Introduction

The Qinghai–Tibet Plateau, situated in the southern part of Asia and the western part of China, spans an estimated 3 million square kilometers in size, earning it the nickname “Roof of the World “ and “Third Pole.“ It serves as a crucial ecological barrier in China and Asia and plays multiple significant roles in water source conservation, biodiversity preservation, and soil and water retention (Yan et al., 2013). Due to its distinctive geographical and climatic conditions, the plateau witnesses restricted industrial activity and minimal human interference. The urbanization rate in 2020, at 54.95%, is lower than the average urban utilization rate of 63.9%; the urban land area of the Tibetan plateau is 796.59 km2, accounting for only 0.03% of the total area of the plateau (Fu et al., 2022). These factors differentiate the material sources in the Qinghai–Tibet Plateau from those in other regions.

Recently, scientists have conducted research on the origins of materials in the Qinghai–Tibet Plateau. Wang et al. (2023) utilized PCA to examine the dispersion and source attributes of heavy metals within the soil of the Huan River wetlands on the Qinghai–Tibet Plateau. Du et al. (2023) employed geostatistics and positive matrix factorization to assess the spatial distribution patterns of seven heavy metals in the Qinghai–Tibet Plateau. Xing et al. (2022) utilized PCA–APCS–MLR to evaluate the potential hazards linked to heavy metals in the Qinghai–Tibet Plateau. However, traditional statistical models assume linearity and are sensitive to multicollinearity when calculating interactions, making them incapable of accurately quantifying the interactions between various sources of soil heavy metals (Zeng et al., 2023).

In contrast, the Geodetector developed by Wang & Xu (2017) can uncover the influence of individual factors on dependent variables, as well as the interactions between two factors, without the need to assume linearity. This approach effectively mitigates the impact of multicollinearity, overcoming the limitations of traditional methods when analyzing categorical variables (Qiao et al., 2017; Zhao et al., 2020). This approach allowed for the quantitative evaluation of the impact of various factors on the spatial diversity of soil elements. It has been successfully applied in the field of soil heavy metals (Guo et al., 2022; Qiao et al., 2022; Xu et al., 2023). Nonetheless, in the application of the Geodetector, researchers often prioritize factor detection and interaction detection; risk detection and ecological detection are less studied in the field of soil heavy metals (Gong et al., 2023). This poses a challenge to achieving a comprehensive understanding of the spatial distribution patterns of heavy metals in soil. Furthermore, prior research has predominantly concentrated on regions with high population densities and economic development, with limited investigations on the Qinghai–Tibet Plateau, which is characterized by minimal human impact. Therefore, further research and development are needed to investigate the material sources on the Qinghai–Tibet Plateau.

Hence, this study takes Dulan County, Qinghai Province, as an example and utilizes PCA to identify the origins of soil heavy metals. Acknowledging that PCA results can only reflect the types of sources contributing to the accumulation of soil heavy metals, this study extends the analysis using Geodetector to investigate the driving factors behind the spatial accumulation and distribution of heavy metals in soil, considering environmental factors. This study aimed to verify the applicability of geodetector analysis in the Qinghai–Tibet Plateau region and elucidate how different environmental factors collectively influence the accumulation of soil heavy metal sources. The objective of this study was to establish a basis for the preservation and management of soil environmental quality in Dulan County, Qinghai Province, China.

Materials and methods

Study region

The study area was located on the southern periphery of the Qaidam Basin within the southern section of the Eastern Kunlun Mountains, particularly in the Bulhan-Budashan Region (Fig. 1). Administratively, it falls under the jurisdiction of Dulan County in Qinghai Province. The terrain generally exhibits a trend of higher elevation in the south and lower elevation in the north. Additionally, this area lies within the northern Kunlun tectonic magmatic subzone of the Eastern Kunlun tectonic magmatic zone, with magmatic intrusion activity primarily dominated by the Indonesian magmatic activity. Granite formations are extensive and characterized by multiple cycles, multifactorial origins, and regional zoning. The elevation in the research area ranges from 2900 to 4950 m above sea level, leading to a high-plateau arid continental climate.

Fig. 1
figure 1

Location of the Study Area

Full size image

The land cover types in the research area primarily include sandy land, bare rock, gravel, and bare soil. The soil types are mainly gray-brown desert soil, saline brown calcareous soil, and high-altitude desertification grassland soil. Geological formations encompass Quaternary alluvial deposits, seven spring formations from the Late Pleistocene, Late Pleistocene flood deposits, Ordovician Qimantage Group, Silurian Qiujidonggou Group, Silurian Xiaomiao Group, Devonian Jingshuikou Formation, angle plagioclase granulite of the Jingshuikou Formation, black cloud plagioclase granulite of the Jingshuikou Formation, dolerite of the Jingshuikou Formation, Late Triassic syenogranite, Middle Triassic quartz diorite, Middle Triassic granite diorite, Middle Triassic monzonitic granite, Middle Triassic porphyritic monzonitic granite, Late Permian plagioclase granulite, Early Devonian pyroxenite, and Early Devonian granite diorite.

Sample collection

The layout of the geochemical sampling points for this 1:25,000 geochemical survey was carefully designed, taking into consideration several factors, such as the geological background, land cover types, and soil types in the study area. The objective was to achieve both uniformity and representativeness. The average sampling density was 19.7 points per square kilometer (19.7 points/km2), and the sampling arrangement was adapted as required to accommodate specific conditions. Large alluvial plains and peripheral floodplain areas were excluded from sampling, as were areas near factories, mines, villages, and highways that may have introduced anthropogenic pollution. This led to an increased density of sampling points in the river basins, prospective mineralization areas, and remote sensing anomaly areas. In total, 6274 samples were collected.

The samples were analyzed at the Qinghai Geological and Mineral Testing and Application Center. The As and Hg contents were assessed using atomic fluorescence spectrometry depending on the analysis parameters and sample characteristics necessary for this project. Soil Ni, Cu, Sn, Sb, Mo, and Be contents were measured using Inductively Coupled Plasma Mass Spectrometry, whereas soil Pb, Zn, and Cr contents were determined using X-ray Fluorescence Spectrometry. To ensure precision and accuracy, the quality of the analysis was controlled using national first-grade soil standards, repeatability tests, outlier checks, and blank tests. The detection limits for As, Be, Cr, Cu, Hg, Mo, Ni, Pb, Sb, Sn, and Zn in the samples were 0.33, 0.03, 3, 1, 0.00046, 0.11, 1.6, 1.2, 0.046, 0.84, and 1 mg/kg, respectively.

Geoaccumulation index (Igeo) method

The Igeo index is employed to compare the concentrations of various heavy metals in soil and evaluate their pollution levels. It is typically denoted as Igeo, as described by Jiang et al. (2020):

$${\text{Igeo}}={{\text{log}}}_{2}\left({\text{Ci}}/\left(K\times {\text{Bi}}\right)\right)$$
(1)

where Igeo denotes the geoaccumulation index for heavy metal i, and Ci represents the measured concentration of a particular heavy metal in the soil. Bi is the reference value, which is often chosen as the background value for soil in Qinghai Province. K is the modification factor, which is typically set to 1.5. The pollution levels are classified into seven grades according to the following Igeo values: Igeo < 0, no pollution; 0 ≤ Igeo < 1, low pollution; 1 ≤ Igeo < 2, moderate pollution; 2 ≤ Igeo < 3, moderate to heavy pollution; 3 ≤ Igeo < 4, heavy pollution; 4 ≤ Igeo < 5, heavy to extremely heavy pollution; and Igeo ≥ 5, extremely heavy pollution.

Geodetector

The geographic detector (geodetector) primarily detects spatial heterogeneity. The fundamental principle of this approach is that, if a specific independent variable exerts a significant influence on a dependent variable, there should be spatial congruence or a correlation between that independent variable and the dependent variable (Wang & Hu, 2012; Wang et al., 2010). We utilized geodetector for factor, interaction, risk, and ecological detection to comprehensively and objectively analyze the impact of three environmental factors (land cover type, soil type, and geology) on the accumulation of soil heavy metals.

Factor detection was employed to investigate the spatial variation in the four material sources and the extent to which various environmental factors explained these sources. It was quantified using the q-value and the following formula:

$$q=1-\frac{\sum_{{\text{h}}=1}^{{\text{L}}}{N}_{h}{\sigma }_{h}^{2}}{N{\sigma }^{2}}=1-\frac{{\text{SSW}}}{{\text{SST}}}$$
(2)

where h = 1,…, and L represents the number of categories for independent variable X; Nh and N are the quantities of units in category h and the entire area, respectively; σh2 and σ2 are the variances of dependent variable Y in category h and the entire region, respectively; SSW and SST represent the sums of variances for all groups or categories of the independent variable X and the overall variance within the area, respectively.

Interaction detection aims to identify the influence of interactions among different influencing factors, that is, whether the explanatory power of influencing factors increases or decreases when they interact, or whether these factors independently affect the material sources. The judgment formulas are listed in Table 1.

Table 1 Expression of factor interactions
Full size table

The risk detection primarily detects whether there are significant differences in the spatial distribution of the contribution rates of the influencing factors to various material sources. It involves an element of risk and is evaluated using t-statistics. The formula for this calculation is as follows:

$${t}_{{\overline{Y} }_{h=1}-{\overline{Y} }_{h=2}}=\left({\overline{Y} }_{h=1}-{\overline{Y} }_{h=2}\right){\left[\frac{{\text{Var}}\left({\overline{Y} }_{h=1}\right)}{{n}_{h=1}}+\frac{Var\left({\overline{Y} }_{h=2}\right)}{{n}_{h=2}}\right]}^{-1/2}$$
(3)

where Yh represents the mean value of the attribute within subregion h, which, in this study, represents the contribution rate of a certain material source. Var represents variance. nh represents the number of samples within sub-region h. The t-statistic followed an approximate student’s t-distribution. A larger t-value indicates that the spatial differentiation effect of the influencing factor on the contribution of the material sources is greater.

Ecological detection employs an F-statistic to measure whether there are significant differences in the spatial distribution of the contribution rates of two influencing factors to material sources.

Selection of factor indicators and data processing

Based on methods proposed by Tao et al. (2020) and Huang et al. (2021) for selecting factor indicators, along with the results of the soil element PCA source analysis, three independent variables were selected as follows: land cover type, soil type, and geology. Additionally, the four principal component variables obtained from the source analysis, Ypc1, Ypc2, Ypc3, and Ypc4, were selected as the dependent variables. Because geodetector analysis requires categorical variables when analyzing influencing factors, continuous variables must be discretized into categorical variables. Following the approach of Qiao et al. (2019), the following discretization criteria were applied:

  • 1. Land cover type—classification based on current land-use categories according to established standards.

  • 2. Soil type—classification based on the HWSD dataset.

  • 3. Geology—classification based on rock formation age and rock type.

Descriptive statistics were calculated using Microsoft Excel to better understand the data. Furthermore, SPSS (version 25.0) was employed to conduct correlation analyses. Geodetector analysis was conducted using Geodetector software, and graphical representations were generated using Origin 2022.

Results

Properties and characteristics of soil elementals

The soil in the study area exhibited various physicochemical properties for 6274 samples (Table 2). The mean Pb concentration in the study area was 23.698 mg/kg. The value exceeded the soil background values (BGVs) for Qinghai Province, being 1.17 times higher than the BGVs. This indicates significant enrichment of the element in the soil. Although the mean values of the other elements did not exceed the BGVs, their maximum values surpassed the background values, suggesting the enrichment of these elements in certain areas. The CV is a crucial metric for quantifying the extent of variability, reflecting the spatial variability of soil elements. The CV values for As, Cr, Cu, Hg, Mo, Ni, and Sb were 0.898, 0.992, 0.975, 0.978, 0.884, 0.964, and 0.987, respectively, all of which exceed 0.8, indicating a high level of variability. This suggests significant differences in elemental content among the different sampling points, potentially due to the local enrichment or accumulation of minerals. The CV for Sn was 0.698, indicating moderate variability, whereas the CVs for Be, Pb, and Zn were 0.242, 0.318, and 0.419, respectively, indicating weak variability. This suggests relatively small regional differences in these elements. Except for Hg, the mean values of the nine soil heavy metals in the study area were below the soil pollution risk screening values (GB 15618–2018).

Table 2 Descriptive statistics of soil sample physicochemical properties
Full size table

Correlations of soil elements

The results of the correlation analysis depicted in Fig. 2 indicate significant correlations between most elements. At the p = 0.01 level, significant positive correlations were evident among the following pairs: As–Cr, As–Cu, As–Mo, As–Ni, As–Pb, As–Sb, As–Sn, As–Zn, Be–Mo, Be–Pb, Be–Sn, Be–Zn, Cr–Cu, Cr–Mo, Cr–Ni, Cr–Sb, Cr–Sn, Cr–Zn, Cu–Mo, Cu–Ni, Cu–Sb, Cu–Sn, Cu–Zn, Mo–Ni, Mo–Pb, Mo–Sb, Mo–Sn, Mo–Zn, Ni–Sb, Ni–Sn, Ni–Zn, Pb–Sb, Pb–Sn, Pb–Zn, Sb–Sn, Sb–Zn, and Sn–Zn. Meanwhile, significant negative correlations were observed at the p = 0.01 level for the following pairs: Be–Cr, Be–Cu, Be–Hg, Be–Ni, Cr–Pb, Cu–Pb, Hg–Pb, and Ni–Pb. Positive correlations between elements may imply similar sources or geochemical behaviors related to enrichment and migration. Conversely, negative correlations may suggest different sources or even antagonistic behavior between the elements (Liu et al., 2019).

Fig. 2
figure 2

Correlation coefficient of heavy metals in soil. Ellipses in the graph represent the degree of correlation, with flatter ellipses indicating stronger correlations. Red indicates a positive correlation, and blue signifies a negative correlation. *p ≤ 0.01

Full size image

Quality evaluation of soil heavy metals

The mean Igeo values for the nine soil heavy metals calculated using the geoaccumulation index method are as follows: As (− 2.769), Cr (− 2.101), Cu (− 1.594), Hg (0.036), Ni (− 2.032), Pb (− 0.354), Sb (− 1.930), Sn (− 1.375), and Zn (− 0.869). Overall, except for Hg, all other measured heavy metals were at an unpolluted level, whereas Hg was at a low polluted level. When looking at the sampling points, 1.62% of the sampling points were in the category of low pollution (for Cr, 3.47% for Cu, 1.91% for Ni, 6.15% for Pb, 3.27% for Sb, 1.23% for Sn, and 1.35% for Zn). In the category of moderate pollution, there were 0.51% for Cu, 0.53% for Pb, 0.16% for Sn, and 0.18% for Zn. Some sampling points fell into the category of moderate-to-heavy pollution, with 0.05% for Pb, 0.03% for Sn, and 0.05% for Zn. Additionally, Sn and Zn accounted for 0.02% and 0.01%, respectively, of the sampling points in the heavy pollution category. For Hg, 9.72%, 6.07%, 2.25%, and 1.01% of the sampling points were lightly, moderately, moderately to heavily, and heavily polluted, respectively.

PCA source analysis of soil heavy metal contamination

The Kaiser–Meyer–Olkin test statistic for soil element analysis in the Bulhan Budashan Mountains of Dulan County was 0.642, which is greater than the threshold of 0.6, indicating suitability for factor analysis. Additionally, the p-value was 0.000, which was highly significant, confirming its appropriateness for the analysis.

The PCA of the soil elements in Dulan County retained the first four principal components based on the criterion of eigenvalues greater than one. The rotated component matrix for the four components showed contribution rates of 25.883%, 18.455%, 11.385%, and 9.117%, with a cumulative contribution rate of 64.84%. The distribution of factor scores is shown in Fig. 3.

Fig. 3
figure 3

Distribution of factor scores

Full size image

PC1 explained 25.883% of the data variance and was characterized by significant loadings for Cr (0.898), Ni (0.879), Cu (0.653), and Zn (0.461). The results indicated that these elements exhibited highly significant positive correlations with each other (p < 0.01), with a correlation coefficient of 0.835 between Cr and Ni. This suggests that they have a similar spatial distribution. PC2 explained 18.455% of the data variance and was characterized by high loadings for As (0.862), Sb (0.842), Mo (0.530), and Cu (0.390). The results indicated highly significant positive correlations between these elements (p < 0.01). Specifically, the correlation coefficient between As and Sb was 0.635, suggesting that they have a common source. PC3 explained 11.835% of the data variance and was characterized by significant loadings for Pb (0.711), Be (0.673), Zn (0.540), and Sn (0.530). Descriptive statistics showed that the CVs for Pb, Be, Zn, and Sn were relatively low, and their average values were mostly below the background values, indicating minimal anthropogenic influence. PC4 explained 9.117% of the variance in the data, and it contained only one major component, Hg, with a loading value of 0.931. The correlation results indicated a significant negative correlation between Hg and Be, as well as between Hg and Pb, suggesting different sources or strong inhibitory effects. The factor scores for PC4 were the highest in the eastern section of the study area, which corresponds to geological units such as the Neoproterozoic Great Wallian Xiaomiao Formation, Paleoproterozoic Jinshuikou Rock Group, Triassic gray quartz diorite, and Permian diorite.

The impact of different land cover types, soil types, and geological backgrounds on elemental content

We examined variations in the concentrations of 11 elements in the soil across different land cover types, soil types, and geological backgrounds by using ANOVA. The results, depicted in Table 3, highlight substantial differences in the concentrations of As, Be, Cr, Cu, Hg, Mo, Ni, Pb, Sb, Sn, and Zn influenced by various factors.

Table 3 Variance analysis
Full size table

On average, As, Mo, and Cr exhibit higher concentrations in bare rock and gravel areas. Be, Cu, Ni, Sn, and Zn demonstrate elevated average concentrations in sandy areas, while Pb and Hg reach their highest average concentrations in bare soil, at 26.65 mg/kg and 15.33 mg/kg, respectively. The average concentration of Sb does not exhibit significant differences across the three soil types.

Analyzing from the perspective of soil types, Be shows a higher average concentration in gray-brown desert soil, at 1.98 mg/kg. As, Cr, Cu, Mo, Ni, Sb, and Zn demonstrate higher concentrations in saline brown calcareous soil, while Pb, Sn, and Hg exhibit higher concentrations in the soil of high-mountain desertification grassland. Regarding the geological background, Be, Cu, Mo, Ni, Pb, Sn, and Zn exhibit the highest average element concentrations in the Changchengian Xiaomiao Formation and Early Devonian pyroxenite Formation. As and Sb demonstrate the highest average concentrations in the Ordovician Qimantage Group, and Hg shows a higher average concentration in the Qiujidonggou Formation, significantly surpassing other formations at 30.26 mg/kg. Cr exhibits the highest average concentration in the Early Devonian Olivine Formation, at 53.66 mg/kg, and its average concentrations in the Changchengian Xiaomiao Formation and Qiujidonggou Formation are also significantly higher than in other formations. The qualitative analysis of the analysis of variance elucidates the impact of land cover types, soil types, and geological factors on the content of various elements, while the geodetector enables a quantitative assessment of the influence of diverse factors on the accumulation of soil elements.

Geodetector analysis

Factor detection

Factor detection reveals how various environmental factors explain the spatial distribution of source contributions. The results of factor detection in Fig. 4 show that except for the soil type factor in PC4 with a p-value of 0.106, which failed to meet the significance threshold at a confidence level of 0.05, all factors passed the significance test, indicating a significant impact on the distribution of source contributions in the study area. The q-values for the ability of each factor to be explained are shown in Fig. 4. For PC1, the q-values for the various environmental factors were as follows: geology (0.1632) > land cover type (0.0241) > soil type (0.0230). For PC2, the q-values were geology (0.1140) > land cover type (0.0096) > soil type (0.0148). For PC3, the q-values were geology (0.1756) > soil type (0.0148) > land cover type (0.0063). For PC4, the q-values were geology (0.0335) > land cover type (0.0083).

Fig. 4
figure 4

Q-value of the contribution of the impact factor to the source of substances

Full size image

Risk detection

Risk detection can reveal whether there are significant differences in source contributions between subregions and the high- and low-value areas of each source contribution within each factor. The results of risk zone detection are shown in Fig. 5.

Fig. 5
figure 5

Risk detection results. PC1, PC2, and PC3 are related to land cover type (1), soil type (2), and geology (3), whereas PC4 is related to soil type (1) and geology (2). In terms of land cover type, the first category was sandy land, the second category was bare rock and gravel land, and the third category was bare soil. In terms of soil type, the first category was gray desert soil, the second category was saline brown calcareous soil, and the third category was high-mountain desertification grassland soil. Geologically, for PC1, the first category was Middle Triassic quartz diorite, the second category was the Changchengian Xiaomiao Formation, the third category was Early Devonian olivine rock, the fourth category was Late Permian diabase, and the fifth category was Jinshuikou Formation hornblende schist. The sixth category was the Jinshuikou Formation black cloud schist. For PC2, in terms of geology, the first category was the Changchengian Xiaomiao Formation, the second category was Late Triassic orthoclase granite, the third category was Late Permian diabase, the fourth category was Jinshuikou Formation hornblende schist, the fifth category was the Qibaikou Formation Qiujidonggou Group, and the sixth category was the Ordovician Qimantag Group. Regarding PC3, in terms of geology, the first category was Middle Triassic spotted binary granite, the second category was the Changchengian Xiaomiao Formation, the third category was Late Triassic orthoclase granite, the fourth category was the Qibaikou Formation Qiujidonggou Group, the fifth category was biotite granite in black mica granite, and the sixth category was Ordovician Qimantag. For PC4, the first category was the Changchengian Xiaomiao Formation, the second category was Late Permian diabase, the third category was Quaternary alluvium, the fourth category was Jinshuikou Formation hornblende schist, the fifth category was Jinshuikou Formation black cloud schist, and the sixth category was Jinshuikou Formation dolomite rock

Full size image

For PC1, the results of the risk detection showed that, in terms of land cover type, the average source contribution rates were as follows: sand > bare rock, gravel > bare soil, and PC1 exhibited significant differences between all pairwise comparisons. In terms of soil type, salinized brown calcareous soils have the highest average source contribution rates and significant differences existed among all pairwise comparisons. Among the geological factors, the highest average source contribution was observed in the Changchengian Xiaomiao Formation. The Xiaomiao Formation showed no significant difference from the biotite granite and Ordovician Qimantage Group but exhibited significant differences from the other categories.

For PC2, in terms of land cover type, the average source contribution rates were as follows: bare rock, gravel > sand > bare soil, and PC2 exhibited significant differences between sand and bare soil, bare rock, and gravel. In terms of soil type, salinized brown calcareous soils have the highest average source contribution rates and significant differences existed among all pairwise comparisons. Regarding geological factors, the highest source contribution rate was observed for the Qingbaikou Hongji Donggou Formation, which exhibited significant differences from the other category groups.

For PC3, the results of risk detection showed that, in terms of land cover types, sand has the highest average source contribution rates, and PC3 exhibited significant differences between bare soil and bare rock, gravel, sand, and bare rock gravel. In terms of soil type, the average source contribution rates were as follows: alpine desertified grassland soil > salinized brown calcareous soil > gray-brown desert soil; significant differences existed between all pairwise comparisons. Regarding the geological factors, the Xiaomiao Formation had the highest source contribution.

For PC4, the results of risk detection showed that in terms of land cover types, the average source contribution rates were as follows: sand > bare rock, gravel > bare soil, and PC3 exhibited significant differences between bare soil and bare rock, gravel, sand, and bare rock gravel. Regarding the geological factors, the Changchengian Xiaomiao Formation had the highest source contribution.

Interaction detection

Soil composition is inherently complex and often results from the combined influence of multiple factors. Interaction detection can assess whether environmental factors are independent of or interact with each other, providing clear insights into the types of interactions. This is advantageous for accurately determining the underlying driving forces that influence the spatial distribution of materials of different origins (Su et al., 2020). From the results (Fig. 6), it is evident that the capacity of the interaction to explain any two factors in the spatial differentiation of material sources exceeds that of any individual environmental factor. Most of the interaction results exhibited nonlinear enhancement, with a minority showing bivariate enhancement. There were no instances of weakening or mutual independence among the factors.

Fig. 6
figure 6

Interaction results. 1 is land cover type, 2 is soil type, and 3 is geology

Full size image

For PC1, the strongest interaction was between soil type and geology, with an explanatory power of 0.1872, followed by the interaction between land cover type and geology, with an explanatory power of 0.1740, both of which exhibited bivariate enhancement. For PC2, the strongest interaction was between soil type and geology, with an explanatory power of 0.1221, characterized by nonlinear enhancement, followed by the interaction between land cover type and geology, with an explanatory power of 0.1209, showing bivariate enhancement. For PC3, the strongest interaction was between soil type and geology, with an explanatory power of 0.1990, exhibiting nonlinear enhancement, followed by the interaction between land cover type and geology, with an explanatory power of 0.1870, also displaying nonlinear enhancement. For PC4, the interaction between land cover type and geology had an explanatory power of 0.0358, displaying bivariate enhancement.

Ecological detection

Ecological detection primarily focuses on comparing the notable disparities in the influence of a single factor compared to another on the spatial arrangement of heavy metal accumulation sources.

The results of ecological detection showed that there were significant differences in the impacts of land cover type and geology, as well as soil type and geology, on PC1, PC2, and PC3. However, the differences in the influences of land cover and soil type on PC1, PC2, and PC3 were not significant. This indicates that changes in land cover and soil type, as environmental factors, have significant impacts on the ability of geological factors to explain. Regarding PC4, the differences in the impact of land cover type and geology on PC4 were not significant.

Discussion

Soil element sources

The elements with significant loadings in PC1 are Cr, Ni, Cu, and Zn, with a correlation coefficient greater than 0.8 between Cr and Ni. In PC2, As, Sb, Mo, and Cu exhibit higher loadings. PC3 shows elevated loadings for Pb, Be, Zn, and Sn, with low coefficients of variation for these elements. Notably, in PC4, Hg demonstrates a remarkably high loading of 0.931, with insignificant correlations or negative associations with other elements. The geological formations corresponding to regions with higher factor scores in PC4 include the Neoproterozoic Great Wallian Xiaomiao Formation, Paleoproterozoic Jinshuikou Rock Group, Triassic gray quartz diorite, and Permian diorite.

For PC1, studies such as Sahoo et al. (2020) have found that the combination of Cr-Ni-Cu is associated with underlying rock lithology, primarily influenced by volcanic metamorphic rocks, basic rocks, and ultrabasic rocks. As the study area experienced frequent magmatic activity, the widespread presence of volcanic metamorphic rocks, basic rocks, and ultrabasic rocks rich in Ni, Cr, and other heavy metals contributed to PC1. In addition, some ultramafic outcrops are rich in Cu (van der Ent & Reeves, 2015). Therefore, PC1 was primarily influenced by the geological background and underlying rock lithology.

For PC2, As and Sb primarily reflect the mineralization associated with low-to-moderate-temperature hydrothermal fluids. The study area was located in the Bayankala–Xiongshandi Indosinian mineralization belt, which is characterized by a continental uplift-magmatic arc setting. The strong intrusive activity of acidic magmas in this belt has resulted in the formation of a meso-acidic intrusion-related mineralization system and a superimposed reconstruction mineralization system, with the main ore type being magmatic-hydrothermal. Therefore, the concentrations of As and Sb were mainly influenced by these mineralization processes. Weathering of the bedrock and parent material is also considered a source of As (Shi et al., 2018). Thus, PC2 was also associated with the geological background.

For PC3, studies have shown that heavy metals related to soil parent material components tend to have lower pollution levels (Zhang et al., 2009). This aligns with the findings of the geoaccumulation index, which suggests that Pb, Be, Zn, and Sn have low pollution levels overall. Therefore, it can be inferred that PC3 was primarily influenced by the geological background.

For PC4, among these formations, the Xiaomiao Formation has been significantly disrupted by tectonic movements and magmatic intrusions, while the Jinshuikou Rock Group is incompletely exposed owing to faulting and magmatic intrusions. Hg is the element most sensitive to geological tectonic activities (Migaszewski et al., 2010), and anomalies mainly occur in areas with developed fault structures in the study area, indicating that Hg levels are related to geological structures, such as faults. Therefore, PC4 was inferred to be primarily influenced by the geological background.

Relationships between influencing factors and various element sources

From the results of the factor detection, the most significant factor influencing the spatial distribution of source contribution rates for various substances was the geological factor. The regional geological background, including the composition of the parent rock material and structural features, the type and source of parent rock material, and various geological processes and historical evolutions during formation (Tran et al., 2013), results in different levels of heavy metal enrichment under different geological backgrounds. The values of Hg and Cu in the Jinshuikou Group were higher than the background values, indicating significant enrichment. The Wulonggou East Qingyanggou tungsten mineralization point in Dulan County is exposed in this formation, which is mainly composed of garnet containing scheelite, pyrite, and chalcopyrite, leading to high Hg and Cu contents. The average Pb content exceeded the provincial background level, with the highest average content being in the Qiuji Donggou Formation of the Qingbaikou Period, indicating the possibility of local enrichment and mineralization. Studies have shown that, under different land-use types, soil nutrient levels vary owing to varying degrees of human disturbance, leading to differences in element and nutrient contents (Chi et al., 2018, 2019). The results of the source apportionment showed that PC1 mainly included Cr, Ni, and Cu, reflecting mineralization information related to basic and ultra-basic rocks. PC2 mainly included As, Sb, and Mo, reflecting mineralization information related to medium- to low-temperature hydrothermal activity. Human mining activities can affect the physicochemical properties of soil, thereby influencing the enrichment of heavy metals. Jian-Min et al. (2007) observed elevated concentrations of Cu, Zn, and Cd in mining tailings, reaching levels of 567, 1140, and 2.48 mg/kg, respectively, in the soil. Due to the enduring and non-degradable nature of heavy metals, their long-term accumulation in the soil is inevitable. For PC3, the soil type was the second largest influencing factor. Soil type is a comprehensive expression of the external soil-forming conditions and internal soil properties. Descriptive statistics showed that the elements with high loadings in PC3 had a relatively low CV, indicating that they were derived from geological background sources and were less influenced by external factors.

In terms of land cover types, PC1, PC3, and PC4 had the highest source contribution rates in sand, while PC2 had the highest source contribution rates in bare rock and gravel. Under natural conditions, rocks and parent materials in bare rock and gravel are prone to weathering, resulting in a higher As content. Additionally, these land cover types were significantly affected by human mining activities (Liu et al., 2023), which explains the higher contribution rates for PC2, indicating human influence. In contrast, the formation of sand in the study area was the result of erosion, sedimentation from floods, and the combined effects of arid, cold, and wind-driven climates. The research conducted by Ding et al. (2017) revealed that natural factors such as topography, wind force, and wind direction influence the distribution of heavy metals in the soil. Additionally, climatic factors can regulate soil moisture content and temperature, thereby impacting vegetation cover, which, in turn, alters the distribution of soil organic carbon and subsequently affects the accumulation of heavy metals. Therefore, the higher source contribution rates for PC1, PC3, and PC4 in sand suggest that natural factors have a significant impact on heavy metal accumulation. In terms of soil type, PC1 and PC2 had the highest source contribution rates in salinized brown calcareous soil. Brown calcareous soil is mainly formed by the accumulation of humus and calcium in grassland soils. Humus, an important chelating agent in the environment, strongly adsorbs heavy metals (Zhu et al., 2011). Moreover, mining points in the study area are mostly distributed in salinized brown calcareous Soil, and mineralization processes partly dictate the accumulation of heavy metals in the study area. PC3 had the highest source contribution rate in the alpine desertified grassland soil because this soil type is rich in granite, and Pb and Zn are enriched in acidic igneous rocks such as granite (Hoerlle et al., 2019). Therefore, PC3 made a greater contribution to this soil type. From a geological perspective, PC1, PC3, and PC4 exhibited the highest source contribution rates in the Changchengian Xiaomiao Formation. The rock composition of the Xiaomiao Formation includes gray-white, medium–thick-layered quartzite and feldspathic quartzite intercalated with brown-gray mica-quartz schist in the lower part, and brown-gray mica-quartz schist intercalated with thin-layered quartzite and marble in the upper part. The rock composition indicates that the formation contains metamorphic rocks, and the combination of Cr-Ni-Cu in PC1 was significantly affected by volcanic metamorphic rocks. Additionally, this formation is in contact with the geological layers in the southwest Gold Creek Group and is in parallel unconformity contact with the strata of the Wolf Mountain Formation on the northeast side. Therefore, the Hg content was higher in this rock formation because of structural disturbances and magmatic intrusion (Zeng et al., 2018). PC2 had the highest source contribution rate in the Qing Baikou Qiuji Donggou Formation, predominantly located in the southwestern region of the study area near Baigou-Pingtai. According to previous research, regional metamorphism is significant in the area, with regional dynamic heat flow metamorphism being the main factor, forming the shallow metamorphic detrital rock series of the Qingbaikou Period (Guodong et al., 2016). Consequently, the As, Sb, Mo, and Cu contents were higher in PC2 in this geological context.

Through the interaction of three influencing factors, the interactions between the geology and additional factors significantly influenced the spatial variation of material origins in the study area. For PC1, PC2, and PC3, the interactions between soil type and geological background exhibited the highest explanatory power. Geological factors such as minerals, rock types, and groundwater levels can directly influence the physical and chemical characteristics of soil. These geological characteristics affect soil nutrients, texture, pH, and other factors, ultimately determining the formation of different soil types (Djodjic et al., 2021). This alignment can lead to a stronger association between the geological background and soil type, making the interactions more pronounced. In addition, soil type also exhibits feedback effects. Once a specific soil type has formed, it influences the geological background. Certain soil types may store or release specific elements or compounds that can interact with the geological background (Kaiser & Zech, 1998), thereby affecting mineral transformation and release processes within the geological background. The geological background and soil type collectively influence various physical and chemical processes in the soil. The complex interactions among these factors further emphasize the significance of the interactions between the geological background and soil type, as they jointly shape the soil environment. Concerning PC4, the interaction between land cover type and geology had greater explanatory power than the geological factors alone. The geological environment fundamentally affects land cover conditions, but due to human activities, land cover also significantly impacts geological alterations (Letourneau et al., 2012). The results, which indicate stronger interaction effects than the sole influence of geology, suggest that, although the geological environment is a primary influencing factor for source four, human activities also play a role in the accumulation of heavy metals. The q-values for the combination of land cover categories and soil type with geology were both greater than the q-values for geology alone. This indicates that both land cover and soil type significantly enhanced the impact of geology on various material sources. Therefore, when studying and analyzing the distribution and impact of material sources in the study area, and when addressing the contamination of heavy metals in the soil, it is essential to strengthen research on land cover and soil types.

Conclusion

In this study, we conducted an assessment and source analysis of soil elements (As, Cr, Cu, Hg, Ni, Pb, Sb, Sn, Zn, Be, and Mo) in Dulan County, Qinghai Province, China. Additionally, a geodetector was employed to comprehensively and objectively reveal how various environmental factors jointly influence the accumulation of soil heavy metals.

The results indicated that, except for Hg, the nine soil heavy metals in Dulan County, Qinghai Province showed no pollution. The study area was influenced by four sources, all of which are geological background sources. The ANOVA results revealed significant differences among various elements in land cover type, soil type, and geology. ANOVA can qualitatively describe differences but falls short in quantifying them, the geodetector, through quantitative analysis, effectively addresses this limitation. The geodetector results indicated that geology exhibited the most substantial explanatory capability regarding the contribution rates of the four substance sources. Various land cover types, soil compositions, and geological backgrounds exhibited distinct contributions to the sources. The interplay of these three influencing factors heightened their collective impact on the sources. The geodetector not only serves as a valuable supplement to traditional statistical methods but can also be extended to similar study areas.

In conclusion, the geodetector is applicable even to the less human-influenced Tibetan Plateau regions. PCA can be used to extract primary elements and identify primary sources, whereas the geodetector can specifically examine the impact of environmental factors on the spatial dispersion of soil components. This study combined the two methods to provide a more comprehensive approach to the source analysis of soil elements in Dulan County, Qinghai Province. This study provides a scientific basis for the conservation and management of this research area.