Study on Spatial Distribution of Soil Heavy Metals in Huizhou City Based on BP--ANN Modeling and GIS
Yin Li,Chao-kui Li,Jian-jun Tao,Li-dong Wang
DOI: https://doi.org/10.1016/j.proenv.2011.09.306
2011-01-01
Procedia Environmental Sciences
Abstract:Using BP neural network model and GIS for heavy metal descript the spatial dynamics of distribution in Huizhou City, Guangdong Province, The Results indicate:(1)BP neural network model can learn the relationship between spatial location of sampling points with content of them and be able to forecast soil heavy metal content. by interpolating more sampling points,such as Fig.3 and Fig.4, Cd from the original refuses the normal distribution to approximate submit to normal distribution, with the increase in sampling points, the data can be used in geostatistical analysis. the range of Cu was changed from the original 136.24km to 25.20km, as TABLE 2. more realistic conditions.(2)Import the coordinates of the interpolation data and heavy metal content data into Arcgis, the most significant pollution is Pb, most of the area level over the natural background value .Zn, Cd have a high aggregation in some areas, the majority does not exceed the natural background value. Cu does not exceed the background, as shown in Figure 11 to 14.Therefore, the most noteworthy is the Pb pollution. Use the single factor index method, the four metal did not meet pollution threshold, indicating the study area have not been heavy metal pollution of soil.(3) Compared original data and interpolated data of the background index, as shown in Table 2 and Table 3, the nugget and sill ratio of Cu and Zn decreased, indicating that as interpolation points increase, Cu, Zn by the random factor decreases, increased spatial correlation, Pb nugget and sill ratio increases, the spatial correlation decreases, Cu by the random factors the strongest, spatial correlation the weakest, Zn and Pb is mainly affected by structural factors.Soil is not a homogeneous body, rather a variant of spatial continuity, this highly complex spatial heterogeneity of soil allows us to study the spatial dynamics of change becomes very difficult, caused not only by nature but also by human activities. These changes are the spatial dynamics of Soil heavy metals(SHM) can become extremely complex and, thus, the description on spatial distribution of SHM and spatial correlation of quantitative is very difficult. In the past there have been many models trying to determine the relationship between them, but these models are more or less the existence of defects[1]. In this paper, an unconventional Modeling Artificial Neural Networks method, combined with GIS technology to the spatial correlation of SHM and its spatial distribution were studied. Using neural network model to determine spatial distribution and pollution of SHM.