基于地面实测热红外光谱和电磁感应技术的土壤盐渍化建模研究
发布时间：2018-10-13 19:30
【摘要】：土地荒漠化、土地沙化以及土地退化的原因众多,其中最主要就是土壤盐渍化。新疆作为我国最为重要的粮食基地和农业后备资源战略基地,其绿洲农业的稳定和可持续发展对于保障国家的粮食安全具有重大的战略意义。而当前灌溉农业的快速发展、土地的不合理开发以及土地利用方式地改变,都破坏了新疆绿洲的水盐平衡。新疆绿洲的农业生产和生态环境在土壤盐渍化、次生盐渍化的威胁下变得岌岌可危。目前,世界范围内资源调查、环境监测常用的方法是遥感技术。它具有很多明显的优势,如速度快、精度高、花费少又能综合宏观的对区域范围进行不同时段的多次监测。基于上述优点,它已经被用在大面积监测土壤盐渍化上。但相对应的地面实测实验也在近年快速发展,取得了丰硕的成果,电磁感技术和实测热红外发射率就是典型代表,其具备明显的优点。电磁感应技术(EM38电磁感应仪)可以直接获取土壤表观电导率数据,具有实时、快速、精度高的特点,但其对于细节的表达较为粗略,不能够完整地反映局部小区域的土壤盐渍化信息。而热红外发射率光谱能够获取地物的连续光谱信息,具有更为精细的、微米级的光谱分辨率,能够反映地物细微的光谱特征。本研究将这两种技术相结合,应用于土壤盐渍化这一实际问题上,以期为热红外遥感技术实现区域大尺度盐渍化监测提供基础。本研究在实验室详实、细致的野外工作的基础上,建立基于电磁感应技术与实测数据的轻、中、重三种土壤盐渍化解译模型。采用ISSTES方法对土壤热红外发射率和温度进行了分离。以尽量减小外界的各种干扰因素(仪器、环境、人为)为目的,通过相邻平均值法(Adjacent-Averaging)、Savitzky-Golay滤波法、分位数滤波(Percentile Filter)以及FFT Filter四种方法,分别对样点光谱数据进行了平滑去噪,通过最终分析,Savitzky-Golay滤波法的效果是最好的。用一阶微分变换、对数变换、倒数变换、对数的倒数变换、二阶微分变换5种数学变换形式对平滑降噪后的光谱与含盐量模型结合进行变换处理,分析它们之间的特定关系,提取敏感波段。最后进行了模型的拟合工作,运用了线性(简单线性模型)、非线性(对数模型、幂指数模型、抛物线模型、指数模型)、逐步多元回归模型以及偏最小二乘等方法。之后分别对7个模型进行了精度评价,通过不同情况的对比分析,找到最佳模型。
[Abstract]:There are many reasons for land desertification, desertification and land degradation, among which soil salinization is the most important. Xinjiang is the most important food base and agricultural reserve resource strategic base in China. The stability and sustainable development of oasis agriculture is of great strategic significance to ensure the national food security. However, the rapid development of irrigated agriculture, the unreasonable development of land and the change of land use mode have destroyed the balance of water and salt in Xinjiang oasis. Under the threat of soil salinization and secondary salinization, the agricultural production and ecological environment of Xinjiang oasis become precarious. At present, remote sensing technology is commonly used in worldwide resource survey and environmental monitoring. It has many obvious advantages, such as fast speed, high precision, low cost and comprehensive macro monitoring of different periods of time. Based on the above advantages, it has been used to monitor soil salinization in a large area. However, the corresponding experiments on the ground have developed rapidly in recent years and have achieved fruitful results. The electromagnetic induction technology and the measured thermal infrared emissivity are typical examples, which have obvious advantages. Electromagnetic induction technology (EM38 electromagnetic induction instrument) can directly obtain soil apparent conductivity data, which has the characteristics of real time, fast and high precision, but the expression of details is relatively rough. The soil salinization information of local small area can not be completely reflected. The thermal infrared emissivity spectrum can obtain the continuous spectral information of the ground object, and has a more fine, micron spectral resolution, which can reflect the fine spectral characteristics of the ground object. This study combines these two techniques and applies them to the practical problem of soil salinization in order to provide the basis for the realization of regional large-scale salinization monitoring by thermal infrared remote sensing technology. On the basis of detailed and detailed field work in laboratory, three kinds of soil salinization interpretation models based on electromagnetic induction technology and measured data were established. The soil thermal infrared emissivity and temperature were separated by ISSTES method. In order to minimize all kinds of external interference factors (instrument, environment, artificial), the spectral data of sample points are smoothed and de-noised by four methods: adjacent average method (Adjacent-Averaging), Savitzky-Golay filtering method, quantile filtering (Percentile Filter) method and FFT Filter method, respectively. Through the final analysis, the effect of Savitzky-Golay filtering method is the best. The first order differential transform, logarithmic transformation, reciprocal transformation, logarithmic reciprocal transformation and second-order differential transformation are used to transform the smooth denoising spectrum and salt content model, and the special relationship between them is analyzed. The sensitive band is extracted. Finally, the fitting work of the model is carried out. The methods of linear (simple linear model), nonlinear (logarithmic model, power exponential model, parabola model, exponential model), stepwise multivariate regression model and partial least squares are used. After that, the accuracy of the seven models was evaluated, and the best model was found through the comparative analysis of different cases.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S156.41
本文编号：2269672
[Abstract]:There are many reasons for land desertification, desertification and land degradation, among which soil salinization is the most important. Xinjiang is the most important food base and agricultural reserve resource strategic base in China. The stability and sustainable development of oasis agriculture is of great strategic significance to ensure the national food security. However, the rapid development of irrigated agriculture, the unreasonable development of land and the change of land use mode have destroyed the balance of water and salt in Xinjiang oasis. Under the threat of soil salinization and secondary salinization, the agricultural production and ecological environment of Xinjiang oasis become precarious. At present, remote sensing technology is commonly used in worldwide resource survey and environmental monitoring. It has many obvious advantages, such as fast speed, high precision, low cost and comprehensive macro monitoring of different periods of time. Based on the above advantages, it has been used to monitor soil salinization in a large area. However, the corresponding experiments on the ground have developed rapidly in recent years and have achieved fruitful results. The electromagnetic induction technology and the measured thermal infrared emissivity are typical examples, which have obvious advantages. Electromagnetic induction technology (EM38 electromagnetic induction instrument) can directly obtain soil apparent conductivity data, which has the characteristics of real time, fast and high precision, but the expression of details is relatively rough. The soil salinization information of local small area can not be completely reflected. The thermal infrared emissivity spectrum can obtain the continuous spectral information of the ground object, and has a more fine, micron spectral resolution, which can reflect the fine spectral characteristics of the ground object. This study combines these two techniques and applies them to the practical problem of soil salinization in order to provide the basis for the realization of regional large-scale salinization monitoring by thermal infrared remote sensing technology. On the basis of detailed and detailed field work in laboratory, three kinds of soil salinization interpretation models based on electromagnetic induction technology and measured data were established. The soil thermal infrared emissivity and temperature were separated by ISSTES method. In order to minimize all kinds of external interference factors (instrument, environment, artificial), the spectral data of sample points are smoothed and de-noised by four methods: adjacent average method (Adjacent-Averaging), Savitzky-Golay filtering method, quantile filtering (Percentile Filter) method and FFT Filter method, respectively. Through the final analysis, the effect of Savitzky-Golay filtering method is the best. The first order differential transform, logarithmic transformation, reciprocal transformation, logarithmic reciprocal transformation and second-order differential transformation are used to transform the smooth denoising spectrum and salt content model, and the special relationship between them is analyzed. The sensitive band is extracted. Finally, the fitting work of the model is carried out. The methods of linear (simple linear model), nonlinear (logarithmic model, power exponential model, parabola model, exponential model), stepwise multivariate regression model and partial least squares are used. After that, the accuracy of the seven models was evaluated, and the best model was found through the comparative analysis of different cases.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S156.41
【参考文献】
相关期刊论文 前4条
1 夏军;塔西甫拉提·特依拜;买买提·沙吾提;张飞;韩桂红;;热红外发射率光谱在盐渍化土壤含盐量估算中的应用研究[J];光谱学与光谱分析;2012年11期
2 侯艳军;塔西甫拉提·特依拜;张飞;买买提·沙吾提;依力亚斯江·努尔麦麦提;;荒漠土壤全磷含量热红外发射率光谱估算研究[J];光谱学与光谱分析;2015年02期
3 刘福汉,王遵亲;潜水蒸发条件下不同质地剖面的土壤水盐运动[J];土壤学报;1993年02期
4 李洪义;史舟;程街亮;李艳;;基于EM38的土壤剖面电导率预测研究[J];中国农业科学;2008年01期
，本文编号：2269672
本文链接：https://www.wllwen.com/kejilunwen/nykj/2269672.html
