Extraction Of Soil Salinization Information In Yutian Oasis Based On Polarized Decomposition Information Of PALSAR Data

In this paper the Yutian Oasis in southern section of the Taklimakan Desert which has long-term research foundation,was chosen as the study area,using GIS and remote sensing techniques analysis the soil physical and chemical properties.Polarimetric information obtained from quad-polarized PALSAR-2?Phased Array type L-band Synthetic Aperture Radar?data in a variety of target polarization decomposition treatment,and SVM classification trained with 11 less noise polarized parameter's as the best classification features selecting through visual interpretation.Wishart maximum likelihood classification,support vector machine?SVM?classification and polarimetric decomposition methods are combined,and then different levels of soil salinization information extracted.Classification results analyzed and validated by Landsat-8 OLI image of the study area combined with field investigation.The four-polarization back-scatter coefficient values corresponding to the sampling points were extracted based on the previous results by the spatial analysis module of ArcGIS.Analysis the correlation between backscatters of sampling points and soil salt content,water content,pH value.On the basis of,the multiple linear regression?MLR?,geographically weighted regression?GWR?and back propagation artificial neural network?BP ANN?adopted to establish the quantitative inversion models of soil salt content.Polarization entropy?H?and scattering angle?alpha?extracted from quad-polarized PALSAR-2 data using H/A/?target polarization decomposition treatment.The spatial distribution of the ground features obtained by wishart-H/A/?Classification method,and then H-?feature space was established and determine the target features and the saline soil types.The main findings are as follows:1)Polarimetric information obtained from quad-polarized PALSAR-2 data in a variety of target polarization decomposition treatment.SVM classification trained with 12 less noise polarized parameter's which included T₁₁?anisotropy_lueneburg?entropy_shannon?Span?Freeman2_Vol?entropy_shannon_Inorm?TSVM_alpha_s?Yamaguchi3_Odd?H/A/a_l1?VZ3_Odd?Krogager_kd?Cloude_T11 as the best classification features selecting through visual interpretation.Wishart maximum likelihood classification,support vector machine?SVM?classification and polarimetric decomposition methods are combined,and then different levels of soil salinization information extracted.Classification results analyzed and validated by Landsat-8 OLI image of the study area combined with field investigation.The results showed that,by comparing the confusion matrix,the classification accuracy of SVM methods higher than wishart classification.The overall accuracy and Kappa coefficient was increased from 80.48%to 88.00%,0.73 to 0.83,respectively.The overall accuracy improved 8.78%,Kappa coefficient increased 0.14.2)Considering the distribution of saline soil spatial variability,68 points were designed as sampling points,Hand-held GPS?global position system?receiver was used to record the coordinates of sampling points,0-10 cm topsoil samples were collected in the field and Soil total soluble salt content was measured in the indoor.The four-polarization back-scatter coefficient values corresponding to the sampling points were extracted based on the previous results by the spatial analysis module of ArcGIS.Total salt content was taken as dependent variable Main soil parameters as independent variables,the multiple linear regression?MLR?,geographically weighted regression?GWR?and back propagation artificial neural network?BP ANN?adopted to establish the quantitative inversion models of soil salt content.Results shows that the inversion accuracy of the MLR and GWR models was weak and their standard error?SE?was 0.531and 0.482 g.Kg^-1,respectively.For the ANN?BP?model,the SE and R₂ was only 0.276 and 0.549,and its accuracy better than the other two models.The established ANN?BP?model in this paper can reduce the smoothing effect compared with the two traditional models and improve the accuracy and reliability of model predictions,which meets the needs of soil salinity monitoring to a certain extent,and can promote and develop the application of microwave remote sensing in the soil salinity monitoring.3)Polarization entropy?H?and scattering angle?alpha?extracted from quad-polarized PALSAR-2 data using H/A/?target polarization decomposition treatment.The spatial distribution of the ground features obtained by wishart-H/A/?Classification method,and then H-?feature space was established.The results showed that:the method of target polarization decomposition have the more useful information to some extent,which is convenient for visualizing interpretation and extracting target information.The wishart-H/A/?Classification results indicated that saline soil mainly distribute in ecotone between oasis and bare land area.For validation,the results quantitatively analyzed by field data and confusing matrix,and the classification accuracy are 85.90%.As the change of H and?value,the ground objects have significant differences in H-?feature space,the saline soil mainly belongs to Z₉ region with polarization entropy H of 0⁰.67 and?<42.5°,and the scattering types are low-entropy surface scattering.The vegetation belongs to Z₁?Z₂?Z₄ and Z₆ regions.