盐渍化土壤水分微波雷达反演与验证

Inversion and verification of salinity soil moisture using microwave radar

土壤介电常数是微波遥感进行土壤含水率测量的物理基础,尤其介电常数实部是必须解决的问题,土壤介电特性的研究显得尤为重要。该文目的是试验与评价C波段RADARSAT-2 SAR(synthetic aperture radar)数据模拟土壤介电特性,进而反演土壤水分的性能。以受盐渍化影响较严重的内蒙古河套灌区解放闸灌域为试验区,首先回归分析了介电常数实部与SAR四极化后向散射系数、地表粗糙度的复杂关系,并与Oh经验模型对照,其决定系数R2为0.859 7,模拟精度较高;然后验证常用的2个介电常数模型,Dobson半经验模型、Hallikainen简化实部经验模型模拟的介电常数实部与实测值的决定系数R^2分别为0.935 9、0.869,表明2个模型均能模拟地表土壤水分与介电常数实部的密切关系;最后构建了Dobson模型、Hallikainen简化实部模型反演土壤含水率的模型,并与统计回归模型比照,其模拟数值与土壤实测值的决定系数R2分别为0.803 8、0.737 4、0.842 1,均方根误差RMSE分别为5.2%、5.7%、5%。Dobson模型与统计回归模型反演结果与实地土壤墒情分布较为吻合,具有良好的精度和适用性,从而建立了一个较为完整的土壤介电特性研究体系,为微波遥感监测土壤水分奠定了基础。

Soil dielectric constant is the physical basis for soil moisture simulation based on microwave remote sensing,and especially the real part of the dielectric constant is of great significance to the research of the soil dielectric characteristics.Main aim of this study was to investigate capability of C-band RADARSAT-2 SAR（synthetic aperture radar） data applied in the soil dielectric characteristics monitoring and the soil moisture inversion over agricultural fields.Bare area of Jiefangzha sub-district of Hetao Irrigation District in Inner Mongolia of China was selected as the study region,which was influenced by soil salinization seriously.In order to achieve above purposes,an image of Radarsat-2 SAR was bought in April 2016,which has a kind of four fine polarization SLC（single look complex） format,covering an area of 25 km×25 km with 8 meter ground resolution.Taking spatial uneven distribution of the saline soil into account,100 sampling points were designed in the study area,and soil digging depth was 10 cm.Hand-held GPS（global positioning system） receiver was used to record coordinates of the sampling points.The experiment data included the soil dielectric real constant,surface roughness,surface temperature,percentages of clay and sand particles,soil bulk density and soil moisture.Agilent microwave network analyzer was used to measure the real part value of soil dielectric constant with coaxial probe method.Surface roughness was measured using centimeter grid profile plate to calculate the value of RMS（root mean square） height and the correlation length,and then composite roughness was got to represent the surface roughness in later research.Real-time ground temperature of the sampling points was measured by geothermometer.Particle analysis was fulfilled with laser particle size analyzer named Helos/B,obtaining the percentage content of clay and sand particles.Soil bulk density was measured by ring cutter.Soil moisture was measured by way of drying.SAR scape module of ENVI software was mainly used to perform the radar image processing,including radiometric calibration,geometric correction,slant range turning and filtering.Four polarization back scatter coefficient values corresponding to the sampling points were extracted based on previous results by spatial analysis module of Arc GIS software.In order to analyze complex relationship between the real part of the dielectric constant with SAR four polarization back scattering coefficients and surface roughness,firstly Oh empirical model was established,for which the relative relationship was significant between simulated and measured soil moisture,and the value of R^2 was 0.8209.Results showed that Oh model can offer precise real part value of the dielectric constant to inverse the soil moisture based on the soil dielectric model by means of the remote sensing and surface roughness data.Then Dobson semi-empirical dielectric models and simplified Hallikainen real part experience model were verified,and the R^2 between the measured and simulated real part values was 0.935 9 and 0.869 respectively,which indicated that the 2 models can simulate close relationship of the surface soil moisture and the real part of the dielectric constant.Finally Dobson model and Hallikainen simplified real part soil moisture inversion model were constructed.Compared with the statistical regression model,it looked like that relative relationship between simulated and measured value was significant,and the value of R^2 was 0.803 8,0.737 4,and 0.842 1,respectively,for the former 2 models and the statistical regression model,the RMSE（root mean square error） value was 5.2%,5.7%,and 5% respectively.The inversion results of Dobson model and statistical regression model were similar with the field soil moisture distribution,so they had good precision and applicability.Without considering the surface roughness,the simulation result of Hallikainen simplified real part model was then slightly worse than the other 2 models.The soil dielectric characteristics researching system and the moisture retrieval models established in this study can promote the application of the microwave remote sensing in the soil moisture monitoring.