Baseline and Quality Reference Values for Natural Radionuclides in Soils of Rio de Janeiro State, Brazil

A first large-scale systematic survey of natural radioactivity contents of soils of the state of Rio de Janeiro is presented, focused on the establishment of Quality Reference Values (QRVs). Undisturbed soil samples were collected from 243 areas and analyzed by gamma spectrometry. The activity contents varied largely, ranging from 12.2 to 1,029 Bq kg for K (geometric mean of 111.1 Bq kg), from 3.5 to 99.8 Bq kg for Ra (geometric mean of 29.7 Bq kg), and from 5.4 to 314.5 Bq kg for Ra (geometric mean of 67.1 Bq kg). The highest contents of radium isotopes were found in soils developed on igneous rocks (Leptosol), and the lowest in a soil of sedimentary origin (Podzol). Among the different soil types, the radioisotope contents differed substantially. Separate QRVs were calculated for each radionuclide by the 75 and 90 percentile approach, and the QRVs were estimated for each soil type. The results emphasized the restrictiveness of QRVs based on the 75 percentile or of a single overall QRV for all soils. Therefore, rather than estimating a separate QRV for each radionuclide for the State, we suggest the use of an upper threshold value, defined as the 90 percentile, and a specific QRV for each soil type area.


INTRODUCTION
Natural radionuclides as 40 K and those originated in the decay chain of 238 U and 232 Th are present in soils and contribute with about 80 % to the dose of background radiation received by human beings (Unscear, 2000).Among these natural radionuclides, 226 Ra and 228 Ra are of particular interest, due to their environmental mobility and consequent capability of entering in the food chain and also because of the high gamma energy of their decay products, and the radon production.Therefore, knowledge about the contents of natural radionuclides in soils is essential for an accurate assessment of possible radiological risks to human health in any region.
Even though 40 K and radium isotopes occur naturally in soils, their contens may be increased by the disposal of liquid, air, and solid effluents, or storage of wastes from industries that generate naturally occurring radioactive material (NORM), e.g., oil and gas production, and many conventional mining and milling industries (IAEA, 2014).Moreover, the extensive use of phosphate fertilizers in cultivated soils was mentioned as a potential source of natural radionuclides, affecting soils and crops (Saueia and Mazzilli, 2006;Sheppard et al., 2008;Hamidalddin, 2014;Todorović et al., 2015).Therefore, to assess the effect of soil contamination by natural radionuclides and resulting increase of human exposure to ionizing radiation, a reference database of the natural background in soils is necessary.In this way, quality reference values (QRVs) for radionuclides can be established, using the same methodology as for heavy metal QRVs.The QRVs can be a guide to evaluate soil contamination and to make decisions regarding remediation actions for a more efficient environmental management (Ballesta et al., 2010;Alfaro et al., 2015;Lima et al., 2016).
The National Council of Environment of Brazil (Conama) established a list of QRVs for heavy metals and other substances listed in Resolution 420 (Brasil, 2009), determining soil-quality guidelines of chemical substances.However, this resolution did not include radionuclides, and decisions about remediation of radionuclide-contaminated areas are based on risk analysis (Lauria and Rochedo, 2005;Peres, 2007).For this analysis, local levels of background radionuclides should be available, but in the absence of this information, an approach based on international data is suggested (Marssim, 2000).However, this method is rather uncertain because of the differences among regions.The lack of these data justifies a survey for a baseline evaluation and the development of QRVs of natural radionuclides for specific soils of each region.
The Rio de Janeiro State (RJ) comprises an area of 43,781.6 km 2 (IBGE, 2016), with different soil types, geology, and climates (Ceperj, 2014).The differences between these variables can give rise to varying levels of natural radionuclides in soils.As a result, the heterogeneity of the environmental variables requires extensive surveys to determine a list of QRVS for these soils.Some major facilities for nuclear energy generation in Brazil are located in Rio de Janeiro: two nuclear power plants and a nuclear-fuel enrichment facility.In the northern region of the State is located a facility for processing monazite sand, which also contains thorium and uranium.Rio de Janeiro has extensive agricultural areas, mainly small-scale vegetable production and sugarcane plantations, with intensive use of fertilizers.Altogether, these potential sources of contamination reinforce the importance of investigating the natural radiation background and establishing reference values for natural radionuclides for the State.
This study surveyed the background contents of natural radionuclides in soils, to investigate the relationships among soil properties and radioactive isotopes and to establish a list of QRVs for natural radionuclides in soils of Rio de Janeiro.

Study area
The level of industrial development in Rio de Janeiro ranks second nationwide and it is the second most populated State in the country, with around 16 million inhabitants and 92 municipalities (IBGE, 2016).It is divided into eight administrative regions: Northwestern, Northern, Seacoast plains, Metropolitan, Mountain region, Center-South, Medium Valley of the Paraíba do Sul River, and Green Coast (Ceperj, 2014).The Seacoast region comprises a sequence of sedimentary plains of fluvial or marine origin, from the south to the north of the State, with some granite and gneiss mountain massives from the Precambrian.In the center of the State, there is a large chain of granite, migmatite and gneiss mountains, divided into three segments: Serra das Araras, Serra dos Órgãos, and Serra do Desengano, with a maximum altitude of 2,316 m a.s.l.Behind the mountain chain, there is a sedimentary plateau in the direction of the neighbor Minas Gerais State.On the western border with the states of Minas Gerais and São Paulo, another chain of igneous and metamorphic mountains occurs together with a sedimentary plateau, formed by sediments from these hills.In the Northern region of the State, there is a vast plain formed by sediments from the river Paraíba do Sul, and in the Northwestern region, another high plateau with granite, migmatite, and gneiss formations.Alkaline rocks with uranium and thorium associated with minerals are spread in small areas of the State (Silva and Cunha, 2001).
A variety of soils can be found in the State.The main soils types are Red Yellow Ferralsols, Red Yellow Acrisols, Cambisols, Red Acrisols, Fluvisols, Podzols, Gleysols, and Leptosols (Carvalho Filho et al., 2003;WRB, 2014).Its ecosystems are heterogeneous, with lagoons, mangroves, swamps, wetlands, sandbank vegetation, forests, and grassland areas.The climate is also varied, classified according to the Köppen System as Aw, Am, Af, BSh, Cfa, Cfb, Cwb, and Cwa.The rainfall is strongly influenced by the geographical localization and orographic factors.In the mountain region, precipitation is higher than in the Metropolitan and Northern regions and than in the Medium Valley of Paraíba do Sul (Carvalho Filho et al., 2003).

Sample collection and preparation
A total of 243 samples of topsoil (0.00-0.20 m) were collected in a previous survey performed by Lima (2015) to establish the QRVs for heavy metals in soils of Rio de Janeiro (RJ).Undisturbed soils or soils with minimum anthropic interference (natural vegetation or pasture with no fertilization) were sampled in areas of all regions of the State.The sampling points were selected based on a combined analysis of soil, geology, land cover, and land use maps (scale 1:500,000).A road map of the Rio de Janeiro State was used to orient the projection of the sampling areas.The maps were superimposed using the program ArcGIS Desktop 10, produced by the Environmental Systems Research Institute -ESRI (Redlands, CA).To define the locations of sampling points, the cLHS program -Conditioned Latin Hypercube System was used.The sampling areas were defined to be at least 200 m away from roads to avoid any contamination.The sampled soils were those with largest area of occurrence in the State: Red Yellow Ferralsols, Red Yellow Acrisols, Rev Bras Cienc Solo 2018;42:e0170146 Cambisols, Red Acrisols, Fluvisols, Podzols, Gleysols and Leptosols (WRB, 2014).According to Brazilian Classification System, the above described soils are respectively Latossolo Vermelho Amarelo, Argissolo Vermelho Amarelo, Cambissolo, Argissolo Vermelho, Neossolo Flúvico, Espodossolo, Gleissolo, and Neossolo Litólico (Santos et al., 2013).The samples were georeferenced and labeled in the field.
The samples were air-dried and sieved through 2 mm mesh at the Laboratory of Soil Pollution of the Universidade Federal Rural do Rio de Janeiro, where the chemical and physical-chemical properties were also determined and the soil types classified according to the Brazilian Classification System (Santos et al., 2013) and the FAO/WRB Classification System (WRB, 2014).The particle-size distribution analysis (sand, silt, and clay), pH(H 2 O), cation-exchange capacity (CEC), and organic matter were determined by the methodology described by Donagema et al. (2011).

Analytical methods
The soil samples were filled in 300 mL pots, sealed and stored for at least 30 days to ensure the equilibrium between 226 Ra and its decay products.Afterwards, the radionuclides in the samples were determined by gamma spectrometry with hyperpure germanium (HPGe) detector systems (Canberra Inc.Meriden, CT, USA) at the Laboratory of Gamma Spectrometry (LSG) of the Institute of Radiation Protection and Dosimetry -IRD (Rio de Janeiro).The activity content of 228 Ra was derived from the 338.3 keV, 911.6 keV, and 969.1 keV peaks of 228 Ac (Bé et al., 2010) and gamma-ray peaks at 351.9 keV ( 214 Pb) and 609.3 keV ( 214 Bi) were used to determine the activity of 226 Ra (Bé et al., 2006).To determine the 40 K content, its photopeak of 1,460.8keV was used.The counting time of the samples was 60,000 s and the background was measured during 230,000 seconds to decrease uncertainty.The detectors were calibrated efficiently with a standard aqueous solution containing various radionuclides in a nitric acid medium, supplied by the Brazilian Laboratory of Metrology of Ionizing Radiations, certified by the Bureau International des Poids et Measures (BIPM), France.The energy calibration was routinely performed with a 152 Eu source.The activity contents and associated uncertainties were determined according to the statistical uncertainty of peak area, using Genie2000 software (Canberra Inc, Meriden, USA).
The Minimum Detectable Activity (MDA) value was around 12 Bq kg -1 for 40 K, 2 Bq kg -1 for 226 Ra, and 3 Bq kg -1 for 228 Ra, for a counting time of 60,000 s (IAEA, 1989).To ensure the analytical quality, a customized certified reference material of soil-089 produced by ERA Inc. (Colorado, USA, certified by the National Institute of Standards and Technology -NIST) was analyzed several times as a blind sample among the others.The Laboratory of Gamma Spectrometry (LSG) had a good performance in intercomparisons with that of the Mixed Analyte Performance Evaluation Program, United States Department of Energy DOE-Radiological and Environmental Science Laboratory.The Brazilian Laboratory of Metrology organizes national intercomparison exercises in which the LSG also obtained a good agreement of results.

Statistical data treatment and map design
The results obtained by sample analysis were statistically evaluated with software Statgraphics Centurion XVII for descriptive statistics, frequency histograms, and testing of distribution fitting.The QRVs for radionuclides were established acccording to the Brazilian legislation (Brasil, 2009).The outliers for each radionuclide were identified and removed from the dataset, the frequency distribution of the activity content was tested, and the QRVs were calculated.Software SPSS version 20 (IBM Inc, USA) was used for the multivariate and correlation analyses.The geographical location of the sampling points is shown on the soil map of the State (Figure 1), adapted from Carvalho Filho et al. (2003).The map was designed with the free download software Qgis v.2.12.2 Lyon (Open Source Geospatial Foundation Project, 2016).

Analysis results of the certified reference material
The certified activity contents of the certified reference material (Soil 089/ERA) and the means of the results of the six detection systems used are listed in table 1.The differences were evaluated statistically (confidence interval 95 %).The Z-score values were calculated according to equation 1, in which X ob is the content obtained in this study, X cert is the certified value, and σ is the standard deviation (IAEA, 2012).According to the protocol, the laboratory performance is considered satisfactory if |z score| ≤2 and questionable for 2< |z score| ≤3.The Z-score values showed satisfactory results for all detection systems and radionuclides.

Soil physical and chemical characterization
Table 2 shows the statistical summary of the sample analyses, performed by Lima (2015).
Most of the soils are acidic [pH(H 2 O) arithmetic mean of 5.4] and have a medium organic matter content (5.3-69.3g kg -1 ).The CEC varied from 2.6 to 463 cmol c dm -3 (average 76 cmol c dm -3 ) and the average sand content was higher than that of silt and clay.

Descriptive statistics of radionuclide content in all soil samples
Descriptive statistics of the radionuclide activity contents were performed for all soil samples (Table 3).The goodness-of-fit (GOF) test indicated a log-normal distribution of the results and in this case, the geometric mean represents the main tendency.

Comparison of activity content with other areas in Brazil and worldwide
Table 4 shows comparative data found in the literature for Brazilian soils.Most results were based on arithmetic means, but only the results of authors who reported geometric means  (1) Arithmetic mean.
(3) Organic matter: determined by oxidation with K 2 Cr 2 O 7 .pH in water (1:2.5, v:v).H+Al: acidity potential extracted with calcium acetate 0.5 mol L -1 . Sand, silt clay determined by total dispersion method.CEC: cation exchange capacity at pH 7.0.will be compared.Thus, the geometric means for 226 Ra and 228 Ra found in this study are lower than those reported by Amaral (1992) for the areas with highest background radiation of Poços de Caldas (Minas Gerais), and quite similar to those reported by Lauria et al. (2009) for the agricultural area of Paty de Alferes and Teresópolis, also in Rio de Janeiro.
Comparing the range of values, the soils of São Paulo State (Hiromoto et al., 2007) reached lower values than those found in this survey, although our values are lower than those reported by the Brazilian HBRA of Poços de Caldas (Minas Gerais), Caetité (Bahia), the phosphate region of Pernambuco (Amaral and Mazzilli, 1997;Santos Júnior et al., 2006;Cardoso et al., 2009), and in the area rich in monazite sands in São Francisco de Itabapoana, also in Rio de Janeiro (Lauria et al., 1999).For 40 K, this study found higher values than those reported by Veiga et al. (2006) for the sand beaches in the southern part of Rio de Janeiro and for São Paulo and Bahia.
In conclusion, most of Rio de Janeiro State was found to be Normal Radiation Background Area, however with a higher radiation level than in São Paulo State.The presence of igneous rocks in a significant part of the State may be responsible for this higher radiation than in São Paulo.
Of the sample data set (n = 243) only seven samples (2.88 %) had higher 226 Ra than 228 Ra values.This is a typical result for Brazilian soils, known to contain more 232 Th than 238 U (Pfeiffer et al., 1981;Linsalata et al., 1989).The 228 Ra/ 226 Ra ratio varied between 0.79 and 8.77, with an average of 2.52 and geometric mean of 2.28.
The United Nations Scientific Committee of the Effects of Atomic Radiation (Unscear), in Annex B of the publication "Sources and effects of ionizing radiation" (Unscear, 2008), reported a range of 40 K and 226 Ra contents (Table 5) for various countries.The 228 Ra is not mentioned, since the available data for this radionuclide are very restricted.Comparisons with 238 U and 232 Th data were not performed because the radioactive equilibrium of the decay chains is not usually found in open environmental systems (Ivanovich and Harmon, 1992;Berkowitz et al., 2014).It is noteworthy that the 40 K contents in this study were lower than those reported for Cuba, China, Republic of Korea, Estonia, Finland, Sweden, Germany, Luxembourg, Ireland, Portugal, Spain, Czech Republic, Slovakia, Slovenia, and Greece.On the other hand, the 40 K values were similar to the values of Poland and Switzerland and higher than in Argentina.The 226 Ra activity content ranged from 3.5 to 99.8 Bq kg -1 , i.e., the highest value is below the range reported for Algeria, USA, Costa Rica, Cuba, China, Sweden, Germany, Ireland, Spain, Switzerland, Bulgaria, Czech Republic, Poland, Slovakia, Slovenia, Cyprus, and Montenegro, and similar to the values reported for Japan, Malaysia, Iran, and Lithuania.Overall, this comparison showed that the natural radionuclide contents of RJ soils are within the worldwide range.

Evaluation of radionuclide content of each soil type
The frequency distribution and descriptive statistical analyses were performed for each soil type (Table 6).Less than 1.7 % of the total samples were outliers, identified by the Grubb's test and removed from the data set.As commonly observed in environmental data distribution, the results fit a lognormal distribution, and their geometric mean values represent better the main tendencies.
The highest median 40 K value was observed in the Gleysol, which may be ascribed to the presence of primary minerals such as mica and K-feldspar, leading to high total K contents and high non-extractable K in lowland soils (Britzke et al., 2012).The highest medians for 226 Ra and 228 Ra were found in the Leptsol, a shallow and young soil derived from granite and gneisses that are part of those rocks for which above-normal levels of natural-series radionuclides were detected (Leal and Lauria, 2016).The lowest values for the three radionuclides were found in Podzols, which is a generally sandy soil derived from either quartz-rich sands and sandstones or sedimentary debris from magmatic rocks.

Correlation between soil parameters and radionuclides
Correlation analysis was performed to provide an overview of the relationships among radionuclides and soil properties.Table 7 shows the correlations, examined by   The low loading values of the radium isotopes within the groups 1, 2, and 3 may indicate a quasi-independent behavior regarding these groups.Therefore, the independence from the other groups of variables and strong correlation between them suggests a probable common natural source (Table 6).Figure 2 shows the principal component in a spatial manner.

Quality Reference Values (QRVs) of radionuclides
The QRVs for the 75 th and 90 th percentiles of the radionuclides (Table 8) were calculated for each one of the main soils of Rio de Janeiro State, corresponding to more than 80 % of the State territory (Carvalho Filho et al., 2003).
Only for the states of Minas Gerais and São Paulo, the QRVs were established for natural radionuclides in soils, both based on the 75 th percentile.Peixoto (2013) studied soils of Minas Gerais and established the QRVs for 40 K, 226 Ra, and 228 Ra (438.9Bq kg -1 , 66.8 Bq kg -1 , and 89.9 Bq kg -1 , respectively).In a comparison with the values determined in this  The establishment of a single QRV for each radionuclide and State is a common approach in many studies (Peres, 2007;Peixoto, 2013).However, the results of this research showed a large variety of QRVs for the different soil types: e.g., considering the 75 th percentile, the QRV for Red Yellow Ferrasol was estimated at 142.6 Bq kg -1 , and at 584.8 Bq kg -1 for Cambisols (Table 6).Mainly for 40 K and 226 Ra, a rather large difference was observed between the single overall value and the value for the specific soil types (Table 6).Five of the eight soil types studied had a higher QRV for 40 K (75 th percentile) than the single overall value for the radionuclide (291 Bq kg -1 ).This trend was observed for all radionuclides, regardless of the percentile used.
In general, the most weathered soils (Red Yellow Ferralsol, Red Yellow Acrisol, and Red Acrisol) had lower QRV than the single overall QRV.As they represent the majority of soils in the State area (≈ 56 % of the area) (Carvalho Filho et al., 2003), the establishment of a single QRV would lead to the conclusion that most Rio de Janeiro soils are contaminated.
The differences among the QRVs for each soil show how challenging the establishment of a single and generic QRV for such diverse environments will be.The results clearly showed the need for determining a soil-type specific QRV, as similarly recommended for heavy metals in soils by Santos and Alleoni (2013).
The Brazilian legislation allows establishing QRVs based on the percentiles 75 th or 90 th .Some authors suggested QRVs for heavy metals in soils based on the 75 th percentile (Cetesb, 2001;Caires, 2009;Santos and Alleoni, 2013;Preston et al., 2014), while few studies suggested QRVs based on the 90 th percentile (Paye et al., 2010;Almeida Júnior, 2016).Variations between the 75 th and 90 th percentile were observed, ranging from 21 to 105 % for 40 K, from 9 to 136 % for 226 Ra, and from 7 to 139 % for 228 Ra (Table 7).The vast difference between the percentiles highlights that the establishment of QRVs based on the 75 th percentile can significantly restrict the assessment.Therefore, the use of the 75 th percentile may cause misinterpretations regarding contamination by overestimating the environmental pollution.Thus, considering the data variability, we suggest that the QRVs for the studied natural radionuclides should be based on the 90 th percentile, which is a less restrictive value.
Figure 3 shows the relative deviation for the 90 th percentile of each soil with regard to the use of a single QRV for all soils.The data indicate that the adoption of a single QRV can lead to an overestimation of the radionuclide contents if applied to all soils, e.g., in the case of Podzol, and underestimate the consequences of pollution in other soils.

CONCLUSIONS
A baseline data set of radionuclides in Rio de Janeiro was established in a systematic and extensive research.The activity contents of 40 K, 226 Ra, and 228 Ra for the main soils of Rio de Janeiro are within the range of normal background radiation areas worldwide.
The radium isotopes are strongly correlated with each other, which confirms their common and natural source, while 40 K is more related to pH, CEC, and silt.The values of radionuclide contents varied among the soil classes.The radium isotope contents were highest in Leptosol and lowest in Podzol, where the 40 K content was also lowest.A list of QRVs for 40 K, 226 Ra, and 228 Ra was established for the eight main soils of the State, based on the 75 th and 90 th percentiles of their contents.Also, a single overall QRV (for the 75 th and 90 th percentiles) was estimated for 40 K (291 and 580), for 226 Ra (43 and 63) and 228 Ra (101 and 130).The large differences between the single overall QRV and specific QRVs for each soil type are noteworthy.Furthermore, notable differences were also observed between the 75 th and 90 th percentile values.Considering the adoption of a single QRV for the State and the establishment of QRV based on the 75 th percentile, the adopted QRV would lead to a false positive assessment of contamination.
We recommend the establishment of QRVs based on the less restrictive value of the 90 th percentile.Moreover, we also suggest the estimation of QRVs for each soil class, instead of a single value for the region, mainly in areas with highly heterogeneous environmental conditions, as found in Rio de Janeiro.

Figure 3 .
Figure 3. Relative deviation of QRVs for 40 K, 226 Ra, and 228 Ra activity contents for each soil class with respect to the single QRV for all soil types.

Table 1 .
Certified content and results of the analysis of the reference material of soil (089/ERA) Figure 1.Soil map of the state of Rio de Janeiro (adapted from Carvalho Filho et al., 2003).White points indicate the soil sampling locations.Rev Bras Cienc Solo 2018;42:e0170146 The 40 K content ranged from 12.2 to 1,029 Bq kg -1 , 226 Ra from 3.5 to 99.8 Bq kg -1 , and 228 Ra from 5.4 to 314.5 Bq kg -1 .The geometric means were, respectively, 111.1, 29.7, and 67.1 Bq kg -1 .Compared with the worldwide median values for 40 K,226Ra, and 228 Ra (400, 35, and 30 Bq kg -1 , respectively) (Unscear, 2000), the 40 K median value in Rio de Janeiro is nearly four times lower, the 226 Ra median value is comparable, and the 228 Ra median value is nearly twice as high as the value cited by the UN institution.

Table 3 .
Summary statistics for the radionuclides analyzed in pooled soil samples(n = 243) Radionuclides ativities concentration were determined by Gamma Spectrometry.

Table 2 .
Summary of statistical analysis of the physical and chemical properties of soil samples (n = 243)

Table 4 .
Activity content of 40 K, 226 Ra, and 228 Ra reported in other studies for Brazilian soils -: not determined.Rev Bras Cienc Solo 2018;42:e0170146

Table 5 .
Ranges of40 K and 226Ra in soils worldwide [adapted fromUnscear (2008)] The 226 Ra is negatively correlated with pH(H 2 O) and positively with 40 K and 228 Ra.The test also showed that 228 Ra has a negative correlation with sand, but a positive one with clay, 40 K, and 226 Ra.The high content of 226 Ra in acidic soil may be related to the accessory minerals in the acidic bedrock and mechanisms of radium retention in soil.The correlation between the two radium isotopes indicates a common source of the radionuclides.
The statistical treatment by principal component analysis (Varimax rotation method) detected four components, responsible for 75 % of the data variance.The first component

Table 6 .
Statistical summary of activity concentrations (in Bq kg -1 ) of radionuclides in the different soil types

Table 7 .
Matrix of Spearman's correlation between the soil properties and the studied radionuclidesRev Bras Cienc Solo 2018;42:e0170146 combines silt, clay, and OM; the 2 nd component OM and acidity (H+Al); and the 3 rd component pH(H 2 O), OM, silt, CEC, and 40 K.The radium isotopes are associated with the 4 th component. *

Table 8 .
Peres (2007)rence Values of 40 K,226Ra, and228Ra for each soil class QRVs for 40 K in Rio de Janeiro soils are lower, except for Cambisols, Gleysols, and Fluvisols.For 226 Ra, except for Leptosols, the values estimated here are lower than those for Minas Gerais; and the QRVs for 228 Ra for Rio de Janeiro are higher than those for Minas Gerais, except for Podzols, Gleysols, and Red Yellow Acrisols, which are quite similar.For soils of São Paulo State,Peres (2007)reported geometric means of 40 K,226 Ra,  and 228 Ra, respectively, as 86.7, 17.1, and 27.8Bq kg -1 .