a remote weather station. For both the monthly and 3-day data, most of the variation is explained by temperature, using mean daily maximum temperature as its measure. This accounts for 72᎐94% and 52᎐75% of the variation for monthly and 3-day variation, respectively. The explanation for this strong negative relationship is that external temperature is inversely related to the outdoor᎐indoor temperature difference, when the indoor temperature is kept constant. Such constancy, at a relatively high temperature, is the aim of occupants, being obtained by control of heating and ventilation. The underlying radon transport mechanism will be the stack effect, which occurs when a higher indoor than outdoor temperature generates a pressure difference which then draws in small volumes of soil gas with high radon concentrations ŽNazaroff et al., 1988.. However, Hubbard et al. Ž1996. demonstrated that, although this was the normal condition for Swedish houses, some showed the opposite effect, with reduced indoor radon level as temperature difference increased, due to increased ventilation by outdoor air with low radon concentrations. Other meteorological variables, which appeared to have a contributory, but lesser effect on radon concentration, are wind speed, rainfall and barometric pressure. These have also been previously identified as influencing radon transport from the ground into houses. Wind pressure on building surfaces is widely assumed to have a positive effect on radon entry from the soil by generating an over-pressure on the surface of the building shell ŽNazaroff et al., 1988. or by developing an under-pressure by the Bernoulli effect of wind blowing over open chimneys ŽKendall et al., 1994.. Conversely, Riley et al. Ž1996. suggest that soil gas radon levels could be diluted by wind
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
induced inflow of air. Rainfall can also have an influence on the radon concentration, by causing relative differences in permeability and soil air pressures below and beyond the house, as can barometric pressure where the ground is differentially saturated or frozen ŽNazaroff et al., 1988.. The temporal scale of the periodicity in the radon levels, and in the meteorological variables with which the radon levels are correlated, indicates a relationship with weather systems of a large spatial scale. In this region of north-west England, the weather systems are dominated by the passage of the North Atlantic depressions, which give major variations in air mass temperature together with frontal precipitation, elevated wind speeds and reduced pressure. In such a climate, it is obvious that single charcoal adsorption measurements cannot be used to estimate mean radon concentration for periods much longer than the measurement interval. The variability in the degree of association of radon concentration with weather system variables shown between rooms and between houses will be due to the additional effects of house construction and occupancy factors. Within houses A and C, the close similarity of radon concentration between floors suggests a high degree of air exchange between the two floors. In contrast, for house B, the upper room appears to be relatively isolated from the lower one, giving radon levels typically half those of the ground floor. Although the regression models identified have provided good approximations to the radon concentration time-series measured in the three houses, only linear, constant-parameter models have been considered. Initial work using the recursive regression technique ŽYoung, 1999. for the radon series and meteorological variables for house A has shown that the parameters of the regression model exhibit significant time variability throughout the year. This could indicate that either there is an additional regressor influencing the measured radon concentration, which has not yet been identified, or that one of the meteorological regressors is in fact influencing the radon concentration in a non-linear way. Therefore, future work should focus on identifying the nature
165
and character of the time varying parameter sequence, which when measured or described mathematically can be incorporated in the model structure, resulting in a much improved model fit.
Acknowledgements We would like to thank the occupants of the houses for their patience and help with the radon sampling. References Appleton JD, Ball TK. Radon and background radioactivity from natural sources: characteristics, extent and relevance to planning and development in Great Britain. Technical Report. 1995; WPr95r2, British Geological Survey. 1995:93 pp. Cliff KD, Wrixon AD, Green BMR, Miles JCH. Concentrations of radon in dwellings in the United Kingdom. In: Hopke PK, editor. Radon and its decay products. Washington, DC: American Chemical Society, 1987:536᎐559. Furrer D, Crameri R, Burkart W. Dynamics of Rn transport from the cellar to the living area in an unheated house. Health Phys 1991;60:393᎐398. George AC. Passive, integrated measurement of indoor radon using activated carbon. Health Phys 1984;46:867᎐872. Hubbard LM, Mellander H, Swedjemark GA. Studies in temporal variations of radon in Swedish single-family homes. Environ Int 1996;22:S715᎐S722. Kendall GM, Miles JCH, Cliff KD, Green BMR, Muirhead CR, Dixon DW, Lomas PR, Goodridge SM. Exposure to radon in UK dwellings. Report NRPB-R272. Chilton: National Radiological Protection Board, 1994 43 pp. Lomas PR, Green BMR, Miles JCH. Radon atlas of England. Report NRPB-R290. Chilton: National Radiological Protection Board, 1996 16 pp. Miles JCH, Green BMR, Lomas PR. Radon affected areas: England. Documents of the NRPB 1996;7Ž2.:2᎐9. Nazaroff WW, Fuestal H, Nero AV, Revzan KL, Grimsrud DT, Essling MA, Toohey RE. Radon transport into a detached one-storey house with a basement. Atmos Environ 1985;19:33᎐46. Nazaroff WW, Moed BA, Sextro RG. Soil as a source of indoor radon: Generation, migration and entry. In: Nazaroff WW, Nero AV, editors. Radon and its decay products in indoor air. Wiley-Interscience, 1988:57᎐112. Porstendorfer J, Butterweck G, Reineking A. Daily variation ¨ of the radon concentration indoors and outdoors and influence of meteorological parameters. Health Phys 1994; 67:283᎐287. Riley WJ, Gadgil AJ, Bonnefous YC, Nazaroff WW. The effect of steady winds on radon-222 entry from soil into houses. Atmos Environ 1996;30:1167᎐1176.
166
J.E. Rowe et al. r The Science of the Total En¨ ironment 284 (2002) 157᎐166
Rowe JE, Kelly M, Hewitt CN. The occurrence of high indoor radon levels in Carboniferous bedrock areas of NW England. Radiat Prot Dosim 1993;26:201᎐205. Wilson DL, Gammage RB, Dudney CS, Saultz RJ. Summertime elevation of 222 Rn levels in Huntsville Alabama. Health Phys 1991;60:189᎐197. Wrixon AD, Green BMR, Miles JCH, Cliff KD, Francis EA,
Driscoll CMH, James AC, O’Riordan MC. Natural radiation exposure in UK dwellings. Report NRPB-R190. Chilton: National Radiological Protection Board, 1988 188 pp. Young PC. Nonstationary time series analysis and forecasting. Prog. Environ. Sci. 1999;1Ž1.:3᎐48.