Study On Pedo-transfer Function Of Saturated Hydraulic Conductivity Of Purple Soil In Hilly Area Of Chongqing

Purple soil is widely distributed in hilly areas in southwestern and southern China,especially in Sichuan and Chongqing.Purple soil accounted for more than 70% of the arable land in Three Gorges Reservior area,mainly being the sloping land.Purple soil has soft texture,loose structure and poor anti-erosion characteristics,with the occurrence of severe soil erosion.Meanwhile,the rainfall in the Three Gorges Reservoir area was unevenly distributed spatially and temporally,resulting in abnormal soil nutrient and soil erosion in the reservoir area.Therefore,this study took the 0-10 cm,10-20 cm and 20-30 cm purple soil in seven districts of the northeastern Chongqing as the research object.According to the different properties of soil layers in the study area,the distribution characteristics of soil physical and chemical properties were systematically analyzed,especially the main factors affecting soil saturated hydraulic conductivity.Hence,to establish the relationship between the saturated hydraulic conductivity and the influencing factors is the main purpose of this study.The adaptability of the traditional transfer function model in the research area was studied,and based on the multiple nonlinear regression and BP neural network were established for different soil layers.This can provide some suggestions and decision support recommendations for migration of solute,soil moisture forecasting and regional water consumption,soil and water conservation,agricultural non-point source pollution control.The conclusions are as follows:(1)With the increase of soil depth,soil saturated hydraulic conductivity,organic matter content,saturated water content were gradually decreased,while soil bulk density increases with increasing soil depth.Soil particles are mainly silt,followed by sand,clay,mainly concentrated between 0.05~0.002 mm.The saturated hydraulic conductivity is closely corelated with the content of organic matter and saturated water content,but no clear relationship with the bulk density and soil texture.(2)The saturated hydraulic conductivity increases with the increasing organic matter content for the three soil layers,and they showed good correlation,the curves between them exhibit an exponential function relationship.The saturated hydraulic conductivity decreases with the increasing soil bulk density for the three soil layers,also showed the exponential form,the effect of soil bulk density on saturated hydraulic conductivity is not significant.There is significant positive correlation between the saturated water content and saturated hydraulic conductivity for the upper layer soil.While the correlation of 20-30 cm soil layer is not significant,with a quadratic curve relationship.The effect soil texture on saturated water ratio has little effect,only 20-30 cm clay layer had a significant negative correlation with saturated hydraulic conductivity,and other correlations were not significant.(3)It is not ideal to estimate the saturated hydraulic conductivity by using the transfer function of Campell,Cosby,Saxton,Wosten1997,Wosten1999 and Puckett,and the prediction results of large deviation.It indicates that the existing transfer function model can not be applied to the prediction of saturated hydraulic conductivity in this study area.(4)The study established the multiple nonlinear regression model,which can predict well the saturated hydraulic conductivity for all the three layers.The predicted values have good fit with measured ones.It is feasible to forecast the soil saturated hydraulic conductivity through multivariate nonlinear transfer function model by using organic matter content,saturated water content,soil texture.(5)Using the MATLAB working platform to establish the saturated hydraulic conductivity BP neural network pedo-transfer model,the predicted values from the BP model is basically the same as the measured value,the model with the highest prediction accuracy in this paper.(6)The soil saturated hydraulic conductivity established from multivariate nonlinear regression model and BP neural network transfer model have high prediction accuracy,the error is smaller in BP neural network compared with multivariate nonlinear model.Howerver the construction of BP neural network process is complex,while the multivariate nonlinear regression model is simple and easy to operate.In practical application,the forecast model should be selected reasonably according to the realsituation.