Trace Metal Pollution Of Agricultural Surface Soil And Its Multi-scale Relationship With Landscape Pattern At Multiple Scales In Guangzhou

Trace metal contamination in agricultural soils has become a severe problem in China due to rapid industrialization and urbanization. Under the interference of human activities, the trace metal contaminations in the soil combined point- with non-point source pollution. The migrations and accumulations of trace metals into the soil were influenced by the composition and configuration of landscapes(i.e. landscape pattern) at a series of scales from a soil plot to a region. Thus, it is essential to explore the spatial pattern of soil contamination and its relationship with landscape pattern at multiple scales. However, previous studies mainly focused on a single spatial or temporal scale. With regards to the high spatial variability of soils and landscape features, it is necessary to conduct a multi-scale approach to explore soil contamination-landscape pattern relationship. Based on 641 agricultural top soil samples, land use and cover map in Guangzhou of 2005, this study explored the spatial patterns of agricultural soil contaminations for Cd, Pb, As, Cr, and Ni and their relations with landscape pattern at multiple spatial scales. Research methods included multi-scale nested model, landscape pattern analysis, pollution indices, Pearson or Spearman correlation analysis, geographically weighted regression analysis and partial redundancy analysis. Our findings were as follows:(1) Out of 641 samples, most samples were uncontaminated and only 5-18% samples were slightly contaminated with Cd, As and Ni. Soil trace metal contamination levels in different agricultural land types from high to low were in the order of orchard/vegetable> paddy/dry land>forest, following a descending amount of agrochemical inputs from different agricultural practices. In most cases, there were no significant differences of soil trace metal contamination levels among different agricultural land types due to a probable lithogenic source of most trace metals.(2) The spatial variation of trace metal in agricultural soils varied with the scales studied. Compared with the single-scale ordinary kriging interpolation, the multi-scale nested model was more effective to explore the spatial variability structure of concentrations of Cd and Pb in the soil indicated by higher interpolation accuracy. The interpolation results of the multi-scale nested model showed that the concentrations of Cd and Pb in agricultural soils of Guangzhou were generally exceeded the background values of Guangdong Province, but most samples did not exceed China’s maximum allowable concentrations for agricultural soils. High concentrations of Cd in agricultural soils were primarily found in vegetable fields near city center and partly in districts of Conghua, Nansha, Huadu and Baiyun, of which the soil Cd probably resulted from traffic activities, sewage irrigation and chemical fertilizer.(3) Pearson or Spearman correlation analysis, geographically weighted regression analysis(GWR) and partial redundancy analysis(PRDA) showed that the relationships between soil trace metal contamination and landscape pattern were scale-dependent. As indicated by three model parameters of adjusted R2, the corrected Akaike information criterion(AICc) and spatial-autocorrelation tests of model residuals(Moran’s I indices), the GWR model was more effective than OLS model in exploring the correlations between soil contamination and landscape pattern at the three spatial scales(2km×2km, 5km×5km, 10km×10km) with the highest explanatory power at the 5km×5km scale. The correlation analysis and GWR model results showed that the soil contamination- landscape pattern correlations varied with spatial locations with either positive or negative correlation coefficients. Most soil trace metal contamination levels were significantly correlated with parent material composition(the percentage of deposit or granite) and land use pattern(the percentage of different land use types especially for water) regardless of scales, implying that alluvial soils and pollutants migrated with rivers or sewage irrigations were important input ways of trace metals. Moreover, the significant correlations were primarily found in districts of Baiyun, Panyu, and Nansha. The PRDA results showed that landscape patterns explained 12.8-29.1% of soil trace metal contamination variations, and the explanatory powers increased with the scale studied(10km × 10km>5km × 5km>2km × 2km). On different spatial scales, the effects of the three subsets of landscape pattern variables(i.e. parent materials-PM, distance-density variables-DD, land use-LU) on the variance of soil trace metal contaminations varied with spatial scales. More than half of the variation in soil trace metal contaminations was explained by two or more subsets of landscape pattern variables. The PM and DD generally showed stronger influences at a large scale than a small scale, whereas the influences of LU decreased with the scales employed.