Ecological Load and Migration of Heavy Metals in Soil Profiles in Wheat–Corn Rotation Systems

Abstract

Heavy metal contamination in agricultural soils is a critical global concern, threatening ecosystem safety and food security. The wheat–corn rotation system, vital for food production in regions like Northern China, is particularly vulnerable. However, comprehensive studies investigating vertical migration, future dynamics under climate change, and predictive modeling of heavy metals within this system are still limited. This study combined field sampling of soil profiles (0–200 cm) with geochemical modeling (the PROFILE and SSCL models) and machine learning techniques (Multiple Regression, Neural Networks, and Random Forest). Key findings revealed that atmospheric deposition was the primary input source for most heavy metals, contributing 49.50–93.27%. The release rates (Rm) of heavy metals were significantly higher during the corn season than the wheat season and are projected to increase by 1.2–1.5 times under the RCP4.5 climate scenario. Vertical distribution analysis showed a significant accumulation of heavy metals in the middle soil layer (20–120 cm), with Arsenic (As) and Cadmium (Cd) exhibiting the strongest migration potential, posing a threat to groundwater. The Random Forest model demonstrated superior performance (R² > 0.95) in predicting heavy metal behavior, identifying Fed and soil TOC as the dominant controlling factors. This study provides a unique and significant contribution by integrating geochemical fate modeling with climate projections and advanced machine learning to offer a predictive, multi-faceted risk assessment framework, thereby supplying a scientific basis for targeted pollution control and sustainable soil management in wheat–corn rotation systems under a changing climate.

1. Introduction

Heavy metals in soil possess a significant global environmental risk, threatening ecological safety and human health [1]. Soil serves as a critical medium in the Earth’s surface environment, and plays an essential role in the cycling of elements in the biosphere. Its quality is directly related to sustainable development and human health. The soil-crop system, an inseparable entity, is consistently associated with human health issues through the food chain, rendering heavy metal pollution in this system an urgent global environmental concern [2]. The increase of socio-economic activities and intensified human endeavors have led to the discharge of various heavy metals into the soil, resulting in a severe situation of heavy metal pollution [3]. Due to their high toxicity, persistence, difficulty in degradation, and sustainable utilization by living organisms, heavy metals are widely recognized as a global environmental threat [4], and have become a hot spot of studies [5]. Soil characteristics govern the bioavailability and mobility of heavy metals, directly affecting their absorption and accumulation in crops [6]. The sources, translocation, and output of heavy metals in the soil comprise a complex dynamic biochemical process influenced by various factors. Hence, analyzing the extent of heavy metal pollution in local soils provides a data foundation for future investigations into the impacts of soil heavy metals on human health.

Crop rotation systems such as legumes–paddy, legumes–wheat or corn–wheat, etc., are recognized as sustainable agricultural management practices that significantly improve soil physicochemical properties, enhance crop yields, and elevate soil health levels [7]. The interaction between crop rotation and tillage practices significantly affects soil pH and organic carbon content, particularly in the surface soil layer [8]. Existing studies have indicated that implementing crop rotation fosters the balanced utilization of soil nutrient elements, enhances soil fertility and resource efficiency, and ultimately leads to higher food production [9]. Corn, as the highest-yielding cereal crop globally, supplies a vital source of feed and industrial raw materials. As the foundation of the food chain, corn absorbs heavy metals from the soil, then transports them to higher trophic levels, particularly impacting human health [10]. Some researchers have documented significant heavy metal accumulation in corn. For example, corn harvested in northern Ningxia exhibited Pb and Cr levels exceeding established standards [11] while corn cultivated in Zambia demonstrated a greater mobility of Cd than Zn and Pb [12]. While serving for over half of the global population, wheat has undergone research regarding the pollution, transport, and toxicity of heavy metals within the soil-wheat system [13]. Previous studies have evaluated the accumulation and transfer of heavy metals within this system [14]. Recent findings indicate that the extent of Pb and Cd pollution in wheat is gradually expanding and worsening annually [15]. Furthermore, the levels of heavy metal exposure in wheat vary across regions in China. For example, wheat in Xi’an showed Cr levels of 0.533 mg/kg, As of 0.930 mg/kg, Cd of 0.044 mg/kg, and Pb of 0.218 mg/kg [16]. In Henan Province, Cd levels range from 0.102 to 0.168 mg/kg, exceeding national standards [17]. While a study in Pakistan identified Cd concentrations in wheat that exceeded permissible limits [18]. The agricultural model of wheat and corn rotation not only holds significance for global food production but also plays a crucial role in maintaining ecological balance and improving soil quality.

Soil is a fundamental resource for agricultural production. With rapid population growth, resource depletion, and environmental degradation, per capita arable land is steadily declining [19]. Wheat and corn, as major global food crops, occupy a central role in agricultural crop production. The practice of wheat–corn rotation is widely regarded as an effective method to enhance land use efficiency and improve soil fertility, thereby promoting sustainable agricultural development [20]. With the economy and society increasing, various human activities-such as mineral exploitation, industrial production, agriculture, and commerce, have generated environmental damage. One of the direct consequences of these activities is the accumulation of heavy metals in surface soils [21]. Heavy metals in farmland soil can enter the food chain through crops like vegetables, rice, and wheat, thereby posing threats to both animals and humans [22]. Research on the wheat–corn rotation system in North China indicates that Cd, Cr, and Ni in soils are the most likely to exceed safety standards [21]. Furthermore, assessments of the ecological and health risks associated with heavy metal pollution have been conducted in wheat–corn planting systems characterized by high geological backgrounds [23]. The increasing environmental issues, such as climate change and ecosystem degradation, have drawn global attention [24]. Agricultural soils are intricately linked to climate change, forming a significant component of agricultural systems influenced by climatic variations, which can severely impact food security [25]. The impact of climate change on heavy metal pollution has been a topic of discussion. In the context of increasing environmental change, research on heavy metal pollution has emerged as a critical focus within current ecological and environmental studies. Numerous studies have reported on heavy metal contamination in farmland soils worldwide through various assessment methods [26,27]. Previous research indicates that the vertical distribution of metal elements exhibits certain regularities, with the migration and leaching phenomena of heavy metal elements occurring in the soil weathering process, showing varying degrees of migration and accumulation for different elements [28]. Human actions are exacerbating global climate change, leading to a rise in the frequency of extreme weather events such as droughts and high temperatures. Research on the geochemical behaviors of heavy metals under these extreme environmental conditions remains relatively rare [29].

This study focuses on the soil profiles in wheat–corn rotation farmland, exploring the vertical leaching and migration accumulation of heavy metals. At present, the longitudinal migration and deep accumulation of heavy metals under the crop rotation system are still unclear, which limits the accurate assessment of the long-term environmental behavior of heavy metals in farmland soil and the risk of groundwater pollution. Given the complexity and diversity of climate change and rotation systems, current research has some limitations; thus, future investigations must integrate multiple research methodologies to assess the influences of climate change on rotation systems, thereby promoting and ensuring their sustainable development.

2. Materials and Methods

2.1. Study Area and Sample Collection

This study is located between 117°14′ E and 117°31′ E, and 32°48′ N and 33°08′ N, Anhui province (Figure 1), located in the transition zone between a subtropical humid and a warm temperate semi-humid monsoon climate. The area is characterized by a mild climate with distinct seasons, moderate precipitation, pronounced monsoon influences, dry winters, and hot, rainy summers with frequent extreme weather events. Significant variability exists in annual precipitation and temperature, as illustrated by the monthly averages presented in Figure 2. Approximately 60% of the yearly precipitation occurs between June and September, and rainfall from November to February contributes only about 10% to the annual total. The average annual water surface evaporation is 984 mm, with 39% occurring from June to August, and a humidity coefficient of 0.7 indicates moderate humidity. The annual average temperature is 15.1 °C, and the frost-free period ranges from 200 to 220 days. Notably, there is considerable variability in precipitation distribution, with heavy rains concentrated during the flood season each year.

According to the Köppen–Geiger climate classification system, the study area is classified as Cwa.

The soils in our study area are classified as “Shajiang Black Soil” according to the Chinese Soil Genetic Classification system, which corresponds to Endocalcic Stagnic Luvisols (Clayic, Aric, Humic) in the World Reference Base for Soil Resources (WRB, 2022) [30]. This classification reflects the characteristic sticky black surface horizon, subsurface calcium carbonate concretions, and seasonal waterlogging of the region.

Soil profiles and corresponding crop samples from 16 farmlands with wheat–corn rotation in the study area were collected. The detailed soil characteristics for each of the 16 individual sampling sites are provided in Supplementary Table S1. A systematic sampling method was employed. After removing surface debris such as dead branches and leaves from the selected farmland. A soil pit was excavated to dimensions of 150 cm × 100 cm × 200 cm. Profile subsamples were taken at varying depths using a shovel, with intervals determined by the actual soil morphology (0–20, 20–60, 60–120, 120–160, 160–200 cm). Most profiles extended to a depth of 0–200 cm. The samples were stored in plastic bags, each weighing about 500 g, and labeled according to profile and depth. In the selected farmland, soil and plant samples of a 50 cm × 50 cm area were randomly collected, including straw and seeds. For irrigation water samples, the sampler was washed with the corresponding irrigation water from each site 3–4 times before each sample collection. After filtration, the collected irrigation water samples were stored in acid-washed plastic bottles and taken to the laboratory for storage at 5 °C. Fertilizer samples were obtained from local farmers. A part of each soil sample was reserved for weighing and air drying. Then the dried samples were prepared for subsequent laboratory analysis. Details of the sampling procedure, including plant, fertilizer, and irrigation water, were described in our previous studies [31].

2.2. Sample Processing and Calculation Formula

2.2.1. Elemental Chemical Analysis

A small solid sample (200.00 g), including soil, fertilizer and wheat and corn cereal, was weighed and add 10 mL of hydrochloric acid (ρ = 1.19 g/mL), followed by the addition of 10 mL of nitric acid (ρ = 1.42 g/mL), 10 mL of perchloric acid (ρ = 1.68 g/mL), and 10 mL of hydrofluoric acid (ρ = 1.49 g/mL). The mixed solution was digested in a microwave digestion system (Ethos Touch Control, Milestone Inc., Bergamo, Italy) at 180 °C. After digestion, the solution was diluted with ultra-pure deionized water for subsequent chemical analysis.

Irrigation water samples were digested using hydrogen peroxide and nitric acid, followed by dilution with ultra-pure deionized water for further analysis.

Arsenic (As) and mercury (Hg) concentrations were measured using the AFS230E cold atomic fluorescence spectrometer (HG-AFS) (xgy-1011a, IGGE, Langfang, Hebei, China); cadmium (Cd) was measured with a graphite furnace atomic absorption spectrophotometer (AAS ZEEnit6o, Analytic Jena AG, Jena, Germany); chromium (Cr), copper (Cu), lead (Pb), and zinc (Zn) were analyzed using the Thermo ICP-MS XSERIES (Thermo Fisher Scientific, Waltham, MA, USA); and the quantities of major elements in the soil were determined using powder X-ray fluorescence (XRF, ZSX primus II, Rigaku Corporation, Tokyo, Japan).

Soil pH was measured with the Delta 320 pH meter (Mettler Toledo Delta 320, Mettler Toledo, Greifensee, Switzerland). A mixture was prepared using 5 g of air-dried soil and 12.5 mL of distilled water, which was agitated and allowed to sit for 30 min before measuring the pH of the supernatant with a pH meter.

Quality assurance and quality control (QA/QC) procedures were implemented to assess the accuracy and precision of the analytical data, including reagent blanks, replicate samples, and certified reference materials. To prevent sample contamination, high-purity deionized water was used to rinse glassware, prepare standard solutions, and dilute samples.

2.2.2. Other Parameters

Soil bulk density, moisture content, particle size distribution, and cation exchange capacity were sourced from the soil science database (http://vdb3.soil.csdb.cn/; accessed on 15 November 2023). Selective extraction of iron oxides was performed on the collected soil samples: acid ammonium oxalate-extractable Fe (Feo) was extracted in darkness at pH 3 to target poorly crystalline phases, and citrate-bicarbonate-dithionite-extractable Fe (Fed) was determined following the standard DCB method to represent total free Fe oxides. Wheat and corn yields, average annual temperatures, average annual rainfall, irrigation water, and fertilizer consumption data were obtained from the National Bureau of Statistics (http://www.stats.gov.cn/sj/ndsj/2020/indexch.htm; accessed on 20 November 2023). The mineralogical characteristics of the soil samples were calculated based on major element contents using the Norma program (Standardized Mineral Distribution Model) [32]. Climate data from 2024 to 2100 were sourced from WMO, ECA&D, and KNMI forecasts (https://climexp.knmi.nl/start.cgi; accessed on 15 April 2024).

2.2.3. Weathering Model and Leaching Calculation

In agricultural ecosystems, soil leaching is generally considered a weathering process of soil minerals. Currently, the PROFILE model has been applied as the most reliable method for calculating soil weathering rate successfully [33]. Hence, this study evaluates soil leaching using the PROFILE model. According to the PROFILE model, the leaching rate in farmland soil is the sum of the weathering rates of all minerals in the soil [34]:

$$R_w={\sum }^j{r}_j·x_j·A_w·z·\theta$$

(1)

where R_w denotes the total weathering rate of soil layers (kmol/(m²·s)), r_j indicates the decomposition rate of mineral j (kmol/(m²·s)), x_j represents the content of mineral j in the soil (%), A_w is the specific surface area of the mineral (m²/m³), θ is the soil water saturation (%), and z is the thickness of the soil layer (m). The parameters used in this study are shown in Table S1.

The release rate (R_BC) of exchangeable cations (Ca²⁺, Mg²⁺, Na⁺, and K⁺) can be calculated as follows:

$$R_{BC}={\sum }^j{r}_j({\mathrm{y}}_{\mathrm{C}\mathrm{a},\mathrm{j}}+{\mathrm{y}}_{\mathrm{M}\mathrm{g},\mathrm{j}}+{\mathrm{y}}_{\mathrm{N}\mathrm{a},\mathrm{j}}+{\mathrm{y}}_{\mathrm{C}\mathrm{a},\mathrm{j}})$$

(2)

y_Ca,j, y_Mg,j, y_Na, y_K,j are the stoichiometric coefficients representing the elements Ca, Mg, Na, and K in minerals.

Release rate of heavy metals from soil leaching:

$$R_m=R_{BC}{\displaystyle \frac{C_t}{C_{BCt}}}$$

(3)

where R_m is the leaching rate of heavy metals (mg/m²·yr); R_BC is the release potential of base cations (mol/ha·s); C_BCt is the total content of base cations in soil (mol/kg); and C_t is the concentration of heavy metals in soil (mg/kg). The concentration of total base cations was calculated according to the percentage of element composition in soil:

$$C_{BCt}=10({\displaystyle \frac{CaO\%}{56}}+{\displaystyle \frac{MgO\%}{40}}+{\displaystyle \frac{K_{2}O\%}{47}}+{\displaystyle \frac{{Na}_{2}O\%}{31}})$$

(4)

The release rates of heavy metals, the flux of plant absorption of heavy metals, and the external heavy metal input flux are calculated according to the procedures outlined in the SSCL model [33], followed by calculations based on the specified heavy metal concentration equations. This study’s mass balance is predicated on the assumption that the dynamic processes of heavy metals in the soil can be described as a black box model. Thus, the accumulation of heavy metals is based on dynamic changes in heavy metal concentrations in the soil. In agricultural systems, the accumulation of heavy metals in the topsoil may result from a series of subprocesses related to mass balance, including atmospheric deposition, irrigation and fertilizer inputs, crop absorption, and leaching outputs [33]:

$$\rho \mathrm{z}{\displaystyle \frac{\partial C_t}{\partial t}}=Q_{in}-{\displaystyle \frac{\partial U}{\partial t}}-R_m$$

(5)

where ∂C_t/∂_t is the changes in heavy metals; ∂U/∂t (g/ha·a) is the output flux of heavy metals absorbed by plants. They are the dependent variables that change with time where Q_in means external input; Z is the thickness of the target soil layer (m); and ρ is the bulk density (kg/m³).

According to the above formula, we can get the equation:

$$\rho \mathrm{z}{\displaystyle \frac{\partial C_t}{\partial t}}=Q_{in}-\rho \mathrm{z}{K}_u{C}_t-R_{BC}{\displaystyle \frac{C_t}{C_{BCt}}}$$

(6)

The uptake of Heavy metal uptake by plants (∂U/∂t) is was calculated as follow (K_u is the plant absorption coefficient of plant) [35]:

$${\displaystyle \frac{\partial U}{\partial t}}=\rho \mathrm{z}{K}_u{C}_t$$

(7)

External heavy metal input (Q_in) is mainly made up of fertilizer, irrigation, and atmospheric deposition [36]:

$$Q_{in}=Q_a+Q_f+Q_i$$

(8)

Q_a, Q_f and Q_i are atmospheric deposition, fertilizer, and irrigation flux, respectively.

The concentration of heavy metals in topsoil (C_t) is a function of the time variable (t):

$$C_t=C_0{e}^{-(K_u+{\displaystyle \frac{R_{BC}}{{\rho zC}_{BCt}}})t}+{\displaystyle \frac{Q_{in}[1-e^{-\left(K_u+\frac{R_{BC}}{{\rho zC}_{BCt}}\right)t}]}{\rho z{K}_u+\frac{R_{BC}}{C_{BCt}}}}$$

(9)

The input in the lower soil is considered from only the leaching of the upper soil, and the output is the leaching of the lower soil:

$$\rho \mathrm{z}{\displaystyle \frac{\partial C_t}{\partial t}}=R_m^{upper}-R_m^{lower}$$

(10)

According to the above formula, we can get that heavy metals (C_t) in the lower soil are a function of the time variable (t):

$$C_t=C_0+{\displaystyle \frac{(R_m^{upper}-R_m^{lower})t}{\rho z}}$$

(11)

When C_t reaches the limit value (C_l), the SSCL, the maximum allowable heavy metal flux for external input at the control time, can be derived. Based on the above derivation, we can give the functional relationship between the SSCL value and time t:

$${SSCL}_{tupper}={\displaystyle \frac{\left(\rho z{K}_u+\frac{R_{BC}}{C_{BCt}}\right)[C_l-C_0{e}^{-\left(K_u+\frac{R_{BC}}{{\rho zC}_{BCt}}\right)t}]}{1-e^{-\left(K_u+\frac{R_{BC}}{{\rho zC}_{BCt}}\right)t}}}$$

(12)

$${SSCL}_{tlower}={\displaystyle \frac{(C_l-C_0)\rho z)}{t}}+R_m^{lower}$$

(13)

2.3. Data Processing

Statistical analyses using the SPSS software package (SPSS Inc., Chicago, IL, USA, Version 26, 2019) including Pearson correlation, multiple linear regression, and predictive analysis of neural networks using a multilayer perceptron (MLP). Predictive modeling of random forest algorithms using SPSS Modeler (IBM Corporation, Armonk, NY, USA, version 18.0, 2017). Monte Carlo simulation using the Crystal Ball plug-in for Microsoft Excel (Microsoft Inc., Redmond, WA, USA, Version 18, 2019). Mapping and spatial analysis were performed using Origin (Origin Pro Inc., Northampton, MA, USA, Version 9.1, 2020) and SURFER software (Golden Software, Inc., Golden, CO, USA, Version 21, 2021).

3. Results and Discussion

3.1. Distribution and Leaching Release of Heavy Metal Elements in Soil

3.1.1. Vertical Distribution Characteristics of Heavy Metals in Soil Profiles

Figure 3 shows that the concentrations of As, Hg, Cu, Pb, Zn, Cd, and Cr in rhizosphere soil profiles range from 11.42 to 13.00, 0.02 to 0.04, 22.61 to 28.18, 22.50 to 29.04, 58.50 to 75.88, 0.11 to 0.19, and 69.98 to 74.59 mg·kg⁻¹, respectively. None of these heavy metal concentrations exceeds the limits specified in China’s Soil Environmental Quality Standards (GB 15618-2018) [37]. The distribution of these elements is uneven: except for As, all heavy metals are enriched in the surface soil and show a decreasing trend with depth. Surface-enriched metals are susceptible to inputs from surface pollution sources, and their elevated concentrations likely reflect anthropogenic contributions. Mineral colloids in surface soil can act as carriers for heavy metals, adsorbing and fixing them and thereby reducing vertical migration [38]. With the exception of As, concentrations of heavy metals in the surface soil exceed those in the subsoil. The increase in As and Cr concentrations within the 20–60 cm soil layer indicates downward migration of these elements. The soil pH in the study area is above 7. Under alkaline conditions, many heavy metals tend to be immobilized by forming low-solubility compounds and by adsorption onto clay minerals or iron oxides, which collectively reduce their leaching to groundwater [39]. However, the mobility of As may increase under alkaline conditions, posing a potential accumulation risk in the ecosystem.

Correlation analyses between initial heavy metal concentrations and TOC (total organic carbon), Feo (Oxalate-extractable Fe), Fed (Citrate-Bicarbonate-Dithionite extractable Fe) and Feo/Fed (ratio) were performed and their trends and R² values were examined. Notably, TOC is positively correlated with all heavy metals (As, Hg, Cu, Pb, Zn, Cd, and Cr) (Figure 4). Both Feo and Fed show positive correlations with all heavy metals, whereas the Feo/Fed ratio is negatively correlated, indicating that organic matter and iron oxides are significant factors controlling heavy metal concentrations in soils. Hg and Cd display similar distribution patterns in the profile, but their correlations with iron oxides and TOC are relatively low, suggesting that their migration and enrichment are only weakly regulated by iron oxide adsorption and TOC complexation. Previous studies suggest Hg can bind to methylated organic matter and undergo speciation changes [40], while Cd, with a small ionic radius and low charge density, is more prone to migrate with soil solutions. Cu and Pb show different behavior: the R² of both Cu and iron oxides was greater than 0.5, indicating a strong influence of iron oxides on their soil occurrence. The abundant reactive sites on amorphous iron oxides enable adsorption and co-precipitation of Cu and Pb [41]. However, the R² of TOC is only 0.2–0.3, indicating that TOC has a limited regulatory effect on TOC. This may be related to the high charge of Cu and Pb ions and strong selectivity for complexing with organic matter–only specific functional groups can bind effectively, resulting in a weakening of the overall TOC contribution [42], when remediating Cu and Pb contaminated soil, priority should be given to adjusting the form of iron oxides to achieve fixation. As and Cr exhibit highly similar distribution and response patterns: both correlate with iron oxides (R² ≈ 0.5 for As, R² > 0.8 for Cr) but show very weak correlations with TOC (R² < 0.1). This is consistent with their predominantly anionic forms, which are less prone to complexation with organic matter [43]. Organic matter enhances soil adsorption capacity and can increase heavy metal accumulation in soils with higher organic content [44]. The content and composition of soil organic matter affect heavy metal accumulation and influence migration and transformation through complex formation [45]. Due to their loose structure, large specific surface area, and high reactivity [46], amorphous iron oxides provide more adsorption sites, facilitating heavy metal fixation. The Feo/Fed ratio suggests a greater proportion of amorphous iron oxides, enhancing their ability to adsorb and stabilize heavy metals, which significantly reduces their mobility. Strong fixation of heavy metals in soil diminishes their capacity to migrate to other layers, explaining the negative correlation between the Feo/Fed ratio and heavy metal concentrations. Over time, Feo can gradually convert into crystalline iron oxides. In soils with a lower Feo/Fed ratio, crystalline iron oxides may arise from long-term accumulation, indicating prolonged weathering of the soil environment, which enriches heavy metals. Beyond iron oxides and organic matter, the influence of clay minerals and other soil constituents is also critical to metal behavior. These factors are integrated into the comprehensive discussion that follows.

3.1.2. Heavy Metal Weathering and Leaching Rates

Based on the PROFILE model and the data presented in Table S1, we calculated the weathering and base cation release rates in wheat–corn rotation soil profiles. The release rates decreased in the order of Ca²⁺ > Mg²⁺ > Na⁺ > K⁺ (Table S3). Surface soils exhibited higher weathering and release rates than subsoils, primarily due to greater exposure to the atmosphere and more intense biological and chemical activity. Surface processes such as biological activity, precipitation, and temperature variations promote weathering [38], whereas subsoils, with limited environmental exposure, have reduced biological activity [47], and slower water movement, exhibit slower rates. In comparison to other research areas, the weathering rate in the forests of Pennsylvania, USA, ranges from 0.12 to 9.20 keq/(ha·a) [48], while the northern boundary mountains of Switzerland display a weathering rate from 0.01 to 25.00 keq/(ha·a) [49]. The wheat–corn rotation soil weathering rates in the present study fall within the same ranges.

Using Equation (2), the weathering release rates of heavy metals in the wheat–corn rotation soil profile were calculated, decreased in the order (

Table 1. Leaching rates (Rm) of heavy metals in soil (g/(ha·a)).

		Surface Layer (0–20 cm)			Lower Layer (20–200 cm)
		Minimum	Maximum	Average Value	Minimum	Maximum	Average Value
wheat	As	1.386	6.217	3.685	3.407	33.612	12.742
	Hg	0.004	0.043	0.013	0.004	0.065	0.023
	Cu	4.290	12.889	8.228	6.978	59.615	24.562
	Pb	4.882	15.103	8.447	6.872	58.955	24.359
	Zn	8.827	37.000	22.393	19.236	164.817	64.701
	Cd	0.024	0.141	0.057	0.023	0.380	0.131
	Cr	11.446	34.114	21.263	19.050	177.091	74.364
corn	As	1.711	7.889	4.716	4.397	42.068	15.954
	Hg	0.005	0.056	0.017	0.005	0.081	0.029
	Cu	5.291	16.740	10.521	9.007	74.766	30.731
	Pb	6.268	19.502	10.796	8.871	71.951	30.467
	Zn	10.896	48.057	28.664	24.166	206.704	81.125
	Cd	0.030	0.182	0.073	0.028	0.475	0.164
	Cr	14.111	44.022	27.184	24.589	222.094	93.213

): Cr > Zn > Cu > Pb > As > Cd > Hg, demonstrating significant differences in release rates among various elements. Previous studies in Eastern Finland indicated that heavy metal release rates are substantially dependent on the soil’s base cation content and release rates [50]. The average release rates of heavy metals in surface soil are lower than in deeper layers. This indicates that the effect of soil profile weathering leaching carries cations into the human underlying soil [51]. During the wheat–corn rotation, heavy metal release rates when corn is planted are higher compared to when wheat is planted. This is attributed to the warmer, wetter summer climate during corn growth and its more extensive root system. Root system differences significantly influence heavy metal dynamics: corn’s deeper taproot system efficiently accesses subsoil minerals [12] while secreting more exudates that form soluble complexes with heavy metals, enhancing their mobility [52]. Additionally, in the corn-wheat rotation system, the substantial organic acids secreted by corn roots during its growth season activate heavy metals, while the reduced exudate input and stronger nutrient absorption during the wheat season enhance heavy metal fixation, creating a cyclical activation-fixation process throughout the year [53].

The correlation between heavy metal release rates and soil properties (Table S4) underscores the collective role of soil composition in metal mobility. Soils rich in clay can effectively capture heavy metals and reduce their mobility [54], potentially increasing absorption by wheat and corn. The positive correlations of Feo and Fed with heavy metals further indicate the iron oxides, which may affect the release of heavy metals in soil through complex formation or altering their chemical forms. Meanwhile, the negative correlation with Feo/Fed suggests that heavy metal release during leaching and weathering is predominantly impacted by crystalline iron oxides hydrating into Feo, with Feo mainly adsorbing heavy metals in colloidal forms [55]. Total Organic Carbon (TOC) shows a mitigating effect through complexation, reducing metal release.

3.2. Translocation of Heavy Metals in Soil Profiles

3.2.1. Input and Output Flux of Heavy Metals in Surface Soil

Heavy metals can enter agricultural soils from various sources, including atmospheric deposition, irrigation water, and fertilizers, with distinct pathways for output, such as crop absorption and leaching. The input and output fluxes of heavy metals in the surface soil of the wheat–corn rotation in the study area are calculated (Table S5).

Figure 5 illustrates the relative contributions of inputs and outputs pathways for heavy metals (Hg, As, Cd, Cr, Pb, Cu, and Zn). Atmospheric deposition constitutes the dominant input source for Hg, Cu, Pb, Zn, Cd, and Cr, accounting for 49.50% to 93.27% of the total. In contrast, irrigation water serves as the primary input pathway for As (72.65%), while fertilizers make a minor contribution (2.57–6.73%). Regarding outputs, straw absorption is the major pathway for Hg, Cu, Zn, and Cd in wheat, contributing 50.90–86.31%. For As, Pb, and Cr in wheat, however, weathering predominates (68.15–93.53%). Grain uptake in wheat plays a negligible role (0.60–27.90%). In corn, weathering is the overwhelmingly dominant output mechanism across all metals, with contributions exceeding 90% (93.78–99.53%), while accumulation in stems and grains is minimal. This discrepancy is attributed to differences in the shape and structure of various plant species and their distinct mechanisms for absorbing and accumulating heavy metals, leading to notable disparities in accumulation [56].

This research identifies atmospheric deposition as the primary source of heavy metals in the local agricultural soils, a finding consistent with reports from Heilongjiang and Hunan provinces. However, a related study in Hainan Province found that irrigation water is the primary pathway for heavy metals in agricultural soils [57]. In Poland, the predominant source of heavy metals in soil is the usage of organic fertilizers [58]. Thus, it is evident that the relative contributions of input pathways vary significantly across different regions. Additionally, crop harvesting constitutes a major output pathway for soil heavy metals. Some studies have suggested that crop straw can serve as an economical biological adsorbent for heavy metal-contaminated soil, with soil pH and organic carbon improving through the application of biochar and straw [59]. As a key interface of heavy metal environmental behavior, the weathering leaching of surface soil has a significant impact on the vertical migration of As, Hg, Cd and Pb, because the weathering of the surface soil is actually mainly reflected in downward leaching. Driven by exogenous factors such as tillage, rainfall and changes in oxidation-reduction conditions, heavy metals in the surface soil are easily migrated to the lower layer through the dissolution-leaching process [60]. Currently, considering economic, soil quality, and yield factors, the agricultural return of straw is recommended. Therefore, reducing the external input of metals could reduce the accumulation of heavy metals for a long time and improve the soil’s carrying capacity [61].

The critical loads of the upper and lower soil heavy metals were calculated based on Equations (12) and (13) of the steady-state critical load (SSCL) model, and the long-term variation in critical load of heavy metals in wheat and maize rotation soils could be determined (Figure 6). The critical loads for most heavy metals exhibit an exponential decline over time. After approximately 40–50 years in the surface layer and 55–60 years in the subsoil, the decline plateaus, and the critical loads stabilize. In contrast, the critical loads of As and Cd approach zero much earlier—within 15–20 and 5–10 years for the surface layer, and 25–35 and 15–25 years for the subsoil, respectively. This means that due to the continuous input of heavy metals, the critical load value of heavy metals in the soil will become zero or negative, that is, higher than the maximum carrying capacity of the soil for environmental pollutants, which will cause excessive enrichment of heavy metals and threaten the safety of the farmland ecosystem. This is linked to the long-term, large-scale application of fertilizers, pesticides, manure, and agricultural films in the region, which has already resulted in significant heavy metal accumulation. In the initial stage of the study, the surface soil may have been contaminated with heavy metals to a certain extent, and the accumulation of heavy metals reduces its critical load. The degree of pollution in the lower soil is relatively light, and the accumulation of heavy metals is less, so the critical load is higher, which also reflects the characteristics of heavy metal pollution in the vertical direction of the soil; the surface layer is affected first, and the degree of pollution is relatively deeper and more serious. Heavy metals can persist in the soil for a long time and can even stay in the soil for thousands of years [62]. Therefore, measures to prevent heavy metal pollution in agricultural soils need to be developed, and standards to limit the input of heavy metals into the soil by various anthropogenic routes based on the assessment of critical loads of heavy metals.

3.2.2. Translocation and Accumulation of Heavy Metals in Soil Profiles at Different Depths

The accumulation of heavy metals in surface wheat–corn rotation soil results from atmospheric deposition, irrigation, and fertilizer inputs, as well as outputs from leaching. However, heavy metals in the subsoil primarily include inputs from leaching of the upper layers and outputs to deeper soil from leaching. Therefore, through calculations, the overall accumulation of heavy metals in the entire soil profile can be assessed. Figure 7 reveals a distinct vertical pattern of heavy metal accumulation. In surface soil (0–20 cm), significant positive accumulation occurs, with corn exhibiting a notably higher Zn accumulation. Conversely, during wheat season, Cu and Zn show negative accumulation, indicating a net loss potentially due to plant uptake, harvest removal, and leaching [63]. This negative accumulation phenomenon suggests that trace elements in the soil may be at risk of loss, necessitating the consideration of reasonable fertilization and soil management measures to maintain the balance of Cu and Zn. This negative accumulation trend extends to the 20–120 cm layers; this finding is closely related to leaching effects, root distribution, soil physicochemical characteristics, and external agricultural management practices [64]. However, a re-accumulation peak is observed at 120–160 cm, suggesting temporary retention by clay minerals and colloids [65]. However, this retention in the 120–160 cm layer is temporary. Continued slow percolation of soil solution, combined with the lower adsorption capacity or strong hydrodynamic conditions in deeper layers, can leach heavy metals past the 160–200 cm layer, ultimately resulting in negative accumulation in this layer [66]. Heavy metals may also directly enter deeper layers, even the groundwater system with flowing water, leading to a decrease in heavy metal content in the soil [67]. This layered pattern underscores a continuous leaching risk and the necessity for management practices that mitigate the long-term vertical migration of heavy metals.

3.2.3. Spatio-Temporal Changes in Heavy Metal Concentrations in Soil

Flux-based calculation of heavy metal concentrations (Equations (9) and (11)) demonstrated significant vertical stratification. Compared with prior research, the concentrations of heavy metals in surface soil exceeded local background values [68], they still did not surpass the second-level standards set forth in China’s Soil Environmental Quality Standards, indicating that the soils in the study area are indeed affected by heavy metal pollution. Overall, heavy metal concentrations exhibit an increase at depths of 0–20 cm and 120–160 cm, with a greater increase in concentration observed in the surface layer. In contrast, the concentrations of heavy metals at other depths decrease to varying degrees, with notable changes in Cr concentrations across different soil depths; this may be related to the adsorption and redox reactions of iron oxides [69], which is consistent with the high correlation between Cr and iron oxides in Figure 4. The significant changes in the surface soil are likely attributable to the input of external heavy metals, whereas the changes at other depths may result from leaching effects. Additionally, this may be attributed to the differing adsorption capacities of organic matter and clay minerals for heavy metals [70]. Due to its polyvalent characteristics, Cr may be adsorbed more complexly than other heavy metals, resulting in greater variability in its adsorption, precipitation, and leaching at different depths in the soil profile [71].

The vertical distribution of heavy metals revealed two distinct patterns: Hg, Pb, and Cr were primarily enriched in the surface layer (0–20 cm), while As, Cd, Cu, and Zn exhibited significant accumulation in the middle and deeper layers (20–200 cm). The substantial migration capacity of As and Cd, driven by reductive dissolution of iron oxides and facilitated transport by dissolved organic carbon, poses a particular threat to groundwater, highlighting that pollution risks extend beyond the surface and require layered remediation strategies [72,73]. Long-term projections (2024–2100) reveal a clear vertical redistribution of heavy metals (Figure 8), with surface concentrations decreasing as mid-layer (20–120 cm) accumulation of As, Cd, Zn, and Cu increases. Critically, the significant migration of As, Cd, and Zn into deep layers (120–200 cm) underscores a growing potential threat to groundwater, providing essential evidence for stratified soil remediation and groundwater protection strategies.

3.3. Heavy Metal Concentrations Under Different Climate Scenarios

3.3.1. Leaching and Enrichment of Heavy Metals

Climate change significantly impacts chemical weathering processes, especially through fluctuations in temperature and precipitation [74]. The Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) in 2013 adopted four climate scenarios (RCP) from CMIP5. RCPs represent comprehensive scenarios for emissions and concentrations, serving as inputs for models predicting climate change effects stemming from human activities in the 21st century, on greenhouse gases [75], reactive gases, and aerosol emissions, as well as shifts in atmospheric components associated with future population growth, socioeconomic development, technology enhancements, energy consumption, and land use [76]. Existing research suggests that climate change scenarios affect soil carbon sequestration and other physicochemical properties [77], but studies on heavy metal leaching and enrichment in the crop rotation system are limited under climate change. In this study, the weathering rate of chemical leaching of cations and heavy metals during the wheat–corn rotation from 2024 to 2100 was predicted by temperature and precipitation under RCP4.5.

The calculated results indicate that the release rates of heavy metals under the RCP4.5 scenario are higher than those in 2024, driven by increased temperatures, heavy rainfall, soil acidification, and extreme weather. This accelerated release and migration underscores the need for stringent CO₂ emissions control and low-carbon policies to mitigate the root causes of climate change. Strengthened soil monitoring and management are also essential to address ensuing pollution risks.

Figure 9 shows significant heavy metal leaching at 60–120 cm, corresponding to the pronounced accumulation observed at 120–160 cm in Figure 7. The surface soil (0–20 cm), which is significantly influenced by plant roots and organic matter, rich in organic material, can form stable complexes or chelates with heavy metals, enhancing their fixation in the soil and reducing their migration and release [78]. Due to the presence of clay minerals and organic matter, surface soil exhibits a strong adsorption capacity, effectively stabilizing heavy metals and diminishing their responsiveness to environmental changes. At intermediate depths (20–60 cm, 60–120 cm), soil is considerably affected by precipitation infiltration, with leaching effects owing to moisture movement facilitating heavier metal migration with water flow [73]. Furthermore, the intermediate soil layer may demonstrate elevated hydraulic conductivity, increasing the sensitivity of heavy metal migration and release to climate change [79]; this may also stem from the greater influence of microbial and soil animal activities, fostering the transformation and migration of heavy metals. Deeper soil (120–160 cm) typically experiences more compaction and lower porosity, restricting vertical percolation of moisture and solutes, thus reducing changes in migration and accumulation of heavy metals. Deeper soil minerals are relatively stable after long-term geological processes and thus respond slowly to external environmental changes. The closed deep-soil environment, with its stable chemical and biological processes, also reacts gradually to climate change. In the lowest layer (160–200 cm), leaching can still be significant, potentially due to groundwater influence. Accelerated metal release here may result from the downward movement and subsequent re-release of heavy metals previously adsorbed in overlying soil, leading to increased leaching amounts.

Under the high-emission scenario (RCP4.5), climate change modifies the soil chemical environment by raising temperatures and increasing precipitation, significantly promoting the release and migration of heavy metals (Figure 9). These changes not only affect the exposure levels of heavy metals at plant roots, but also may directly impact the vertical distribution of heavy metals in the soil profile and even in groundwater [80]. The substantial accumulation of heavy metals in the middle layer of the soil profile needs particular attention, as it may become a prospective source for secondary release of heavy metals, especially amid further environmental alterations [81]. Therefore, studying the long-term effects of climate change on heavy metal behavior in soils is crucial for formulating effective land management and environmental protection measures.

3.3.2. Leaching of Heavy Metals Under Extreme Climate Conditions

The increasing levels of carbon dioxide in the atmosphere, global warming, and the enhanced greenhouse effect, along with the growing frequency and intensity of extreme weather and climate change, are now widely recognized to occur more frequently and severely [82]. Climate change will lead to changes in soil quality and physicochemical parameters, such as pH, organic carbon, etc., which will affect the migration activity and distribution of soil heavy metals [83], so it is important to study soil heavy metal leaching and vertical, respectively, in extreme climates.

This study simulated the impact of two extreme climate scenarios on heavy metal release rates in wheat–corn rotation soil. Results show that under high temperature conditions (Scenario A: ≥30 °C with no rain), the release rates of base cations and soil weathering rates increased by approximately 57.90% and 62.69% in the surface layer, and by 52.08% and 57.87% on average in the lower layer, respectively. Under heavy rainfall (Scenario B: ≥25 mm·day⁻¹, approximately 25 °C), the increases were more pronounced, reaching approximately 83.51% and 85.44%, with average lower layer increases of 79.36% and 81.76%. The results demonstrate that both high temperatures and heavy rainfall accelerate soil chemical weathering, with a significantly greater increase under heavy rainfall (Scenario B) than under high temperature (Scenario A). Furthermore, weathering and base cation release rates were higher in surface soil than in lower layers across both scenarios, confirming the dominant role of precipitation [24] and the particular vulnerability of surface soil to climate-driven weathering processes.

Under extreme climate conditions, heavy metal release rates increased significantly at all soil depths, with a more pronounced effect from heavy rainfall than from high temperature. This difference is attributed to the dominant role of leaching and associated physicochemical changes induced by rainfall [84], whereas temperature-driven increases are limited by water availability, consistent with the observed patterns in soil weathering and cation release.

3.4. Transfer Prediction of Heavy Metals in Soil Profile Based on Leaching Model

3.4.1. Multiple Regression Prediction Model

Multiple regression analysis identified Feo, Fed, Clay, TOC, pH, and release rate (Rm) as key factors controlling heavy metal concentrations across soil depths, with all models showing high goodness of fit (Table S6). Multiple regression analysis quantified how key factors influence heavy metal concentrations (CT) variably with depth. In surface soils, larger coefficients for Feo and Fed indicate iron oxides dominate metal through coprecipitation or surface complexation [85]. With increasing depth, the growing influence of TOC—despite its lower absolute content—highlights its heightened role in metal complexation [45], while pH and release rate (Rm) further modulate metal behavior, reflecting the combined impact of initial levels and dynamic release processes.

A comparison of the regression equations for different heavy metals reveals distinct controlling factors. Using Hg and Cu as examples, the large coefficients for RmHg and TOC in shallow layers indicate that Hg migration is highly sensitive to its release rate, while TOC exerts a strong complexation and fixation effect. For Cu, the coefficients for Feo and Fed remain consistently high across all depths, underscoring the persistent influence of iron oxides throughout the soil profile. In contrast, the coefficients for pH and other factors fluctuate with depth, reflecting the varying impact of the soil acid-base environment on Cu’s chemical speciation and the adsorption and desorption process. From the perspective of the overall soil profile, the depth change in each factor coefficient in the multiple regression equation is the result of the interaction between the vertical differences in soil physical and chemical properties and the geochemical behavior of heavy metals (migration, transformation, etc.). These equations not only quantify the influence of each factor on heavy metal content at different depths, but also provide data support for an in-depth understanding of the complex coupling mechanism of soil-heavy metal systems in the vertical direction, which is helpful for the precise formulation of heavy metal pollution prevention and control strategies for different soil depths.

A comparison of the multiple regression equations for corn and wheat soils reveals both universal patterns and crop-specific differences in heavy metal behavior. Iron oxides consistently emerged as a core factor governing the CT of most heavy metals across all depths in both soil types, underscoring their universal role in adsorbing and immobilizing heavy metals across different crop environments. Other factors, including TOC, pH, and clay content, also featured prominently in the regression models for both crops, indicating that their influences on heavy metal content are common and not fundamentally constrained by crop species. Nevertheless, significant differences were observed. Taking CTHg as an example, the coefficients for RmHg and TOC differed between corn and wheat soils. In the shallow corn soil, root exudates may foster a distinct microbial environment that promotes Hg migration and transformation, thereby amplifying the influence of RmHg on CTHg. Wheat roots, by contrast, exert a different influence, leading to a varied impact of RmHg. Furthermore, the decomposition of corn residues—rich in lignin and cellulose—alters the content and composition of soil TOC, enhancing its complexation with Hg and resulting in a larger TOC coefficient in the regression. The return of wheat residues modifies TOC composition differently, changing its complexation strength with Hg and thus yielding a different coefficient. These distinctions highlight how crop-specific root activities and residue return reshape the soil environment, thereby modulating the factors influencing heavy metals—an effect captured in the regression coefficients. These findings are crucial for formulating targeted heavy metal pollution strategies in different cropping systems, emphasizing the need to account for crop type in managing soil-heavy metal interactions.

Monte Carlo simulations were used to evaluate the main factors affecting heavy metal concentrations. Heavy metal concentrations are the result of the interaction of multiple environmental factors (Figure 10), wherein iron oxides and organic carbon play crucial roles in heavy metal fixation, while fluctuations in pH and heavy metal release rates further influence their mobility and bioavailability. This analysis aligns with prior discussions regarding correlations, providing a scientific basis for understanding the forms of occurrence and migration mechanisms of heavy metals in the environment while offering theoretical support for risk assessment and remediation strategies for heavy metal pollution.

The influence of various factors on heavy metal content exhibited distinct stratification with soil depth. For As, the influence of Fed and Feo in the upper soil is large, and the influence of TOC, pH, and other factors increases with the increase in depth. The effects of TOC and RmHg on Hg were significant in the shallow layer, and the effects of pH and Clay were enhanced at the deep level. The proportion of influencing factors of Cu and Pb changed with depth, and Feo and Fed were always the key factors, while the contributions of TOC and pH also changed with depth. The proportion of sensitive factors of Zn, Cd, and Cr was also adjusted with depth, and the influence weights of iron oxides and TOC at different depths fluctuated. On the whole, iron oxides, as the core factor of the influence of heavy metal CT, make important contributions at all depths, while the influences of TOC, pH, clay, and Rm show complex changes with soil depth, reflecting the coupling effect of soil physicochemical properties and occurrence states during the vertical leaching process of heavy metals. For some heavy metals, the sensitivity ratio of each factor varied between crops. For example, the proportion of RmHg and TOC in corn soil was different from that in wheat soil, reflecting the different effects of root activities and residue returns on the soil environment of different crops, which in turn led to different effects of each factor on heavy metals. From the perspective of the specific manifestations of depth differentiation, there were differences in the details of the influence of various factors with depth in wheat and corn soils. For example, the proportion of the influence of Feo on certain heavy metals differed between the two crop soils, which may be related to the vertical differences in soil formation process, root distribution, and metabolic activity of the two crops, reflecting the influence of crop type on the vertical law of soil-heavy metal interaction. In conclusion, the influences of various heavy metal total contents in wheat and corn soils share common core factors, while also exhibiting specificity due to differences in crop types, providing a targeted basis for soil heavy metal pollution prevention and control in different crop cultivation areas.

In the Monte Carlo simulation, the scatter plots of heavy metal concentrations and corresponding Rm at different depths were made (Figure S1), indicating that multiple regression combined with Monte Carlo simulation can effectively characterize the linear correlation between the two, which provides a reliable basis for analyzing the longitudinal migration mechanism of heavy metals in soil.

The high R² values observed for most heavy metals indicate a strong linear relationship between their total concentration (CT) and the release rate (Rm), confirming Rm as a robust predictor for characterizing longitudinal leaching. The slope of the regression line exhibits a gradual increase with soil depth. This trend can be attributed to depth-dependent soil processes: in surface layers, active biological processes and organic matter enhance metal fixation, thereby constraining the leaching potential driven by Rm. In deeper layers, reduced biological activity and progressive saturation of adsorption sites moderately enhance Rm-driven leaching. However, the absence of a pronounced leaching “jump” is likely due to limited disturbance from wheat roots at depth. Biological disturbances of corn roots (such as root interspersion and secretion release) can change soil pore structure and chemical microenvironment, which in turn affects the relationship between Rm and CT. The shallow root system is dense, and the secretions promote microbial activity, which may enhance the fixation of heavy metals and weaken leaching. Deep root systems are reduced, microbial activity is reduced, and heavy metals are more likely to be leached under the Rm drive. This “root-leaching” coupling highlights a need in corn cultivation to monitor mid-depth metal migration. Mitigation strategies could include managing soil organic matter and improving soil structure to suppress Rm-driven leaching, thereby reducing the risks of subsoil contamination and groundwater pollution.

3.4.2. Neural Network Prediction for Heavy Metal Enrichment

The multilayer perceptron (MLP) is a type of artificial neural network model capable of learning appropriate and effective features from complex data to describe the intricate relationships between input and output variables [86]. Compared to traditional predictive methods, MLP has shown higher accuracy for nonlinear issues [87]. This paper uses six variables: Feo, Fed, TOC, Rm, PH, and Clay to make predictions. For the various heavy metals in this study, quantifying identical qualitative variable categories varies, necessitating the establishment of seven distinct neural network models to simulate heavy metal enrichment.

The neural network models demonstrated high predictive accuracy for most heavy metals (Cu, Pb, Cr, Zn, etc.) in both wheat and corn soils, confirming their robustness and generalizability in capturing the complex relationships between soil properties and heavy metal content. As shown in the actual vs. predicted scatter plots (Figure 11), the high R² values (mostly above 0.9) across all heavy metals indicate that the models effectively learned the underlying nonlinear interactions. Furthermore, the fitted slopes for most elements were close to 1, suggesting minimal systematic prediction bias. However, the slope of Hg and Cd is slightly lower, reflecting that the occurrence mechanism of these heavy metals is more complex (volatility of Hg and high mobility of Cd), and there is still room for improvement in the model’s capture of subtle changes.

From the perspective of depth change, the model fitting accuracy of shallow soil is higher than that of the deep layer. This is because shallow soils are affected by tillage and biological activities, and the spatial variability of physical and chemical properties is more easily recognized by the model. However, the deep soil environment is stable, and the factor interaction is more hidden, which increases the prediction difficulty of the model. Cu, Pb and Cr remained at high levels of R² at all depths, indicating that the vertical distribution of these heavy metals was regulated by soil factors with strong consistency, and their occurrence morphology (such as Cu and Pb were strongly adsorbed to clay minerals and organic matter, and Cr was dominated by insoluble oxides) was relatively stable, and the vertical migration resistance was large, and the distribution law was easily captured by the model. However, the accuracy of Hg and Cd deep fitting is slightly lower, indicating that they have stronger vertical mobility, and the morphological transformation caused by deep soil environmental changes is more complex, which increases the uncertainty of model prediction.

The difference between R² in shallow and deep wheat soil was relatively more obvious, while the depth difference in corn soil was relatively flat. This is related to the distribution of crop roots: the root system of wheat is shallow, the shallow soil is strongly disturbed by organisms, the physical and chemical properties are highly variable, and the model is easy to identify. The deep variability is weak, and the accuracy is significantly reduced. The deeper the root system of corn, the stronger the biological disturbance to the deep soil, and the vertical variability of the physical and chemical properties is gentler, so the depth difference in model accuracy is small. For heavy metals with strong migration such as Hg and Cd, the slope change in corn soil model was gentler than that of wheat, indicating that the vertical distribution of these heavy metals in corn soil was more “uniform” affected by crops, and it may be that corn root exudate or microbial activity was more likely to promote its vertical morphological transformation to be stable. The vertical distribution of these heavy metals in wheat soil is more dependent on the soil’s intrinsic physicochemical gradient, resulting in more obvious changes in the slope of the model.

The neural network models of corn and wheat soils provide reliable tools for heavy metal content prediction and can be used to quickly assess pollution risks at different depths. For corn soil, it is necessary to focus on the uncertainty of deep prediction of Hg and Cd, and optimize the model by combining the microscopic processes such as redox and microbial community, in deep soil. For wheat soil, it is necessary to take advantage of the high precision of the shallow layer to strengthen the prevention and control of surface pollution. The differences in soil between the two crops provide a basis for the precise control of soil heavy metal pollution under different planting systems: corn planting areas can affect the occurrence of deep heavy metals by regulating root activities. Wheat planting areas can focus on surface soil management to inhibit the vertical migration of heavy metals. MLP demonstrates flexibility in utilizing various activation functions [88], enabling it to better address nonlinear relationships among multiple parameters. However, the model still requires extensive training and optimization of parameters to achieve improved prediction results.

A standardized importance analysis of different variables for multiple target variables was carried out (Figure S2). Iron oxides have always occupied an important position in the variable importance of most heavy metals, which is closely related to the stability of heavy metals through adsorption and co-precipitation, and their role as a key host carrier of heavy metals is significant regardless of soil depth. For some heavy metals (such as As, Hg, Zn, Cd, etc.), the importance of Rm is clearly reflected at different depths and changes with depth, which intuitively reflects the dynamic influence of the release rate on the total heavy metal content during the vertical leaching process of heavy metals. The importance of TOC fluctuated with depth, and the upper soil was affected by biological activities, and the complexation and fixation effect of TOC on heavy metals was prominent and important. The biological activity of deep soil decreased, and the TOC content decreased; and its effect changed accordingly, but the total heavy metal content could still be affected by complexation with heavy metals. pH affects the chemical morphology and adsorption behavior of heavy metals, and Clay relies on the specific surface area and interlayer structure to fix heavy metals and the importance of these two changes with soil depth, reflecting the influence of soil acid-base environment and vertical differences in particle composition on the occurrence of heavy metals. In predicting soil heavy metal content, this study found that variables such as total organic carbon (TOC) and clay content (Clay) substantially influence model predictions [84].

3.4.3. Random Forest Models for Heavy Metals Accumulation Under Leaching Process

Random Forest (RF) is an ensemble learning algorithm that aggregates various decision trees. Except artificial neural network model, this study utilizes the random forest model to identify the importance of predictive variables. Variable importance reflects the relative contributions of feature variables within the model [89]. The following figure illustrates the variable importance of heavy metals.

Random Forest demonstrated robust variable selection, with importance frequently concentrated on a few key drivers. Specifically, Fed and TOC were identified as critical factors for most heavy metals in both corn and wheat soils (Figure 12). Fed immobilizes metals through adsorption and co-precipitation, while TOC regulates metal activity via complexation with functional groups—a common mechanism across cropping systems. The importance of these factors, however, varied with depth and crop type. The influence of Rm increased in deeper soil layers, where a more stable redox environment accentuates its role in controlling metal speciation. The shallow layer is strongly affected by human interference, such as tillage and fertilization, and the influence of Fed and TOC is more significant. The importance of Fed to As, Cu, Pb, and Cr in corn soil was generally lower than that of wheat soil, indicating that wheat root activity or residue decomposition had a stronger activation/fixation effect on Fed, which in turn had a more significant impact on the occurrence of heavy metals. The importance of TOC for Zn and Cd in corn soil was higher than that of wheat, indicating that the accumulation or turnover of soil organic matter in corn planting areas was more active, and the complexation of Zn and Cd was more prominent. In conclusion, the concentration of heavy metals in corn soil is jointly driven by Fed, TOC, and Rm, and there are crop-specific differences between the driving effect and wheat soil, which provides a target for the precise regulation of heavy metals in corn planting areas.

Taken together, multiple regression, neural network, and random forest models all showed good prediction performance for heavy metal concentrations in wheat and corn soils, among which neural networks and random forests could accurately capture nonlinear and multi-factor interactions, and the prediction accuracy related to heavy metal release rate (Rm) in wheat soil was slightly better. From the perspective of variable importance, Fed and TOC were the common core drivers of heavy metals in soil of the two types of crops, but there were crop-specific differences: the regulation of heavy metals such as As and Cu by Fed was more prominent in wheat soil, and the synergistic effect of TOC on Zn and Cd in corn soil was stronger. The release rate of heavy metals was positively correlated with the concentration and was nonlinear and the release was slow due to the constraints of the soil matrix at low concentrations, and the release was accelerated due to the high proportion of active forms at high concentrations. In the longitudinal dimension, shallow soil was affected by anthropogenic input and soil matrix fixation, heavy metals were mainly enriched, and leaching migration was weak. Due to the reduction in substrate content and large hydraulic gradient, leaching migration is dominant, and the improvement of pore structure by deep root activities of corn further enhances the leaching dynamics and the high content of deep Fed in wheat is more conducive to fixing heavy metals such as Pb. These differences stem from the different root characteristics of crops and soil management methods, which provide a basis for layered prevention and control: the shallow focus is on the fixation of heavy metals by Fed and TOC, the deep strengthening of soil structure improvement to slow down leaching, while wheat focuses on Fed activation management, and corn focuses on TOC turnover optimization.

It is evident from Figure S3 that the random forest model shows good fitting ability in all heavy metal predictions, with the R² of most heavy metals above 0.95, demonstrating the strong ability of the random forest model in modeling nonlinear relationships between multiple variables. Random Forest not only possesses strong predictive performance but also demonstrates good robustness and generalization capabilities, making it suitable for environmental risk assessment research focused on identifying pollution factors and estimating concentrations of heavy metals [90].

The high R² values for most heavy metals confirm the excellent predictive accuracy of the random forest model in estimating their concentrations its capability to capture complex soil factor heavy metal relationships. However, the R² of Hg and Cd is slightly lower than that of other heavy metals, reflecting that the occurrence mechanism of these heavy metals is more complex (such as the volatility of Hg and the high mobility of Cd), and the model has certain challenges in capturing their subtle changes. The slope of the fitted line is close to 1, indicating that the deviation between the predicted value and the actual value is very small, and there is no obvious systematic deviation of the model.

From the perspective of depth change, the model fitting accuracy of the deep soil is slightly lower than that of the shallow layer. This is because the spatial variability of deep soil physical and chemical properties is weaker, but the interaction between factors is more hidden and less interfered with by external interference, and the occurrence of heavy metals depends more on the subtle differences in soil intrinsic properties, which increases the difficulty of model prediction. Cu, Pb, and Cr remained at high levels of R² at all depths, indicating that the vertical distribution of these heavy metals was consistently controlled by soil factors. Its occurrence morphology (such as Cu and Pb are strongly adsorbed to clay minerals and organic matter, and Cr is dominated by insoluble oxides) is relatively stable, the vertical migration resistance is large, and the distribution law is easy to capture by the model. However, the deep fitting accuracy of Hg and Cd is slightly lower, indicating that their vertical migration is stronger, and the morphological transformation caused by deep soil environmental changes is more complex, which increases the uncertainty of model prediction.

The partial depth R² of Hg and Cd in corn soil was slightly higher than that in wheat soil, indicating that the occurrence and migration mechanism of Hg and Cd in corn soil was relatively easier to capture by the model. The difference between R² in shallow and deep wheat soil was relatively more obvious, while the depth difference in corn soil was relatively flat. This is related to the distribution of crop roots: the root system of wheat is shallow, the shallow soil is strongly disturbed by organisms, the physical and chemical properties are highly variable, and the model is easy to identify. The deep variability is weak, and the accuracy is significantly reduced. The deeper the root system of corn, the stronger the biological disturbance to the deep soil, and the vertical variability of the physical and chemical properties is gentler, so the depth difference in model accuracy is small. In conclusion, the prediction performance of the random forest model on the heavy metal concentration of corn soil is good, and the difference between the model accuracy and depth response of wheat soil reflects the profound impact of crop types on the soil-heavy metal system, and provides scientific support for targeted pollution prevention and control.

According to the random forest model, the heavy metal concentrations in wheat and corn soil from 2024 to 2100 were predicted and compared with the actual values (Figure 13), and the soil at different depths showed obvious differences.

In the shallow soil of 0–20 cm, the actual and predicted concentrations of heavy metals in the soils of the two types of crops showed an overall upward trend, which was related to the fact that the shallow layer was more affected by human agricultural activities, and the heavy metal input and enrichment continued. In the deeper soils of 20–60 cm and 60–120 cm, the concentration changes are relatively flat or even have a downward trend, indicating that the deep soil is less disturbed by external interference, and the migration and transformation of heavy metals may be relatively stable due to the adsorption and fixation of the soil itself. Compared with wheat and corn soils, the rate of heavy metal concentrations in shallow wheat soil was slightly faster than that of corn, which may be due to the more significant impact of root activity and residue return on the morphology and distribution of heavy metals in shallow soil during wheat planting, which promoted the enrichment of heavy metals. The predicted concentration fluctuations of heavy metals (such as CTCd and CTCr) in the deep soil of corn were more obvious, or related to the disturbance of the deep soil environment by the deeper roots of corn, which changed the occurrence state of heavy metals. On the whole, the random forest model can better capture the changing trend of heavy metal concentration in soil at different depths and crops, which provides a prediction basis for the future prevention and control of heavy metal pollution in agricultural soil.

4. Conclusions

• In the wheat–corn rotation soil in the study area, atmospheric deposition is the main source of input for heavy metals such as Hg, Cu, Pb, Zn, Cd, and Cr in surface soil; irrigation water is the primary source of input; and fertilizers contribute relatively little to soil heavy metals. Straw absorption is the primary output pathway for heavy metals such as Hg, Cu, Zn, and Cd in wheat, while weathering is the dominant output pathway for As, Pb, and Cr in wheat, with a low contribution rate of grains in wheat to the output of heavy metals from soil. In corn, weathering is also the primary output pathway, consistently contributing above 90%, with minimal accumulation occurring in the stems and grains. Consequently, corn is potentially safer than wheat regarding heavy metal pollution, although its remediation capabilities are relatively weak.
• This study integrates the Steady-State Critical Load (SSCL) model with an analysis of spatiotemporal dynamics to systematically reveal the risks and evolutionary trends in heavy metal pollution in the wheat–corn rotation system. The SSCL model predicts that the soil critical loads for As and Cd are projected to approach zero within just 15–20 years and 5–10 years, respectively, signaling that external inputs will soon exceed the soil’s maximum carrying capacity, posing an extremely urgent ecological risk. Concurrently, spatiotemporal simulations of heavy metal concentrations confirm that these readily overloaded elements (particularly As and Cd) are undergoing significant vertical migration from the topsoil, showing pronounced accumulation in the middle layer (20–120 cm) and demonstrating a clear tendency to penetrate deeper layers (120–200 cm), even threatening groundwater.
• This study demonstrates that climate change is a critical amplifier of heavy metal release rates (Rm). Under the RCP4.5 scenario, the overall Rm is projected to be 1.2–1.5 times higher than current levels. Extreme weather events, particularly heavy rainfall (≥25 mm/day), drastically accelerate this process, increasing weathering and cation Rm by over 80% in surface soils—a far greater effect than high temperature alone. This climate-induced enhancement of Rm directly governs the increased vertical migration of heavy metals, elevating long-term ecological risks. To predict these complex dynamics, machine learning models were employed. The Random Forest model achieved exceptional accuracy (R² > 0.95), with Rm itself emerging as a variable of high importance in predicting future heavy metal concentrations and behavior. This underscores Rm as a pivotal, quantifiable link between climate forcing, geochemical response, and predictive environmental risk assessment. Finally, our study supports the insightful perspective that the unsuitability of staple crops for phytoremediation is ultimately beneficial for food security, directing future focus toward preventative measures.