南疆干旱区土壤盐渍化的高光谱响应与定量反演模型构建研究

Soil salinization is the main obstacle factors of agricultural development and sustainable development, is the one of the root cause of the ecological environment deterioration in the southern of xinjiang too. It is imperative that Rapid and accurate grasp the soil salinization information and do corresponding management plan. Hyperspectral remote sensing is a new type of method for monitoring soil salinization, and carries on the corresponding ground remote sensing test is the important basis for guaranteeing its monitoring accuracy. Therefore, the research object of the project is the southern of Xinjiang, which salinization area is wide and salinization degree is heavy,use ASDFieldSpec-FRTM spectrometer to collect different degree of soil salinization spectrum information after field and indoor standardized treatment, indoor spectrum information of the different artificial configuration humidity and different soil salt content is collected, combined with the soil physical and chemical properties data, using neural network analysis and principal component analysis, the partial least squares method etc. statistical analysis method, to discusses the soil salinization spectral features representation; Reveals the law of soil moisture effecting soil salinization spectrum characteristics; Establish the soil salinization spectrum evaluation index and standards that is propitious to this area; Construct hyperspectral quantitative inversion model of the content of soil salt. Thus high precision ground data and the theoretical basis is presented for hyperspectral remote sensing technology used in soil salinization monitoring in the area.

土壤盐渍化是南疆农业开发和可持续发展的主要障碍因素，也是生态环境恶化的根本原因之一。快速、准确的掌握该区土壤盐渍化信息而进行相应治理和规划势在必行。高光谱遥感技术是监测土壤盐渍化的一种新型手段，而进行相应的地面遥感试验是保证其监测精度的重要依据。因此，本项目以盐渍化面积广、盐渍化程度重的南疆土壤为研究对象，利用ASDFieldSpec-FRTM光谱仪野外实地、室内标准化处理后采集不同程度盐渍化土壤的光谱信息，室内采集人工配置的不同湿度与不同盐分含量土壤的光谱信息，结合土壤理化性质数据，采用神经网络分析、主成分分析、偏最小二乘法等统计分析方法，探讨土壤盐渍化的光谱特征表征；揭示土壤水分对盐渍化土壤光谱特征的影响规律；建立适宜于该区土壤盐渍化光谱评价的指标与标准；构建土壤盐分含量的高光谱定量反演模型。从而为高光谱遥感技术在此地区土壤盐渍化监测的应用提供高精度的地面数据与理论依据。

土壤盐渍化是南疆农业与生态环境可持续发展的主要障碍因子。本项目以南疆盐渍化土壤为研究对象，对盐渍化土壤的高光谱特征及土壤盐分的高光谱定量反演模型构建等内容进行了深入研究。通过分析不同盐渍化程度土壤的光谱反射率数据，表明盐分的特征波段在400~800nm 波段范围内。比较含盐量与电导率的光谱建模精度后表明含盐量的光谱定量反演精度明显高于电导率。利用模糊k-均值聚类方法将土壤反射光谱特征分为四种光谱类型，结合PLSR方法进行分类建模，可以显著提高模型的预测精度和稳定性。对盐基离子的光谱定量反演研究结果表明，不同离子的最佳反演模型所使用的预处理方法不尽相同，选用最佳的方法对光谱数据进行预处理有益于模型的建立和反演精度的提高，不同离子的反演精度也不同，除了K+、Na+与光谱反射率的相关性较差且几乎不能反演，其他5种离子（HCO3-、Cl-、SO42-、Ca2+、Mg2+）均有较好相关性，可以获得较好的反演精度。利用室内测定高光谱数据模拟Landsat8多光谱数据后构建的盐分光谱指数结合SVR模型可高精度的反演土壤盐分含量，模型的预测决定系数达到0.77。