3. Results and discussion
3.1. Descriptive statistics of soil heavy metal concentrations in farmland
Representative statistical data on heavy metal levels in the samples of topsoil in agriculture and the respective context data are shown in Table 1. The mean soil heavy metal concentrate in both periods was above the Chengdu soil element context values (CNEMC, 1990). Geological conditions determine to some extent the concentration of heavy metals in soils, but humans can also influence their content in soils(G. Liu et al., 2013) . From 2008 to 2017, the average contents of Cd, As, and Cr in soil increased from 0.18, 6.85, and 42.73 mg·kg-1to 0.28, 9.18, and 88.41 mg·kg-1, indicating that human activities affected and caused a build-up of these heavy metals during the rapid development of the city, while the mean levels of Pb and Hg increased from 42.71 mg·kg-1 and 0.083mg·kg-1 to 39.24mg·kg-1 and 0.057mg·kg-1, which indicates that some heavy metal accumulation in the soil has been controlled during the urban development. According to the coefficient of variation (CV) classification of Nielsen and Bouma(Nielsen & Bouma, 1985), we observed that most of these five heavy metals from the two periods mentioned above belong to moderate and high variability, except for soil Cr, CV of the other four heavy metals increased, among which the coefficient of variation of Pb content increased the most in the two periods, reaching 227%, indicating that Pb content was most disturbed by human activities. The results indicate that urban modernization significantly affects the heavy metal content of arable soils and its variability increases.
3.2. Spatial and temporal distribution of heavy metals in soil
The spatial structure of regional variables was analyzed by calculating and fitting the variance functions of five heavy metals from arable soils of the research area in 2008 and 2017 using GS+9.0 geostatistical software, and we use the residuals (RSS) and the coefficient (R2 ) of determination to judge the degree of model fit. The goodness of the model fit depends on whether the residuals are close to 0 and whether the coefficient of determination is close to 1.(Sakata, Ashida, & Tanaka, 2010) . In this paper, the coefficient of determination as well as the residuals meet the criteria for a better model fit. Therefore, it is possible to carry out interpolation in Kriege’s space. Normality tests were performed using the K-S method before performing the fit. After model fitting, the fits of the five heavy metals in the two periods are shown in Table 2. The block-base ratio [C0 /(C0 +C )] is a measure of the strength of correlation in space between samples, and comparing the block-base ratio values between two periods, the block-base ratios between the same elements showed an increasing trend, indicates that the distribution of heavy metals on the soil in the area studied has been disturbed by some anthropogenic factors(Zou, Dai, Gong, & Ma, 2015).
The results of the spatial distribution of soil heavy metal contents in the study area are shown in Fig. 3. The overall spatial distribution of soil Pb, Cr, and Hg contents in 2008 were similar, with the high-value areas mainly at study area in the east, they all showed a gradual decrease from east to west, While Cd and As are mainly concentrated in the second circle of the study area, their heavy metal contents gradually decrease from the second circle to the remaining two circles.
In 2017, the western part of the study area was the main area of soil Pb, Cr, and Hg accumulation, and the trend of the 10-year time scale change was from the east to the west, while Cd and As were similar to the spatial distribution in 2008, mainly concentrated in the second circle. The space distributed content of Pb, Cr and Hg in soils may be related to the ”electronic information industry cluster” and the ”one area and two belts” development planning strategy in the study area, while the spatial distribution of Cd and As is similar in both periods. This could be due to the effect of prolonged human interference, which is related to the industrial clustering in the second circle of the study area. Therefore, they become mainly influenced by structural variability factors.(Zou et al., 2015) .
After nearly 10 years of rapid urbanization, mean contents of Cd, As and Cr in the arable soils of research area showed an increasing trend, and the spatial variation of As and Cr was large, and the results were analyzed by the spatial variation of the content of these three heavy metals as shown in Figure 4. where: the absolute increase (AI) is calculated as X2017 -X2008; the relative increase (RI) is calculated as (X2017-X2008 )/X2008×100%, and where X is the raster layer corresponding to the distribution of heavy metals (He et al., 2019). The spatial distribution of the high values of AI and RI of soil Cd and As contents were similar and mainly concentrated in the second ring, showing that soil Cd showed a serious accumulation trend, while the high values of AI of soil Cr contents showed a point distribution in the central part, but the RI values of Cr show a large surface distribution in the middle of the study area, which indicates that Cr shows a strong spatial and temporal variability in the area. This indicates a significant accumulation of soil Cr in the central part of the study area with urbanization.
3.3. Source analysis of heavy metals
3.3.1. Correlation analysis
To further investigate the causes driving the variation of five heavy metals in arable soil and their relationships with environmental variables, correlation analysis was conducted for the above five heavy metals (B. Wang, Xia, Yu, Jia, & Xu, 2012). The results showed that in 2008, Cd and Pb were significantly positively correlated (P<0.01) in the arable soil, suggesting that these two heavy metals probably share a common supplier, while Cr and As were significantly negatively correlated, indicating that their sources in arable soils are different, while As and Hg were not significantly correlated to any other element. Therefore, the source of As and Hg is probably unique. and In 2017, the correlation between Cd and Pb in arable soil in the study area was weakened (0.01<P<0.05) and there was no significant correlation with Cr, Pbwas significantly positively correlated with Cr, so the two may have the same source, As and Hg had no significant correlation with each element in 2008 became significantly positively correlated, Cd had no significant correlation with other elements The correlation between Cd and other elements was not significant, so they may have a single source (Figure 5). The above results indicate that after 10 years of rapid urbanization, the sources of element in arable land have been changed by human activities.
3.3.2. Quantitative source with PMF
In order to get the quantification of the contributing of various pollution sources to the heavy metal content in arable land, we used the PMF model to quantify it(Qingyu Guan et al., 2018). This study was based on EPA PMF5.0 software, with factor numbers setting to different numbers respectively. The number of optimal factors was determined by comparing the Qrob/Qexp different factor counts. We found the smallest difference between Qrobust and Qtrue when the factor number was 4, with most residuals between -3 and 3. Figures 6 and 7 show the fitting performance for the two period models. In 2008, the determining Co-efficient of the five heavy metals were Cd(0.89), Pb(0.42), As(0.97), Cr(0.99), and Hg(0.99); in 2017, the determining Co-efficient of the five heavy metals were Cd(0.99), Pb(0.73), As (0.99), Cr (0.62), and Hg (0.99). The determining Co-efficient of the five elements in both periods were above 0.6 except for Pb in arable soil in 2008, showing an excellent fit of this model. Thus, the model can achieve the purpose of the study and the fitted results can adequately contain the messages of the raw data (Z. Chen et al., 2022).
The source resolution of the five farmland soil heavy metals shows that both 2008 and 2017 have four main sources (Figure 8). In the source resolution in 2008, factor 1 had a relatively high contribution of 92.1% to Hg, and Hg was primarily located at the east of the study area (Fig. 3), which had a clear spatial correlation with the construction of efficient agricultural production base in the study area in 2008, so that humans were the major source of mercury, not a nature origin. Studies have shown Hg is a key ingredient of agrochemical fertilizer, which is evaporative and migratory in nature. (Giersz, Bartosiak, & Jankowski, 2017). Thus, the high usage of pesticides and fertilizers will pollute the arable soil. It has been reported that more than 50 million tons of fertilizers find their way onto arable land annually of China due to the massive and irrational use of chemical fertilizers. For this reason, factoring 1 could be regarded as a source of agricultural activity.
Factor 2 represented 65.8% the As concentration. The areas with high As content were spread over the south part (Figure 3), and the southern part of the study area was distributed with the most high-tech enterprises and key pollution source sewage treatment plants in the whole Chengdu city, and the structure of the river water system and irrigation canal network was similar to the high-value area of As. Therefore, it was speculated that the soil As pollution in the high-value area might be caused by long-term river sewage irrigation. Hence, factor 2 could be treated in terms of a sources of industrial activity.
Factor 3 was responsible for 66.3% with well above other factors of total chromium concentration. It is generally accepted that heavy metals associated with soil parent materials are usually low contaminating elements, It could be explained by the overall low-lying topography over the area and influence from the river (G. Liu et al., 2013). In addition, as can be seen in Table 1, the Cr content is below the local background value. Many studies point out that chromium arises primarily in the maternal matrix (Nanos & Rodríguez Martín, 2012). Consequently, factor 3 could be assumed to stand for maternal material, i.e., natural origin.
The contribution of factor 4 to the concentration of Cd and Pb was 59% and 45%, respectively. One correlation analysis (Figure 5) showed that Cd and Pb were remarkably related to each other, indicating they might have the same source. It is shown that Pb is the main contributor to transportation releases, which could result from vehicle tire friction and fuel combustion in vehicle engines (Qingyu Guan et al., 2018; Venkatalaxmi, Padmavathi, & Amaranath, 2004). During the present research, great numbers of agricultural fields were scattered alongside transportation roads, and this may have resulted in significant levels for Pb to enter the cultivated soils. This is especially true in the most congested traffic areas of Wuhou and Jinjiang. In addition, Cd is also associated with vehicles and is present in high levels of vehicle emissions. Such contamination is accumulated by a series of atmospheric activities in the land (Pardyjak, Speckart, Yin, & Veranth, 2008). As a result, factor 4 may be recognized to be an origin of transportation.
The same four potential sources were resolved in the 2017 source resolution. The contribution of factor 1 to As concentration was 83.5%, an increase of 17.7% relative to the contribution to As concentration in 2008. Based on the comparison of AI and RI values of As concentrations (Figure 4), As showed a trend of severe accumulation in the second rim of the study area, and the spatial location distribution was similar to that of 2008, so factor 1 was considered to have the same source as in 2008, i.e., agricultural sources.
The contribution of factor 2 to Pb and Cr concentrations was 50.1% and 46.1%, respectively. Similar distributions of the two heavy metals were found spatially (Figure 3), with two high values of point pollution zones at the northern center, whose spatial distribution is related to the ”electronic information industry cluster” and the ”one area and two zones” key development planning strategy in the urban tourism planning of the study area, and there is obvious The spatial correlation is obvious. In the source analysis in 2008, Cr was considered as a ”natural source” related to the parent material of soil formation, and Pb was considered to come from vehicle fuel combustion and tire wear. The planning and construction of the tourist area will certainly accelerate the weathering of the parent material and the change of soil properties, and according to the survey, from the completion of the scenic area to 2017, the reception of tourists increased by 11% year-on-year, which greatly enhanced the regional vehicle circulation. Therefore, it can be assumed that factor 2 represents the origin of human intervention.
Consideration of factor 3 represents 85.5% for Cd concentration. Table 1 shows that the mean content of Cd increased by 55.56%. Cd is widely used in various chemical industries and is also a major by-product of paint production, and some studies have shown that these heavy metals can cause enrichment of Cd in the soil through the emission of waste gas, wastewater, and sludge, through atmospheric deposition, surface runoff and solid waste piles(Huang et al., 2020). Meanwhile, the spatiotemporal variation on Cd shows an accumulation characteristic along a circle of urban areas, which may be related to the industrial upgrading in a sub-ring in the rapid urbanization stage of the study area and the out-migration of industrial parks. Therefore, factor 3 represents industrial sources.
The contribution of factor 4 to the Hg concentration was as high as 98.1%, but its mean concentration decreased to 0.057 mg·kg-1 compared to 2008, a decrease of 31.32%. Moreover, the spatial distribution of Hg has changed significantly, with the high-value area migrating from the eastern part of the study area to the western part, where medical device manufacturing enterprises such as Chengdu Medical City and Chengdu Medical Device Creation Center are located according to the survey, while some studies have shown that the great deal of healthcare facilities can be a significant source of Hg as well. Examples include clinical equipment like thermometers and sphygmomanometers that carry Hg (Y. Li et al., 2016). Unless properly handled, this would lead indirectly to soil by several pathways. Accordingly, factor 4 would qualify as a contributor to medical devices.
In summary, we can know that heavy metal pollution of arable soils is becoming increasingly serious due to the impact from mankind’s activity, and pollution sources tend to be complex and diversified. In the urbanization stage carried out in China, taking Chengdu city as an example, while the economy and infrastructure are developing rapidly, the environmental problems brought about by it should be given full attention, because when it is enriched in the soil it will not only seriously affect the growth of crops, but also further endanger human health and ecological environment through the food chain. Therefore, it is essential to understand the spatial and temporal characteristics of heavy metal pollution changes during urbanization construction for sustainable socio-economic development and construction of green cities.