Total selenium content of agricultural soils in Japan

Abstract

To evaluate the selenium (Se) level in agricultural soils in Japan and to investigate its determining factors, 180 soil samples were collected from the surface layer of paddy or upland fields in Japan and their total Se contents were determined. Finely ground soil (50 mg) was wet-digested with HNO₃ and HClO₄ solution and the released Se was reduced to Se(IV). The concentration of Se(IV) was then determined by high-performance liquid chromatography with a fluorescence detector after treatment with 2,3-diaminonaphthalene and extraction with cyclohexane. The total Se content ranged from 0.05 to 2.80 mg kg⁻¹ with geometric and arithmetic means of 0.43 and 0.51 mg kg⁻¹, respectively. The overall data showed a log-normal distribution. In terms of soil type, volcanic soils and peat soils had relatively high Se content and regosols and gray lowland soils had relatively low Se content. In terms of land use, upland soils had significantly higher Se content than paddy soils. Among regions, soils in the Kanto, Tohoku, Hokkaido and Kyushu regions had relatively high content. The total Se content had a significant positive correlation with the organic carbon content (P < 0.01) and the equation for the estimation of total Se content with organic carbon suggested that on average approximately 48% (0.24 mg kg⁻¹) of the total Se was in inorganic forms and approximately 52% (0.25 mg kg⁻¹) was in organic forms. Soil pH, on the contrary, did not show a significant relationship with the total Se content. In conclusion, the organic matter content, in combination with volcanic materials, was the main determining factor of the total Se content of agricultural soils in Japan.

Key words:

• 2,3-diaminonaphthalene
• agricultural soils
• Japan
• organic matter content
• selenium (Se)

Introduction

Selenium (Se) is an element belonging to group 16 with the atomic number of 34 and an atomic weight of 78.96. It is widely distributed in the environment (Inhat 1989; Jacobs 1989) and is one of the essential elements for animals and plays a major role in the activity of the glutathione peroxidase enzyme, of which it is a component (Rotruck et al. 1973). Several endemic diseases are caused by Se deficiency, for example, Keshan disease observed as cardiomyopathy in humans and white muscle disease in the livestock of northeastern China is closely related to a low level of Se in food grains and drinking water (Yang et al. 1983). In this region the total Se content in the soil is extremely low (as low as 0.05 mg kg⁻¹) (Kang 1989; Yang et al. 1983). In contrast, Se can be harmful to animals when the total content in forage crops is high, for example, chronic blind staggers and alkali disease are observed in many seleniferous areas around the world (Miller et al. 1991; Yang et al. 1983). In this case, again, the Se content in plants is considered to reflect the Se content of the soil where the plants grow. In this context, information on the Se content in soil, especially in agricultural soil, is vital to prevent Se deficiency or toxicity for humans and livestock.

The Se content in soil can be quite variable, ranging from <0.1 to as high as 8000 mg kg⁻¹ (Berrow and Ure 1989; Lakin 1973). Ure and Berrow (1982) reported a Se concentration range of 0.03–2.0 mg kg⁻¹ and a mean of 0.40 mg kg⁻¹ for 1623 soils throughout the world. Bowen (1979) reported a median of 0.40 mg kg⁻¹ and a range of 0.01–12 mg kg⁻¹. In Japan, several studies have examined the total Se content in soil and the arithmetic mean of the total Se content has been reported to be 0.78 mg kg⁻¹ for 10 soils in Hyogo Prefecture (Tsuge and Terada 1949), 1.11 mg kg⁻¹ for 95 grassland soils (Asakawa et al. 1977), 0.47 mg kg⁻¹ for 14 soils in Kyoto Prefecture with a variety of geological origins (Yamada et al. 1987), 0.46 mg kg⁻¹ for 44 paddy soils and 0.89 mg kg⁻¹ for 25 grassland soils (Kang et al. 1990), 0.30 mg kg⁻¹ for 72 agricultural soils in Kyoto Prefecture (Yamada et al. 1990), 0.62 mg kg⁻¹ for 247 soil samples from 20 profiles in the Kanto region (Terashima et al. 2005) and 0.54 mg kg⁻¹ for 61 upland soils and 0.45 mg kg⁻¹ for 50 paddy soils (Uchida et al. 2007a,b). In these reports, however, the soil samples were generally limited in number, region or soil type and comprehensive information on the total Se content in soils in Japan has not been obtained. Thus, the objectives of this research were: (1) to evaluate the total Se content of agricultural soils collected in sufficient number from all over Japan, (2) to investigate its determining factors in relation to the physico-chemical properties of the soils.

Materials and methods

Soil samples

One hundred and eighty soil samples were used, which were collected from the surface layer of agricultural fields all over Japan, that is, from 38 prefectures from Hokkaido to Okinawa (Fig. 1). Ninety-six samples were collected from paddy fields and eighty-four samples were collected from upland fields. The number of samples corresponded to approximately one per 270 and 250 km² for paddy and upland fields, respectively. The soils investigated were classified into 15 soil types (Cultivated Soil Classification Committee 1995). The numbers of samples for each soil type and land use are listed in Table 1.

Figure 1 Location of the sampling sites. •, volcanic soils; ○, non-volcanic soils.

For chemical analysis, the samples were air-dried, disaggregated using a porcelain pestle and mortar and sieved to <2 mm before analysis. The general physico-chemical properties of the soils are shown in Table 2. The means of the soil pH, total C, total N, cation exchange capacity (CEC), sand, silt and clay content were 5.75, 31.7 g kg⁻¹, 2.51 g kg⁻¹, 20.1 cmolc kg⁻¹, 47.7%, 28.3% and 24.0%, respectively. As the average value of the total N, 2.51 g kg⁻¹, was very close to 2.62 g kg⁻¹, which was calculated based on 2272 data of Japanese agricultural soils by the Soil Conservation Project (Oda et al. 1987), the soil samples used in the present study are believed to adequately represent agricultural soils in Japan (Sano et al. 2004).

Analytical methods

The total Se content of the soil was determined using the method proposed by Yamada et al. (1987). A 50 mg finely ground soil sample was decomposed with a mixture of nitric and perchloric acids on a hot plate and heated to strong fumes of perchloric acid (approximately 200°C). A preliminary experiment suggested that decomposition without hydrofluoric acid gave a similar recovery rate to that with hydrofluoric acid: on average 99% for eight different soil samples. To the resultant solution, 2 mL of 6 mol L⁻¹ hydrochloric acid was added, and the solution was heated to reduce all the Se in the solution to the quadrivalent state: Se(IV). The solution was then filled up to 20 mL with deionized water. To 8 mL of the above solution, 0.5 mL of 0.1 mol L⁻¹ ethylenediaminetetraacetic acid–sodium fluoride solution and 1 mL of 20% acetic acid were added, and the pH was adjusted to 1.0 with hydrochloric acid or ammonia. Then 2 mL of 0.2% 2,3-diaminonaphtalene (DAN) in 0.1 mol L⁻¹ HCl was added, and the solution was warmed at 50°C for 30 min to produce a Se–DAN complex (4,5-benzopiazselenol), as shown in Fig. 2. After being cooled to room temperature, the solution was extracted with 4 mL of cyclohexane by shaking for 5 min. A 200 μL aliquot of the cyclohexane extract was analyzed by high-performance liquid chromatography with a fluorescence detector (the wave lengths for excitation and detection were 380 and 525 nm, respectively) on a silica gel column (Cosmosil 5SL-2, Nakarai tesque, Kyoto, Japan, 100 mm × 4.6 mm) using a mixture of cyclohexane and ethyl acetate (95:5, v/v%) as the mobile phase at a flow rate of 1 mL min⁻¹. In the analysis, one standard soil sample was incorporated in each lot. The data suggested that the coefficient of variation of the total Se content of the standard sample was 5.6% (n = 14). All analyses were carried out in duplicate.

Table 1 Number of samples for each soil type and land use

Soil type^†	Relevant soil type^‡	Paddy	Upland	Total
Peat soils	Hemists	5	3	8
Sand-dune Regosols	Psamments	0	1	1
Volcanogenous Regosols	Vitrands	1	4	5
Wet Andosols	Aquands	3	2	5
Non-allophanic Andosols	Udands	4	3	7
Andosols	Udands	5	36	41
Lowland Paddy soils	Aquepts	47	1	48
Gley Lowland soils	Aquents	25	3	28
Gray Lowland soils	Aquents	3	3	6
Brown Lowland soils	Ochrepts	0	14	14
Gray Upland soils	Aquepts	0	4	4
Dark Red soils	Udalfs	0	2	2
Red soils	Udults	0	1	1
Yellow soils	Udults	3	6	9
Brown Forest soils	Ochrepts	0	1	1
Total		96	84	180
^†By classification of cultivated soils in Japan (3rd approximation). ^‡By soil taxonomy (dominant type is indicated).

Table 2 General physico-chemical properties of the soil samples

Total	No.	pH	Total C (g kg^−1)	Total N (g kg^−1)	CEC (cmolc kg^−1)	Sand (%)	Silt (%)	Clay (%)
	180	5.75	31.7	2.51	20.1	47.7	28.3	24.0
Soil type^†
Peat soils	8	5.42	47.7	3.42	25.4	38.0	29.6	32.4
Sand-dune Regosols	1	6.71	6.3	0.53	3.3	95.9	2.2	1.9
Volcanogenous Regosols	5	5.66	27.2	2.15	13.4	72.3	16.1	11.5
Wet Andosols	5	5.59	62.3	4.02	31.4	49.3	27.4	23.3
Non-allophanic Andosols	7	5.59	64.4	4.63	27.7	46.9	29.3	23.8
Andosols	41	5.83	51.4	3.75	27.5	51.2	26.5	22.3
Lowland Paddy soils	48	5.73	22.2	1.97	15.4	49.7	28.6	21.8
Gley Lowland soils	28	5.56	24.1	2.05	18.6	41.8	30.3	27.9
Gray Lowland soils	6	5.71	18.3	1.74	13.1	56.5	24.5	19.1
Brown Lowland soils	14	5.79	18.1	1.63	18.8	47.0	32.6	20.5
Gray Upland soils	4	6.85	13.3	1.17	19.3	25.8	37.2	37.0
Dark Red soils	2	7.29	11.4	1.45	17.5	7.7	40.1	52.2
Red soils	1	3.96	1.8	0.47	17.9	16.1	17.8	66.0
Yellow soils	9	5.86	13.8	1.36	13.1	47.4	27.2	25.4
Brown Forest soils	1	5.52	9.0	0.81	18.2	31.0	34.4	34.5
Land use
Paddy	96	5.67	27.9	2.31	17.6	47.2	29.0	23.8
Upland	84	5.84	36.1	2.74	22.9	48.2	27.5	24.3
^†By classification of cultivated soils in Japan (3rd approximation). CEC, cation exchange capacity.

Figure 2 Reaction of selenite with 2,3-diaminonaphtalene.

Statistical analysis

Descriptive statistics were calculated for the overall dataset of the total Se in soil. A correlation analysis was carried out between the total Se content and the physico-chemical properties of the soils, and a regression analysis was also carried out.

Results and discussion

Descriptive statistics of the total Se content of agricultural soils in Japan

Table 3 shows the descriptive statistics, that is, minimum, maximum, median, arithmetic mean and geometric mean, of the total Se content of agricultural soils in Japan. For the overall 180 samples, the total Se content ranged from 0.05 to 2.80 mg kg⁻¹, with arithmetic and geometric means of 0.51 and 0.43 mg kg⁻¹, respectively. The distribution was considerably skewed with a skewness of 2.68 and the median (0.42 mg kg⁻¹) was within the range of 0.30–0.45 mg kg⁻¹, as shown in Fig. 3. The distribution was regarded as log-normal and accordingly the geometric mean of 0.43 mg kg⁻¹ was regarded as representative of the dataset. Using log-transformed values of the total Se content, the confidence limit (95%) of the mean was calculated to be between 0.40 and 0.47 mg kg⁻¹. Similarly, confidence limit (95%) of the distribution was calculated to be between 0.14 and 1.34 mg kg⁻¹.

Table 3 Minimum, maximum, median, arithmetic mean and geometric mean of the total selenium content of agricultural soils in Japan

Total	No.	Minimum (mg kg^−1)	Maximum (mg kg^−1)	Median (mg kg^−1)	Arithmetic mean (mg kg^−1)	Geometric mean (mg kg^−1)
	180	0.05	2.80	0.42	0.51	0.43
Soil type^†
Peat soils	8	0.31	0.96	0.61	0.61	0.57
Sand-dune Regosols	1	0.05	0.05	0.05	0.05	0.05
Volcanogenous Regosols	5	0.33	0.50	0.35	0.39	0.39
Wet Andosols	5	0.56	2.80	0.72	1.10	0.88
Non-allophanic Andosols	7	0.44	1.60	0.89	0.87	0.80
Andosols	41	0.25	1.32	0.70	0.72	0.66
Lowland Paddy soils	48	0.11	0.59	0.32	0.33	0.31
Gley Lowland soils	28	0.18	1.35	0.38	0.46	0.41
Gray Lowland soils	6	0.14	0.41	0.26	0.27	0.25
Brown Lowland soils	14	0.25	0.66	0.46	0.46	0.44
Gray Upland soils	4	0.31	0.39	0.34	0.35	0.34
Dark Red soils	2	0.34	0.37	0.36	0.36	0.35
Red soils	1	1.04	1.04	1.04	1.04	1.04
Yellow soils	9	0.14	0.50	0.35	0.35	0.33
Brown Forest soils	1	0.53	0.53	0.53	0.53	0.53
Land use
Paddy	96	0.11	2.80	0.36	0.46	0.38
Upland	84	0.05	1.35	0.50	0.57	0.50
^†By classification of cultivated soils in Japan (3rd approximation).

Figure 3 Frequency distribution of the total selenium content of agricultural soils in Japan.

The arithmetic and geometric means of 0.51 and 0.43 mg kg⁻¹ were comparable to reported values for soils in Japan, that is, 0.78 mg kg⁻¹ (Tsuge and Terada 1949), 1.11 mg kg⁻¹ (Asakawa et al. 1977), 0.47 mg kg⁻¹ (Yamada et al. 1987), 0.46 and 0.89 mg kg⁻¹ (Kang et al. 1990), 0.30 mg kg⁻¹ (Yamada et al. 1990), 0.62 mg kg⁻¹ (Terashima et al. 2005), 0.54 and 0.45 mg kg⁻¹ (Uchida et al. 2007a,b) as arithmetic means, and 0.46 and 0.38 mg kg⁻¹ (Uchida et al. 2007a,b) as geometric means. These values were also comparable to the mean value for world soils: 0.40 (0.03–2.0) mg kg⁻¹ (Ure and Berrow 1982) and 0.40 (0.01–12) mg kg⁻¹ (Bowen 1979), as Kang et al. (1990) suggested that the total Se contents of various Japanese soils were comparable to, or even slightly higher, than those reported from other countries. Accordingly, it can be concluded that the total Se content of agricultural soils in Japan was within the ordinary range of the contents observed for soils worldwide, although there may be some exceptions, for example, Mizutani et al. (1996) reported values as high as 148 mg kg⁻¹ for a hot spring area in Toyama prefecture.

Total Se content of agricultural soils with reference to soil type, land use and region

The total Se content was relatively high for Wet Andosols (arithmetic mean: 1.10 mg kg⁻¹), Red soils (1.04 mg kg⁻¹), Non-allophanic Andosols (0.87 mg kg⁻¹), Andosols (0.72 mg kg⁻¹) and Peat soils (0.61 mg kg⁻¹), and relatively low for Sand-dune Regosols (0.05 mg kg⁻¹), Gray Lowland soils (0.27 mg kg⁻¹), Lowland Paddy soils (0.33 mg kg⁻¹), Gray Upland soils (0.35 mg kg⁻¹) and Yellow soils (0.35 mg kg⁻¹) (Table 3). The fact that Andosols generally had higher total Se content suggested that organic matter and/or volcanic materials were related to the accumulation of Se in soil; Asakawa et al. (1977) and Terashima et al. (2005) suggested that volcanic ash soils had relatively higher total Se contents than non-volcanic ash soils.

In relation to land use, the geometric and arithmetic means of the total Se content of paddy and upland soils were 0.38 and 0.50 mg kg⁻¹, and 0.46 and 0.57 mg kg⁻¹, respectively (Table 3). A t-test indicated that the total Se content of the upland soils was significantly higher than that of paddy soils, regardless of whether the data were log-transformed (P < 0.01) or not (P < 0.05). This result was in accordance with the finding by Uchida et al. (2007b) and is related to the fact that waterlogging management can induce a decrease in the total Se content in soils because Se can be translocated downward in the soil profile in paddy fields either as inorganic Se or as Se compounds in fulvic acids (Kang et al. 1991). In addition, it would reflect the fact that the upland soils were composed of larger numbers of volcanic soils, which have a relatively high Se content.

The means of the total Se content were further calculated for each of the nine regions in Japan. The arithmetic means for the Hokkaido, Tohoku, Kanto, Hokuriku-Chubu, Kinki, Chugoku, Shikoku, Kyushu and Okinawa regions were 0.52, 0.62, 0.75, 0.39, 0.29, 0.38, 0.29, 0.57 and 0.44 mg kg⁻¹, respectively. These data suggested distinct regional variation for the total Se content in Japan. Relatively higher values for the Hokkaido, Tohoku, Kanto and Kyushu regions can be ascribed to the relative abundance of Andosols in these regions (Saigusa and Matsuyama 1998).

Relationship between the total Se content and the physico-chemical properties of the soils

A correlation analysis indicated that the total Se content was positively correlated with total C, total N, CEC and clay content (P < 0.01) and negatively correlated with sand content (P < 0.01) (Table 4). No significant relationship was observed between the total Se content and soil pH. These relationships were generally observed both for paddy soils and upland soils. Judging from the fact that CEC is mainly regulated by the contents of both organic matter and clay, and that a higher clay content often induces accumulation of organic matter by the formation of an organo–mineral complex, it would be reasonable to suppose that the organic matter content was the main determining factor of the total Se content of agricultural soils in Japan.

To investigate whether the total Se content was influenced more strongly by organic matter or volcanic materials, a correlation analysis was carried out between the total Se content, total C content and Alo content (acid ammonium oxalate extractable Al) for 145 samples investigated. The correlation coefficient between the total Se content and total C content was 0.56 (P < 0.01) and the coefficient between the total Se content and Alo was 0.54 (P < 0.01); thus, it was not possible to distinguish between the effects of organic matter and volcanic materials. This would be because the total C and Alo were highly correlated, with a correlation coefficient of 0.67 (P < 0.01). It was therefore assumed that organic matter contributes considerably to the accumulation of Se in soil in combination with volcanic materials.

Figure 4 shows the relationship between the total C content and the total Se content for the 180 soil samples. There was an outlier, 2.80 mg kg⁻¹, which had the highest Se content. The outlier was a paddy soil sample of Wet Andosol collected from the Aso district, Kumamoto prefecture. The extraordinary high Se content in relation to the C content may result from the volcanic activity of Mt Aso. A regression analysis using the data without the outlier indicated that the total Se content can be estimated by the equation below:

Figure 4 Relationship between the total selenium content and the total carbon content.

Table 4 Correlation coefficients between the total Se content and the physico-chemical properties of the soils

	No.	pH	Total C	Total N	CEC	Sand	Silt	Clay
Total	180	−0.03	056**	056**	0.58**	−0.20**	0.06	0.26**
Paddy	96	−0.04	0.60**	0.59**	0.55**	−0.17	0.00	0.26**
Upland	84	−0.08	0.57**	0.57**	0.60**	−0.26*	0.18	0.28**
*P < 0.05; **P < 0.01. CEC, cation exchange capacity.

This equation suggests that, on average, 1 g kg⁻¹ of C corresponded to 7.95 × 10⁻³ mg kg⁻¹ of Se, that is, one Se atom was observed for each 8.3 × 10⁵ C atoms in soil organic matter. It also suggests that, based on the mean of the total C content of 31.7 g kg⁻¹, the mean of the total Se content is estimated to be 0.489 mg kg⁻¹, which is reasonably close to the measured arithmetic mean of 0.51 mg kg⁻¹. The equation further suggests that the total Se content is estimated to be 0.237 mg kg⁻¹ if we assume no organic matter in the soil (total C content of zero). This value can be regarded as the mean of the total inorganic Se in Japanese soils. Accordingly the mean of the total organic Se in Japanese soils was calculated to be 0.252 mg kg⁻¹. From this result, it can be concluded that on average approximately 48% of the total Se was in an inorganic form and approximately 52% was in an organic form in the agricultural soils of Japan. Soils with higher organic matter content would have a higher organic Se percentage and those with lower organic matter content would have a lower organic Se percentage. This corresponded with the finding that more than 60% of the total Se was organically bound for soils with 16–150 g kg⁻¹ of carbon (Kang et al. 1993).

Conclusion

The total Se content of agricultural soils in Japan ranged from 0.05 to 2.80 mg kg⁻¹, with geometric and arithmetic means of 0.43 and 0.51 mg kg⁻¹, respectively. The Se content was relatively high for volcanic soils and peat soils among soil types, and for upland soils compared with paddy soils. Geographically, soils in the Kanto, Tohoku, Hokkaido and Kyushu regions had a relatively high content. The total Se content had a significant positive correlation with organic carbon content and a regression equation suggested that on average 0.24 and 0.25 mg kg⁻¹ or 48 and 52% of the total Se content was derived from inorganic and organic Se, respectively. In conclusion, the organic matter content, in combination with volcanic materials, was the main determining factor of the total Se content of agricultural soils in Japan. Further research is needed to understand the mechanism of accumulation of Se in soils, particularly in relation to the dynamics of Se compounds in the soil and in the environment. An evaluation of the available fraction of agricultural soils in Japan is also needed to estimate the risk of the occurrence of Se deficiency and toxicity by animals through the food chain by eating Se-deficient or Se-accumulating plants.

Acknowledgment

The authors wish to thank Dr Shuji Sano, Research Institute of Environment, Agriculture and Fisheries, Osaka Prefecture, for his collaboration in collecting soil samples and for valuable information on the soil samples investigated.