基于高光谱技术的盐渍化土壤水盐含量估测研究

发布时间：2017-12-28 13:37

本文关键词：基于高光谱技术的盐渍化土壤水盐含量估测研究　出处：《石河子大学》2016年硕士论文　论文类型：学位论文

【摘要】：土壤盐渍化是最常见的土地退化过程,自然因素和人类不合理的使用使得土壤盐渍化、次生盐渍化问题突出,对环境构成风险。新疆属于典型的干旱半干旱地区,由于温度高、降雨稀少,土壤蒸发强烈,土壤中盐分随水分蒸发带到地表,水分变成水蒸气进入了空气,而大量盐分在土壤表层累积,形成盐渍化土壤。盐渍化土壤不仅低产、低肥、作物存活率低而且可持续利用差,因此,对土壤实行盐渍化监测是十分必要的。传统的监测方法不仅需要大量的人力、物力、财力得出的结果既表面又不具有代表性,随着高光谱技术的诞生和发展逐渐弥补了这些缺陷。本文以新疆典型盐渍化灌区为研究对象,运用便携式高光谱仪获取了不同质地的土壤水盐含量和八大离子含量的光谱曲线,采用一阶微分、二阶微分、连续统去除的数据处理方法对土壤原始光谱进行变换,研究利用高光谱技术同步获取盐渍化土壤水分、盐分和八大离子的含量。主要结论如下:(1)从不同水分含量梯度看,两种质地的土壤的高光谱反射率均随水分含量的升高呈现下降趋势且砂土的反射率变化幅度明显大于壤土;从不同盐分含量来看,随着盐分含量的增加两种质地的高光谱反射率大体呈现增加趋势;从不同水盐含量梯度来看,原始的水盐高光谱反射率与水盐含量并没有表现出明显的相关性。(2)原始水盐光谱数据经过一阶微分、二阶微分和连续统去除处理后相关系数均得到明显提高,但对于不同水盐含量的原始光谱三种数据处理后的相关系数的增加幅度具有显著差异。在壤土中,0%含水量和10%含水量的水盐光谱曲线经过连续统去除,15%含水量的水盐光谱曲线采取一阶微分、19%含水量水盐光谱曲线经过二阶微分后更有利于相关系数的提高。而在砂土中,0%含水量的水盐光谱曲线通过连续统去除,10%含水量、15%含水量、19%含水量的水盐光谱曲线则经过二阶微分处理后的相关系数更显著的增加。(3)选取一阶微分、二阶微分和连续统去除后的显著波段建立了多元线性回归模型,当壤土含盐量6.38 m S/cm时对其的水分预测模型的R2分别为0.85、0.91、0.89、0.93、0.71,砂土含盐量7.45 m S/cm时对其的水分预测模型的R2分别为0.83、0.95、0.87、0.80、0.74、0.78,达到显著相关;在不同水分的盐分预测模型中,壤土含水量12%时对其盐分含量预测的R2分别为0.89、0.92、0.82、0.73,砂土含水量16%时对其盐分含量预测的R2分别为0.82、0.90、0.78、0.86、0.68,达到显著相关;在水盐同步预测模型中,含水量10%,含盐量5.83 m S/cm时壤土水盐含量同步预测的R2达到0.84,砂土含水量13%,含盐量7.45m S/cm时,水盐含量同步预测R2为0.90,有较好的预测结果。(4)盐渍土土壤中阴离子中含量最高的是Cl-,占总阴离子的百分比为51.21%,其次是SO42-、HCO3-,分别占阴离子总量的45.86%和2.93%;在阳离子中,含量最高的是Na+,占阳离子总量的60.51%,其次是K+、Ca2+和Mg2+,占阳离子总量的24.87%、13.74%和0.88%。说明该地区土壤类型为硫酸盐-氯化物盐土。对原始盐分离子与光谱反射率进行相关性分析表明原始七个离子的原始相关系数与光谱反射率并没有显现出良好的相关性,因此对其进行了一阶、二阶微分和连续统去除的处理,结果表明,K+、Na+、Cl-、HCO3-、Ca2+、Mg2+通过连续统去除后、SO42-一阶微分处理后更有利于敏感波段的提取。采用一阶微分、二阶微分和连续统去除处理后的七大离子敏感波段分别建立多元线性回归模型,在模型中SO42-、HCO3-、Cl-、Na+、K+、Mg2+、Ca2+的R2分别为0.84、0.70、0.71、0.79、0.74、0.85、0.76,P值分别为0.046、0.037、0.039、0.048、0.044、0.034、0.023,P0.05均达到良好的预测结果。
[Abstract]:Soil salinization is the most common process of land degradation. Natural factors and unreasonable use of human make salinization and secondary salinization prominent. Xinjiang is a typical arid and semi-arid regions, due to high temperature and rainfall, strong evaporation, soil salt to the surface with the water evaporation, water into water vapor into the air, and a lot of salt accumulation in the surface soil, saline soil formation. The salinized soil is not only low yield, low fertilizer, low crop survival rate and poor sustainable utilization, therefore, it is very necessary to carry out the soil salinization monitoring. The traditional monitoring method not only needs a lot of manpower, material resources, and financial resources, but also results from both surface and non representativeness. With the birth and development of hyperspectral technology, these defects are gradually made up. This paper takes Xinjiang typical salinization as the research object, using the portable spectrometer with high spectral curves obtained different texture of soil water and salt content and ion content of eight, the first order differential, two order differential, continuum removal method of data processing to transform the original spectral content of the soil, using hyperspectral technology, salinization of soil moisture salt and eight ions synchronous acquisition. The main conclusions are as follows: (1) from the gradient of different water content, high spectral reflectance of the two kinds of soil texture were increased with water content showed a downward trend and the variation of reflectivity is significantly greater than the sand loam; from different salt content, salt content increased with high spectral reflectance in two texture generally showed an increasing trend different from the gradient; water salt content, the original high spectral reflectance and water salt water and salt content did not show significant correlation. (2) after the first order differential, two order differential and continuum removal, the correlation coefficients of the original water and salt spectral data have been significantly improved. However, the correlation coefficients of the three spectra processed with different water and salt contents are significantly different. In loam soil, the spectral curve of water and salt with 0% water content and 10% water content was removed by continuum. The first order differential and 19% water content of water salt spectrum curve of 15% water content took the first derivative and water salt spectral curve after two order differentiation, which was more conducive to the improvement of correlation coefficient. In sandy soil, the spectral curve of water and salt of 0% water content is removed by continuum. The correlation coefficient of water and salt spectrum of 10% water content, 15% water content and 19% water content is increased more significantly after two order differential treatment. (3) to select significant band first-order differential, two order differential and continuum removal after established a multiple linear regression model, the forecasting model of water on R2 were 0.85, 0.91, 0.89, 0.93, 0.71 when the salt content of 6.38 m S/cm loam, sandy soil salt content of the water forecast model the R2 were 0.83, 0.95, 0.87, 0.80, 0.74, 0.78, 7.45 m S/cm, significant correlation; in different water salinity prediction model, soil water content of 12% when the prediction of the salt content of R2 were 0.89, 0.92, 0.82, 0.73, 16% when the soil water content prediction of the salt content of R2 were 0.82, 0.90, 0.78, 0.86, 0.68, significantly correlated in water and salt; synchronous prediction model, 10% moisture content, salt content of 5.83 m S/cm soil water salt content synchronization prediction of R2 reached 0.84, sand 13% moisture content, salt content of 7.45m S/cm, water and salt the content of the same The prediction of R2 is 0.90, and the prediction results are better. (4) soil in saline soil anion was the highest content of Cl-, the percentage of total anions was 51.21%, followed by SO42- and HCO3-, respectively, 45.86% and 2.93% of total anion; the cation, the content of Na+ was the highest, accounting for 60.51% of the total cation, followed by K+, Ca2+ and Mg2+, accounting for cation a total of 24.87%, 13.74% and 0.88%. It shows that the soil type of this area is sulfate chloride saline soil. The original salt ions and spectral reflectance correlation analysis shows that the original correlation coefficients of the original seven ions and spectral reflectance and did not show a good correlation, so the first and two order differential and continuum removal treatment, the results show that the K+, Na+, Cl-, HCO3-, Ca2+, Mg2+ by continuous after the removal of SO42-, the extraction of first order differential treatment is more conducive to the sensitive band. Multiple linear regression models are established by the first order differential, two order differential and continuum removal after seven ion sensitive bands, SO42-, HCO3-, Cl- in the model, Na+, K+, Mg2+, Ca2+ and R2 were 0.84, 0.70, 0.71, 0.79, 0.74, 0.85, 0.76, P respectively 0.046, 0.037, 0.039, 0.048, 0.044, 0.034, 0.023, P0.05 achieved good prediction results.
【学位授予单位】：石河子大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：S156.41

【相似文献】