Spatial autocorrelation analysis of monitoring data of heavy metals in rice in China

Food Control 89 (2018) 32e37

Contents lists available at ScienceDirect

Food Control journal homepage: www.elsevier.com/locate/foodcont

Spatial autocorrelation analysis of monitoring data of heavy metals in rice in China Gexin Xiao a, Yingli Hu b, Ning Li a, Dajin Yang a, * a b

China National Center for Food Safety Risk Assessment, Beijing, 100022, China School of Agricultural Economics and Rural Development, Renmin University of China, Beijing, 100872, China

a r t i c l e i n f o

a b s t r a c t

Article history: Received 26 October 2017 Received in revised form 29 January 2018 Accepted 30 January 2018

The aim of this study was to analyze the spatial distribution of heavy metal monitoring data in paddy rice to provide a scientific basis for food safety risk assessment and suggestions for possible risk management. In this study, the spatial distribution data of heavy metals (cadmium, lead, total arsenic, total chromium, and total mercury) in rice collected in the main production provinces of China were analyzed by multidimensional visualization and aggregation analysis. The spatial correlation of elemental contamination was also identified. Monitoring data of cadmium, lead, and arsenic content in different varieties of rice was compared with their mean confidence intervals. Results showed that cadmium content in rice was higher than the limit value in some areas of Hunan, Sichuan, Guangxi and Anhui Provinces in China. With respect to other heavy metals, a small area of Sichuan Province experienced lead levels in rice higher than the limit value. Also, the arsenic level in rice was higher than the limit value in Jiangxi Province, a northern area of Liaoning Province and most parts of Guangzhou and its surrounding areas. In contrast, chromium was only detected at excessive levels in southern Sichuan Province. In addition, a small part of the eastern Sichuan Province was found to have excessive levels of arsenic. Moran's I index of cadmium, arsenic, chromium, lead, and mercury in rice was 0.50, 0.55, 0.21, 0.09, and 0.05, respectively, which revealed a spatial autocorrelation. Overall, there was moderate aggregation of cadmium and arsenic in the monitoring areas, while lead, chromium and mercury showed low aggregation. Geographically for the provinces, the high aggregation of cadmium in rice was evident in Hunan and Jiangxi Provinces and Guangdong border areas. The arsenic in Jiangxi Province and border areas of Jiangxi and Guangdong Provinces also showed high level of aggregation. Meanwhile, the parameter testing of the samples showed that the concentration of cadmium and lead were significantly higher in late Indica rice compared to early Indica rice, while the arsenic and chromium showed the opposite effect. In view of the high levels of certain heavy metals in rice in some provinces, more refined dietary intake assessments of rice as consumed are necessary to determine if populations are exposed to levels that exceed the health-based guidance levels. © 2018 Elsevier Ltd. All rights reserved.

Keywords: Rice Elemental contamination Cadmium Lead Mercury Arsenic Chromium Spatial autocorrelation analysis Geographic information system China

1. Introduction Rice is the main food of more than half of the world's population, especially in developing countries in Asia, for which rice provides more than 70% of the daily caloric intake from food (Qian et al., 2010). Therefore, the quality and safety of rice are closely related to health and the quality of life. Pollution and natural contamination of rice with heavy metals or other potential toxins are important factors that may threaten health and safety, i.e.,

* Corresponding author. No.37, Guangqu Road, Chaoyang District, Beijing, China. E-mail address: [email protected] (D. Yang). https://doi.org/10.1016/j.foodcont.2018.01.032 0956-7135/© 2018 Elsevier Ltd. All rights reserved.

whether or not the rice can be consumed without risk. Therefore, strict monitoring and control of potentially toxic elements in rice are very important. Over the past three decades, rice cultivation in China has been expanding, reaching 303,117 million hectares by 2013 (Dong et al., 2015; Zhong, 2015). However, assessing heavy metal contamination of rice is complicated, because the distribution patterns of contamination of the elements in rice are different depending on both the kinds of rice and the region in which it is grown. The heavy metal content in the soil and irrigation water, pollution from surrounding industry and agriculture practices can significantly affect the content of heavy metals in rice (Duan, Gao, Jiang, & Wu, 2005;

G. Xiao et al. / Food Control 89 (2018) 32e37

Zhao et al., 2009, 2010). The absorption characteristics of elements in rice also vary greatly due to different elements and different varieties of rice. Under certain conditions when there is a mixture of pollutants, the absorption of heavy metals in rice will increase (Wang & Liang, 2000). Some of these challenges have led to difficulties in the formulation of rice pollution management policies, which, in turn, makes it difficult for the relevant regulatory authorities to develop a localized management and monitoring policies. Heavy metals enter the body in a variety of ways, with the primary routes being inhalation (breathing), dermal contact, and diet intake. Relative to respiratory and dermal contact, dietary intake is considered the major source of exposure due to large amount of food consumed, dietary types and content of heavy metals in food, and other factors (Crasnuck & Scholz, 2005). Heavy metal pollution in the environment has a strong correlation with the presence of these harmful elements in rice and has become the focus of ecological and health concerns. The intent of this research work was to use monitoring data of heavy metals in rice from the main production areas of China to analyze the contents of cadmium, lead, arsenic, chromium, and mercury in paddy rice both by location (i.e. space) and by species. Furthermore, we try to provide clues for the sources and trends of rice contamination and provide a scientific basis for risk assessment and the formulation of heavy metal management and monitoring policies.

33

space; xi represents the attribute value in the i-th region, xj represents the attribute value in the j-th region, x represents the average value of the attribute value of the studied region; xij represents the spatial weight matrix, and i is not equal to j. The Z-score of Moran's I is:

I  EðIÞ Z ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi VarðIÞ

(2)

where E(I) represents the expectations of Moran's I, Var(I) represents the variance of Moran's I. When jZ j> 1.96, p < 0.05, we reject the null hypothesis, and spatial autocorrelation (Getis & Ord, 1992) is present. Moran's I coefficient is [-1, 1]. When the value is greater than 0, it indicates that there is a spatial positive correlation between the study area. Values indicate a stronger spatial autocorrelation with numbers closer to 1. When values less than 0, values approaching 1 indicates that the spatial negative autocorrelation is stronger. Random distribution exists when the value is closer to zero. (2) Localized Moran's I coefficient (also called LISA-local indicator of spatial autocorrelation) and local Getis coefficient (Gi) were used to reflect the specific accumulation area and spatial aggregation of harmful elements in rice. The localized Moran's I coefficient provides the determination of the correlation of each spatial unit. For the i-th region, LISA of Moran's I is defined as follows (Zhang, Luo, Xu, & Ledwith, 2008):

2. Material and methods 2.1. Data source The data contained in this paper were collected from a 2015 monitoring study in China that covered 18 cities in 19 provinces. The main monitoring data included cadmium, lead, arsenic, chromium and mercury of rice and its products. The monitoring results included a total of 2151 samples and a number of rice varieties, including early Indica Rice, late Indica Rice and Japonica Rice. 2.2. Method ArcGIS (geographic information system software; Esri, Redlands, California, USA) was used to produce a thematic map showing areas where potentially harmful contamination occurred (Mitchel, 2005). To effectively monitor the concentration of heavy metals in rice and reduce possible adverse health impacts, it is necessary to understand the geographical distribution of the compounds of interest. This study uses the following methods: (1) Global spatial autocorrelation analysis mainly uses Moran's I (Griffith, 1987) coefficients to reflect the degree of spatial clustering of attribute variables in the whole study area. The application software OpenGeoDa (open source; geodacenter.github.io) is used for cluster analysis. The global spatial autocorrelation analysis examines whether a given region is a cluster region. Moran's I coefficient is used to reflect the degree of clustering. Moran's I is expressed as (Harry & Prucha, 2001):

n I¼

n P n P

  wij ðxi  xÞ xj  x

i¼1 j¼1 n P n P i¼1 j¼1

! wij

n P

(1) ðxi  xÞ

2

Ii ¼

n   xi  x X wij xj  x S2 j¼1

(3)

P P where isj, and s2 ¼ 1n ni¼1 ðxi  xÞ2 ; x ¼ 1n ni¼1 xi . The LISA statistic of Moran's I was tested using the Z-score test:

Ii  EðIi Þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z¼p VarðIi Þ

(4)

The LISA coefficient is used to determine the presence or absence of spatial clustering in the heavy metal elements. A LISA coefficient >0 indicates the existence of a spatial positive correlation between the local space unit and the adjacent spatial unit, which is expressed as “high-high” or “low-low”; a LISA coefficient <0 indicates negative correlation between the performance of the “low-high” or “high-low” aggregation. The Getis coefficient is used to detect the “hot” or “cold” areas of the spatial distribution of harmful elements. The local Getis coefficient is defined as: n P

G*i

j¼1

wij xj  w*i x

¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   ns1i  w*2 s ðn  1Þ i

(5)

P P where w*i ¼ nj¼1 wij ; s1i ¼ nj¼1 w2ij . * When Gi > 0, the aggregation area is a hot spot area with high value (a hot zone with respect to one of the classes of harmful elements). If G*i < 0, the aggregation area is a low-value area; i.e., a socalled “cold region” with respect to harmful element pollution. Values close to or equal to zero indicate no accumulation in a given area.

i¼1

where n represents the number of regions of the study object

(3) The distribution of detection values for different heavy metal elements in rice was estimated. Heavy metal content of many

34

G. Xiao et al. / Food Control 89 (2018) 32e37

samples are below LOD (limit of detection), and therefore can not be expressed in exact figures, these samples of elemental contamination detection will appear to be left-censored data in this situation. These unquantifiable data need to be considered during distribution fitting. In this paper, we use the method of maximum likelihood to estimate distribution parameters. In the log likelihood function, the term corresponding to the non-censored sample is the probability density function with parameters, and the corresponding term of the censored sample is the parameter distribution function value. The (log) likelihood function is maximized as the target, and the estimated value of the parameter is obtained. The function is expressed as:

ln LðqÞ ¼

X

ln Fðsd ; qÞ þ

X

ln pðsx ; qÞ

(6)

Fig. 2 showed the spatial distribution of cadmium, lead, arsenic, chromium, and mercury. The cadmium level exceeded national limited value of 0.2 mg/kg in most parts of Hunan Province, eastern Sichuan Province, and in some areas of Guangxi Province and Anhui Province. Lead was detected in most of the monitoring areas, but the detected values were low, with only a small part of Sichuan Province having lead that exceeded the prescribed value limit. The distribution of arsenic was more concentrated in Jiangxi Province; most of Guangzhou and its surrounding region also had high arsenic levels, as did northern Liaoning Province. Chromium content was low in each monitoring area with the exception of Sichuan Province. The last picture in Fig. 2 showed that the mercury in central and eastern China was within the standard, and was exceeded in only a small part of eastern Sichuan Province. 3.2. Global spatial autocorrelation analysis results

s2g

s2F

where q is the distribution parameter, F is the distribution function, p is the probability density function, F is a set of censored samples, g is a non-censored sample set, Sd is the sample detection limit, and Sx is the sample detection value. 3. Results 3.1. Spatial distribution of cadmium, lead, arsenic, chromium, and mercury contamination

As seen in Fig. 3, the Moran's I indexes of cadmium, lead, arsenic, chromium, and mercury in rice were 0.50, 0.10, 0.55, 0.21, and 0.05, respectively. The corresponding p values were <0.0001, 0.00014, <0.0001, <0.0001, and 0.02380. It can be seen from the figure that significant spatial autocorrelation of the five elements occurred and that the autocorrelation of the three heavy metals (cadmium, arsenic, and chromium) was particularly significant. These details were useful to determine whether the need existed to perform ongoing local autocorrelation analyses. 3.3. Local spatial autocorrelation In the first picture of Fig. 4, the majority of Hunan Province and Jiangxi Province and its critical area revealed “high” (local “highhigh” area) local spatial autocorrelation, the region of cadmium content in rice showed high aggregation; especially in Jilin Province and Heilongjiang Province. In the region of “low-low” (low local low), the cadmium content of rice in this region was low and with agglomeration. A small range of low-value clustering existed in the central provincial areas, including Guizhou, Shandong, Jiangsu Provinces and the southern parts of Fujian Province. The second picture in Fig. 4 showed that in most areas of Jiangxi Province and critical areas of Guangdong Province, the arsenic content was high and with aggregation, but most areas of Fujian Province showed a low accumulation of arsenic; in Jilin Province and Heilongjiang Province, the arsenic content in the paddy rice was low and aggregated. The third picture in Fig. 4 showed that the low-value accumulation of the chromium content appeared in the majority of Fujian Province, Hunan Province, Guangdong Province and Anhui Province. In addition, a small range of highly contaminated areas appeared in the belt of Sichuan and Guizhou Provinces.

1.0

3.4. Parameter testing

0.0

0.5

Content(mg/kg)

1.5

2.0

The boxplot showed the distribution of the heavy metal elements in rice. As seen in Fig. 1, the distribution of cadmium in rice was positively skewed, which gathered between 0 and 0.15 mg/kg, and about 20% exceeded the national limit (0.2 mg/kg). The distribution of lead in rice was positively skewed, 50% of the data gathered between 0 and 0.1 mg/kg, the maximum was close to the national limit of 0.2 mg/kg; the chromium distribution in rice was also positively skewed, which ranged from 0 to 0.6 mg/kg, and gathered from 0 to 0.2 mg/kg, below the national limit of 1 mg/kg; mercury distribution in rice was highly aggregated in the vicinity of 0.01 mg/kg, below the national limit 0.02 mg/kg; the distribution of arsenic in rice was close to normal distribution, the range of the content was 0e0.35 mg/kg, generally aggregated in 0.1e0.2 mg/kg.

As

Cd

Cr

Hg

Pb

Fig. 1. Box plot of content of heavy metals (cadmium, lead, arsenic, chromium, and mercury) in rice.

Because different rice species have different rates of absorption of heavy metal elements, this study tested for differences in elemental contamination of rice species. The first step in performing these tests was to fit the distribution for the detected harmful elements in rice. Because the detection value for each element showed obvious characteristics (i.e., “a small amount of detection value was very high, a large number of detection value was very low”), logarithmic normal distribution is appropriate for distribution fitting, as logarithmic normal distribution tend to be obvious negatively skewed. Higher detection values were the focus when the values were analyzed. Higher detection values were close to the mean sample value due to the negatively-skewed shape of the distribution. The

G. Xiao et al. / Food Control 89 (2018) 32e37

35

Fig. 2. Spatial distribution of heavy metals (cadmium, lead, arsenic, chromium, and mercury) in rice.

purpose of the parameter test was to test the significance of the mean difference. Under the assumption of lognormal distribution, this significance test examined the degree of significance in the difference of the logarithmic mean of the distribution. When choosing varieties of rice for comparison, those varieties with sample size less than 30 were removed during distribution, and the remaining two rice varieties were middle and late Indica Rice and early Indica Rice. Additionally, since mercury content was generally low, no comparisons had been drawn for any of the samples. Cadmium and chromium levels were significantly higher in late Indica Rice, and arsenic and chromium levels were significantly higher in early Indica rice (Table .1.). Given this, the content of harmful elements in different rice varieties was significantly different across varieties.

4. Discussion In this paper, data related to rice containing heavy metals in the main paddy growing areas in mainland China in 2015 were analyzed from the aspects of geographical scope, severity, type of elemental contamination, and correlations with rice species. Spatial statistical analysis revealed that cadmium and arsenic

were in excess in the majority of paddy rice production areas. The finding that cadmium content was mostly high in Hunan Province and Jiangxi Province may be related to the rich mineral resources of Human, a township famous for non-ferrous metal mining and the large amount of waste that was not adequately disposed of (Li, Liu, Lu, Ouyang, & Chen, 2006). Additionally, Hunan exhibited high levels of waste water, industrial residue and waste gas, significantly increasing the presence of heavy metals in the environment and potentially causing the contamination of rice. Arsenic content was elevated in Jiangxi Province and Guangdong Province, but Fujian Province showed a low arsenic content aggregation. The high arsenic content in paddy fields in Jiangxi and Guangdong was likely to be related to the higher levels of arsenic in the soil in the region. Due to the abundant water network in the area, arsenic in the soil easily flowed into rivers and was transported to rice through irrigation (Chen, Zhang, Zeng, & Liu, 2011; Li, Wang, Zhou, Ge, & Hu, 2015). The content of arsenic in the soil of Fujian Province was low, which was possibly explained by the river running through small parts of the region, reduced arsenic exposure (Guo et al., 2011). Therefore, the arsenic content of rice in Fujian was low. Inorganic arsenic was much more toxic than organic arsenic, which is the form prevalent in fish. Consequently,

36

G. Xiao et al. / Food Control 89 (2018) 32e37

Fig. 3. Moran's I of cadmium, lead, arsenic, chromium, and mercury.

Fig. 4. Local autocorrelation of cadmium, arsenic and mercury.

further studies need to differentiate between these two arsenic forms. Also arsenic in rice is concentrated in the bran so that polished white rice has less arsenic than paddy rice. Through the distribution fitting and the mean parameter test, significant differences were found in the content of elemental pollution between middle and late Indica Rice and early Indica Rice. Monitoring for middle and late Indica Rice should focus on cadmium and lead contamination; for early Indica Rice, its monitoring should be related to arsenic and chromium contamination.

Knowledge of the correlation between heavy metal varieties in the soil and rice varieties allowed suitable rice varieties to be planted and accompanying reduction of heavy metal content. The policy moving forward may combine the current prevention and treatment measures together, and increase monitoring efforts for cadmium and arsenic levels in rice on a national scale. In particular, Hunan and Jiangxi provinces may be the focus of cadmium monitoring, with the Jiangxi, Guangdong area as a focus of arsenic monitoring. The reasons for the emergence of

G. Xiao et al. / Food Control 89 (2018) 32e37

37

Table 1 Compilation of harmful elements in early Indica Rice, middle and late Indica Rice. Elements

rice

cadmium

middle early middle early middle early middle early

lead total arsenic total chromium

and late and late and late and late

Log mean

SD

90%Lower confidence interval

90%Upper confidence interval

2.33 2.75 5.17 4.42 1.88 1.07 5.41 4.87

0.08 0.06 0.18 0.13 0.03 0.05 0.14 0.3

2.46 2.85 5.46 4.64 1.93 1.15 5.64 5.37

2.2 2.66 4.88 4.21 1.83 0.99 5.18 4.37

contamination in the aforementioned areas may be explored for possible soil remediation or other environmental management intervention. Simultaneously, a corresponding epidemiological investigation should be considered to investigate the possible correlation of contamination of heavy metals in paddy fields in key areas, and health effects of the population, especially in the elderly, infants and children. These results should be considered when conducting dietary studies and should be augmented in some cases with biomarker testing. In populations consuming rice with elevated levels of heavy metals, any real health concern can only be assessed in the context of the total diet. In a broader sense, the future course of action should be to strengthen agriculture, water resources, land resources along with food protection and health care. Cooperation between departments in provinces responsible for remediating heavy metal pollution and pollution in general should also be strongly encouraged. Acknowledgements The present study was supported by a scholarship from the China Food Safety Talent Competency Development Initiative: CFSA 523 Program. References Chen, J. J., Zhang, H. H., Zeng, X. D., & Liu, J. M. (2011). Spatial variation and environmental indications of soil arsenic in Guangdong province. Ecology and Environment, 20(5), 956e961. Crasnuck, D., & Scholz, R. W. (2005). Risk perception of heavy metal soil contamination by High-Exposed and Low-Exposed inhabitants: The role of knowledge and emotional concerns. Risk Analysis, 25(3), 611e622. Dong, J. W., Xiao, X. M., Kou, W. L., Qin, Y. W., Zhang, G. L., Li, L., et al. (2015). Tracking the dynamics of paddy rice planting area in 1986e2010 through time

series Landsat images and phenology-based algorithms. Remote Sensing of Environment, 160, 99e113. Duan, F. Z., Gao, J. X., Jiang, H. E., & Wu, Y. H. (2005). Impact of irrigation water quality on heavy metals' concentrations in surface soil of paddy field. Journal of Agro-environmental Science, 24(3), 450e455. Getis, A., & Ord, J. K. (1992). The analysis of spatial association by use of distance statistics. Geographical Analysis, 24(3), 189e206. Griffith, D. A. (1987). Spatial autocorrelation: A primer. Association of American Geographers, 19(2), 308e316. Guo, Y. L., Wang, G., Luo, D., Ge, H. L., Wang, G. F., Chen, J. H., et al. (2011). Evaluation on the environmental quality about lead, cadmium, arsenic, chromium, mercury, copper and fluorine in the soils of Tieguanyin tea plantation in Fujian Province. Chinese Journal of Eco-agriculture, 19(3), 676e681. Harry, H. H., & Prucha, I. R. (2001). On the asymptotic distribution of the Moran I, test statistic with applications. Journal of Econometrics, 104(2), 219e257. Li, H., Liu, Y., Lu, H. W., Ouyang, S., & Chen, J. M. (2006). On the effects influencing the growth of the two rice varieties due to Cd contaminates soil passivation in Hunan. Journal of Safety and Environment, 16(6), 298e302. Li, Y. H., Wang, M. L., Zhou, X. L., Ge, G., & Hu, C. H. (2015). Sources and physicochemical properties of arsenic in Poyang Lake. Journal of Nanchang University (Natural Science), 39(3), 280e285. Mitchel, A. (2005). The ESRI Guide to GIS analysis, Volume 2: Spartial measurements and statistics. U.S.A.:Esri Guide to Gis Analysis. Qian, Y. Z., Chen, C., Zhang, Q., Li, Y., Chen, Z. J., & Li, M. (2010). Concentrations of cadmium, lead, mercury and arsenic in Chinese market milled rice and associated population health risk. Food Control, 21(12), 1757e1763. Wang, X., & Liang, R. (2000). Interaction and ecological effect of combined pollution for heavy metals on soil-rice paddy system. Chinese Journal of Ecology, 19(4), 38e42. Zhang, C. S., Luo, L., Xu, W. L., & Ledwith, V. (2008). Use of local Moran's I and GIS to identify pollution hotspots of Pb in urban soils of Galway, Ireland. The Science of the Total Environment, 398(1e3), 212e221. Zhao, K., Liu, X., Xu, J., & Selim, H. M. (2010). Heavy metal contaminations in a soilrice system: Identification of spatial dependence in relation to soil properties of paddy fields. Journal of Hazardous Materials, 181(1e3), 778e787. Zhao, K. L., Zhang, W. W., Zhou, L., Liu, X. M., Xu, J. M., & Huang, P. N. (2009). Modeling transfer of heavy metals in soil-rice system and their risk assessment in paddy fields. Environmental Earth Sciences, 59(3), 519e527. Zhong, D. X. (2015). Planting area and yield of grain and rice in national and provincial (city and district) in China. China Rice, 16(6), 73e75.