Study On Prediction Method Of Spatial Distribution Of Heavy Metal Cadmium In Plain Soil

In recent years, soil heavy metal pollution has caused wide public concern along with the rapid development of economy and society. Accurate information of the spatial change of soil heavy metal within the region is important for prevention of soil heavy metal pollution. It is an effective means and inevitable way to improve the predicting accuracy of spatial distribution of soil property use related environmental factors as auxiliary variables. However, it is hard to use the commonly used environmental factors as auxiliary variables in regions with a flat terrain. This study takes the soil cadmium content in the hinterland area of Chengdu plain of Sichuan Province as a case, analysze the influence factors of soil cadmium content in this plain region, employs radial basis function neural network ?RBFNN? and high accuracy surface modeling ?HASM? to to build a prediction method of spatial distribution of soil cadmium content, and simulates the spatial distribution pattern of soil cadmium content in the study area. The main conclusions are as follows:?1? Results indicated that the cadmium content in soils of this study area ranged from 0.13 to 0.61 mg.kg^-1, with an average value of 0.21 mg.kg^-1. This result suggested that the content of cadmium not exceeded the Grade II of soil environmental quality standard ?GB 15618-1995?. The contents of cadmium in west plain area are much higher than that of east area. The proportion of soil samples exceeded the standard of cadmium were 6.87%, most of which were distributed along the jingma river.?2? The nugget/sill ratio was 43.7%, which suggests their spatial variability were determined by the co-effects of structural and random factors. The variance analysis and regression analysis showed that the distance from the sampling sites to river ?jingma river?, NDVI, road density of high level road ?G\S\X? and low level road ?Y\C\Z? played important roles in controlling the content of soil cadmium. The above four factors can respectively explained 22.0,12.8,5.7 and 1.4% of soil cadmium variability. This result revealed that the heterogeneity of soil cadmium was determined by both natural geological factors and human activity factors, and natural geological factors were more important than human activity factors in controlling soil cadmium variability. Moreover, the impact range of jingma river were 10 km, and the the impact ranges of high level road and low level road were 1.2-1.5km and 150-200m.?3? a combined method ?RBF1_HASM? using the distance from the sampling sites to river, NDVI and road density as auxiliary variables was constructed to predict the spatial distribution of soil cadmium content across study area. The results show that this combined method obtains a much lower prediction bias. Compared with other three methods, the mean relative error ?MRE?, the mean absolute error ?MAE?, and the root mean squared error ?RMSE? are reduced by 5.56%-17.65%, and can presents more realistic details of the spatial variation of soil cadmium. Therefore, the proposed method obtained the high accurate spatial distribution information of soil Cd across study area.Based on a full analysis of influence factors of soil cadmium content in the plain region, this study rationally chooses and describes the influence factors of spatial distribution of soil cadmium content in the plain region, and used artificial neural network model and high accuracy surface modeling to built a combined method to predict spatial distribution pattern of soil cadmium content in a plain region, and obtained a relatively high accurate spatial information of soil cadmium content. This method provides an efficient approach to accurately obtain spatial distribution patterns of soil properties in other plain area.