Radiocesium transfer from Andosols to brown rice in the northern and northwest areas of Tochigi Prefecture, in the first 3 years following the 2011 Fukushima Daiichi nuclear power plant accident

ABSTRACT

Following the Fukushima Daiichi Nuclear Power Plant accident of 2011, the potential for radiocesium transfer from contaminated soils, such as Andosols, to agricultural crops became a significant concern. Andosols account for up to 70％ of paddy soils in the northern and northwest areas of Tochigi Prefecture, where the radiocesium concentration is 1000 Bq kg⁻¹ or greater in the soil of some fields. The present study was carried out in order to determine the phytoavailability of radiocesium in Andosols by comparing it with that of gray lowland soils in the first 3 years following the accident. The transfer factor (TF) tended to be higher in Andosols than in gray lowland soils, leading to higher radiocesium concentrations in brown rice grown in Andosols. The exchangeable potassium (Ex-K₂O) in Andosols was highly and negatively correlated with TF, followed by clay. The Ex-K₂O value was positively correlated with the clay/total carbon (T-C) value, suggesting that a high T–C ratio could weaken K₂O adsorption on clay mineral sites; hence, the low clay/T-C values can partially explain the relatively large TF values of Andosols. Samples with Ex-K₂O contents less than 200 mg kg⁻¹ and with low clay/T-C values showed striking decreases in TF values from 2011 to 2012. However, the decrease from 2012 to 2013 was quite small; radiocesium in these samples was potentially available for rice uptake for a long time, likely due to the reversible adsorption and fixation characteristics of allophane. Most gray lowland soil samples showed very low TF values over the 3 years of the study, except for those with TF values greater than 0.1 due to low Ex-K₂O and clay contents; the geometric mean (GM) value of TF was below 0.01 in 2012. The extraction of exchangeable radiocesium (Ex-Cs) with a 1 mol L⁻¹ ammonium acetate solution may not be an appropriate method for explaining the variability in radiocesium TF in Andosols. This is because the Ex-Cs value was significantly correlated with Ex-K₂O in Andosols, but not in gray lowland soils, indicating that Ex-K₂O explained this variability in relation to Ex-Cs.

KEY WORDS:

• Andosols
• clay/total carbon
• exchangeable potassium
• radiocesium
• rice

1. Introduction

Following the Fukushima Daiichi Nuclear Power Plant accident in March 2011, a large amount of radionuclides were released into the atmosphere and deposited in the soil, resulting in contamination over a wide area, including Tochigi Prefecture, adjacent to Fukushima Prefecture. The main radionuclides were cesium-134 (¹³⁴Cs) and cesium-137 (¹³⁷Cs), both of which have relatively long physical half-lives (2.06 and 30.17 yr, respectively). These radionuclides will, therefore, pose a substantial risk to not only ecosystems, but also to human health via food chain contamination, for years to decades. The measured amount of radionuclides abruptly decreased at the end of April 2011 (Ohse et al. 2015). Radiocesium absorption by plants has been almost exclusively via root uptake from contaminated soils since the decrease. Rice (Oryza sativa L.) is a significant vector of radiocesium delivery to the human body because rice is a dominant staple crop in Japan. Because of these factors, an urgent soil survey was carried out across 715 fields in Tochigi Prefecture by the MAFF (Takata et al. 2015) and the Tochigi Prefecture Agricultural Department (2012). The highest total radiocesium (¹³⁴Cs+¹³⁷Cs) concentration was detected in the soil of a field in the northern area of Tochigi Prefecture, in Nasu Region. The measured value was about 4000 Bq kg⁻¹, which is below the upper safety limit of 5000 Bq kg⁻¹ for radiocesium concentration in agricultural soil. This regulation value was based on the following factors: 1) a 0.1 transfer factor (TF), defined as the ratio of radiocesium concentration in brown rice to that in soil (the value was estimated using data collected before the accident), and 2) the provisional regulation value of radiocesium in food, including brown rice; this was revised from 500 Bq kg⁻¹ in 2011 to 100 Bq kg⁻¹ in April 2012.

The TF value is the most basic, practical, and widely used metric in estimating the phytoavailability of radionuclides. However, the TF value can vary widely depending on environmental conditions, and soil properties. Potassium (K) competitively suppresses radiocesium uptake by plants because of the physicochemical similarity to Cs (Shaw et al. 1992; Shaw 1993; Zhu and Smolders 2000). Following the accident, urgent field experiments in Japan were intensively carried out to study the effectiveness of K fertilization in reducing radiocesium uptake by various crops: rice (Kato et al. 2015), buckwheat (Fagopyrum esculentum Moench; Kubo et al. 2015), and forage crops (Harada, Sunaga, and Kawachi 2015). Lasat, Norvell, and Kochian (1997) showed that the radiocesium concentrations of several crops increased with increasing applications of ammonium nitrogen, which is another competitor of radiocesium. Clay minerals are known to be the dominant adsorbents of radiocesium (Cornell 1993). Even more important is the role that different types of clay minerals play in reducing the phytoavailability of radiocesium as described in a later section. In contrast, phytoavailability increases with increasing organic matter content (Van Bergeijk et al. 1992). A review by Staunton, Dumat, and Zsolnay (2002) determined that organic matter could play an important role in radiocesium mobility and bioavailability by preventing radiocesium access to the adsorption sites on clay minerals.

Andosols account for about half of total upland fields in Japan (MAFF 1979). These soils are characterized by properties such as high organic matter content, which is associated with cation exchange capacity (CEC) dominated by variable negative charges, and a high phosphate retention capacity (Shoji 1984). Allophanic Andosols, in which noncrystalline colloidal materials such as allophane and imogolite predominate account for the majority of Andosol soils in Kanto District, including Tochigi Prefecture (Matsuyama, Saigusa, and Abe 1994). The radiocesium adsorption and fixation characteristics of soils rich in organic matter are different from those of mineral soils. The former can quickly adsorb radiocesium due to high CEC, and therefore high adsorption capacity. However, the capacity to selectively adsorb radiocesium over K is small in organic soils; therefore, the adsorbed radiocesium remains reversibly exchangeable, and available for plant uptake for extended periods (Valcke and Cremers 1994). In contrast, mineral soils in particular weathered micaceous minerals such as illite and vermiculite predominated by 2:1 phyllosilicate contain sites which selectively adsorb radiocesium over K; these are referred to as frayed edge site (FES). Although radiocesium takes some time to reach the FES (Comans, Haller, and De Preter 1991), irreversible adsorption followed by fixation occurs on this site (Comans Haller, and De Preter 1991; Comans and Hockley 1992). The radiocesium interception potential (RIP) proposed by Cremers et al. (1988) is an intrinsic parameter which can be used to categorize soils in terms of their capacity to selectively adsorb radiocesium via FES. Vanderbroke et al. (2012) showed that the RIP values of Andosols are smaller than those of other soil groups worldwide. Yamaguchi et al. (2017) showed that the RIP values of Andosols in northeast Japan tended to be smaller than those of mountainous/terrace soils and lowland soils, decreasing with increasing total carbon content. Delvaux, Kruyts, and Cremers (2000) found a positive correlation between the RIP and TF values. For the above reasons, Smolders and Tsukada (2011) pointed out that Japanese soils would sustain phytoavailability for a longer time than those in Europe, mainly due to the soil properties of Andosols. In 2011 the radiocesium concentrations were about 1000 Bq kg⁻¹ or more in some fields in the northern and northwest areas (Nikko Region) of Tochigi Prefecture, suggesting that the brown rice from these areas may not be below the regulation value of 100 Bq kg⁻¹.

The information on radiocesium transfer from contaminated soil to rice plants in Japan has come mainly from studies carried out in Fukushima Prefecture (Kohyama et al. 2015; Yagasaki et al. 2019a, 2019b), where lowland soils account for large portions of paddy fields (MAFF (Ministry Agriculture, Forestry and Fishers) 1979). The results of Kondo et al. (2015) were obtained from a pot experiment using lowland soils from Fukushima Prefecture. In contrast, Andosols and gray lowland soils are common in the arable land of Tochigi prefecture (MAFF (Ministry Agriculture, Forestry and Fishers) 1979). The former accounts for up to 70％ of paddy soils in the northern and northwest areas (Tochigi Prefecture Agricultural Department 2017).

As described above, the RIP value is the most descriptive indicator of the TF value. An RIP map, along with a soil radiocesium concentration map of northeast Japan have allowed evaluation of the long-term behavior of radiocesium in agricultural soils (Yamaguchi et al. 2017). The analytical procedure is, however, complex and elaborate. The purpose of the present study is to provide essential insight into the behavior of radiocesium by examining soil properties which are generally analyzed in the laboratory. We specifically focus on the phytoavailability of radiocesium in Andosols by comparing it with that in gray lowland soils. This will provide the basic information with which to carry out appropriate agricultural practices to further reduce the risk of high radiocesium concentrations in brown rice grown in Andosols over the long-term.

2. Materials and methods

2.1. Sampling of rice and soils

From September to October 2011–2013, brown rice samples were collected from the northern and northwest regions of Tochigi Prefecture. In 2011 and 2012, brown rice was exclusively obtained from bulk samples harvested by farmers. Some of the samples obtained in 2012 and all of those obtained in 2013 were collected from five locations in each field. The samples were combined, air-dried, and passed through a 1.8-mm sieve to be used for analysis. Most of the rice cultivars were ‘Koshihikari.’ Following the harvest, paddy soil samples were collected from the plow layer using a hand shovel. The five subsamples from each field, were combined, air-dried, and passed through a 2-mm sieve to be used for further analysis.

In total, 25 sample sets of brown rice and soil were collected in 2011. As described below, the counting time for measurement with a germanium semiconductor detector was set to be maximally sensitive. However, in order to accommodate an urgent need for immediate measurement of radiocesium concentrations in brown rice in some samples, the counting time for these was set to 1000 s, with a detection limit of about 4 Bq kg⁻¹. This low level of sensitivity failed to detect radiocesium in seven samples, which were, therefore, excluded from statistical analysis, although they were used in the discussion of interannual changes in TF values. Eleven samples, including the seven described above, were collected for during each of the 3 years in this study. Eighty-four samples sets of brown rice and soil were collected in 2012 and 16 in 2013.

2.2. Sample analysis

Soil pH was measured in water (soil: solution = 1:2.5 [w/v]) after periodic stirring for 1 h. Total carbon (T-C) and total nitrogen (T-N) contents were determined with a CN analyzer (SUMIKA Chemical Analysis Service, NC-220 F, Tokyo, Japan). Cation exchange capacity (CEC) was determined by the semimicro Schollenberger method. Exchangeable calcium (Ex-CaO), magnesium (Ex-MgO), potassium (Ex-K₂O), and radiocesium (¹³⁴Cs+¹³⁷Cs, Ex-Cs) were extracted by using a 1 mol L⁻¹ ammonium acetate solution (soil: solution = 1:10 [w/v], pH 7.0); the former three cations were measured with an atomic spectrometer (Hitachi High-Tech Service, Z-5310, Tokyo, Japan), and radiocesium was measured with a germanium semiconductor detector as described below. Available nitrogen (N) was determined by the incubation method under anaerobic conditions at 30°C for 4 weeks. Particle size distribution categories included clay (<0.002 mm), silt (0.002–0.02 mm), fine sand (0.02–0.2 mm), and coarse sand (0.2–2 mm); these were determined by a sieving and a sedimentation method. The phosphate adsorption coefficient, which is one of the criteria by which soils are classified as Andosols, was determined by the method of phosphate ammonium solution. Soil classifications were based on a soil survey, a soil classification map for each sampling location (National Institute for Agro-environmental Sciences n.d.), and the phosphate adsorption coefficient.

For the determination of the total radiocesium (¹³⁴Cs+¹³⁷Cs, T-Cs) and exchangeable radiocesium (¹³⁴Cs+¹³⁷Cs, Ex-Cs) concentrations in soil, and the total radiocesium (¹³⁴Cs+¹³⁷Cs, R-Cs) concentration in brown rice, the soil samples, and the brown rice samples were each placed in a 1 L and a 2 L of Marinelli beaker, respectively, and measured with a germanium semiconductor detector (relative efficiency 21.4％; SEIKO EG＆G ORTEC, GEM15-70) connected to a multichannel analyzer (MCA7600, SEIKO EG＆G, Tokyo, Japan). The counting time was set to maximize detection (with the exception of the brown rice samples from 2011, which were measured at 1000 s, as described above). The highest counting time for radiocesium concentration in brown rice in 2011 and 2012 was 10000 s, with a detection limit of about 0.8 Bq kg⁻¹, leading to nondetection in some samples. The detection limit value was used for the statistical calculation of the nondetected samples.

The TF was calculated as follows:

$\mathrm{T}\mathrm{F}=\mathrm{R}-\mathrm{C}\mathrm{s}/\mathrm{T}-\mathrm{C}\mathrm{s}$

where R-Cs and T-Cs equal the total (¹³⁴Cs+¹³⁷Cs) radiocesium concentration in brown rice and soil on a dry weight basis, respectively. Analysis was conducted using Gamma Studio software. Radiocesium concentrations were corrected to the collection date based on the radioactive decay.

2.3. Statistical analysis

Statistical analysis was carried out using the statistical software Statcle 3 (Yanai 2011) and Microsoft Excel programs. For normal value distributions of both Andosols and gray lowland soils, differences in arithmetic mean (AM) values between the groups were compared using Student’s t-test or Welch’s t-test at a 5％ level. When the values of at least one group were not normally distributed, differences in medians between the groups were compared using Mann–Whitney’s U test at the 5％ level.

3. Results

3.1. Radiocesium concentrations in brown rice and TF values among the soil groups

Table1 shows the statistical values of radiocesium concentrations in brown rice and TF values among the soil groups. Both categories show striking variations, with relative standard deviations of about 1.0 or greater. The AM of radiocesium concentration and median TF value were significantly higher in rice grown in Andosols than in gray lowland soils in 2011 and 2012, respectively. An examination of AM values reveals that both parameters tended to be greater in Andosols than in gray lowland soils in all 3 years of the study.

Table 1. Statistical values of radiocesium concentrarions in brown rice and TF values among the soil groups.

		R-Cs
	Statistical values	Bq kg^−1							TF
Year	Soil groups	A		G			R	W	A		G			R	W
2011	number	10		8				18	10		8				18
	arithmetic mean	32.0	a	14.0	b	*		24.0	0.061		0.031		-		0.048
	min.	3.2		<0.8				<0.8	0.001		0.001				0.001
	max.	64.9		38.1				64.9	0.231		0.170				0.231
	RSD	0.63		0.95				0.81	1.10		1.81				1.32
	median	35.6		8.2				22.5	0.042		0.010		ns		0.024
	geometric mean								0.031		0.010				0.019
	95％CI (lower)								0.010		0.003				0.009
	95％CI (upper)								0.090		0.038				0.042
2012	number	66		17			1	84	66		17			1	84
	arithmetic mean	19.0		11.5		-	76.5	18.2	0.019		0.013		-	0.257	0.020
	min.	<0.8		<0.8				<0.8	0.001		0.001				0.001
	max.	65.9		62.0				76.5	0.115		0.117				0.257
	RSD	1.01		1.53				1.10	1.12		2.11				1.68
	median	11.0		5.1		ns		9.5	0.012a		0.004b		*		0.010
	geometric mean								0.010		0.005				0.009
	95％CI (lower)								0.007		0.002				0.007
	95％CI (upper)								0.014		0.009				0.012
2013	number	10		5			1	16	10		5			1	16
	arithmetic mean	7.1		1.2		-	37.1	7.1	0.016		0.003		-	0.212	0.024
	min.	0.5		0.2				0.2	0.0008		0.001				0.0008
	max.	33.4		4.1				37.1	0.054		0.007				0.212
	RSD	1.52		1.30				1.67	1.09		0.889				2.18
	median	1.5		0.6		ns		1.1	0.009		0.002		ns		0.004
	geometric mean								0.0080		0.003				0.007
	95％CI (lower)								0.0030		9E-04				0.003
	95％CI (upper)								0.021		0.005				0.015

R-Cs: Total (¹³⁴Cs+¹³⁷Cs) radiocesium concentrations in brown rice. The detection limit in 2011 and 2012 was about 0.8 Bq kg^−1. The value was used for statistical calculation of the nondetected samples.

TF:Transfer factor.

Soil groups were based on the classification of cultivated soils in Japan (third approximation), by the Cultivated Soil Classification Committee (1995). A: Andosols, G:ｇray lowland soils, R: regosolic lowland soils, W: Whole the samples.

RSD: Relative standard deviation, The ratio of the standard deviation to arithmetic mean.

95％CI:95％ confidential intervals.

The arithmetic mean values between Andosols and gray lowland soils were not compared using Student’s t-test or Welch’s t-test because the values of at least one group were not normally distributed. In this case, differences in medians between the groups were compared using Mann–Whitney’s U test.

Values within a column followed by different letters are significantly different at the 5％ level. *: significant at the 5％ level. ns: Not significant at the 5％ level.

The radiocesium concentrations in brown rice did not exceed 100 Bq kg⁻¹ throughout the 3 years. The maximum values were 64.9 Bq kg⁻¹ in 2011, 76.5 Bq kg⁻¹ in 2012, and 37.1 Bq kg⁻¹ in 2013; the two latter values were detected from a field categorized as regosolic lowland soils. The TF values were 0.257 in 2012 and 0.212 in 2013, resulting in relatively high radiocesium concentrations in brown rice.

As also reported by Uchida and Tagami (2007), the TF value exhibited a lognormal distribution throughout all 3 years of the study; therefore, the geometric mean (GM) value was a more appropriate measure than the AM value. The GM values of TF with 95％ confidence intervals (CI) are shown in Table1. In 2011 the GM value for Andosols was 0.031 with 95％ CI [0.010, 0.090]. In 2012 the GM value decreased 0.010, with a narrowed range of 95％ CI [0.0073, 0.014]. The GM value decreased only slightly in 2013 to 0.0080 with 95％ CI [0.0030, 0.021]. The GM values of gray lowland soils were lower than those of Andosols throughout all 3 years, with much smaller 95％ CI. In 2011, the GM value was 0.010 with 95％ CI [0.0027, 0.038]; in 2012 and 2013, GM values were 0.0046 with 95％ CI [0.0023, 0.0094] and 0.0029 with 95％ CI [0.0009, 0.0049], respectively.

3.2. Comparison of radiocesium concentrations in soil, and soil properties of Andosols and gray lowland soils

Table 2 shows the statistical values of radiocesium soil concentrations, and soil properties along with plow depth among the soil groups in 2012 (more samples were collected in this year than in the other two). The statistical values obtained for 2012 are approximately representative of the differences in properties between the two soil groups for all 3 years of the study. Neither T-Cs nor Ex-Cs were significantly different between Andosols and gray lowland soils. As expected from the known characteristics of Andosols, the phosphate adsorption coefficient, T-C, T-N, and CEC of Andosols were all greater than those of gray lowland soils in all 3 years. This was also the case for Ex-CaO and Ex-MgO, which was due to the high CEC of Andosols; however, this trend was not observed in 2013. Similarly, fine particles such as clay were higher in Andosols than in gray lowland soils for all 3 years, whereas coarse sand showed the opposite pattern in 2011 and 2012. Although Ex-K₂O and plow depth differed significantly between the two soil groups in 2012, these differences were not observed in the other 2 years. Both pH and available N of Andosols were significantly greater than those of gray lowland soils only in 2011.

Table 2. Statistical values of radiocesium soil concentrations, and soil properties along with plow depth among the soil groups in 2012.

	Soil groups	T-Cs	Ex-Cs				Ｔ－Ｃ		Ｔ－Ｎ		ＣＥＣ		Ex-ＣaＯ		Ex-ＭgＯ		Ex-Ｋ2Ｏ		Av-N	Clay		Silt
	Statistical values	Bq kg^−1	Bq kg^−1	ｐH (Ｈ２O)	Phosphate adsorption coefficient		g kg^−1		g kg^−1		cmolc kg^−1		mg kg^−1		mg kg^−1		mg kg^−1		mg kg^−1	％		％	Fine sand ％		Coarse sand ％		Plow depth cm
Andosols (n = 66)
	arithmetic mean	1098	44.3	6.1	1944a		81.9		5.96		36.1a		3963		491	a	275		154	12.2a		36.7	32.6		18.5b		17
	min.	112	10.0	5.6	1540		25.0		2.50		14.9		1245		100		133		19	4.4		22.2	25.1		5.5		13
	max.	2931	213.2	7.0	2360		135.8		9.30		52.7		10,628		1422		593		266	21.5		49.8	47.9		31.5		21
	RSD	0.55	0.66	0.05	0.11		0.30		0.23		0.22		0.38		0.46		0.41		0.32	0.27		0.16	0.14		0.31		0.12
	median	977	37.1	6.0	1960		80.3a		5.89a		34.9		3645a		481		234b		148	12.0		37.1	32.0b		18.9		17a
gray lowland soils (n = 17)
	arithmetic mean	1095	53.7	6.0	1145b		45.1		3.51		24.6b		2758		367b		328		129	10.6b		27.8	37.1		24.5a		16
	min.	111	14.5	5.7	740		14.3		1.43		12.7		590		28		158		55	5.5		5.6	25.9		7.5		10
	max.	2195	122.0	7.0	1490		101.0		7.01		46.2		4232		631		542		215	22.4		44.0	57.5		46.1		21
	RSD	0.57	0.59	0.06	0.23		0.61		0.51		0.40		0.41		0.47		0.34		0.33	0.34		0.42	0.21		0.40		0.18
	median	980	45.1	5.9	1140		36.7b		2.95b		21.3		2835b		350		321a		127	10.7		28.1	37.0a		24.1		15b
	arithmetic mean	-	-	-	*		-		-		*		-		*		-		-	*		-	-		*		-
	median	ns	ns	ns			*		*				*				*		ns			ns	*				*
regosolic lowland soils (n = 1)
		298	38.2	6.2	900		32.8		2.49		16.6		2108		156		172		76	3.2		14.9	33.6		48.3		15
Whole the samples (n = 84)
	arithmetic mean	1087	46.1	6.1	1770		73.9		5.42		33.5		3697		462		285		148	11.8		34.6	33.5		20.1		16
	min.	111	10.0	5.6	740		14.3		1.43		12.7		590		28		133		19	3.2		5.6	25.1		5.5		10
	max.	2931	213.2	7.0	2360		135.8		9.30		52.7		10,628		1422		593		266	22.4		49.8	57.5		48.3		21
	RSD	0.56	0.64	0.05	0.23		0.40		0.33		0.29		0.41		0.48		0.40		0.33	0.30		0.24	0.17		0.38		0.13
	median	959	37.9	6.0	1880		74.4		4.58		33.4		3563		457		241		143	11.6		36.5	32.5		19.8		16

Arithmetic mean values between Andosols and gray lowland soils were not compared using Student’s t-test or Welch’s t-test because the values of at least one group were not normally distributed. In this case, differences in medians between the groups were compared using Mann–Whitney’s U test.

Values within a column followed by different letters are significantly different at the 5％ level. * significant at the 5 ％ level.ns: Not significant at the 5％ level.

3.3. Relationship among radiocesium concentrations in brown rice, TF, and main soil properties

The present study was carried out with the aim of investigating the phytoavailability of radiocesium in Andosols by comparing it with that in gray lowland soils. Pearson’s correlation coefficient test was therefore carried out separately for the two soil groups. As small numbers of samples in 2011 and 2013 made statistical significance difficult to detect, only the results from 2012 are shown in Tables 3 and 4.

Table 3. Pearson’s correlation coefficients (r) among radiocesium concentrations in brown rice, TF, and main soil properties for Andosls in 2012.

R-Cs,TF Soil properties	R-Cs		TF		Ex-Cs		Ｔ－Ｃ		Ex-Ｋ2Ｏ		Clay
TF	―		―
Ex-Cs	0.091		−0.157		―
Ｔ－Ｃ	−0.004		0.027		−0.445**		―
Ex-Ｋ2Ｏ	−0.353**		−0.426**		0.317**		−0.278*		―
Clay	−0.334**		−0.334**		0.084		−0.054		0.330**		―
Coarse sand	0.412**		0.264*		−0.013		−0.019		−0.197		−0.579**

** and * Significant at the 1％ and 5％ levels, respectively.

n＝66.

Table 4. Pearson’s correlation coefficients (r) among radiocesium concentrations in brown rice, TF, and main soil properties for gray lowland soils in 2012.

R-Cs,TF Soil properties	R-Cs	TF	Ex-Cs	Ｔ－Ｃ	Ex-Ｋ2Ｏ	Clay
TF	―	―
Ex-Cs	−0.246	−0.260	―
Ｔ－Ｃ	0.406	0.044	−0.190	―
Ex-Ｋ2Ｏ	−0.301	−0.400	0.121	−0.033	―
Clay	−0.345	−0.396	−0.092	0.019	0.085	―
Coarse sand	0.265	0.522*	−0.199	−0.318	−0.126	−0.632*

* Significant at the 5％ level.

n＝17.

Of all Andosol soil properties, Ex-K₂O shows the greatest negative correlation with both radiocesium concentrations and TF in brown rice, followed by clay. In contrast, coarse sand is positively correlated with these parameters. None of the other soil properties were significantly correlated with TF. We found that Ex-K₂O was positively correlated with clay (r = 0.330, p < 0.01) and negatively correlated with T-C (r = −0.278, p < 0.05), resulting in a positive correlation between Ex-K₂O and clay/T-C (r=0.405, p < 0.01), as shown in Figure 1. However, some points fell outside the regression line. Phosphate adsorption coefficient, CEC, Ex-CaO, and T-N are not shown in Table 3. The former three properties are significantly and positively correlated with T-C but not with Ex-K₂O. We found that T-N was negatively correlated with Ex-K₂O, whereas it was highly and positively correlated with T-C (r=0.962, p < 0.01). These results indicate that T-C can explain some of the relationships among other soil properties, such as the positive correlation between Ex-K₂O and clay/T-C in relation to T-N.

Figure 1. Relationship between the Clay/T-C value and Ex-K₂O content in the soil after the harvest in 2012.

As shown in Figure 2(a), an exponential relationship (R²=0.320, p < 0.01) was observed between Ex-K₂O and TF. However, the regression equation could not fully explain the variations in TF. In particular, TF values tended to increase substantially for Ex-K₂O levels below about 200 mg kg⁻¹. For such low values of Ex-K₂O, the TF values of some samples deviated greatly from the regression line. In contrast, TF values decreased small for Ex-K₂O concentrations above about 200 mg kg⁻¹, and showed relatively small variations.

Figure 2. Relationship between Ex-K₂O content in the soil after the harvest and TF in 2012.

The Ex-Cs value was negatively correlated with T-C (r=0.445, p < 0.01). As shown in Figure 2 (a) and 3(a), the relationship between Ex-K₂O and TF is similar to that between Ex-Cs and TF, which reflects that Ex-Cs is significantly and positively correlated with Ex-K₂O (r=0.317, p < 0.01). It is noteworthy that the TF values of some samples are relatively high despite their low Ex-Cs concentrations. Clay is significantly and negatively correlated with coarse sand (r＝-0.579, p < 0.01).

Figure 3. Relationship between Ex-Cs concentration in the soil after the harvest and TF in 2012.

For gray lowland soils, none of the properties were significantly correlated with radiocesium concentrations in brown rice. Only coarse sand was significantly correlated with TF (r=0.522, p < 0.05). Pearson’s correlation coefficients among radiocesium concentrations in brown rice, TF, and the corresponding properties of Andosols in 2012 are shown in Table 4. None of the properties were significantly correlated with Ex-K₂O. As shown in Figure 1, Ex-K₂O was independent of clay/T-C. As shown in Figure 2(b), for Ex-K₂O values greater than about 200 mg kg⁻¹, both TF values and their variations were distinctly smaller than in Andosols. However, in two out of three samples (one of which was regosolic lowland soil) with Ex-K₂O values below 200 mg kg⁻¹, the TF values were greater than 0.1. As shown in Figure 3(b), Ex-Cs was also independent of TF. The TF values of the above two samples were greater than 0.1 despite their low Ex-Cs concentrations. As in Andosols, clay was significantly and negatively correlated with coarse sand (r=-0.632, p < 0.05).

As shown in Figure 4, TF was independent of T-Cs for the two soil groups.

Figure 4. Relationship between T-Cs concentration in the soil after the harvest and TF in 2012.

3.4. Comparison of TF values between Andosols and gray lowland soils with the same Ex-K₂O content

As shown in the above results, Ex-K₂O was most highly and negatively correlated with TF in Andosols. Therefore, to confirm whether TF values of Andosols are greater than those of gray lowland soils, we compared TF values between the two soil groups with the same Ex-K₂O content. The results, along with the values of clay/T-C, T-C, and clay are shown in Table 5.

Table 5. Comparison of TF values between Andosols and gray lowland soils with the same Ex-Ｋ₂Ｏ content.

	Soil groups	Ｋ2Ｏ < 200 mg kg^−1							Ｋ2Ｏ ～400 mg kg^−1							Ｋ2Ｏ > 400 mg kg^−1
		Ex-Ｋ2Ｏ				T-C	Clay		Ex-Ｋ2Ｏ				T-C	Clay		Ex-Ｋ2Ｏ				T-C	Clay
Year	Statistical values	mg kg^−1		TF	Clay/T-C	g kg^−1	％		mg kg^−1	TF		Clay/T-C	g kg^−1	％		mg kg^−1	TF		Clay/T-C	g kg^−1	％
2011	Andosols	n = 4							n = 3							n = 3
		132 ± 27		0.102 ± 0.089	1.34 ± 0.33	101.6 ± 22.3	13.7 ± 4.6		295 ± 39	0.046 ± 0.034		1.68 ± 0.45	103.7 ± 15.2	17.2 ± 3.8		449 ± 33	0.020 ± 0.029		1.85 ± 0.23	97.3 ± 34.1	17.6 ± 4.5
	gray lowland soils	n = 2							n = 4
		178 ± 31		0.057 ± 0.076	2.89 ± 1.10	45.2 ± 14.7	11.9 ± 0.9		310 ± 83	0.0059 ± 0.0064		3.76 ± 0.82	34.3 ± 9.5	12.5 ± 2.8
				-						-							-
2012	Andosols	n = 15							n = 41							n = 10
		168 ± 18		0.034 ± 0.033	1.27 ± 0.40	87.6 ± 18.5	10.8 ± 3.1		250 ± 51	0.017 ± 0.015		1.61 ± 0.83	83.6 ± 24.1	12.2 ± 3.1		498 ± 66	0.0049 ± 0.0063		2.86 ± 2.20	66.4 ± 30.6	14.2 ± 3.9
										0.012	a						0.0022
	gray lowland soils	n = 2							n = 11							n = 4
		159 ± 1		0.067 ± 0.071	1.91 ± 0.56	33.2 ± 5.0	6.2 ± 0.9		303 ± 55	0.0053 ± 0.0056		3.83 ± 2.83	44.7 ± 27.2	11.8 ± 3.7		482 ± 64	0.0076 ± 0.0080		2.48 ± 1.25	52.2 ± 37.5	9.7 ± 2.7
										0.0036	b						0.0060
									(n = 3	0.011 ± 0.080)						(n = 2	0.0092 ± 0.012)
				-						*							ns
2013	Andosols	n = 4							n = 3							n = 3
		97 ± 53		0.032 ± 0.015	1.30 ± 0.32	97.4 ± 19.3	12.8 ± 4.5		253 ± 32	0.0071 ± 0.0058		3.06 ± 1.32	65.6 ± 18.0	18.5 ± 3.3		604 ± 232	0.0022 ± 0.0012		4.21 ± 3.43	65.8 ± 43.0	18.5 ± 2.3
	gray lowland soils								n = 2							n = 3
									383 ± 14	0.0015 ± 0.0005		2.98 ± 0.45	36.8 ± 6.1	10.8 ± 0.2		684 ± 242	0.0034 ± 0.0030		4.83 ± 0.92	29.1 ± 9.9	14.4 ± 7.1
				-						-							-

Because TF values were not normal distributed, the differences in medians between Andosols and gray lowland soils were compared by using Mann–Whitney’s U test. -: the test was not carried out because of the small sample numbers.

Values denote arithmetic mean ± standard deviation, and medians were additionally shown below the values when Mann–Whitney’s U test was carried out.

Values within a column followed by different letters are significantly different at the 5％ level. *significant at the 5％ level. ns: Not significant at the 5％ level.

TF values of gray lowlands soils samples with T-C contents greater than 60 g kg⁻¹ were shown in parentheses.

Because TF values were not normally distributed, as described above, the differences in medians between the two soil groups were analyzed using the Mann–Whitney U test. Only two cases were compared mainly because of the small sample numbers of gray lowland soils. In 2012, TF values in samples with Ex-K₂O content greater than 400 mg kg⁻¹ did not significantly differ between the two soil groups. However, TF values of Andosols with an Ex-K₂O content ranging from 200 mg kg⁻¹ to 400 mg kg⁻¹ were significantly greater than those of gray lowland soils in 2012. An examination of AMs shows that the TF values of Andosols tended to be greater than those of gray lowland soils with the same Ex-K₂O content. In 2012, five samples of gray lowland soils contained T-C contents greater than 60 g kg⁻¹. The TF values of these samples tended to be slightly greater than in the other gray lowland soil samples with almost identical Ex-K₂O content.

The clay/T-C, T-C, and clay values of Andosols were not significantly different among the three soil groups based on Ex-K₂O content. An examination of AM values shows that the clay/T-C value of the groups with less than 200 mg kg⁻¹ of Ex-K₂O was lower than the other two groups, whereas T-C content was higher, as shown in 2012.

3.5. Change in the relationship between Ex-K₂O and TF on a yearly basis

Figure 5(a,b) shows the relationship between Ex-K₂O content and TF for the 3 years of the study. The exponential relationship between Ex-K₂O content and TF was generally observed in Andosols for all 3 years. Some samples from 2011 greatly deviated from the regression line, reflecting the wide range of the 95％ CI, as described above. The regression lines for all 3 years indicated that TF values decreased substantially from 2011 to 2012, with only a small decrease in 2013; this reflects the changes in the GM values, as described above. Additionally, the regression lines showed that the decrease in TF values from 2011 to 2012 was greater for Ex-K₂O content below 200 mg kg⁻¹.

Figure 5. Relationship between Ex-K₂O content in soil after the harvest and TF for the 3 years of the study.

The TF in some gray lowland soil samples was greater than 0.1. The TF values in the rest of the samples were quite small throughout the 3 years of the study.

3.6. Effect of Ex-K₂O, clay, and T-C on TF on a yearly basis

The changes in TF values and Ex-K₂O in samples collected in each of the 3 years are shown in Table 6. The values of clay/T-C, T-C, and clay, and the samples with TF exceeding 0.1 are also shown in Table 6. The samples in which TF exceeded 0.1 were: Andosols, sample A; gray lowland soils, samples L and M; and regosolic lowland soils, sample N. These samples were similarly low in Ex-K₂O content, which was below 200 mg kg⁻¹. Additionally, all except for sample L showed similarly small clay content.

Table 6. Changes in TF values and Ex-Ｋ₂Ｏ in samples collected in each of the three years, and the samples with TF exceeding 0.1.

Year					2011			2012			2013
TF, Soil properties			T-C	Clay	Ex-Ｋ2Ｏ			Ex-Ｋ2Ｏ			Ex-Ｋ2Ｏ
Soil groups	Samples	Clay/T-C	g kg^−1	％	mg kg^−1	TF		mg kg^−1	TF		mg kg^−1	TF
Andosols	A	0.94	81.0	7.6	95	0.231		147	0.115		51	0.035
	B	1.76	99.4	17.4	145	0.081		150	0.027		164	0.021
	C	1.34	115.9	15.5	340	0.085		310	0.021		115	0.021
	D	1.82	82.0	14.9	285	<0.02*		295	0.008		239	0.005
	E	3.29	58.5	19.2	407	<0.007*		241	0.002		552	0.003
	F	4.77	45.5	21.5	405	<0.03*		342	0.010		290	0.003
	G	7.87	25.9	20.4	460	<0.006*		416	0.003		402	0.003
gray lowland soils	H	4.03	22.1	8.9	376	0.015		310	0.006		963	0.007
	I	2.75	39.8	10.9	486	<0.03*		321	0.006		373	0.002
	J	7.20	17.6	12.1	529	<0.005*		234	0.002		560	0.001
	K	6.74	35.9	22.4	662	<0.04*		358	0.004		528	0.002
	L	3.46	35.6	12.3	194	0.170		-	-		-	-
	M	1.50	36.7	5.5	-	-		158	0.117		-	-
regosolic lowland soils	N	0.98	32.8	3.2	-	-		172	0.257		63	0.212

The values of clay/T-C, T-C and clay were used average values for three (two) years.

* the seven samples of which radiocesium in brown rice was not detected because of short counting time for measurement with a germanium semiconductor detector. These data were not used for the statistical analysis.

-:Not collected.

The 2011 TF values of Andosol samples A, B, and C were about 0.1 or greater. The Ex-K₂O contents of samples A and B were less than 200 mg kg-¹. All three samples shared relatively small clay/T-C values of less than 2.0. The three samples showed marked decreases in TF values from 2011 to 2012; however, the decrease between 2012 and 2013 was much smaller than in the years prior, resulting in that the TF values did not decrease down to 0.01 in 2013. In contrast, the TF values of samples E, F, and G from 2011 were less than 0.03. The Ex-K₂O content of these three samples was higher than 400 mg kg-¹; they also showed clay/T-C values greater than 3.0, which was similar to the gray lowland samples H, I, J, and K. The TF values of the seven samples decreased to 0.01 in 2012.

Although we did not have access to the amount of K₂O applied to each field, a yearly fluctuation in Ex- K₂O content was observed in both Andosol samples and gray lowland soil samples. The Ex- K₂O content of gray lowland soil samples (J, and K) decreased by about 300 mg kg⁻¹ or more from 2011 to 2012, whereas that of gray lowland soil samples (H, J, and K) increased by about 200 mg kg⁻¹ or more from 2012 to 2013. A similar increase was observed in the Andosol sample (E).

4. Discussion

4.1. Differences in TF values and soil properties between Andosols and gray lowland soils

The TF value was independent of T-Cs, which is consistent with a previous report by Nisbet and Woodman (2000). They concluded that this independence validates the use of the TF approach in predicting radionuclide uptake by plants. Therefore, the following discussion will mainly focus on TF.

Delvaux, Kruyts, and Cremers (2000) found a positive correlation between the RIP and TF values. The reports of Vanderbroke et al. (2012) and Yamaguchi et al. (2017) show that the RIP values of Andosols are lower than those of other soil groups. However, to the best of our knowledge, there have been no reports which clearly show that the TF values of brown rice grown in Andosols are greater than those for rice grown in other soil groups. Statistical significance was difficult to demonstrate between the TF values of Andosols and gray lowland soils due to the lognormal distribution. However, the median TF value for Andosols was significantly greater than that for gray lowland soils in 2012. Additionally, GM values of TF for Andosols were greater than those for gray lowland soils throughout the 3 years of our study. The soil property we found to be most highly correlated with the TF value of Andosols was Ex-K₂O (r = −0.426, p < 0.01, 2012). In just one case, the TF value of Andosols was significantly greater than that of gray lowland soils for similar values of Ex-K₂O, confirming the above results. Radiocesium concentrations in brown rice grown in Andosols were also greater than those of gray lowland soils.

The soil property most correlated with TF value Andosols after Ex-K₂O was clay, which showed a negative correlation (r = −0.334, p < 0.01, 2012). The TF values of four samples with Ex-K₂O concentrations below 200 mg kg⁻¹, regardless of soil type, exceeded 0.1; the clay content of three contained of these was less than 10％. These results are consistent with the fact that clay is an essential adsorbent of radiocesium (Cornell 1993). In both Andosols and gray lowland soils, coarse sand was significantly and positively correlated with TF, and negatively correlated with clay; hence, coarse sand may play an indirect role in the adsorption of radiocesium by decreasing clay content.

Organic matter plays an important role in radiocesium mobility and bioavailability by preventing its access to adsorption sites on clay minerals (Staunton, Dumat, and Zsolnay 2002). According to Yamaguchi et al. (2017), the RIP values of Andosols decrease with increasing T-C content. The present study provides new insight into the important role of T-C in radiocesium bioavailability in terms of K₂O mobility. The positive correlation we found between Ex-K₂O and clay/T-C for Andosols (r=0.405, p < 0.01, 2012) suggests that a high level of T-C could weaken K₂O adsorption on clay mineral sites; hence, low clay/T-C values can partially explain the relatively high TF values of Andosols. Moritsuka (2009) pointed out that Andosols, in which negatively charged sites are derived from allophane and humus, are susceptible to K₂O losses compared with lowland soils. Although we did not determine the clay mineral composition of the Andosols we sampled, the main clay mineral was probably allophane, as described above. Although further investigation is needed, we observed a fluctuation in Ex-K₂O content on an annual basis both in gray lowland soils and in Andosols. In order to decrease radiocesium concentrations in brown rice, the target level of Ex-K in soils in which rice is cultivated has been set at 200 mg kg-¹, which is equivalent to about 250 mg kg-¹ Ex-K₂O (Kato et al. 2015). For instances in which excessive radiocesium concentrations are still detected in brown rice, the criteria for soil Ex-K₂O content will require further revision.

A relatively large amount of NH₄-N is expected to be released from soils with high available N concentrations over an extended period of time under anaerobic conditions. Tensho, Yeh, and Mitsui (1961) showed that the radiocesium uptake by paddy rice is enhanced by NH₄-N under anaerobic conditions. However, the effect of available N on TF values was not observed in the present study.

4.2. Differences in changes in TF values on a yearly basis between Andosols and gray lowland soils

In 2011, the GM values of TF for both Andosols and gray lowland soils were one order of magnitude greater than those reported before the accident. Uchida and Tagami (2007) reported a GM value of 0.0033, while Komamura et al. (2005) reported a GM value of 0.0047 with 95％ CI [0.00057, 0.033]; the values of the latter from 1990 to 2000 were calculated by Tsukada, Yamaguchi, and Takahashi (2011). For gray lowland soils, except for those samples with TF values greater than 0.1 due to low levels of Ex-K₂O and clay, TF values were relatively low throughout the study. This resulted in a 2012 GM value below 0.01, which is equivalent to that reported from before the accident as described above. The drastic decrease in the GM value of Andosols in 2012 suggests that radiocesium may have migrated from the humus surface to clay minerals such as allophane. The samples with Ex-K₂O contents of less than 200 mg kg⁻¹ and with low clay/T-C values showed the most substantial decreases. Radiocesium in these samples is potentially available for rice uptake over an extended time period due to the reversible adsorption and fixation characteristics of allophane; this is reflected in the small decrease between 2012 and 2013.

4.3. Validity of Ex-Cs extraction with a 1 mol L⁻¹ ammonium acetate solution in Andosols

The effect of Cs availability on rice plants has recently been investigated in combination with K availability using soils from Fukushima Prefecture. In the present study, Ex-K₂O content was most highly correlated with TF; however, the r-value was only −0.426 in 2012. To improve the predictive power of TF, multiple regression analysis was carried out. However, it failed to add any explanatory variables, such as clay, coarse sand, or Ex-Cs at the 5 ％ level. In contrast, a model with two explanatory variables, Ex-¹³⁷Cs and Ex-K₂O, explained the variations in ¹³⁷Cs concentrations in rice in Fukushima Prefecture very well (Yagasaki et al. 2019a). The results by Kondo et al. (2015) imply that the Ex-¹³⁷Cs (Bq kg⁻¹)/Ex-K₂O (mmol kg⁻¹) can potentially serve as a soil index to estimate Cs accumulation in rice plants by using four lowland soils in Fukushima Prefecture. In both reports, Cs accumulation in rice plants increased with increasing Ex-¹³⁷Cs concentrations. As shown in Figure 6, the value of Ex-Cs/Ex-K₂O was not significantly correlated with TF for either Andosols or gray lowland soils. The difference in the results between the present study and the two previous reports from Fukushima Prefecture may be partly due to the following reasons: 1) the Ex-Cs concentrations in the present study were lower than those of the report (median Ex-Cs [¹³⁴Cs+¹³⁷Cs] in 2012 was 37.9 Bq kg⁻¹ in the present study as shown in Table 2, and median Ex-Cs (¹³⁷Cs) was 28–80 Bq kg⁻¹ from 2012 to 2015 (Yagasaki et al. 2019b); and, 2) the Ex-Cs/Ex-K₂O values obtained in the present study only reached 20, which is substantially smaller than those reported by Kondo et al. (2015), which up to 250. However, Ex-Cs was significantly correlated with Ex-K₂O in Andosols, but not in gray lowland soils. This simply indicates that Ex-K₂O explains the variability in TF values of Andosols in relation to Ex-Cs; hence, extraction of Ex-Cs with a 1 mol L⁻¹ ammonium acetate solution may not be an appropriate method for explaining the variability of radiocesium TF in Andosols.

Figure 6. Relationship between Ex-Cs/Ex-K₂O in the soil after the harvest and TF in 2012.

Sugiyama and Ae (2000) found that upland rice could take up K more readily than hot nitric acid-extractable K, which is regarded as a non-exchangeable form, from an Andosol. Further investigation is needed to determine the generality of the ammonium acetate extraction method for Ex-Cs and Ex-K₂O in terms of soil groups. A more appropriate extraction method for Andosols will be helpful in implementing feasible agricultural practices to further reduce the risk of high radiocesium concentrations over the long-term.

5. Conclusions

The TF values tended to be higher in Andosols than in gray lowland soils, leading to higher radiocesium concentrations in brown rice grown in Andosols. The present study provides new insight into the important role of T-C in radiocesium bioavailability in terms of K₂O mobility. The positive correlation between Ex-K₂O and clay/T-C was observed in Andosols but not in gray lowland soils; hence, the low clay/T-C values can partially explain the relatively high TF values of Andosols. For instances in which excessive radiocesium concentrations are still detected in brown rice, the criteria for soil Ex-K₂O content will require further revision. The Andosols with Ex-K₂O contents of less than 200 mg kg⁻¹ and with low clay/T-C values are those most potentially vulnerable to radiocesium transfer to rice for a long time. Extraction of Ex-Cs with a 1 mol L⁻¹ ammonium acetate solution may not be appropriate for explaining the variability in radiocesium TF in Andosols. A more appropriate extraction method for Andosols will be helpful in implementing feasible agricultural practices which will further reduce the long-term risks of radiocesium contamination in brown rice.

Acknowledgments

We thank the farmers and staff members of Tochigi Prefectural Agriculture Promotion Offices for sampling and the providing useful information. We also would like to thank Professor Seiju Ishikawa, Hosei University, for his useful advice in improving this article over the long-term.

Disclosure statement

No potential conflict of interest was reported by the authors.