Comprehensive Perspective on Contamination Identification, Source Apportionment, and Ecological Risk Assessment of Heavy Metals in Paddy Soils of a Tropical Island

Abstract

The closed-loop material and energy cycles of islands increase the susceptibility of their internal ecosystem components to heavy metal accumulation and transfer. However, limited research on the island scale hinders our understanding of heavy metal environmental geochemistry in this unique environment. This study focused on assessing a tropical island’s ecological risk by investigating heavy metal contamination and potential sources. The results revealed elevated cadmium and nickel concentrations in 0.44–1.31% of soil samples, particularly in coastal plains and developed areas. Using the absolute principal component score-multiple linear regression (APCS-MLR) model assisted by GIS mapping, we identified three contamination sources: geological factors, agricultural activities, and traffic emissions. Network analysis indicated direct exposure risks of vegetation and soil microorganisms to contaminated soil (0.4611 and 0.7687, respectively), with nickel posing the highest risk, followed by Zn, Cd, Pb, Cu, and Cr with risks transferring across trophic levels. These findings provide crucial insights for mitigating ecological risks associated with heavy metals by controlling priority pollutants and their sources in tropical island environments.

1. Introduction

Soil, a precious natural resource, forms the foundation of sustainable agricultural development. In China, soil resources are notably scarce. Despite covering only 9% of the world’s arable land, cultivated land in China must support approximately 21% of the global population [1]. Therefore, contamination of soil resources severely hampers food security and sustainable economic development. Long-term exposure to polluted soil environments can cause serious damage at all trophic levels, including producers, consumers, and decomposers, ultimately affecting humans [2,3,4].

Effective control and management of heavy metal contamination is crucial for mitigating risks [5]. Identifying potential sources of soil heavy metals and quantitatively assessing their contributions are essential steps in this process, aiding in the prevention of soil contamination and minimizing health risks [6,7,8]. However, the challenge lies in discerning the sources of heavy metals in surface soil due to their natural presence in the Earth’s crust [2]. Various tools, including principal component analysis (PCA) [9], chemical mass balance (CMB) [10], edge analysis (UNMIX) [11,12], positive matrix factorization (PMF) [5,13,14], and absolute principal component score-multiple linear regression (APCS-MLR) [15,16] have been widely employed for heavy metal source apportionment. The most popular receptor models among them are APCS-MLR and PMF, which are used to identify contamination sources and develop sustainable policies to prevent and reduce heavy metal contamination.

Soil contamination poses significant risks to agricultural productivity, ecological sustainability, and human health [17]. Acting as the primary sink for pollutants from both natural and anthropogenic sources [18,19], soil accumulation of trace metals can have detrimental effects on the environment, especially in areas where levels surpass national soil standards and local background values [20]. This issue is particularly acute in paddy soils, where heavy metal accumulation poses severe ecological and health risks globally. Soil organisms and vegetation are directly impacted by heavy metal-rich soils due to their heightened sensitivity [21]. Moreover, heavy metal contamination can directly influence soil microbial communities by impeding their growth and metabolic activity, consequently affecting organic matter decomposition and nutrient cycling. Vegetation plays a crucial role in ecosystems as it absorbs nutrients and pollutants from the soil and transfers energy through the food chain. Furthermore, the risks posed to vegetation can be transmitted to and harm subsequent consumers (e.g., herbivores and carnivores) via the ecological food chain [22]. Carnivores and herbivores are indirectly influenced by heavy metal contamination through bioaccumulation and biomagnification after consuming contaminated prey or vegetation. The process of bioaccumulation and biomagnification increases heavy metal levels as trophic levels increase, which poses potential ecological risks to animals at higher levels, such as carnivores and herbivores. Heavy metals also negatively affect human health through accumulation, transformation, and uptake (oral intake, inhalation, or dermal contact) [23,24,25]. Therefore, assessing the ecological risks to food web components resulting from direct and indirect environmental interactions is crucial for formulating contamination control measures [26].

Hainan Island is an important area for studying the contamination status, sources, and influence of heavy metals on the ecosystem due to its unique geographical environment [27,28]. This tropical province in China comprises 42.5% of the country’s tropical land area, with a per capita land area of approximately 0.47 hectares. Benefiting from favorable conditions, including ample sunlight, heat, and water availability, Hainan Island supports the cultivation of three rice seasons annually, thus serving as a significant rice planting area and a crucial crop breeding base in China. Given its insular nature, Hainan operates within a closed-loop system of material and energy processes, rendering it susceptible to the accumulation and transfer of soil heavy metals [29]. Understanding the contamination levels of heavy metals in paddy soils, identifying their sources, and clarifying the risks posed by soil contamination between ecological components through substances and energy processes is crucial for sustainable development and conservation efforts. However, limited studies on the island scale have hindered our understanding of the environmental geochemical behavior of heavy metals in such unique environments. Therefore, this study aims to: (1) explore the concentration and spatial distribution of heavy metal contamination using two calculation methods derived from a contamination index; (2) identify and apportion potential contamination sources of seven heavy metals in paddy soils using the APCS-MLR model; (3) quantify direct and indirect ecological risks of heavy metal consumption across four trophic levels (producer: vegetation; consumers: herbivores and carnivores; decomposer: soil microorganisms) using a network environment analysis (NEA) framework. Our findings can provide guidance for lowering the ecological risks posed by soil heavy metals through the management of high-risk sources, thereby ensuring the long-term preservation of the distinctive ecological environment of the tropical agricultural ecosystem and human welfare.

2. Materials and Methods

2.1. Study Area

Hainan Island is located northwest of the South China Sea (108°36′ E–111°3′ E, 18°10′ N–20°10′ N). It has a gross area of 354 km², and the region is characterized as a tropical oceanic monsoon with ample light and heat energy [29]. The island’s topography features low-lying areas surrounding a central highland, resulting in a dome-shaped mountainous terrain. Landforms include mountains, hills, plateaus, and plains, forming distinct tiered structures. Favorable climatic conditions have given rise to valuable mineral resources on the island, including bauxite, coastal sand minerals, gem sand minerals, and leached-type rare-earth minerals, shaped over time by the maritime tropical monsoon climate. Hainan Island is an important agricultural production area in China, with frequent population movements and a concentration of agricultural related activities; at the same time, it has a unique natural environment and abundant medical, tourism, health, cultural, and other resources, making it a world-famous longevity island and base for international tourism and health demonstration. Industrial activities in Hainan Island are relatively weak, especially as the introduction of industries with serious pollution issues such as those that are “heavy metal-related” has been restricted, which creates a solid foundation for the development of the tourism and health industries.

2.2. Data Sources and Chemical Analysis

A total of 229 paddy soil samples were collected from the rice-growing regions of Hainan Island (Figure 1). The sampling locations were chosen based on changes in land use type and terrain factors. During sampling, contiguous farmland was selected to ensure the samples were representative. Topsoil samples (0–20 cm deep) were collected at each sampling site using the double diagonal method (with a weight > 1.000 g). The collected soil samples were placed in polyethylene bags and transported to the laboratory. All soil samples were naturally air-dried, and small objects, such as gravel and plant debris, were removed manually [30]. The samples were ground and filtered through 0.15-mm nylon sieves [4]. Subsequently, they were digested with a mixture of HNO₃, HClO₄, and HF (with a ratio of V1:V2:V3 = 8:1:8) in 100 mL Teflon crucibles, which were left standing for 12 h. Then, the digestion solution was placed on an electric heating plate in turn. The initial temperature of the electric hot plate was controlled at 100 °C, increased to 120–130 °C after two hours, increased to 140–160 °C for two hours, and then kept constant at 170 °C until the solution became colorless and transparent or light yellow, with only one drop remaining. Finally, the resulting clear digestion solutions were then directly diluted with a 1% HNO₃ solution and stored at 4 °C until analysis.

Heavy metals, including cadmium (Cd), lead (Pb), chromium (Cr), nickel (Ni), copper (Cu), zinc (Zn), and manganese (Mn) were measured using an ELAN DRC II inductively coupled plasma mass spectrometer (ICP-MS; PerkinElmer Ltd., Waltham, MA, USA) [31]. Fe and P were measured using an OPTIMA 5300DV inductively coupled plasma optical emission spectrometer (ICP-OES) (PerkinElmer, Waltham, MA, USA). The detection limits for the elements are as follows: Mn is 1 ppb (1 × 10⁻⁶ mg kg⁻¹), Pb, Cr, Ni, Cu, and Zn are 0.1 ppb (1 × 10⁻⁷ mg kg⁻¹), and Cd is 1 ppt (1 × 10⁻⁹ mg kg⁻¹). Three blank and three soil standard reference material samples (GBW07410) were used to ensure quality control. A set of parallel samples for every 10 samples was also used in the analysis. The recoveries in soil samples were 96.1–101.3% for soil Cd content, 89.5–109.3% for soil Pb content, 85.8–107.7% for soil Cr content, 87.5–109.4% for soil Ni content, 87.5–110.8% for soil Cu content, 90.8–112.6% for soil Zn content, and 85.8–104.9% for soil Mn content, respectively. The tested substances’ relative standard deviation (RSD) is <15%, which satisfies the USEPA’s requirement [32] of an RSD <30%.

Soil physicochemical properties, such as pH, organic carbon content (OC), and cation exchange capacity (CEC), for each paddy soil sample were sourced from the China soil map-based harmonized world soil database (HWSD; version v1.1, 2009), accessed through the National Tibetan Plateau Data Center [33]. Land-use type (1000 × 1000 m, Figure S1a) and road type data (Figure S1b) were acquired from the Resources and Environmental Sciences Data Center, Chinese Academy of Sciences (RESDC, 2020, http://www.resdc.cn, accessed on 10 March 2024), and OpenStreetMap, respectively. Metal mine data were obtained from China’s Mineral Resources Report for 2021 (Figure S1c). The data for metal-related factories (mining, metallurgy, and manufacturing) (Figure S1d) were obtained from points-of-interest gathered from Gaode Maps for 2023.

2.3. Contamination Level of Heavy Metals in Paddy Soils

The single factor pollution index (SFPI) and Nemerow composite pollution index (NCPI) were employed to evaluate heavy metal contamination levels in paddy soils. SFPI aids in identifying primary heavy metal pollutants and assessing the extent of damage with seven categories [34,35]. Meanwhile, NCPI evaluates the risk of heavy metal contamination in paddy soils. Due to the varying impacts of heavy metals on paddy soils and ecological environments, the mean NCPI value was derived using a weighted calculation method. To establish weights, this study adopted Swaine’s proposal to assign values of three, two, and one based on the importance of heavy metals to the environment (i.e., Cd: 3, Pb: 3, Cr: 3, Ni: 2, Cu: 2, and Zn: 2) [36]. The NCPI classification method defined by Förstner et al. (1990) was utilized to categorize the integrated contamination risk into five classes, enabling an evaluation classification [37]. The formulas of the single factor pollution index (SFPI) and Nemerow composite pollution index are as follows:

$$\mathrm{S}\mathrm{F}\mathrm{P}\mathrm{I}={\displaystyle \frac{C_i}{S_i}}$$

(1)

$$\mathrm{N}\mathrm{C}\mathrm{P}\mathrm{I}=\sqrt{(P_{ave}^2+P_{max}^2)/2}$$

(2)

where SFPI refers to the single heavy mental pollution index, $C_i$ refers to concentrations of Cd in paddy soils, and $S_i$ refers to risk screen value of Cd in paddy soils [38]. NCPI refers to the integrated pollution risk for paddy soils. Pave and Pmax refers to the mean and maximum values of each heavy metal pollution index.

2.4. Contamination Source Apportionment

2.4.1. Regression Model

The regression model describes the relationship between variables by fitting a line to the observed data. Simple linear regression is used to estimate the relationship between two quantitative variables. Generally, linear regression models use straight lines, while logistic regression and nonlinear regression models use curves.

2.4.2. PCA Analysis

PCA analysis reduces the dimensionality of a dataset composed of a large number of correlated variables to minimize the loss of original information [4]. The principal component can be expressed as:

$$Z_{ij}={\alpha }_{i1}{X}_{1j}+{\alpha }_{i2}{X}_{2j}+{\alpha }_{i3}{X}_{3j}+\cdots +{\alpha }_{im}{X}_{mj}$$

(3)

where Z is the component score, α is the component loading, X is the measured value of the variable, i is the component number, j is the sample number, and m is the total number of variables.

The FA can be expressed as:

$$Z_{ij}={\alpha }_{f1}{f}_{1i}+{\alpha }_{f2}{f}_{2i}+{\alpha }_{f3}{f}_{3i}+\cdots +{\alpha }_{fm}{f}_{mi}+e_{fi}$$

(4)

where Z is the measured variable, α is the factor loading, f is the factor score, e is the residual term accounting for errors or other sources of variation, i is the sample number, and m is the total number of factors.

Kaiser–Meyer–Olkin (KMO) and Bartlett’s Sphericity tests were used to assess the applicability of data sets for PCA [39]. The evaluation criteria can be described as follows: when KMO > 0.9, it is very suitable for PCA; when 0.8 < KMO > 0.9, it is suitable for PCA; and when 0.7 < KMO > 0.8, PCA can be performed.

2.4.3. APCS-MLR Model

In the APCS-MLR model, absolute principal component scores and heavy metal concentrations are often used as independent and dependent variables, respectively [40]. Regression coefficients are then utilized to determine the contribution rate of each pollutant source. The specific calculation method is described by Sheng et al. (2022) [5].

2.5. Risk Assessments

2.5.1. Potential Ecological Risk

The potential risk between different ecological components related to soil heavy metals can be evaluated by the potential ecological risk index (PERI) [34]. This approach considers various factors, including heavy metal properties, environmental and behavioral characteristics, concentration levels, biological toxicity, and ecological effects [34]. The formula for PERI is as follows:

$$C_f^i={\displaystyle \frac{C_s^i}{B_n^i}}$$

(5)

$$\mathrm{P}\mathrm{E}\mathrm{R}\mathrm{I}=E_f^i=T_f^i\times C_f^i$$

(6)

$$\mathrm{I}\mathrm{R}\mathrm{I}=\sum _{i=1}^n{E}_f^i$$

(7)

where $C_f^i$ represents the heavy metal pollution coefficient, $C_s^i$ represents the heavy metal concentration (mg kg⁻¹), and $B_n^i$ represents the heavy metal background value, which was taken from the soil environmental background value of Hainan Province in this study (Cd: 0.04, Pb: 28.4, Cr: 76.4, Cu: 19.3, Ni: 30.8, and Zn: 60.0. Units: mg kg⁻¹) [41]. PERI is the potential ecological risk index, $T_f^i$is the heavy metal toxicity response coefficient (Cd: 30, Pb: 5, Cr: 2, Ni: 10, Cu: 5, Zn: 1, Mn: 1), and IRI is the integrated risk index for multiple heavy metals in paddy soils [42,43].

In addition, the mean effect range median quotient (MERMQ) is used to identify and prioritize areas associated with soil quality that may potentially pose a threat to biological organisms [44]. The mean effect range median quotient (MERMQ) represents the 50th percentiles of harmful biological effects, and the specific calculation process is as follows:

$$\mathrm{M}\mathrm{E}\mathrm{R}\mathrm{M}\mathrm{Q}={\displaystyle \frac{\sum _{i=1}^n\left(C_i/{ERM}_i\right)}{n}}$$

(8)

where C_i is the heavy metal concentration, and n is the studied number of heavy metals. ERM_i is the heavy metal’s effect range median (Cd: 0.2, Pb: 31.15, Cr: 41.38, Ni: 10.65, Cu: 18.41, Zn: 81.53, and Mn: 483.68). MERMQ considers the following parameters: low priority (probability of existing toxicity 9%) if MERMQ < 0.1, medium to low priority (probability of existing toxicity 21%) if 0.1 < MERMQ ≤ 0.5, high to medium priority (probability of existing toxicity 49%) if 0.5 < MERMQ ≤ 1.5, and high priority (probability of existing toxicity 76%) if MERMQ > 5.

2.5.2. Risk Allocation among Soil Microorganisms and Vegetation

The risks posed by environmental factors to ecological components in direct proximity to the soil, such as vegetation (V) and soil microorganisms (SM), provide information on the initial risk (R_i). The R_i is contingent upon the variability of the risk factors, the likelihood of their occurrence, and the susceptibility of specific organisms, which can be calculated using the following equations:

$$R_i=∆I_{x }\times P_x\times S_{xi},{0\le R}_i\le 1$$

(9)

$$P_x=P_{ex}\times P_{ox}$$

(10)

$$∆I_{x }={\displaystyle \frac{\left|I_{tx}-I_{ox}\right|}{max\left|I_{tx},I_{ox}\right|}}, 0\le ∆I_{x }\le 1$$

(11)

where R_i is the initial risk (simulated and dimensionless), ΔI_x is the risk intensity caused by the change of risk factor x at time (t) after interference, P_x is the occurrence probability, and S_xi is the sensitivity of component I to indicator x. In this formula, S_i = 0.3 (vegetation) or S_i = 0.5 (soil microorganism) [45]. In Equation (2), P_ox indicates the probability of risk, whereas P_ex indicates the distribution probability. The specific calculation method used was described by Zerizghi et al. (2022) [46]. In Equation (3), I_tx represents the measured value of the indicator x at time (t), and I_ox is the Chinese background value of the indicator without interference [47].

2.5.3. Risk Flow Mechanism among Ecological Components

Network environment analysis (NEA) can quantify the complex ecological risks caused by multiple sources on ecological receptors within an ecosystem [48]. The control allocation (CA) controls the risk intensity from one component to another [16,45,49]. Details regarding the calculation of CA can be found in Supplementary Material S1. Notably, when an ecosystem is subjected to specific disturbances, both the transmission and interpretation of risk information occur between various components.

3. Results and Discussion

3.1. Soil Physicochemical Properties and Heavy Metal Concentrations

Table 1. Descriptive statistics for element concentrations in Hainan paddy soils.

	Max	Min	Average	Median	SD	CV. (%)	Background Value [51]	Risk Screening Value [38]	% of SER
pH	8.90	4.40	5.75	5.20	1.20	20.92	-	-	-
OC	6.74	0.40	1.33	1.13	0.76	56.93	-	-	-
CEC	32.00	1.00	10.80	9.00	5.86	54.24	-	-	-
Cd	1.60	0.02	0.29	0.20	0.28	93.79	0.22	0.3	29.26%
Pb	211.93	1.03	43.05	31.15	36.24	84.18	40.6	80	11.35%
Cr	662.14	1.70	82.94	41.38	112.65	135.82	65.2	250	9.61%
Ni	313.54	1.57	32.23	10.65	54.30	168.48	42.5	60	15.72%
Cu	121.16	0.88	25.82	18.41	24.25	93.93	38.38	50	13.10%
Zn	575.70	3.20	99.15	81.53	74.65	75.29	89.7	200	6.55%
Mn	2481.15	47.27	596.62	483.68	470.92	78.93	626	-	-
Fe	20.82	0.19	3.98	2.67	3.61	90.63	-	-	-
P	5496.43	78.26	1133.91	929.82	881.50	77.74	-	-	-

SD: Standard deviation; CV. (%): Coefficient of variation; and % of SER: % of samples exceeding standard value. Units: OC (dg kg⁻¹); CEC (mmol kg⁻¹); Fe (%); Cd, Pb, Cr, Ni, Cu, Zn, Mn, and P (mg kg⁻¹).

displays the descriptive statistics for pH, OC, CEC, and heavy metal concentrations in the paddy soils. The average pH of paddy soils ranged from 4.40 to 8.9, with a median value of 5.20, indicating that most of the paddy soils in the studied area were slightly acidic. The OC ranged from 0.40 to 6.74% in weight, with a mean value of 1.33%. The mean CEC ranged from 1 to 32 cmol kg⁻¹. This study used Kolmogorov Smirnov to test for normal distribution, ANOVA to analyze differences between data, and Tukey for post hoc testing. The results showed that the mean concentrations of soil heavy metals decreased in the following order: Zn > Cr > Pb > Ni > Cu > Cd (p < 0.01), and there is a significant difference between any two elements of them. These concentrations exceeded screening values regulated by risk control standards (Cd: 0.3; Pb: 80; Cr: 250; Ni: 60; Cu: 50; Zn: 200) [38] by 29.26%, 11.35%, 9.61%, 15.72%, 13.10%, and 6.55%, respectively. This indicates that Cd and Ni were the primary contaminants in Hainan Island. Moreover, Cd has been identified as a significant pollutant in agricultural soils in China [50].

3.2. Contamination Level of Heavy Metals in Paddy Soils

Knowledge about integrated contamination levels of heavy metals in tropical paddy soils is essential for island soil management. In this study, the SFPI and NCPI were selected and results showed that the SFPI values of heavy metals Cd, Pb, Cr, Ni, Cu, and Zn in paddy soils ranged from 0.067 to 5.333, 0.013 to 2.649, 0.007 to 2.649, 0.026 to 5.226, 0.018 to 2.423, and 0.016 to 2.879, with average values of 0.980, 0.538, 0.332, 0.537, 0.516, and 0.496, respectively (Figure 2). The proportions of different heavy metal contamination levels in the paddy soils are shown in Figure S2. According to Zhang et al. (2018) [52], most heavy metal contamination levels fell within the class of 0−3, while Cd and Ni contamination levels were within the class of 0−6. This indicates that the contamination levels of Cd and Ni at some sampling points were extremely high. Approximately 0.44 and 1.31% of the soil samples exhibited extremely high levels of Cd and Ni, respectively. From the spatial distribution maps showing the contamination levels in paddy soil samples (Figure 2a–f), it can be concluded that most of the soil contamination was concentrated in coastal plains and economically developed areas, such as Haikou, Chengmai, Danzhou, Lingao, and Wenchang. This may be because the coastal plain areas and economically developed regions of Hainan Island are the main production areas for rice, and factors such as the extensive use of pesticides and fertilizers may bring heavy metals into the soil during agricultural activities. Lead arsenate (PbHAsO₄) insecticides have been widely used to control pests in different crop systems [53]. Bordeaux mixtures of copper sulfate (CuSO₄) are commonly used to control fungal invasion in various crop systems [54], and phosphate fertilizers may contain 0.1–170, 7–38, and 7–225 milli-grams per kilogram of total Cd, Ni, Hg, and Pb, all of which are important sources of heavy metal elements [55]. Similarly, transportation is also an important source of heavy metals. Therefore, in order to prevent further pollution of heavy metals, crop rotation and fallow modes should be implemented in areas with severe pollution; or fertilizers and pesticides should be applied reasonably to reduce the accumulation of heavy metals; or finally, planting trees and forests on both sides of the road to reduce the opportunity for heavy metals to enter farmland through atmospheric deposition and other means.

The NCPI values ranged from 0.102 to 4.256, with an average of 1.001, indicating that it is slightly polluted (1.0 < NCPI ≤ 2.0) (Table S1). Based on the evaluation standards, around 4.80% of paddy soils were heavily polluted, primarily located in portions of the Haikou, Chengmai, Danzhou, Wenchang, Lingao, and Anding regions (Figure 2g). Additionally, approximately 21.83% of soil samples were slightly polluted, followed by those categorized as precautionarily polluted (17.47%) and moderately polluted (6.55%). The samples with NCPI values < 0.7 occupied most of the study area, accounting for 49.34%.

3.3. Source Apportionment

3.3.1. Correlation Analysis

Correlation analysis can help explore the relationships between these elements. The results showed that Cr–Ni (0.70), Cr–Cu (0.65), Cr–Fe (0.59), Ni–Cu (0.71), Ni–Fe (0.69), Cu–Zn (0.69), Cu–Mn (0.64), Cu–P (0.56), Zn–Mn (0.65), and Zn–P (0.62) had higher correlation coefficients, indicating the same or similar origins and geochemical behaviors (Figure 3).

3.3.2. PCA Analysis

Prior to PCA analysis, the data were tested using the Kaiser–Meyer–Olkin (KMO > 0.800) and Bartlett sphericity tests (p = 0.000) after standardization [9]. In this study, the results (KMO = 0.822 and p = 0.000) showed the suitability of these data. According to the principle of eigenvalues greater than 1, three principal components (PC1, PC2, and PC3) were extracted with an accumulative variance of 82.742% (

Table 2. Composition matrix after rotation for each element load.

	Principal Component Load
	PC1	PC2	PC3
Cd	0.158	−0.034	0.861
Pb	0.054	0.129	0.882
Cr	0.910	0.141	0.038
Ni	0.926	0.090	0.182
Cu	0.740	0.499	0.210
Zn	0.302	0.845	0.137
Mn	0.729	0.472	0.058
Fe	0.876	0.281	0.076
P	0.188	0.907	−0.038
Percentage of variance (%)	54.072	16.261	12.410
Percentage of accumulation (%)	54.072	70.332	82.742

).

According to these results, three principal components can explain 82.742% of the total data variance (Table 2). Among them, PC1 was dominated by Ni, Cr, and Fe. Generally, Fe, Cr, and Ni, which are iron group elements, are both readily bound to soil oxides and have a close relationship with the parent materials [24,56]. Cr and Ni are derived from alkaline and ultra-alkaline rocks. According to Table 1, the average concentration of Ni in paddy soil is lower than its background value, and the average concentration of Cr is 1.27 times the background value, which is very close to the concentration of the element background value. Thus, in this study, Ni and Cr are related to geogenic and pedogenic processes in paddy soils [24,56]. Therefore, Factor 1 can be considered the “geological source” since it reflects the comprehensive impact of the geological background.

PC2 was dominated by P, Zn, Cu, and Mn. The application of fertilizers and pesticides in pursuit of high yield is a common practice in China’s grain system [57]. These chemicals, especially phosphate fertilizers and pesticides, contain chemical mixtures of various heavy metals [58,59]. Phosphorus-containing fertilizers typically contain different heavy metal elements, including Cu and Zn, with concentrations of 0.41–11.6 and 4.87–348.2 mg kg⁻¹, respectively [58]. Based on these reports, China applies approximately 1200 tons of Zn and 5000 tons of Cu-containing fertilizer to farmland every year, respectively [60]. Consequently, the application of these chemicals containing heavy metals gradually increases the pollution level of soil heavy metals. The Bordeaux mixture of copper sulfate (CuSO₄) is often used as a protective fungicide in various crop systems, inhibiting the germination of pathogenic spores or mycelial growth by releasing soluble copper ions [54]. Cu and Zn also exist as additives in livestock feed by providing intestinal antibacterial agents [61] and controlling post-weaning diarrhea [62]. Mn is typically added to feed additives to increase feed conversion rates, boost egg production, and help prevent diseases [63,64]. The application of this manure by local farmers can also lead to the enrichment of heavy metals in paddy soil [65,66]. Therefore, PC2 can be considered as “agricultural activities”.

PC3 was dominated by Cd and Pb with a data variance of 0.861 and 0.882, respectively, which was much higher than the other PCs’ variance. According to relevant reports, the soil along highways often has high concentrations of Cd and Pb, and transportation is the main source of pollution [67,68,69,70]. Hainan Island has a well-developed tourism industry and high traffic volume. The road network diagram illustrates that the areas of paddy soils contaminated with Cd and Pb were primarily located in regions with high traffic flow or near intersections (Figure S1b). This phenomenon could be attributed to the presence of Pb in brake pads, tires, and lubricating oils [40,71], which are released into the environment during vehicle operation and subsequently settle in the soil. Additionally, Cd, a significant additive in automotive tire production, contributes to soil pollution as dust containing cadmium is discharged into the soil from vehicle tires. Despite the complete ban on Pb-containing gasoline, the accumulation of lead along roads remains a concern due to decades of leaded gasoline use [7]. Furthermore, heavy metals exhibit poor degradation in soil, emphasizing the ongoing importance of addressing contamination issues. Hainan Island harbors abundant Pb ore resources (Figure S1c). However, the spatial distribution of soil pollution indicates that the pattern of Pb pollution does not align with the distribution of Pb ore resources (Figure S1c), suggesting that the latter does not significantly contribute to the current soil Pb contamination. Therefore, PC3 can be characterized as “traffic emissions”.

3.3.3. Source Apportionment Based on APCS-MLR Model

Based on PCA results, we obtained the influence degree of each possible source and constructed the functional relationships between each possible source and its variables using the APCS-MLR model. The determination coefficient (R² > 0.75) and significance level (p < 0.001) were used to check the validity of this model (Table S2), indicating better applicability in paddy soil. All sources’ contribution to each element was estimated through MLR equations. However, the sum of the three principal sources is less than 100%, indicating the presence of other sources.

Cr (63.75%), Ni (47.87%), and Fe (40.93%) primarily originated from geological sources (Figure S3). Conversely, P, Zn, Cu, and Mn were predominantly attributed to agricultural activities, with contribution rates of 85.86%, 74.19%, 44.45%, and 48.81%, respectively. Traffic emissions dominated the contribution to Cd and Pb, with contribution rates of 82.60% and 82.04%, respectively. Among the nine studied elements, PC4 (unknown sources) made a relatively weaker contribution, with Ni and Fe being the main constituents, possibly linked to limited industrial activities in Hainan Island. Overall, the APCS-MLR model showed that the main element contributions in paddy soil on Hainan Island were derived from geological sources and human activity–related sources. Among the sources related to human activities, agricultural activities and traffic emissions are the most prominent; this may be due to the developmental positioning of Hainan Island.

3.4. Risk Assessment

3.4.1. Potential Ecological Risk of Heavy Metals in Paddy Soil

Figure 4a shows the classification criteria of PERI proposed by Hakanson (1980) [34]. The results show that Cd exhibited the highest ecological risk. This may be related to the high exceedance rate of cadmium and its pollution characteristics. The National Soil Pollution Status Survey Bulletin shows that the point exceedance rate of Cd in soil nationwide is 7%, which is much higher than the exceedance rate of other pollutants. In addition, soil Cd pollution has characteristics such as concealment, lag, and non-degradability. Once exogenous Cd enters the soil, it gradually accumulates and poses a long-term threat to soil health; moreover, the soil pollution carrying capacity of cadmium is much smaller than that of other heavy metals, that is, a slight increase in cadmium content in soil will significantly increase the cadmium concentration in crops. Among the Cd contaminants, 70.00% of paddy soils had a low contamination level, followed by moderate (20.00%), considerable (6.96%), and high (3.04%) levels. For Ni contaminants, 95.65% of paddy soils showed low contamination levels, with the remainder showing moderate contamination (3.91%). Additionally, R_i values for Pb, Cr, Cu, and Zn were below the lowest range (PERI < 40), indicating no ecological risk between different ecological components (Figure 4a).

The mean MERMQ value of the soil samples ranged from 0.5 to 1.5, with a mean of 1.16, suggesting a 49% probability of potential hazards to biological organisms [44]. Moreover, 46.49% of the soil samples fell within the range of 0.5 to 1.5 (49% probability of toxicity), while 5.02% were within the range of 0.1 to 0.5 (21% probability of toxicity). Furthermore, we conducted a linear regression (Figure 4b) on the relationship between the IRI and MERMQ, and the results revealed a moderate relationship (R² = 0.341), which demonstrates that the risks caused by soil contamination have an impact on ecological species. From this perspective, although ecological risk assessments based on PERI and IRI have no negative impact, they continue to affect biological organisms, which aligns with the findings of Zerizghi et al. (2022) [46]. Therefore, PERI and IRI are commonly used for assessing potential ecological risks; however, their impact on ecological species is underestimated.

3.4.2. Risk Allocation among Soil Microorganisms and Vegetation

Soil contamination is harmful to the organisms in ecosystems [45]. Direct risk receptors for vegetation (producer) and soil microorganisms (decomposer) are considered as the R_i, which are risk inflow sources for each trophic level in the entire ecosystem [47].

Table 3. Heavy metal concentrations of the local background and maximum threshold in tropical paddy soils (mg kg⁻¹) used to assess probability of risk occurrence in soil of Hainan Island.

Heavy Metals	Background Level	Polluted	ΔIx	Probability of Risk Occurrence (Px)
Cd	0.22	0.3	0.393	0.368
Pb	40.60	80	0.394	0.345
Cr	65.20	250	0.514	0.192
Ni	42.50	60	0.712	1.151
Cu	38.38	50	0.544	0.222
Zn	89.70	200	0.358	0.607

presents the data used to assess the probability of risk in the soil on Hainan Island.

In the process of calculating Ri, ΔI_x and P_x must first be evaluated. Using Equation (11), we calculated that the risk intensity (ΔΙ_x) for each heavy metal was 0.393 (Cd), 0.394 (Pb), 0.514 (Cr), 0.712 (Ni), 0.544 (Cu), and 0.385 (Zn). Subsequently, P_x was calculated using Equation (10). We chose the soil background [49] and screening values [52] in China to calculate P_x because of its relation to soil contamination levels [49]. Six categories were proposed to represent the risks associated with contamination levels. Further information on the calculation of P_x and the specific division method can be found in Tang et al. (2017). The results showed that Ni had the highest P_x value (1.515), followed by Zn (0.607), Cd (0.368), Pb (0.345), Cu (0.222), and Cr (0.192) (Table 3).

The average R_i values for Cd, Pb, Cr, Ni, Cu, and Zn in the vegetation were 0.043, 0.041, 0.030, 0.025, 0.036, and 0.065 (Figure 5a), respectively, and those for the soil microorganisms were 0.072, 0.068, 0.049, 0.410, 0.061, and 0.109 (Figure 5b), respectively. According to Hakanson (1980) [34], R_i is regardless of toxicity, which the NEA assessment does not consider. The highest R_i value was observed for Ni and ranged from 0.028 to 0.332 for vegetation and 0.046 to 0.554 for soil microorganisms. Additionally, compared with other metals, Zn and Pb displayed comparatively higher R_i values; however, the R_i values for Ni exceeded those for Zn and Pb by 3.77 and 6.03 times, respectively. In summary, the R_i value of Ni is the highest, followed by Zn, Pb, Cd, Cu, and Cr (for both vegetation and soil microorganisms).

Soil microorganisms are more sensitive to heavy metal stress than animals (including herbivores and carnivores) and vegetation [72]. In this study, the risk to soil microorganisms (R_i = 0.769) was approximately 1.67 times higher than vegetation (R_i = 0.461). However, the actual risk for sensitive soil organisms may be even higher because we assume that the soil risk screen values are a contamination index and that no risk would occur below this threshold during the current calculation process. Furthermore, a linear regression relationship was observed between R_i and MERMQ, indicating relatively higher credibility of the impact of risks on vegetation (R² = 0.282) and soil organisms (R² = 0.282) (Figure 5c,d). Therefore, heavy metal contamination still can pose an ecological risk to soil microorganisms in the Hainan tropical paddy soils, although PERI is below the safe limit (PERI < 40), according to Hakanson (1980) [34].

3.4.3. Risk Flow Mechanism among Ecological Components

The basic ecological components of ecosystems are vegetation (producers), herbivores (primary consumers), carnivores (secondary consumers), and soil microorganisms (decomposers) [73]. In an autotrophic food web, energy enters through vegetation before moving to higher trophic levels [49]. Consequently, through energy relationships, the risks posed to vegetation can impact the subsequent levels of consumers (including primary and secondary consumers) and, ultimately, the entire ecosystem [74,75]. In complex natural ecosystems, the material cycle, energy flow, and information transfer between components are typically expressed in the form of flow patterns, and their paths, directions, intensity, and rates have significant impacts on the ecosystem. The energy flow relationships between different trophic levels can be explained through an energy flow matrix model, as the interaction of these components regarding energy depends on various factors, such as ecosystem structure and function and population density [45,49]. Figure S4 presents the energy flow matrix between four trophic levels and the CA calculated from this matrix (dashed arrow indicates direction of risk flow) [45,76]; the stronger the CA, the higher the inflow risk.

Multiplying the CA and risk magnitude of energy supply components can simulate the risk flow between trophic levels to a certain extent (Figure 6) [47,73]. In this study, the highest R_i value of vegetation input to the ecosystem was for Ni (0.2459), followed by Zn (0.0653), Cd (0.0434), Pb (0.0408), Cu (0.0363), and Cr (0.0296). The highest R_i of soil microbial input into the ecosystem was for Ni (0.4098), followed by Zn (0.1088), Cd (0.0723), Pb (0.0680), Cu (0.0605), and Cr (0.0493). There is no direct connection between heavy metal contamination of soil and herbivores or carnivores; however, risk may be transferred to herbivores or carnivores through energy transfer established by competitive prey relationships among the different trophic levels [45]. Vegetation mostly controls herbivores; therefore, a direct threat to vegetation may signify an associated risk. The average direct risk transfer from vegetation to herbivores ranged from 0.0060 (Cr, lowest risk) to 0.0495 (Ni, highest risk). Additionally, the average risk to carnivores from both direct (herbivore) and indirect (vegetation) interactions ranged from 0.0011 (Cr, lowest risk) to 0.0206 (Ni, highest risk) (Figure 6). Soil contamination posed by heavy metals can affect the diversity and stability of soil microorganisms, which can change the community structure and functions [77]. According to the calculation, the risk allocation for soil microorganisms from vegetation was 5.03 and 16.84 times higher for herbivores and carnivores, respectively.

Soil microorganisms were controlled by four pathways: risk from soil exposure (R_i, external soil environment), vegetation, herbivores, and carnivores. In addition to the external environment, the cumulative risks of Cd, Pb, Cr, Ni, Cu, and Zn transferred from vegetation (direct risk), herbivores (indirect risk), and carnivores (indirect risk) to soil microorganisms were 0.0434, 0.0418, 0.0306, 0.2457, 0.0363, and 0.0653, respectively. Carnivores are controlled by both vegetation and herbivores. The cumulative risks from soil heavy metals for carnivores, transferred from vegetation and herbivores, were 0.0165 and 0.0385, respectively. Based on the integrated risk across the four trophic levels, carnivorous animals exhibited the least harm, with herbivores showing slightly lower harm (Figure S5a). Regarding the proportion of risk attributed to the six heavy metals, it can be concluded that the largest percentage of risk posed to the different trophic levels originated from Ni (53.29–53.37%), followed by Zn (14.11–14.16%), Cd (9.41–9.45%), Pb (8.84–8.92%), Cu (7.74–7.87%), and Cr (6.40–6.44%) (Figure S5b).

4. Conclusions

This study analyzed the concentrations of seven heavy metals and their spatial patterns in paddy soils on a tropical island. The findings revealed that most heavy metal contamination levels fell within the class of 0−3, while Cd and Ni contamination levels were within the class of 0−6. In addition, soil contamination was predominantly concentrated in coastal plains and economically developed areas, including Haikou, Chengmai, Danzhou, Lingao, and Wenchang. Utilizing the APCS-MLR model, three main sources of heavy metal pollution were identified: geological, agricultural activities, and traffic emissions. Within each principal fraction, geological sources contributed the most to environmental pollution levels, followed by agricultural activities and traffic emissions. Notably, among the studied heavy metals, Cd and Ni emerged as significant contributors to ecological risks. Soil contamination can be transferred to higher trophic levels through energy exchange across food webs, and the risk flow follows the order of soil microorganisms, vegetation, herbivores, and carnivores. Ecological species on the island were more vulnerable to the effects of Ni, Zn, Cd, Pb, and Cu, whereas Cr posed the least threat. Therefore, it is important to explore governance measures to reduce the concentrations of Cu, Zn, and Mn released from agricultural activities and Cd and Pb emitted from transportation to prevent the environmental and ecological risks caused by heavy metals. These results provide fundamental and significant information for mitigating the ecological risks caused by heavy metals by controlling priority pollutants and sources on tropical islands in China.