Analysis Of Influencing Factors And The Content Of Soil Available Fe, Mn, Cu And Zn At Agricultural Landscape In Hilly Region

Trace elements in the soil are an important part of soil fertility, available trace elements can be absorbed by pants, it not only affects the normal growth and development of plants, and it is important also affect the quality and yield of agricultural products.In this paper, the study area is located in Jingjin, Chongqing, China; we use classical statistical analysis and geostatistical analysis to discuss the impacts of soil parent material, cropping systems, slope position and other topographic factors on soil available Fe, Mn, Cu and Zn. The categorical variables (soil parent material, cropping systems, and slope position) are taken into the regression model to analyze the influence of trace element in the soil trace elements. A machine learning method-Random Forest (RF) can be used to study the importance of each factor on the soil trace elements. Compare the precision of three predictive models, dummy variable regression, Kring and RF regression. The main conclusions are as follows:a.The range of available Fe, Mn, Cu, and Zn in study area is 1.01-299.00mg/kg, 1.58-300.00mg/kg,0.07-9.98mg/kg and 0.42-17.40mg/kg respectively. The average value is 64.40mg/kg,84.79mg/kg,1.30mg/kg and 3.59mg/kg.According to the standard of the second soil survey, soil trace elements is in a rich level in Jiangjin.b.According to the trend analysis, we found that available Fe is in a quadratic function relationship in the north-south direction, available Zn is in a "U" type in the east-west direction with the ends of the high content and middle of the low content. Available Mn and Cu are no significant trends in east-west and north-south direction. The spatial variability of soil trace elements were studied based on semi-variance function, Available Fe, Mn and Zn can be fitted with Exponential model and available Cu can be fitted with Gaussian model.c.Available Fe, Mn, Cu and Zn showed a significant negative coreelation with soil pH in Jiangjin farmland soil. Available Fe, Mn, Cu and Zn showed a significant positive correlation with soil organic matter and nitrogen. Available Fe showed a significant negative correlation with soil available phosphorus, but available Zn showed a significant positive correlation. Only available potassium was significantly negatively correlated with available Mn.d.Major plantation system in Jiangjin are rice, orange, pepper, corn-potatoes. The results showed a significant effect of cropping system(p<0.05) on the content of soil available Fe, Mn, Cu and Zn. The average content of four trace elements was lowest in the peper soil, respectively 14.33mg/kg,44.00mg/kg,0.66mg/kg,2.49mg/kg.e.Tree soil parents in Jiangjin are Shaximiao group, Ziliujing group, Suining group. The results showed a significant effect of soil parents(p<0.05) on the content of soil available Mn, Cu and Zn, soil parents and available Fe content in the soil had nosignificant effect. The average content of soil avaliable Fe, Mn, Cu and Zn was highest in the soil were developed from Ziliujing group,respectively 71.17mg/kg, 104.2mg/kg,1.88mg/kg and 4.42mg/kg.f.Six slope position in Jiangjin:ridge, upper slope, middle slope, flat slope, lower slope and valley. Slope position only affects content of soil available Mn(p<0.05) and had no effect on soil avaliable Fe, Cu and Zn.j.Build dummy variable regression, the accuracy has been improved when joining parent materials and cropping systems to the regression, raised 4.0%,6.2%, 4.9%and 7.1%respectively.k.Build Random Forest Regression model to determine the importance of different variables on soil trace elements. The most important factors of available Fe are pH, cropping system-Rice and organic matter. The most important factors of available Mn are pH, cropping system-pepper and organic matter. The most important factors of available Cu are organic matter, elevation and pH. The most important factors of available Zn are pH, nitrogen and cropping system-orange,h.commpare ordinary Kriging, dummy variable regression and RF regression accuracy of the prediction soil trace elements and founded that dummy variable regression and RF regression of available Fe and Cu have a higher precision. Kring is the best method to predict available Mn and Zn. Overall, the effect of the random forest regression prediction relatively stable, introducing random forest in soil nutrients prediction has advantages.