Quantitative Retrieval Of Soil Constituents Using Hyperspectrum

Hyperspectral remote sensing is characterized by high spectrum resolution, and has a great potential to quantitatively identify mineral, soil, vegetation and water content through inverse analysis. Soil spectrum is mainly influenced by the following constituents: mineral, water, organic matter, roughness and texture, etc.Through investigating soil reflectance spectroscopy, this dissertation aims to uncover the genetic relationship between soil constituents and their spectroscopy. The best spectral band is selected based on mathematic analysis of soil reflectance. Several multi-component linear regression models are established through cross-correlation analysis. Soil organic matter, its moisture and its iron oxide content could be spectrally mapped with a great success. This dissertation concentrates on the following aspects.1. Through soil spectral and chemical analysis, it is found that the brown soil in Dongling Xiang, Zhaoyuan, Shangdong is sandy loam. Its reflectance ranges from 0.45 to 2.5 μm. With increasing water, organic matter and iron oxide, the soil reflectance spectral curve goes downward. Visible and near infrared reflectance are the best bands to study the soil organic matter content.2. Spectral derivative analysis is a useful method to investigate the soil constituents. Correlation between soil organic matter/iron oxides and soil reflectance is enhanced by the 1^st and 2^stderivative of logarithm of reflectance. On this basis, multi-component linear regression model is established, and a map is prepared to display the content of soil organic matter and iron oxide. Additionally, spectral derivative analysis can also partially eliminate atmospheric effect. Especially, the 1^st derivative analysis can eliminate some linear or near linear noises affecting the target spectra.3. Continuum-removal is an effective method in soil spectral analysis. Because it extracts the peak absorption eigenvalues of some soil constituents with a simple and an accurate way. After the continuum-removal, the reflectance is normalized to a range of 0-1.0. This enhances the soil absorption and reflection spectrum. The eigenvalues of absorption peak could be easily obtained, such as the wavelength, depth, width and area, etc. These results are used to predict the content of some constituents in the soil. In this study, it is observed that there is a strong absorption around 1450 nm due to water present in the soil. This is used to map the soil moisture.4. Quantitative remote sensing analysis is based upon calibration. Radiance calibration is realized by atmospheric correction, and from which relative reflectan-ce could be derived. Three kinds of relative reflectance are investigated in this study: linear relative reflectivity, residual error of logarithmic relative reflectivity and atmospheric radiance transmit relative reflectivity. The first is used in this investigation.5. The essence of remote sensing is a process of quantitative inversion. How to successfully and accurately extract information on atmospheric and soil constituents through inversion analysis is the hotspot and frontier. Besides sensor calibration, radiance calibration and atmospheric correction, BRDF of soil reflectance property should be taken into account. Available geometric optical model and radiance transmit model are used to investigate the soil BRDF property.6. A general discussion is made on the Munsell model and its properties in this study. This model is a three-dimensional model, and it can be used to qualitatively or quantitatively describe the soil properties, such as soil classification and prediction of organic matter content, etc. Previous study shows that there is a strong correlation between soil brightness and its reflectivity. Since the soil color is determined by the visible light, this sets a limit to use the soil color to investigate the soil properties.