| As a dynamic ecological system, soil supports the terrestrial life cycle of all microorganisms and maintain the survival and development of human beings. Thus the further study of soil has become the long-term compulsory subject of humans. The research on soil properties is significant in soil classification, soil genesis, soil mapping etc. Although traditional method of soil properties analysis has high accuracy, it is a waste of time and energy. The technique of soil spectral analysis has the higher speed, lower cost and no destruction. Moreover, it can invert various soil attributes. The combination of spectroscopy and imaging technology can not only obtain the soil information of image and spectral, but also be used to analyze small scales of the continuous distribution of soil properties. The method provides new means and technology for deeper analysis and research of soil. Besides, it also helps to improve soil classification ability and rich soil specimen information.The article uses three axial experiment platform and the HeadWall spectrometer (NIR400-1000nm and VNIR1000-1700nm) to get the hyperspectral images of the dried soil profile and different particle size, different levels of soil samples (0.28-0.90mm,0.20-0.28mm,0.28-0.20mm,<0.15mm). Then PLSR is adopted on the average spectral data extracted from the hyperspectral images to establish the prediction model of organic carbon content of soil. On this basis, it analyses the correlation of soil spectral reflectance and organic carbon of soil. Finally, this model is applied to predict organic matter content distribution of soil profile and analyze vertical distribution patterns of the key properties of different type of soil. In this paper, we obtain the following main conclusions.1, Surface roughness of soil is negatively related to the spectral reflectance. The greater lightness value the soil has, the higher value of spectral reflectance. The research shows that as the increase of soil profile depth, soil reflectance increases. However, the study of two kinds of soil samples indicate that, the spectral reflectance curve of the gley layer (G) is far from the other layers for the paddy soil profile and the distances of the spectral reflectance curves between the tillage layer (A), the prototype layer (B1), and the redox layer (B2), are evenly distributed for the fluvo-aquic soil profile. 2, PLSR method is adopted to establish estimation models of the organic carbon content of soil layered samples for different particle sizes. The estimation model of the undisturbed soil is set up after removing the shadow. The results illustrate that soil particle size affects the precision of the model. Comparison of PLSR organic carbon estimation models of different sizes reveal that the highest precision is0.28-0.9mm particle size of soil samples. Therefore,0.28-0.9mm soil sample estimation model is selected to predict the distribution of organic matter content of the whole section soil profile.3, The correlation between soil organic carbon content and the spectral reflectance in different levels is negative, which means the value of soil spectral reflectance is rising with the decline of the organic carbon content. The tillage layer and the plough layer of organic carbon content in paddy soil are uniform. The surface layer of organic carbon content in fluvo aquic soil is relatively abundant, while the content of deeper layer is low. The study of two kinds of soil samples show that, the gley layer (G) is quite different from the other layers for the paddy soil profile. The distances between the spectral reflectance curves are distributed uniformly from the tillage layer (A) to the prototype layer (B1) to the redox layer (B2) for the fluvo-aquic soil profile. In addition, the distribution of soil organic carbon content is associated with soil genetic horizon, and may be also related to the interior of the soil properties such as soil type, texture, and factors such as farming methods and forming characteristics. |
PDF Full Text Request