Study On Effects Of Soil Environmental Factors On The Cadmium Enrichment In Rice From Typical Producing Areas

As the main food crop in China,the quality and safety of rice directly affect the human health.There is cadmium(Cd)pollution in various degrees in the soil of China.In recent years,the problem of cadmium pollution in rice has attracted much attention.Cd enriched in rice is greatly affected by environmental factors which are various with strong regional variability.Therefore,it is difficult to establish the coupling relationship between Cd pollution in soil and quality/safety of rice.In the previous studies,building a transport model of soil-rice Cd was mainly based on a regression model which was unrepresentative with few modeling parameters.In this work,we choose areas in Sichuan Mianzhu,Guangdong Foshan and Hunan Xiangtan with different risk levels to carry out a research.Using the soil-rice cooperative sampling method,553 pairs of soil-rice samples are collected,followed by the determination of 21 environmental factors of soil.Based on the descriptive statistical analysis of these environmental factors,the method of the random forest regression,generalized linear LASSO regression and mixed linear Cubist are applied to build a transport model of soil-rice Cd.As a result,environmental factors that influence the enrichment of Cd in rice in different regions can be explored.The main results are as follows:(1)Results of descriptive statistics and variance analysis show that pH of soil and average contents of SOM,S,P,Si,Ca,Mn and Zn in soil are various in different regions.The collected soil samples are obvious variations with strong geographical representation.(2)Compared with LASSO and Cubist,the Random Forest analysis can explain the variates better,indicating that the influence of environmental factors on the enrichment Cd in rice is complicated and the modeling efficiency of linear model is lower than that of non-linear model.The results further show that p H of soil and Ca contents in soil are the most important factors that determining the enrichment of Cd in rice.Besides,contents of Cd,Fe,Mn and Si in soil also make a difference.Effects of SOM are not obvious.(3)With the increase of the number of modeling samples,the prediction error of the model gradually decreases and finally remains to be stable.The research shows that the modeling prediction error of Mianzhu in Sichuan Province is unstable while prediction errors of Foshan in Guangdong Province and Xiangtan in Hunan Province tend to be stable around 70~80 and 110~120,respectively.(4)Although environmental factors of soil are various in different investigated regions,results show that the effects of environmental factors on Cd enrichment in rice are universal.Contents of Cd in rice decrease with the increase of p H of soil and contents of Ca,Fe and Mn.However,the degree of the decreasing is various in different regions.With the increase of Cd in soil,contents of Cd in soil remain to be basically unchanged or slightly decreased.Contents of SOM in soil have no obvious effects on the enrichment of Cd in rice.In summary,the results show that the prediction ability of the random forest model is the best,with a high correlation(>90%)between the predicted value and the actual value in three different sampling areas.Using this model,the main factors controlling the enrichment of Cd in rice can be screened and the influence of different environmental factors on enrichment of Cd in rice can be simultaneously determined.The research will provide scientific and technological support for the remedy of pollution and source analysis of Cadmium contaminated areas.