Study On Soil Bidirectional Reflectance Characteristics Andapplication Of Soil Standard Spectral Library

The height variations of earth surface will lead to different inclination of pixel of hyperspectral images under natural conditions. However, bidirectional reflectance study can rectify the pixels of hyperspectral images to the same observation angel. The accuracy of quantitative inversion will improved in this way. At the same time, since soil is under the vegetation, studying soil bidirectional reflectance characteristics can provide background reference for studying vegetation canopy spectral. In addition, the directional distribution of soil reflectance may carry soil property information potentially, such as soil moisture, organic content, and mineral content and so on. Therefore, the study of soil bidirectional reflectance characteristics has important theoretical significance and research value in soil and vegetation quantitative remote sensing. Currently, visible-near infrared spectral analysis technique has successfully applied in soil key properties prediction and the hyperspectral inversion model of soil key properties in local region is mature. Nevertheless, it’s difficult to apply local inversion model to soil samples in other region due to the different weather conditions and parent material between different regions. With the emergence of large-scale soil standard spectral library, it’s possible to solve the above problem by mining useful information of standard spectral library and building soil key properties prediction model based on standard spectral library.The research acquires bidirectional reflectance of three typical soils and analyzes the bidirectional reflectance characteristics and differences of different soils. Then it uses Hapke model to invert the average single scattering albedo and the rough parameter. On this basis, it analyzes the relations of parameters and soil particle composition and simulates the bidirectional reflectance. Furthermore, based on a globally distributed soil spectral library, the dissertation adopts fuzzy C-means clustering to extract subset of soil spectral library that is similar to the samples in study area and use PLSR to build the hyperspectral prediction model of soil key properties of the subset. Finally, it conducts the model uncertainty analysis.Based on the above research, the dissertation draws the following conclusions:1. The variation trends of the bidirectional reflectance of three typical soils and the observation angels is consistent. The reflectance increases as the view zenith angel increases. It reaches the minimum in the forward scattering direction and the maximum in the backward scattering direction. The reason is that with the change of observation angle, the proportion of shadow between the soil particles will change, so the light components received by detector are different.2. The parameters of Hapke model have different sensitivity to their initial values. The soil single scattering albedo curves of similar soil particle composition are similar. As the soil coarse particle content goes up, the rough parameter rises and the average single scattering albedo goes down. What’s more, Hapke model is able to simulate soil bidirectional reflectance well, but the simulation accuracy differs with each other.3. By combining fuzzy C-means clustering and PLSR, it can dig effective spectral information of soil spectral library that is similar to the samples in study area and build soil organic carbon content prediction model which can estimate soil organic carbon content of the study area roughly. The prediction ability of soil organic carbon content hyperspectral prediction model is related with the profile layer of soil samples and the model has good prediction ability of lower samples.