含水量对土壤有机质含量高光谱估算的影响研究

发布时间：2018-07-09 21:04

本文选题：土壤有机质 + 高光谱　；参考：《中国农业科学院》2015年硕士论文

【摘要】：土壤有机质(SOM)是土壤肥力的重要指标,不仅能提供作物养分,改善土壤物理性质,还具有土壤保水和保肥的作用。快速、准确地估算SOM含量,可以为耕地质量评价、培肥地力、提高粮食产量提供重要的决策依据。土壤光谱特性是SOM含量、含水量、氧化铁含量、土壤质地等属性的综合反映,利用土壤的光谱信息,可以快速地估算出SOM的含量。但土壤含水量等因素对高光谱估算SOM含量的精度有很大的影响,因此,研究确定高光谱估算SOM的含水量适宜范围具有理论和实践意义。本研究以东北耕地土壤为研究对象,利用美国ASD Fieldspec Pro FR地物高光谱仪在室内条件下对烘干土、风干土和含水量为5~40%(按5%递增)的土壤进行光谱测量,对光谱数据进行反射率(R)、反射率一阶导数(R’)和反射率倒数对数(Log(1/R))3种光谱数据数学变换,然后对全部样本和不同土壤类型,运用偏最小二乘回归法(PLSR)、支持向量机(SVM)和二者的结合方法建立相应的SOM含量估算模型。研究结论如下:1.用全部土壤样本建立的SOM含量估算模型中,风干土光谱数据在PLSR和PLSR-SVM方法建立的SOM含量模型结果最好。SVM建立的模型中,含水量15%和20%的土样光谱数据建立的SOM含量模型结果最好,风干土结果次之。2.当土壤含水量大于25%时,不适宜利用高光谱数据进行SOM含量估算。3.PLSR和PLSR-SVM建立的SOM含量模型中,光谱数据Log(1/R)变换形式的土壤含水量水平建立的SOM含量模型精度都比较高;SVM建立的SOM含量模型中,光谱数据R和Log(1/R)变换形式的土壤含水量水平建立的SOM含量模型相对较好。4.对于单个土壤类型,进行PLSR和SVM的SOM含量建模时,黑土的估算模型结果最好,而草甸土和黑钙土的SOM含量结果不太理想。
[Abstract]:Soil organic matter (SOM) is an important index of soil fertility, which can not only provide crop nutrients, improve the physical properties of soil, but also have the function of soil moisture and fertilizer conservation. Estimating SOM content quickly and accurately can provide important decision basis for evaluating cultivated land quality, increasing fertility and increasing grain yield. The soil spectral characteristic is the comprehensive reflection of SOM content, water content, iron oxide content, soil texture and so on. Using the soil spectral information, the SOM content can be estimated quickly. However, soil water content and other factors have great influence on the accuracy of estimating SOM content by hyperspectral method. Therefore, it is of theoretical and practical significance to study and determine the suitable range of soil moisture content in hyperspectral estimation. In this study, the soil of northeast cultivated land was used as the research object. The dry soil, dry soil and soil with water content of 5 ~ 40% (increasing by 5%) were measured by ASD Fieldspec Pro FR hyper spectrometer under indoor conditions. The spectral data are mathematically transformed into three kinds of spectral data: reflectivity (R), first derivative of reflectivity (R') and log (1 / R), then all samples and different soil types are transformed. Using partial least square regression (PLSR), support vector machine (SVM) and the combination of the two methods, a corresponding SOM content estimation model is established. The conclusion of the study is as follows: 1. In the SOM content estimation model based on all soil samples, the dry soil spectral data is best in the model established by PLSR and PLSR-SVM. The SOM content model with 15% and 20% soil moisture content data was the best, followed by air-dried soil. When the soil moisture content is greater than 25, it is not suitable to use hyperspectral data to estimate SOM content. 3. PLSR and PLSR-SVM established SOM content model. The SOM content model based on log (1 / R) transformation is more accurate than SOM content model established by SVM. The SOM content model based on spectral data R and Log (1 / R) transformation is better than SOM content model based on log (1 / R) transformation. For a single soil type, when the SOM content of PLSR and SVM is modeled, the estimation model of black soil is the best, but the SOM content of meadow soil and calcareous soil is not very good.
【学位授予单位】：中国农业科学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S153.6;S127

【参考文献】