Heavy metal pollution and ecological risk under different land use types: based on the similarity of pollution sources and comparing the results of three evaluation models

In key areas of ecological protection, it is significant to consider the similarity of pollution sources among heavy metals and the interaction between different sources, especially the ecological risk areas caused by heavy metal pollution. We collected 51 soil samples from five land use types with different soil depths in an industrial area on the Qinghai-Tibet Plateau. Two and three major heavy metal combination types of Cd, Cu, Cr, Pb and Zn in different soil layers were identified using absolute principal component score-multiple linear regression models, and the potential pollution sources corresponding to the different types were quantified using Geo-Detector models. Factor-detector explanatory power of the land use type (q = 0.66) was much higher than that of the other factors of APCS1 in soil layer A, which was the most likely potential sources of Cd and Pb with high levels in urban land and irrigated land. Industrial activities, especially metallurgy and mining, are the most likely potential sources of Cd, Cu and Pb pollution. The downward migration of heavy metals in the study area was inferred from the similar trends of several indicators between soil layers A and B. The new model Nemerow Integrated Risk Index (NIRI) was used to analyse the integrated ecological risk across the study area and under different land use types by comparing with the pollution load index and Nemerow Integrated Pollution Index, and it was found that the risk level was lower in grassland and forest land than under other land use types, while it was higher in urban land and irrigated land. The contribution rate of Cd to NIRI values exceeded 80%, while the contribution rates of the 5 heavy metals to NIPI and PLI values are not significantly different, indicating that NIRI can highlight the impact of high cadmium toxicity factors on the overall risk level and is more accurate and flexible in identifying risk areas.


Introduction
As a significant natural and land resource, soil serves as a crucial building block of the human agricultural output and a priceless resource for human survival and development.Soil also offers a significant and comprehensive ecosystem service base for these purposes (Dominati et al. 2014).Over time, under the background of industrial development and rapid economic growth, the problem of heavy metal pollution in soil is very serious in many areas of the world (Khademi et al. 2019;Pyatt et al. 2002;Rawat et al. 2009).
Problems such as the unreasonable use and high-intensity exploitation of land resources, soil pollution originating from industrial sites (Zhang et al. 2022b), soil pollution in industrial land (Chen et al. 2022) and water pollution (Huang et al. 2020) are also becoming increasingly significant.An excessive content of heavy metals in soil can not only damage the soil structure and reduce the soil quality but can also further increase the ecological risk to animal and human health (Zhang et al. 2019).Some recent studies were not only limited to the concentration of heavy metals in soil, but also focused on the effects of heavy metals on soil biology, soil basic respiration and soil enzyme activities (Hong et al. 2021), and also evaluated the risk of heavy metals to human beings by studying the migration of heavy metals in vegetables and food crops (Singh et al. 2023).
In addition to traditional multivariate statistical methods, more established methods for heavy metal pollution source analysis include principal component analysis (PCA) (Anaman et al. 2022), the positive matrix factorization (PMF) model (Lv and Wang 2019), the geographically weighted regression (GWR) model (Yang et al. 2022) and other methods.In addition, machine learning methods have good predictive results for analyzing heavy metal migration pathways (Wang et al. 2023).It includes support vector machine, random forest(RF) and artificial neural network (Tan et al. 2020).If we wish to analyse the heavy metal content in an area at depth, spatial correlation and simple source identification are not enough.
The absolute principal component score-multiple linear regression (APCS-MLR) model is a receptor model based on PCA with backward traceability (Wang et al. 2022;Yang et al. 2020).PCA is a more mature and common multivariate mathematical analysis method, and the APCS-MLR model can be used to obtain APCS values based on PCA after standardization of raw data, which can then be combined with the MLR model to calculate the contribution of different APCSs to each soil heavy metal index (%) (Wang et al. 2022;Yu et al. 2022) and thus to obtain the main types of heavy metal pollution combinations.To further quantify and express the correlation between each heavy metal combination type and potential pollution sources to identify the drivers of heavy metal accumulations, the Geo-Detector model (GDM) was selected (Huang et al. 2021;Qiao et al. 2022) to further identify the potential pollution sources and reduce the subjectivity of relying on PCA/APCS results alone (Zeng et al. 2022).
The complexity of the ecosystem risk has driven the development of various modelling approaches (Zerizghi et al. 2022).In terms of the evaluation of the environmental quality and ecological risk under the influence of heavy metals, the single factor pollution index (PI), geo-accumulation index (I geo ) (Hu et al. 2021), pollution load index (PLI) (Demkova et al. 2017), Nemerow Integrated Pollution Index (NIPI) and potential ecological risk index (RI) (Chen et al. 2016) are widely used (Men et al. 2018), which can analyse the pollution level of each heavy metal and the comprehensive environmental and ecological effects of various heavy metals to produce relatively objective evaluation results.Wang et al. (2021) used the single factor index (PI) to evaluate the status of heavy metal As; Xu et al. (2021) evaluated the pollution and risk assessment of two smelter sites by Nemerow integrated pollution index and showed severely polluted.However, the NIPI may overly highlight the impact of heavy metals with the highest PI values on the environmental quality and may artificially exaggerate or reduce the impact role of some factors in the evaluation process, making this index insensitive to environmental quality evaluation.In some cases, the NIPI calculation results can hardly distinguish the soil environmental pollution degree.Some scholars have tried to improve the traditional models, such as modifying the potential ecological risk index (MRI) through the I geo (Liu et al. 2021).Therefore, in this study, an improved risk assessment method, i.e., a new integrated ecological risk assessment method referred to as the Nemerow Integrated Risk Index (NIRI), which can reduce the influence of the amount of heavy metals involved in the calculation on the objectivity of the evaluation results, was selected (Men et al. 2020).This method had used to identify the key sources of heavy metals in road dust and could provide a more comprehensive assessment of the integrated ecological risk factors of soil heavy metals in the study area.
The Qinghai-Tibet Plateau is the source region of many rivers in Asia and is also an extremely important area for ecological protection (Li et al. 2020) and has been compared to the Asian Water Tower.Scholars have found that heavy metal pollution of soil also occurs on the Qinghai-Tibet Plateau (Huang et al. 2008;Wu et al. 2018), even radioactive contamination has been detected (Zhao et al. 2022).With policy and regional economic development needs, in recent years, industrial parks have been established in the main gathering areas of humans on the Qinghai-Tibet Plateau, especially in the eastern part of Qinghai Province, to meet industrial clustering and industrial upgrading requirements.Li et al. (2018) used I geo and NIPI to study the north-eastern part of the Qinghai-Tibet Plateau and found that mining area was the region with the most serious pollution; Zhang et al. (2022a) studied on heavy metals in arable soil in a region of the Qinghai Tibet Plateau and they found that Cd contamination was caused by agricultural and transportation activities and Cr, As and Pb were also with high risks; Wu et al. (2018) found in their study on the ecological risks caused by heavy metals in the northeast of the Qinghai Tibet Plateau that industrial source contributed to the main fraction of ecological and health risks and mainly caused by Hg.These study results proved that the Qinghai-Tibet Plateau has indeed been polluted by industry.Therefore, the comprehensive quantitative evaluation of soil heavy metal pollution levels and the identification of soil pollution sources in key pollutionprone areas in the north-eastern Qinghai-Tibet Plateau are of great theoretical and practical significance for the prevention of soil heavy metal pollution, improvement in the soil environmental quality and targeted control of key polluting enterprises and provide some reference for pollution prevention and control of industrial development in ecologically fragile areas.In particular, it is important to capture the differences in the extent of pollution of different heavy metals in different land use types and the integrated ecological risks they pose.
Meanwhile, the emissions of other heavy metals, such as Zn Cu, during agricultural fertilization and metal smelting processes, so in this study, five heavy metals, namely, Cd, Cr, Cu, Pb and Zn, were analysed in soil samples collected within a typical industrial area in the north eastern part of the Qinghai-Tibet Plateau (1) to analyse the content and spatial pattern of each heavy metal in the study area at different soil depths (layers A and B) by using the inverse distance weight (IDW) method based on the ArcGIS10.6platform; (2) then, the receptor model (APCS-MLR model) and a quantitative analysis model (GDM) were used to identify the main types of heavy metal pollution combinations in the study area and quantitatively analyse the potential sources of heavy metals corresponding to the different combination types; and (3) finally, the PIL, NIPI and a new improved ecological risk assessment method, the NIRI, were used to evaluate the comprehensive ecological risk situation of soil attributed to heavy metals in the overall study area and under the different land use types, calculate the contribution of each heavy metal to each evaluation model, and analyse the advantages of the evaluation model in identifying risk areas and risk pollution factors.
The above work aimed to comprehensively evaluate the reality of soil pollution caused by heavy metals, maintain the quality of the soil environment and ecological environment, promote the targeted treatment of key polluting enterprises and the green and high-quality development of other industrial area with the key ecological protection needs such as Qinghai-Tibet Plateau, and provide a practical and invaluable basis for the soil and ecological protection.

Study area
The study area is located in the eastern part of the Qinghai-Tibet Plateau, a typical industrial area (range: 36.715667°-36.943608°N,101.554588°-101.805370°E)(Fig. 1) in a valley, belonging to the Qilian Mountains in eastern Qinghai Province, with an average elevation ranging from approximately 2210-3900 m.The multiyear average temperature is approximately 4.9 °C, and the precipitation reaches approximately 520 mm.The main soil type in the study area is black calcium soil and chestnut calcium soil.Heavy metal pollution, including zinc, lead and cadmium pollution, may be caused by some industries in this area (Khademi et al. 2019).In addition, there are mining, farming and agro-food processing industries, which together constitute an industrial cluster.

Sampling and analysis
The sampling sites were set up according to the principle of a uniform distribution, and the density was appropriately increased around the key polluting enterprises and should further consider the need for sampling in areas with different land use type (Fig. 1).Fifty-one soil samples were collected in the study area in spring (March to May), in which each soil sample included two subsamples at different sampling depth: topsoil (layer A: 0-20 cm) and middle soil (layer B: 20-40 cm).Each sample was stored in a foam box or sample box containing ice bags and sent to the laboratory on the same day of sampling to assess the relevant parameters.
After preliminary processing, such as debridement and sieving, the collected soil samples were treated according to Chinese national standards HJ/T 166-2004 Technical Specification for Soil Environmental Monitoring, NYT 391-2021 Environmental Quality of Green Food Production Land, and Soil Environmental Quality Soil Contamination Risk Control Standards for Agricultural Land (Trial) (GB 15618-2018), and the sample pre-treatment technique involving nitric acid ?hydrochloric acid ?hydrofluoric acid supplemented by microwave digestion was used to determine the five heavy metals of Cr, Cd, Cu, Pb and Zn in the collected soil samples via inductively coupled plasma optical emission spectrometry (ICP-OES).Soil organic matter (SOM) and organic carbon (SOC) were determined via the potassium dichromate (K 2 Cr 2 O 7 ) volumetric method (hydration heat method).The soil pH was determined using water as a leaching agent with a water-tosoil ratio of 2.5:1, followed by composite electrode measurement.Other soil physical and chemical properties (total nitrogen (TN), total phosphorus (TP), total potassium (TK) and bulk weight) were determined in accordance with the Technical Specification for Soil Analysis.
The spatial pattern of the content and heavy metal combination type, integrated pollution risk and ecological risk, were obtained via the IDW interpolation prediction method based on the ArcGIS 10.6 platform.We used the GDM based on R language to identify the optimal parameters (intermittent methods and intermittent intervals) for continuous-type variables such as topographic factors, soil indicators, and distance factors to obtain the maximum q-statistic value and thus fully reflect the explanatory power of the influencing factors.
The spatial distribution maps of heavy metals content were created through the ArcGIS pro platform and the statistical results figures were drawn by Origin 2021.

Types of heavy metal combinations of similar pollution sources
The APCS-MLR model calculation process includes 6 steps: Step 1: Five heavy metal contents were converted into a standardized form.
where C i and C i are the measured and mean value of heavy metal i, respectively, and r i is the standard deviation of the content of heavy metal i.
Step 2: PCA was performed of the standardized soil heavy metals.
where j = 1…, p denotes the sources, and g j and h ij are the factor loading and factor score, respectively.Factor extraction with eigenvalues [ 1 was employed after varimax rotation.Z i must be converted into unstandardized absolute principal component scores (APCSs) before they can be used to analyse the contribution of the PCs to each heavy metal.Next, the APCSs could be calculated.
Step 3: An artificial sample of heavy metal i accumulated to 0 was artificially introduced at the observation sites and standardized.The standardized value of 0 can be calculated as follows: where Z i0 is the standardized value of 0, C i and r i denotes the same meanings as Eq.(1).After calculating the standardized value of 0, i.e., Z i0 is attached to the standardized data of Step 1, PCA was again performed.
Step 4: The APCS value for each heavy metal at the different positions was obtained by subtracting the factor score from the initial Z i value, and Z i0 was subtracted from the factor score (artificial samples).The APCS can be calculated as follows: where Z i is the same as that in Eq. ( 1), Z 0 ð Þ i is the PCS value at a value of 0,S i is the factor score coefficient, and Z i0 is the zero-valued standardized pollutant concentration at the observation sites.
Step 5: The contribution of heavy metal combination type k to the accumulation of heavy metal factor i was determined.Adopting the measured heavy metal content C i as the dependent variable and the APCS as the independent variable, multiple linear regression (MLR) fitting was performed, and regression coefficients were obtained.Then, the linear relationship between heavy metal content C i and the APCS under combination type k (APCS ik ) can be expressed as: where b ki is the coefficient of MLR, b i0 is a constant term, and b ki Á APCS ik is the contribution value of combination type k to heavy metal i.
Step 6: The contribution rate PC ki of combination type k to heavy metal factor i can be calculated as follows: where APCS ik is the mean APCS value of all samples of heavy metal i.
All the statistical figures of the results drawn by Origin 2021.

Geo-detector model for quantitative analysis
We choose two models, the factor-detector and the interaction-detector in the Geo-detector model, to further analyse potential pollution sources of heavy metal combination types.The APCS was chosen as the dependent variable, and environmental influencing factors were selected as independent variables.With reference to the results of previous studies and combining the actual conditions of the study area and sample selection, thereby focusing on the exploration of the influence of factors such as industrial activities, we selected 17 potential pollution sources as indicated in Table A1, to match several combination types identified through the APCS-MLR model (considering the common characteristics of waste and wastewater treatment plants (Pan et al. 2021;Yuan et al. 2022), we combined them into one potential pollution source), of which the distance factor was selected as the shortest distance between a given sampling site and the potential pollution sources, and the soil and land use types were assessed and recorded in the field.
Factor and interaction-detector methods were selected to calculate the driving force (q statistic) of the selected driving factors and potential pollution sources, respectively, on the APCS, and the factor-detector model can be expressed as: where q x is the explanatory power of the potential sources for the independent variable, with q [ (0,1) and h = 1, 2, 3…, L, and L is the number of levels of factor x; N and N h are the study area and the number of samples in each stratum h, respectively; and r 2 and r 2 h are the variance in Y in the whole study area and each level h, respectively.
The interaction detector determines whether there exists an interaction between any two of the potential pollution sources.As such, we determined whether the joint effect of the variables X 1 \ X 2 on the distribution of the soil heavy metal content was enhanced or diminished or whether these variables were independent.The calculation principle entails the calculation of q-statistic values of the effect of variables X 1 and X 2 on the soil heavy metal content, i.e., q(X 1 ) and q(X 2 ), respectively, calculation of the q-statistic value of their interaction, q(X 1 \ X 2 ), and comparison of q(X 1 ), q(X 2 ) and q(X 1 \ X 2 ).The calculation results could indicate the following five modes: The calculation of the Geo-detector model was used by GDM package in R studio, which can select and execute the optimal discretization method of independent variables, automatically select the discontinuous method, and obtain the optimal parameters of q statistic results.

Integrated risk index evaluation model
In this study, a new integrated ecological risk factor evaluation model, the NIRI, was established, and the corresponding evaluation results were compared to those of the PLI and NIPI.
The PLI is mainly used to identify the combined pollution due to all heavy metals which are influence the soil pollution risks.
where n is the quantity of heavy metals;C i and C b are the content and background value of heavy metal I, and P i1 … P in denotes the pollution factors of the individual heavy metal pollutants.The PLI can be used to evaluate the quality of the soil pollution more directly and intuitively.
The NIPI can be calculated as follows: where P n is the Nemerow composite P i value and P i can be calculated via the same procedure as that expressed in Eq. ( 10).P iave and P imax are the mean and maximum values of P i , respectively.NIRI not only refers to the algorithm principle of the NIPI and RI but also captures the most important feature of introducing the degree of the toxicity of the different heavy metal factors, thus preventing extreme evaluation results due to the quantity of heavy metals.The NIRI can be expressed as follows: where E i is the potential ecological risk factor for heavy metal i, and T i , in this study, were chosen as 30, 2, 5, 5, and 1, respectively (Fang et al. 2019) for the toxicity of Cd, Cr, Cu, Pb and Zn; C i is the same as Eq. ( 10), and S i is heavy metal reference level.E imax and E iaverage are the maximum and average values of E i , respectively.The level of PLI, PI and NIRI are shown in Table A2.
To compare the contributions of the different heavy metals to the above two risk evaluation models, the following equation could be used for calculation.
where W P i and W E i are the contribution rates associated with the NIPI and NIRI, respectively, of heavy metal i, and n is the number of heavy metal factors.
The spatial distribution map of 3 evaluated models were created through the ArcGIS pro platform and the statistical results figures were drawn by Origin 2021.
(possibly natural and industrial sources), and the distribution of the high-value areas of the soil Cu content in soil layer B was highly consistent with that in soil layer A. The spatial pattern of the Pb content in soil layer A was very consistent with that of Cd, which is located along the rivers, and the overall content level in the southern part of the study area was higher than that in the northern part, while the high-value areas of the Pb content in soil layer B were more obviously concentrated along the rivers in the central part of the study area.The spatial distribution of Zn in soil layers A and B was very consistent, mainly in the central and southern parts of the study area, but the spatial continuity of the high-value areas was not notable.The spatial continuity of the high-value areas was not considerable, considering that Pb may be affected by the combined effect of multiple pollution sources.3.2 Types of heavy metal combinations based on similar pollution sources

Correlation analysis between the soil heavy metals
According to the Pearson correlation coefficient results (Tables A4 and A5), there existed a significant positive correlation (P \ 0.01) among Cd, Cu, Pb and Zn in soil layer A, demonstrating that there could be a high overlap in the source pathways of these four heavy metals and that Cr was significantly correlated only with Pb and Zn (P \ 0.05) but not with the other two heavy metals, indicating that the sources of Cr and Cd and Cu may differ (Liang et al. 2017).The significant positive correlations between Cr and Pb and Zn and between Cu and Pb and Zn in soil layer B suggest that Pb and Zn in soil layer B are homologous and that their source pathways overlap with those of Cr and Cu, respectively.However, the sources of Cu and Cr, Cd and the other heavy metals in soil layer B may vary.In addition, we calculated the correlation between the heavy metals and different soil indicators in soil layers A and B (Tables A4 and A5).The results showed that in soil layer A, Cd, Cr and Pb were significantly negatively correlated with the pH range, which suggests that higher soil acidification may increase the soil heavy metal mobility between soil crops and the soil environment and reduce the soil concentration of heavy metals (Wen et al. 2018).Other soil indicators, such as SOM, could affect the accumulation of soil heavy metals (Mahmoodi et al. 2016), such as the correlation between the soil Cd level and soil capacitance (positive correlation) and between Cr and the cation exchange capacity (CEC) (negative correlation) observed in this study for soil layer A, but the correlation between the soil indicators and heavy metals was not significant in soil layer B. Therefore, we could tentatively determine that the soils in both soil layers A and B within the study area may have been polluted via the exogenous input of pollutants.

Heavy metal combination types identification based on the APCS-MLR model
Before APCS analysis, PCA is needed, while the PCA results (Table A6) can be combined with further identification and comparative analysis of the obtained heavy metal combination type results.A7 and S8), respectively, indicating that the extracted PCs could present most of the sample feature information.The results for soil layer A showed that PC1 mainly explained the sources of Cd, Cu, Pb and Zn, while PC2 explained the sources of Cr and Zn.The results for soil layer B indicated that PC1 mainly explained the sources of Cr, Pb and Zn, PC2 largely explained the sources of Cd, and PC3 mainly explained the sources of Cu, and the PCA results were basically consistent with the correlation analysis results.
Based on the PCA results, we used the APCS-MLR model to quantify and express the contribution of each heavy metal combination type to each heavy metal in soil layers A and B. R 2 was used to evaluate the fitting effect of the linear regression models, all of which revealed highly satisfactory results, verifying that the model calculation results are very reliable (0.875 C R 2 C 0.974).Therefore, according to the equations in Sect.2.5, it could be concluded that there were two and three major heavy metal combination types corresponding to soil layers A and B, respectively (Fig. 3).
The results for soil layer A showed that APCS1 contributed extremely high levels to Cd (79.49%) and Pb (76.30%), respectively, and 48.70% to Cu, and the high APCS1 values were more notably concentrated in the south of the area and along rivers (Table A2), which is similar to the contributions of Cr (in the centre of the area) and Cr (concentrated in the north of the area).In addition, wastewater, waste gas, slag and coal combustion fumes stemming from mining and smelting enterprises may be the sources of Cd and Pb in soil (Yuan et al. 2019).Therefore, APCS1 could be considered a heavy metal combination type related to industrial activities.APCS2 achieved a high contribution (40.51%) mainly to Cr.However, the mean value of Cr in the study area was much higher than the background value of Qinghai Province.We could consider APCS2 a source of pollution related to natural factors, but its contribution to Cr was less than 50%, and the source of Cr in the study area might be jointly influenced by multiple sources of pollution, as most of our sampling sites occurred close to industrial enterprises, roads and along rivers.Similarly, the contribution of both APCS1 and APCS2 to Zn was not high (26.71%and 21.65%, respectively), although it could explain the source of Zn to a certain extent.This also indicates that there must be other influencing factors of the Zn content.However, the spatial pattern of Zn in soil layer A within the study area showed that compared to the other factors, the high-value areas of its content were mainly concentrated along the lower reaches of the river, but since Zn is the least notable polluted metal, it is less influenced by pollution sources directly related to industrial activities, considering that it may be due to the indirect accumulation of heavy metals in the process of migration with rivers, roads, etc.
Since the number of heavy metal combination types in soil layer B was larger than that in soil layer A (a difference of one), the contribution rates and results for the heavy metals based on the heavy metal combination types in these two soil layers obtained with the APCS-MLR model also differed, among which APCS1 in soil layer B mainly contributed to Cd, Pb and Zn, with contribution rates of 79.09%, 59.71% and 47.37%, respectively.Although the depth of soil layer B is greater than that of soil layer A, the mean values of the heavy metals in soil layer B exceeded the background values, and Cd is a typical factor of exogenous input, indicating that the heavy metal content in soil layer B was also very likely influenced by human activities.APCS2 mainly contributed to Cr at a higher rate of 40.51% and could be combined with the spatial pattern of Cr in soil layer B, which is very consistent with the PCA results and similar to the Cr analysis results for soil layer A. The contribution of APCS2 to Cu was relatively high, but it reached only 35.68%, based on the high value of Cu in soil layer B, which was mainly concentrated in the central and western arable areas (Chen et al. 2018).Moreover, scholars have proposed that Cu may be influenced by factors such as road and traffic emissions (Zeng et al. 2022), and the study area exhibits the possibility of downward migration of heavy metals from the soil surface.It could be tentatively ascertained that APCS3 may be related to agricultural activities and roads and other factors in terms of the heavy metal combination type.
We initially screened several heavy metal combination type categories, including industrial activities (mining, metallurgy, chemical industry, etc.), roads, rivers, agricultural activities, and natural factors.This identification of heavy metal combination types tentatively determined the potential pollution sources, but the influence of the pollution sources on the soil heavy metals and a quantitative expression of the interaction effect of multiple factors on these heavy metals should still be quantified.The GDM was used for further analysis and discussion.

Quantitative analysis of potential similar pollution sources for heavy metal combination types
It was generally expected that a polluted landscape might correspond to only one or a few specific pollution sources, but the above results in this study showed that there might occur more than a limited number of similar sources of the multiple heavy metals under the same heavy metal combination type.
Factor and interaction detectors were used to analyse the degree of influence of the selected potential pollution sources on the two and three heavy metal combination types in soil layers A and B, respectively, and the results were quantitatively expressed.Figure 4 shows the factor detector results because we aimed to identify the potential pollution sources that were strongly correlated with the heavy metal combination type, and only significant (P \ 0.05) pollution sources were therefore considered.
The factor detector results for the heavy metal combination type factor in soil layer A showed that the explanatory power of the land use type and slope were highly significant (P \ 0.01) for APCS1, and the explanatory power of the slope and elevation were also significant (P \ 0.05) for APCS2.For APCS1, the factordedtector explanatory power of the land use type (q = 0.66) was much higher than that of the other factors.The soil samples we collected covered arable land (irrigated land (11), dryland (18)), forestland (4), grassland (6) and urban  12), and the content of each heavy metal under the different land use types is provided in Table 1.
For Cd and Pb, which could be highly attributed to APCS1, their mean values indicated the highest levels in urban land, with the second highest levels observed in irrigated land, and both were much higher than the levels under the other land use types.In contrast, the lowest levels were observed in forestland, and the mean values of Zn also indicated a consistent performance with that of the above heavy metals.The mean values of Cr and Cu were also higher in irrigated land and urban land, while the Cr level was the highest in forestland, and the Cr and Cu levels were the lowest in grassland.The Zn concentration  was correlated with the slope and was mainly concentrated in the residential area around the mine.We also found that there was a significant correlation (P \ 0.05) between the factor-detector explanatory power of mining enterprises and the distance from the road on APCS1, indicating that APCS1 also reflects pollution sources including those stemming from industrial activities, which is very consistent with the spatial distribution of mining enterprises in the vicinity of high-Cd value areas, suggesting that heavy metal pollution may be caused by nearby roads during the transportation of ores in mining areas (Yang et al. 2020) and, moreover, factors such as vehicle exhaust may contribute to Cd deposition in soil (Liu et al. 2022).We also found that the explanatory power of rivers and the distance from roads for APCS2 was highly significantly correlated (P \ 0.01),and the distance from chemical enterprises, farming enterprises and food enterprises are also with highly significant correlations (P \ 0.01) in regard to the explanatory power for APCS2, indicating that the pollution sources corresponding to APCS2 could also originate from industrial activities to a large extent, while the contribution of the other four sources to APCS2, although not high, reflects the notable disturbance of soil layer A by exogenous inputderived pollutants.
According to the interaction detector results (Table 2), the interaction between the land use type and SOM, pH, CEC and distance from rivers is high, indicating that the land use type exerts an enhanced synergistic effect on soil indicators in terms of the effect of APCS1 on heavy metals and that the interaction between the land use type and distance from other industrial activities (chemical, metallurgical and mining enterprises) provides a greater explanatory power than the factor-detector influence.The strongest interaction between the elevation and soil capacity for APCS2 (q = 0.7881) further suggests that Cr, which dominates the heavy metals under the landscape type of APCS2, is influenced by more than a single natural factor, while the interactions of the farm distance from enterprises with the soil capacitance (q = 0.7544) and SOM (q = 0.7557), the interaction of the distance from rivers with the CEC (q = 0.7361), and the interaction between the distance from waste/wastewater treatment plants and SOM (q = 0.7557) are also stronger, suggesting that APCS2 exhibits a synergistic effect between industrial activities, river and soil indicators on the impact of this heavy metal combination type.
The detection results for soil layer B also showed that the interaction between factors was stronger than the single factor effect, but the degree of influence of natural factors on the three heavy metal combination types in soil layer B was very obvious, among which the factor-detector explanatory power of the elevation and slope for APCS1 (P \ 0.01) and APCS2 (P \ 0.05), respectively, was significant, with a greater explanatory power of the elevation factor for APCS1 (q = 0.48).Moreover, the explanatory power of the elevation and slope for APCS1 was enhanced with SOM, and farming, food enterprises, chemical enterprises and waste/wastewater treatment plants also provided a significant factor-detector explanatory power for APCS1, indicating that Cd, Pb and Zn in soil layer B were more significantly influenced by the interaction effect between industrial activities and natural factors than those in soil layer A, suggesting that soil layer B was impacted by natural conditions to a greater extent and was less responsive to exogenous input pollution than soil layer A. The CEC was also highly significant for APCS2, and its interaction with the elevation (q = 0.72) and slope (q = 0.73) was also notable, while the land use type provided the greatest factor-detector explanatory power for APCS2 (q = 0.54) and APCS3 (q = 0.32).Table 1 indicates that the mean contents of Cr and Cu, which were highly attributable to APCS2 and APCS3, were both the highest in irrigated land in soil layer B, consistent with soil layer A, while they achieved the lowest mean values in the grassland.However, Cu in particular may be more significantly influenced by agricultural activities in soil layer B. In addition, the explanatory power of metallurgical enterprises, chemical enterprises, mining and distance from roads for APCS2 exhibited a highly significant correlation (P \ 0.01), which increased the synergistic effect with the soil capacity (q = 0.70-0.73).Moreover, we observed a performance highly similar to that of APCS2 in soil layer A, further indicating that soil layer B, especially Cr, was also polluted by exogenous inputs such as industrial activities.Although the single factor effects of the soil indicators on APCS3 were not significant, the interaction between SOM and CEC provided a notable explanatory power.In addition, the interaction effect between the elevation of rivers, roads, and distance from industrial activity sites such as mining enterprises on APCS3 was highly obvious, which could indicate that natural factors, such as topographic factors, could influence metals under specific conditions, while the interaction between the distance from roads with SOM and distance from food enterprises was also notable, indicating that the mining transportation process is likely to yield exogenous input pollution of soil heavy metals, and heavy metals possibly migrated downwards from the surface soil layer.

Integrated ecological risk evaluation
and comparison of the results

Integrated risk evaluation results
The evaluation results of the PLI, NIPI and NIRI for soil layers A and B are shown in Fig. 5 and Tables A9-A11, respectively, and the patterns of the three models are shown in Fig. 6.Within the same evaluation model, the overall risk of heavy metal pollution in soil layer A was higher than that in soil layer B. All sampling sites in soil layer A exhibited moderate pollution based on the PLI, and more than half of the sampling sites were extremely polluted.The area of extreme pollution in soil layer B was smaller than that in soil layer A, and nearly half of the sampling sites in soil layer B exhibited PLI values at moderate pollution levels.Most of the areas indicated moderately and severely polluted levels.These sampling sites were concentrated in the middle of the study area, which is also very consistent with the locations of higher Cd pollution levels, indicating that the contribution of Cd to the overall pollution level in the study area was high and that the content of one heavy metal may affect the total PLI level (Zhao et al. 2014).This suggests that Cd contributed more notably to the overall pollution level in the study area and reflected that the level of one heavy metal factor may affect the total PLI level.
The NIPI of all samples in both soil layers A and B indicated slightly polluted and higher levels, and 7 sampling sites in soil layer A belonged to the strongly polluted level.According to the NIPI spatial distribution map (Fig. 6), it could be found that the higher-polluted level areas occurred close to industrial activities such as chemical, metallurgical and mining enterprises, while the spatial distribution of the NIPI was generally similar to the spatial pattern of the Cd content, indicating that the influence of Cd on the NIPI values was also notable.The results are very consistent with the PLI results, as they are both based on PI, but the NIPI exhibited less variability in the pollution values and revealed a lower level of the integrated risk than those of the PLI.The NIPI of soil layer B was not located at the strongly polluted sites, and the risk level in most areas mainly suggested slight pollution, indicating that although soil layer B was not yet subject to the accumulation of heavy metals migrating downwards from the surface soil layer, the corresponding level exceeded the warning range.
The results of the new integrated ecological risk index (the NIRI) showed that the risk levels in both soil layers A and B were lower than the PIL and NIPI results under the same level, of which 7 sampling sites in soil layer A were located at the moderate risk level, coinciding with 7 high-NIPI samples, indicating that these sampling sites and the surrounding soil were areas with the highest risk levels in our study area.The NIRI of most samples was moderate.The low-risk areas in soil layer B were more extensive than those in soil layer A, with only 17 samples at the moderate risk level, and these areas were spatially close to industrial activities such as mining and chemical enterprises.The spatial distributions of NIRI high-value areas in soil layers A and B were basically the same, and these areas were spatially close to metallurgy enterprises, but the area in soil layer B was smaller, which further indicates that although soil layer B was subject to the accumulation of heavy metals migrating from the surface soil layer, the range of the area with more serious downward heavy metal migration pollution was more concentrated and smaller.
Scholars have measured the ecological risk levels due to heavy metal pollution in Beijing (Men et al. 2020), Beijing, Hefei (Zhou et al. 2022) and the middle and lower reaches of the Han River (Li et al. 2022) based on the NIRI, showing that there are areas with high and extremely high risks, whereas in our study area, despite the notable influence of industrial activities on the heavy metal content, the comprehensive ecological risk levels of heavy metals were not very high, and the areas with relatively high risks were concentrated.This may also indicate that the NIRI results could suggest a lower overall ecological risk level in the study area than that based on the pollution risk evaluation results alone.

Differences in the integrated risk among the land use types
While it is necessary to conduct a comprehensive areawide analysis to elucidate the overall ecological risk situation in our study area from a macroscopic perspective, we believe that it is also more important to further obtain the ecological risk under the different land use types for subsequent individualized and differentiated pollution prevention measures and land remediation significance considering the specific land use types (e.g., cropland, grassland, forestland, and urban land).Based on the results of the 3 risk evaluation indicators for soil layers A and B, we calculated the average level of the integrated risk under the different land use types, as shown in Fig. 7.
According to the results of the three models, only the PLI yielded a mean risk level at the highest risk level (extremely polluted), located in soil layer A in integrated and urban land.Soil layer B in urban land and soil layer A in dryland attained mean levels close to the threshold of the highest level, soil layer A in forestland and grassland attained mean levels matching the severely polluted level, and the lowest level was observed in grassland, but soil layer B in grassland also attained a mean level that matches the moderately polluted level, suggesting that the PLI shows that both soil layers A and B were polluted at the moderately polluted level or higher.The NIPI and NIPI identified urban land as the land use type with the highest risk, similar to the PLI, with the mean risk based on the

Influence of heavy metal combination types
and land use types of similar pollution sources on heavy metal accumulation.
The results of several previous studies indicate that a high Cd content may be closely related to local copper mining and metal smelting activities (Petit et al. 2022;Wang et al. 2016;Zeng et al. 2022).This is more consistent with the conclusions we reached in this study.Soil type and pH can also largely influenced the Cr (Wang et al. 2019), and Cu, Cr, and Pb were significantly negatively correlated with the slope factor in a study of soil heavy metals in a mining area in China, all showing an increase with decreasing slope, while Zn was slightly correlated with the slope; moreover, Zn and the slope were correlated but mainly concentrated in the residential area around the mine (Chu Chunjie 2014).
Our study results showed that Cr was indeed influenced by certain natural factors (elevation and slope).We confirm that industrial areas have a high probability of receiving soil Cr contamination and show significant differences between land use types, which is likely to be absorbed by crops.But the soil type and SOM did not notably influence the factor-detector explanatory power for APCS2, considering that Cr may also be influenced by other exogenous input factors, such as the accumulation of heavy metals carried by rivers and transportation in the surrounding soil.This is also reflected in the distance from chemical enterprises, farming enterprises, food enterprises, roads and rivers with highly significant correlations (P \ 0.01) in regard to the explanatory power for APCS2, indicating that the pollution sources corresponding to APCS2 could also originate from industrial activities to a large extent, while the contribution of the other four sources to APCS2, although not high, reflects the notable disturbance of soil layer A by exogenous inputderived pollutants.Moreover, organic fertilizers, pesticides (Liu et al. 2018), rivers (Wang et al. 2020) etc., may be Zn sources or the main predictor of Zn.Some scholars also suggested that the role of roads in heavy metal accumulation was very pronounced only under the influence of specific conditions, such as climate and topography (An et al. 2022).
The arable land in our study area is mainly sloping land, while industrial and urban land areas are located along the river with a gentle slope and low elevation, and the high value of Zn was concentrated in the downstream area along the river, which may be due to the accumulation of pollutants in the surrounding soil from the upper reaches to the downstream area.Dai et al. (2019) specifically analysed Cd polluted condition due to its degree of contamination and identified the sources of Cd mainly are energy production as well as chemical fertilizers.Our study also indicated that compared to dryland, the influence of human activities such as the construction of farmland hydraulic facilities, irrigation, and application of fertilizers and pesticides was also greater in irrigated land than that in dryland, which is also the reason for the higher heavy metal content in many irrigated land areas.We evaluated the pollution risk of the study area using three calculation models and compared them with existing research results, as shown in the Table 3.The above results compared the pollution risk model results with other industrial areas.Except for PLI, the average risk levels of NIPI and NIRI were all below moderately risk.The high-risk areas were mainly located in key industrial areas along the river.Considering that the industrial areas were concentrated while the northwest of the study area was mainly mountainous where the impact intensity of industrial activities was relatively low.
In order to compare and discuss the results of the three evaluation models, we further calculated the contribution rates of the five heavy metals to the PLI, NIPI and NIRI results, and the average contribution rates are shown in Figs. 8 and 9 (the results for the contribution of the heavy metals to PLI and NIPI were consistent).It could be found that despite the differences in the risk values between soil layers A and B, the five heavy metals revealed basically similar contribution rates to the PLI/NIPI and NIRI in the two soil layers (Fig. 8).First, the average contribution levels of the five heavy metal factors to the PLI/NIPI were basically similar, with the average contribution of Cd (29.38% in soil layer A and 24.50% in soil layer B) the highest, followed by Cr (20.23% in soil layer A and 21.49% in soil layer B), and the lowest contribution was observed for Zn (15.18% in soil layer A and 16.77% in soil layer B), which indicates that the difference in contribution rate among the five heavy metals was not large.The average contribution rates of Cd to the NIRI (78.54% in soil layer A and 73.54% in soil layer B) were much higher Table 3 Comparative analysis of main research results of risk evaluation models

Main results
The results of this study An integrated approach to quantify ecological and human health risks of soil heavy metal contamination around coal mining area Heavy metals (mean) exceeded local background content with a cumulative of moderately polluted using PLI around a coal mining The average pollution level of PLI in soil layer A is extremely polluted, while in soil layer B it shows moderately and severely polluted Study of soil physicochemical properties and heavy metals of a mangrove restoration wetland Only Cd has a higher pollution level, while the PI and of other heavy metal elements shows low or no pollution risk in a mangrove wetland The PI values of five heavy metals in soil layer A are all high, and the spatial distribution of high-risk areas is concentrated in the central part of the study area, similar to the spatial pattern of Cd pollution Comprehensive exploration of heavy metal contamination and risk assessment at two common smelter sites As the depth of the soil layer increases, the proportion of acid soluble and reducible heavy metal parts decreases, while the proportion of oxidized and residual parts increases; NIPI shows that the pollution level in the mining area is relatively high The content of 5 heavy metals in surface soil layer is higher than the deep layer, and NIPI shows that the pollution risk level in the study area is mainly below moderately pollution than those of the other four heavy metals, and the average contribution rates of Cr (3.79% in soil layer A and 4.67% in soil layer B) were significantly lower or only slightly higher than those of Zn (1.41% in soil layer A and 1.41% in soil layer B).We adopt the high-value area determined via the 3 integrated ecological risk models as the basis to delineate the priority area for heavy metal pollution prevention and control and land ecological restoration measures (Fig. 6), considering soil layer A, which is seriously polluted by heavy metals, as an example, we could determine, by comparing the spatial distributions of the PLI, NIPI and NIRI, that the highest-value area indicated by the NIRI was much smaller than that indicated by the NIPI.The extremely polluted area based on the PLI was very large in particular, and the moderately polluted level area based on the NIPI matched the highest level of the NIRI, which is similar to that based on the PLI.In contrast, the high-value area was smaller than that based on the PLI and NIPI, and it occurred near metallurgical enterprises, wastewater treatment plants and chemical enterprises along the rivers but was more notably concentrated towards the middle of the study area.In soil layer B, we also found that the higher risk levels in the lower reaches of the river were considerable.Similarly, according to the risk classification level of the 3 models, the ecological risk indicated by the NIRI was also lower.Therefore, employing the NIRI evaluation results as a basis for the next step of pollution prevention and control in industrial areas and land remediation, etc., could better reveal the actual heavy metal pollution conditions in the study area, and the accuracy and relevance within the range and context of heavy metal pollution risk identification were higher than those of the NIPI.
In summary, since the number of heavy metals may impact the NIPI results, the contribution of each heavy metal factor to the PLI and NIPI results did not significantly differ, thus limiting the objectivity in the accurate identification of the integrated risk of the integrated influence of heavy metals.The NIRI could eliminate the effect of the amount of heavy metals on the cumulative risk by introducing corresponding parameters for the toxicity of the considered heavy metals, which may vary by a factor of 30 between different heavy metals, and the NIRI could highlight the effect of the more toxic factors on the overall risk, which further demonstrates the scientific flexibility of this new integrated ecological risk index (the NIRI) in obtaining measurements.

Fig. 1
Fig. 1 Location of the study area, with a location of Qinghai-Tibet Plateau, b location of study area in Qinghai province, c elevation, distribution of sampling sites, distribution of key polluting enterprises of study area and d land use types of study area

Fig. 2
Fig. 2 Contents of the five soil heavy metals at each sampling point in soil layers A and B

Fig. 3
Fig. 3 Contribution of the APCS to the heavy metals in soil layers A and B

Fig. 4
Fig. 4 Factor-detector results.(Only significant (p \ 0.05) factors are shown, where a and b indicate detected results for APCS1 and APCS2 in soil layer A. The c-e indicate detected results for APCS1, APCS2 and APCS3 in soil layer B)

Fig. 5
Fig. 5 Statistics of the results of the 3 evaluation models

Fig. 7
Fig. 7 Mean risk evaluation results under the different land use types

Fig. 8
Fig. 8 Contributions of 5 heavy metals f to 3 models in soil layer A

Table 2
The interaction detector results