The study area (23°33′-28°20′N, 115°50′-120°40′E) is located in the southeast coast of China with a total land area of about 1.24 × 105 km², a high northwest and low southeast terrain, and a wide area of mountains and hills (Figure 1). The region has a subtropical marine monsoon climate, with an annual average temperature of 17°C–21°C and average precipitation of 1,400–2000 mm. The area of agricultural land in Fujian is about 1.09 × 107 ha, with red soil and yellow soil as the main soil types. The density of the river network is 0.1 km/km2, and the irrigation water is sufficient. Rice is an important food crop in Fujian, and most of the local population also relies on rice as their main food (Kang et al., 2020). Additionally, Fujian is located in the Pacific Rim metallogenic belt, rich in mineral resources, petrochemical industry, machinery industry development scale, and developed transportation network.

Based on the distribution of farmland in the study area, a total of 260 rice samples with a minimum mass of 1 kg were collected in the year 2015 and 2016, and the geographical coordinates of the sampling points were recorded by GPS. Their distribution positions are shown in Figure 1.

The collected rice samples were inactivated (105°C) and dried (75°C) to constant weight. After that, they were shelled to make brown rice, which was ground with an agate mortar and passed through a 100-mesh nylon sieve. 0.5 g of the rice powder was taken and digested by HClO4-HNO3. The obtained digested solution was filtered by a 0.22 μm aqueous membrane, sealed, and then indicated the sample number (Lü et al., 2021). The contents of Cd, Pb, Cr, Ni, Cu, and Zn in each sample were determined by inductively coupled plasma mass spectrometry (ICP-MS), and the contents of Hg and As were determined by atomic fluorescence spectrometry (AFS). In this study, the reagents used were guaranteed reagents (GR), the quality control was carried out by standard materials and blank samples, and the recoveries were between 80% and 110%. The detection limits for Cd, Hg, As, Pb, Cr, Ni, Cu, and Zn were 0.0005, 0.0003, 0.0005, 0.001, 0.002, 0.001, 0.025, and 0.217 mg/kg, respectively.

Semivariogram and Ordinary Kriging (OK) are important tools for exploring the spatial structure characteristics of regionalized variables and implementing numerical simulations (Zhao et al., 2010). Spherical, exponential, Gaussian, and linear models are often used to fit semivariograms, and parameters such as nugget value and sill value are set to ensure optimal linear unbiased estimation of spatial interpolation (Goovaerts, 1997). The semivariogram is calculated as follows: r ( h ) = 1 2 N ( h ) ∑ i = 1 N ( h ) [ Z ( x i ) − Z ( x i + h ) ] 2 (1) where r(h) is the semivariogram at a lag distance h, N(h) is the number of pairs of sample points separated by the h, Z (x_i) and Z (x_i + h) are the values of the Z at the xi and xi + h location, respectively.

The local Moran’s I index could be used to judge the spatial autocorrelation characteristics of the variables, and then to distinguish the spatial association by identifying spatial clusters and outliers (Anselin, 1995). When the value of Moran’s I is more than 0, it means that the data presents a spatial positive correlation, and the greater the value, the more obvious the spatial correlation. When the value of Moran’s I is less than 0, it means that the data presents spatial negative correlation, and the less the value, the greater the spatial difference. When the value of Moran’s I is equal 0, the spatial distribution is random (Nguyen, 2018; Tao and Wu, 2018).Its results include five types: high-high cluster (HH), low-low cluster (LL), high-low outlier (HL), low-high outlier (LH), and not significant (NS), among which “HH” is also called “hotspot”, “LL” called “coldspot” (Li et al., 2014). The local Moran’s I index is calculated as follows: I i = Z i − Z ¯ σ 2 ∑ j = 1 , j ≠ i n [ W i j ( Z j − Z ¯ ) ] (2) where I_i is the local Moran’s I index at the location I, Z_i and Z_j are the values of the variable Z at the i and j (j≠i), Z ¯ is the average value of Z, σ2 is the variance of Z, W_ij is a distance weighting between Z_i and Z_j.

The “National Standards of Food Safety - Limits of Contaminants in Food (GB 2762-2017)” (NHCPRC, 2017) was used as the pollution evaluation standard. The single factor pollution index (SFPI) and the Nemerow comprehensive pollution index (NCPI) were used to evaluate the degree of contamination risk of heavy metals in rice (Zhang S. F et al., 2022), which are calculated as follows: P i = C i / S i (3) P N = ( P a v e 2 + P max 2 ) / 2 (4) where P_i is the SFPI value of heavy metal i, C_i is the content of i, S_i is the standard evaluation value of i, P_N is the NCPI value, P_ave and P_max are the average and the maximum value of the SFPI. In this study, P_i < 1.0, and P_i ≥ 1.0 indicate that the element is safe, and excessive, respectively (Zhang S. F et al., 2022). P_N ≤ 0.7, 0.7 < P_N ≤ 1.0, 1.0 < P_N ≤ 2.0, 2.0 < P_N ≤ 3.0, and P_N > 3.0 indicate that the element is clean, precautionary, slightly polluted, moderately polluted, and heavily polluted, respectively (Xiang et al., 2019).

The exposure evaluation model was applied to assess the potential health risks for the adult and the child from rice consumption (USEPA, 2011). According to whether heavy metals are carcinogens, they could be assessed from two aspects: non-carcinogenic risk and carcinogenic risk. Among them, the exposure was expressed as the average dietary intake (ADI), the non-carcinogenic risks were expressed as the target hazard quotient (THQ) and the hazard index (HI), the carcinogenic risks were expressed as the cancer risk index (CR) and the total cancer risk index (TCR) (He et al., 2019). Their calculation formulas are as follows: A D I = ( C × I R × E F × E D ) / ( B W × A T ) (5) T H Q = A D I / R f D (6) C R = A D I × S F (7) H I = ∑ i = 1 n T H Q i (8) T C R = ∑ i = 1 m C R i (9) Where ADI is the average dietary intake, C is the content of heavy metal, IR is the rice intake rate, BW is the average body weight, EF is the exposure frequency, ED is the exposure duration, AT is the averaging exposure time. RfD is the oral reference dose, with Cd, Hg, As, Pb, Cr, Ni, Cu, and Zn were 0.001, 0.0007, 0.0003, 0.0037, 1.5, 0.02, 0.04, and 0.3 mg/kg/d, respectively (USEPA, 2013). SF is the carcinogenicity slope factor, with Cd, As, Cr, and Ni being 6.1, 1.5, 0.5, and 0.84, respectively (WHO, 2011). When THQ (HI) ≤ 1.0, it means there is no risk; and when THQ (HI) > 1.0, it means there is a non-carcinogenic risk. When CR (TCR) ≤ 1.0 × 10⁻⁶, it means there is no risk; when 1.0 × 10⁻⁶ < CR (TCR) ≤ 1.0 × 10⁻⁴, it means that the risk is acceptable; and when CR (TCR) > 1.0 × 10⁻⁴, it means that there is a serious carcinogenic risk (USEPA, 2011). The meaning and value of each parameter in the model were shown in Table 1.

The descriptive statistical analysis, normality test, and correlation analysis were proceed by software of SPSS 19 and Excel 2019. The semivariogram was calculated by GS + 9.0, and the local Moran’s I index was calculated by GeoDa 1.14. The spatial interpolation and mapping were performed by ArcGIS 10.6, and other plots were made by Origin 9.0 and Python 3.6.

With regard to the Moran ‘I values in different elements, As, Cr, Cu, Ni and Zn are spatially positively correlated, while Cd, Hg and Pb are spatially negatively correlated, among which Cu is the most correlated. Except for Pb (13.46%), about 20%–40% of heavy metals were in spatial clustering areas (i.e., HH and LL, Table 4), which were the main local spatial pattern types in Fujian Province. The percentage of high-low outliers (HL) and low-high outliers (LH) is relatively low. In Table 4, the percentages of heavy metal hotspots (HH) were Cr (19.23%) > As (14.62%) > Cu (10.77%) > Zn (9.62%) > Hg (8.08%) > Ni (4.62%) > Cd (1.54%) ≈ Pb (1.54%). The distribution of hotspots of Cr, As and Cu in rice was more significant in the study area.

The hotspots of As, Cu, and Cr in rice were mainly distributed in Longyan city, southwest of Fujian Province (Figure 4). There were also sporadic Cr hotspots in Putian city, while their coldspots appeared in the northern region (Nanping, Ningde, and Sanming city). The hotspots of Zn and Hg in rice were concentrated in Zhangzhou city, and some hotspots of Zn and Hg appeared in Quanzhou and Nanping city, respectively (Figure 4). The coldspots of Zn and Hg were located in Longyan and Sanming city. In addition, the hotspots of Ni and Cd in rice were mostly in Fuzhou and Zhangzhou city, while coldspots of Cd, Ni, and Pb were observed in Sanming and Putian cities (Figure 4).

The quality and safety of rice in Fujian Province were relatively good, and the degree of heavy metal pollution was low. The results of this study showed that the average contents of the eight heavy metals investigated were lower than the national standards (GB 2762-2017) (NHCPRC, 2017). Compared with the international standards of contaminants in plants/crops, the average contents of metals, such as Cd, As, Pb, Cr, Cu, and Zn, were also relatively low in this study (FAO, 2015; WHO, 1996).

The extent of heavy metal pollution in rice varied in most regions, especially As, Cu and Cd. Compared with the previous results of heavy metals in rice, the contents of rice As and Cd in Fujian Province were lower than the average levels of As (0.125 mg/kg) and Cd (0.116 mg/kg) in China by 14.40% and 34.48%, respectively (Mu et al., 2019). The average contents of As and Cd were also lower than those of the corresponding metals in the Yangtze River Delta region (As: 0.29 mg/kg, Cd: 0.08 mg/kg), while the average content of Cu was about 1.41 times higher than that of Cu in the above region (2.98 mg/kg) (Hu et al., 2019). Meanwhile, the average Cu content in this study area far exceeded the Cu contamination status in rice in Hunan Province (3.69 mg/kg) reported by Du et al. (2018). Compared with our previous long-term monitoring results of rice pollution in the Western Fujian region (Cd: 0.068 mg/kg, As: 0.157 mg/kg, Cu: 6.647 mg/kg), the average content of Cd in rice in Fujian Province was higher by 11.76%, while As and Cu were relatively lower (Lü et al., 2022).

The more correlated heavy metals showed similar spatial characteristics. The results showed that the significant correlations (p < 0.01, Figure 2) among the Cd, Hg, Ni, and Zn contents in rice, whose high contents were mainly located in southeastern Fujian Province (Figure 3), and their hotspots were spatially concentrated in Zhangzhou city (Figure 4). Significant correlations (p < 0.05) also existed between the As, Cu, and Cr contents in rice in this study, especially their high contents and hotspots were more located in the southwestern Fujian Province (Longyan city). Previous studies showed that the similar spatial patterns of heavy metals at a regional scale usually reflect the possible existence of the same pollution sources (Deng et al., 2020; Shi et al., 2020). This implied that Cd, Hg, Ni, and Zn in rice in Fujian Province might be influenced by the same pollution sources, while As, Cu, and Cr were likely to have homologous sources. In contrast, the correlation coefficients were smaller between Pb and the other seven heavy metals in rice (Figure 2). There were significant differences in distributions in their spatial patterns and hotspots as well (Figures 3, 4). Another study also found that the areas with clustered hotspots had a higher likelihood of heavy metal pollution from specific point sources such as industrial emissions and mineral mining (Han et al., 2022).

Heavy metals in rice in Fujian Province were significantly affected by anthropogenic activities. The results showed that the CV values of Cd, Hg, Pb, and Ni in rice belonged to an extensive variability (CV > 1.0, Table 2). This indicated that the distribution of the above elements in rice was uneven across the region and was affected by external disturbances such as human activities (He et al., 2019). In others, the nugget/sill ratio could better represent the spatial variation degree of regionalized factors. Generally, when the ratio is greater than 75%, it means that random variation dominates with the weak spatial association, which is mainly controlled by human factors; when the ratio is less than 25%, structural variation dominates and the influence of natural factors is greater (Gao et al., 2016; Lü et al., 2021). In this study, the nugget/sill ratio of Hg and Pb in rice was greater than 75% (Table 3), showing a stronger role of human activities in the distribution of Hg and Pb. Combined with the spatial characteristics of heavy metals and field investigations, it was found that the contents of Cd, Hg, and Ni in rice were higher in southeastern Fujian Province, where the number of industrial enterprises was high. It was speculated that the heavy metal contents might be closely related to the wastewater and exhaust gas emissions caused by industrial production. This was similar to the findings in Zhejiang and Jiangsu Provinces by Yang et al. (2020). The contents of As, Cu, and Cr in rice were higher in the southwestern Fujian Provinces, which could be related to the excessive application of chemical fertilizers and pesticides in the local agricultural production process. Wan et al. (2020) also showed that the As and Cu contents in rice were significantly affected by the input of agrochemicals in Changsha city, Hunan Province. In this study, the higher CV value and higher nugget/sill ratio of rice Pb indicated the strong influence of anthropogenic activities. The higher content of Pb in the Northwestern region, overlapped greatly with the spatial distribution of the mineral belt in Fujian Province, and there might be more Influenced by mining and smelting activities (Zhang et al., 2018). However, the CV value and nugget/sill ratio of rice Zn were relatively low, and its high-content area (Southeast) was more consistent with the area of high Zn content in soil in Fujian Province reported by Huang et al. (2018). So, Zn was more affected by soil environmental conditions.

The pollution risk of heavy metals in rice in Fujian Province was low. The results showed that the average NCPI value was lower than 0.7, the average SFPI values of eight heavy metals were all lower than 1.0, and the exceeding rates of samples except As and Cu were all lower than 20% (Table 5). Referring to the Agricultural Industry Standard (NY/T 398-2000), it was also found that the residual degree of heavy metals in rice was relatively small, and only As, Cu, and Cd accumulated significantly (SFPI >0.6). The average SFPI values of As and Cu in rice in this study area were closed to those in the southwest Karst area (As: 0.520, Cu: 0.412) (Zhang C. L et al., 2022), but the samples from Fujian Province had a lower excess rate. Compared with the pollution index of heavy metals in rice in the Yangtze River Delta region (As: 0.66, Cu: 0.52, Cd: 0.32), the average SFPI values of As and Cu in rice in Fujian Province were slightly lower, while the average SFPI value of Cd was about 1.18 times that of the above region (Mao et al., 2019). The average SFPI values of As and Cd were also lower than the results of the study in Hunan Province (As: 1.21, Cd: 1.42), but the average SFPI value of Cu was higher than that of Hunan Province by 109.14% (Zhang et al., 2020). In addition, for each city, the comprehensive pollution risk of rice in southern Fujian Province (Longyan: 0.822, Zhangzhou: 1.096) was higher (Figure 5), which was consistent with the investigation results in southwestern Fujian Province by Lin et al. (2021). To sum up, chemical remediation and bioremediation measures should be carried out in time in heavily polluted areas to reduce the bioavailability of heavy metals in rice. For example, adding passivation materials (biochar, quicklime) or microbial inoculants to the soil according to local conditions (Zhao and Wang, 2020; Wang et al., 2021). The agronomic practice was also one of the effective measures to control the enrichment ability of pollutants (like As, Cd). Some studies showed that the voltaic effect could control the remobilization of Cd in paddy soil during drainage (Huang et al., 2021). Controlling the amount of phosphorus fertilizer in farmland and changing the application of silicon fertilizer could also reduce the content of As in rice (Kumarathilaka et al., 2019). Other, for rice safe production, rice could be grown in coldspots areas to reduce the potential risk of heavy metal pollution.

Consumption of polluted rice has caused significantly non-carcinogenic and carcinogenic hazards to the human body. The results showed that the average HI values of the adult and the child in Fujian Province were greater than 1.0, and the average TCR values were far more than 1.0 × 10⁻⁴ (Figure 6). This was consistent with the conclusion of the fifth residents’ dietary survey that there was a comprehensive risk of Cd, As, and Pb in Fujian Province (Wei et al., 2019). Compared with the existing risk studies, the average HI and TCR values of the child in Fujian Province were 1.07 and 10.05 times higher than those in the Yangtze River Delta region (HI: 5.62, TCR: 0.0008), and the values of the adult were 8.59 and 19.61 times the local risk index (HI: 0.414, TCR: 0.0003) (Hu et al., 2017). The average HI values of the child and the adult were 5.05 and 3.87 times higher than those in Lengshuikeng, Jiangxi Province (child: 1.19, adult: 0.92), and the average TCR values were 1.65 and 1.32 times those of the corresponding population (child: 0.0051, adult: 0.0038) (Huang et al., 2020). For the contribution rates of each element, As and Cd in rice were the key factors for high HI values and high TCR values, which was consistent with the previous research reports by He et al. (2019) and Zeng et al. (2015). This study indicated that the health risks of the child were higher than that of the adult in Fujian Province. Similarly, some studies also showed that children tend to suffer more health hazards than adults from environments threatened by heavy metals (Hu et al., 2020; Baruah et al., 2021). This should be attributable to specific behavioral (e.g., arbitrary grasping) and physiological characteristics (e.g., lower body weight, weaker organ toxicity tolerance) of the child (Kumarathilaka et al., 2019; Guo et al., 2021). Therefore, residents (especially the child) should pay attention to reasonably controlling rice daily intake and adjust the dietary structure in time. Another study found that As was concentrated in the husk, bran, and germ of brown rice, the As content in milled rice was less than 30% of that in brown rice (Wu et al., 2019). Milling also reduced the Cd content in rice by about 23.5% (Gu et al., 2020), which may be one of the effective measures to reduce the health risk of rice.