Remote Sensing Monitoring And Cotton Plantation Benefit Analysis Of Cotton Field Soil Salinization In Alar Reclamation Area

Soil salinization caused the reduction of agricultural productivity and crop yields in arid and semi-arid areas.So monitoring the spatial distribution range,area and degree of salinization in different depths of section farmland,which is of great significance to prevent and control local area`s soil salinization effectively,alleviate the contradiction between the shortage of available water resources and excessive agricultural water demand and improve the local crop yield.At present,the application of remote sensing technology combined with a certain number of soil sample data has shown great potential and advantages in obtaining high efficiency,low-cost and periodic acquisition of soil salinization information in a certain depth of cotton fields and the evaluation of agricultural economic benefits.This study took the Alar reclamation area in southern Xinjiang as the study area,and aimed at the salinized soil in the cotton growing area in the study area.It used Landsat 8 image analysis of 16 types of maximum or minimum composite data for many years,ground survey data,apparent conductivity data and estimated yield data,using Multiple linear Regression(MLR),Partial least Squares Regression(PLSR),Principal Component Regression(PCR),Random Forest(RF),Neural Network(NN),Support Vector Machine(SVM),Cubist and other modeling methods and studied the inversion model monitors of best soil conductivity based on apparent conductivity data at the regional scale.And we carried out remote sensing monitoring and digital mapping of soil salinization in different sections of cotton field to evaluate the negative effects of soil salinization on the cotton yield of the main economic crop.The main results are as follows:(1)Research on Conductivity Inversion Model Based on Electromagnetic Induction Data at Regional ScaleLinear modeling methods such as multiple linear,partial least squares and principal component regression have validation sets of R~2 in the range of 0.84?0.88 for different profile conductivity inversion models at the field scale,while the corresponding R~2 for regional scale models is in the range of0.33?0.51.Linear modeling methods are not applicable to the construction of soil profile conductivity inversion models at the regional scale.The model accuracy was significantly improved when inversion of regional scale soil profile conductivity was performed by nonlinear modeling methods such as random forest,neural network and support vector machine,with the validation set R~2 ranging from 0.54?0.83,among which the RF model had the best inversion of different profile conductivity with the validation set R~2 ranging from 0.79?0.83.The use of RF modeling methods to construct inversion models of different profile conductivity at regional scale can significantly reduce the number of soil profile samples collected,improve the sampling efficiency and reduce the sampling cost.(2)Inversion Study of Soil Profile Conductivity in Cotton Fields Based on Time Series Remote Sensing DataThe random forest modeling method is used to invert multiple apparent conductivity data of the same image element into conductivity and take the average value as the final conductivity of the image element,so that the conductivity of different profiles corresponding to a single image element is more representative.The correlation between vegetation index and conductivity can be improved by the maximum or minimum composite values of vegetation index over the years.When using the multi-year maximum or minimum composite values of 16 vegetation indices to construct Cubist models of conductivity at different profile depths,R~2 was 0.80,0.74,0.72,and the RMSE ranged from 0.86?1.50d S/m.MAE was 0.55?1.08 d S/m,and RPD ranged from 1.83?2.28.The results of soil conductivity mapping in different depths of the profile showed that the soil salinity in the cotton planting area of the reclamation area showed a trend of aggregation in the middle and bottom.About one-third of the cotton fields in the whole reclamation area were not affected by soil salinity,and the cotton fields with more serious salinization damage were mainly distributed in the north,northwest,south and southeast.(3)Analysis of Cotton Planting in Alar Reclamation Area of XinjiangThe multi-temporal cotton yield prediction model based on the MLR modeling method was able to predict cotton yields in the study area better,and the model accuracy R~2 was 0.83,RMSE was 433.25kg/mu.The overall cotton yield in the study area of the northwestern,western,southwestern,and southern showed that they were lower than that in the northern,central,and southeastern areas.The northwestern,northern,southwestern,southern,and part of the southeastern regions had a loss of cotton cultivation throughout the year.Soil salinity showed high values in the low yielding areas of cotton within 1.000 m soil profile depth,while the soil salinity in the high yielding cotton region showed a low value distribution.