【作者】 于薇；

【导师】 潘军；

【作者基本信息】 吉林大学 ， 地球探测与信息技术， 2024， 硕士

【摘要】 土壤盐碱化是指土壤底层或地下水的盐分随毛管水上升到地表,水分蒸发后,使盐分积累在表层土壤中的过程,是严重的土壤退化现象。及时了解并掌控该地区土壤盐碱化的程度以及分布、面积、类型等信息,是治理盐碱化、防止土壤进一步退化的重要前提。传统人工采样监测方式获取信息效率低且成本高,遥感技术可以实现快速获取、实时监控且覆盖面积广,在土壤盐碱化的反演与监测中应用广泛,但是由于遥感影像生成过程中存在许多变量元素,例如降水导致土壤含水量变化、不同季节植被覆盖度变化等导致反演精度有限。因此,土壤盐碱化程度与光谱反射率之间存在怎样的定量数学关系以及如何对变量因素影响下的光谱进行校正是利用遥感数据进行土壤盐碱化程度反演中重要的科学问题。论文以吉林省松原市乾安县为研究区,根据乾安地区实地采样和测量获取的土壤理化性质和实测高光谱数据以及同步的无人机机载高光谱数据、高分二号数据和Landsat 9数据利用敏感波段和盐分指数通过多种回归分析方法建立土壤盐碱化程度的定量反演模型,通过R~2和RMSE筛选出最优的反演模型。并通过仿真实验分别测量不同土壤含水量和植被覆盖度情况下的盐碱土光谱反射率,针对不同土壤含水量和植被覆盖度情况下的盐碱土光谱反射率构建校正模型,以减弱这两项变量因素的影响,从而为乾安地区的土壤盐碱化监测提供更为精确的数据支撑。本研究主要得到以下结论:(1)土壤盐碱化程度与光谱反射率的关联性研究显示野外实测高光谱数据的波段范围内,土壤的光谱反射率随着土壤碱化程度的加重呈上升趋势,其中碱土的光谱反射率明显高于其他程度的盐碱土。利用野外实测高光谱数据、无人机机载高光谱数据、高分二号数据和Landsat9数据的土壤盐碱化程度反演结果显示,乾安地区的土壤盐碱化程度与光谱反射率之间存在一定的定量关系。其中,基于敏感波段的pH值反演结果中无人机光谱数据经过倒数变换后的反演效果最佳,决定系数R~2为0.80,基于敏感波段的土壤电导率反演结果中实测光谱数据经过对数变换后的反演结果最佳,决定系数R~2为0.33。基于盐分指数的pH值反演结果中利用无人机数据的S3盐分指数进行二阶多项式回归的反演效果最佳,决定系数R~2为0.79。基于盐分指数的土壤电导率的反演结果中利用实测光谱数据SI-T盐分指数进行幂函数回归的反演效果最佳,决定系数R~2为0.49。(2)不同含水量的盐碱土光谱校正及模型构建研究发现,不同含水量的盐碱土的光谱特征表现出差异,对基于光谱信息进行土壤盐碱化程度反演造成影响。经过土壤含水量校正的光谱构建的新的土壤盐碱化程度反演模型结果显示,基于敏感波段的土壤pH值反演以及土壤电导率的反演精度高于原始模型的反演精度。因此,该种光谱校正方式较适合基于敏感波段进行土壤盐碱化程度反演,其反演精度得到了有效提高。(3)不同植被覆盖度的盐碱土光谱校正及模型构建研究发现,不同植被覆盖度的盐碱土的光谱特征表现出差异,对基于光谱信息进行土壤盐碱化程度反演造成影响。经过不同植被覆盖度校正的光谱构建的新的土壤盐碱化程度反演模型结果显示,基于盐分指数的土壤pH值反演和土壤电导率的反演相较于原始的反演模型的精度有所提升,说明该种光谱校正方式较适合基于盐分指数进行土壤盐碱化程度反演,其反演精度得到一定的提高。更多还原

【Abstract】 Soil salinization refers to the process in which the salt of the soil bottom or groundwater rises to the surface with the capillary water,and the salt accumulates in the surface soil after the water evaporates,which is a serious soil degradation phenomenon.Timely understanding and controlling the degree,distribution,area,type and other information of soil salinization in this region is an important prerequisite for controlling salinization and preventing further soil degradation.Traditional manual sampling and monitoring methods have low information acquisition efficiency and high cost.Remote sensing technology can achieve rapid acquisition,real-time monitoring and covers a wide area,and is widely used in soil salinization inversion and monitoring.However,due to the existence of many variable elements in the process of remote sensing image generation,For example,the inversion accuracy is limited due to the change of soil water content caused by precipitation and the change of vegetation coverage in different seasons.Therefore,what quantitative mathematical relationship exists between soil salinization degree and spectral reflectance and how to correct the spectrum under the influence of variable factors are important scientific problems in soil salinization degree inversion using remote sensing data.This paper takes Qian’an County,Songyuan City,Jilin Province as the research area,and establishes a quantitative inversion model of soil salinization degree through multiple regression analysis methods based on soil physical and chemical properties obtained from field sampling and measurement,observed hyperspectral data,synchronous UAV airborne hyperspectral data,Gaofen-2 data and Landsat 9data.The optimal inversion model was selected by R~2and RMSE.Moreover,the spectral reflectance of saline-alkali soil under different soil water content and vegetation coverage was measured by simulation experiments,and a correction model was built for the spectral reflectance of saline-alkali soil under different soil water content and vegetation coverage to reduce the influence of these two variables,so as to provide more accurate data support for soil salinization monitoring in Gan’an area.The main conclusions of this study are as follows:(1)Research on the correlation between soil salinization degree and spectral reflectance shows that in the band range of hyperspectral data measured in the field,the spectral reflectance of soil shows an increasing trend with the increase of soil alkalinity degree,and the spectral reflectance of alkaline soil is significantly higher than that of other salinized soils.Inversion results of soil salinization degree based on field measured hyperspectral data,UAV airborne hyperspectral data,Gaofen-2 data and Landsat 9data show that there is a certain quantitative relationship between soil salinization degree and spectral reflectance in Gan’an area.Among them,the inversion result of pH value based on the sensitive band has the best inversion effect after reciprocal transformation of the UAV spectral data,and the determination coefficient R~2is 0.80;the inversion result of soil conductivity inversion result based on the sensitive band has the best inversion result after logarithmic transformation of the measured spectral data,and the determination coefficient R~2is 0.33.Among the inversion results of pH value based on salinity index,the second order polynomial regression using S3 salinity index of UAV data has the best inversion effect,and the determination coefficient R~2is 0.79.Among the inversion results of soil conductivity based on salinity index,the inversion effect of power function regression using the measured spectral data SI-T salinity index is the best,and the coefficient of determination R~2is 0.49.(2)The study on spectral correction and model construction of saline-alkali soils with different water contents found that the spectral characteristics of saline-alkali soils with different water contents showed differences,which affected the inversion of soil saline-alkali degree based on spectral information.The results of a new inversion model of soil salinization degree based on the corrected spectrum of soil water content show that the inversion accuracy of soil pH value and soil conductivity based on the sensitive band is higher than that of the original model.Therefore,this spectral correction method is more suitable for soil salinization degree inversion based on the sensitive band,and its inversion accuracy has been effectively improved.(3)The study on spectral correction and model construction of saline-alkali soils with different vegetation coverage found that the spectral characteristics of saline-alkali soils with different vegetation coverage showed differences,which affected the inversion of soil salinization degree based on spectral information.The results of the new inversion model of soil salinization degree constructed by spectra with different vegetation coverage correction show that the inversion accuracy of soil pH value and soil conductivity based on salinity index is improved compared with the original inversion model,indicating that this spectral correction method is more suitable for inversion of soil salinization degree based on salinity index,and its inversion accuracy has been improved to a certain extent.更多还原