Inversion Of Soil Salinity By Hyperspectral Remote Sensing In Part Of Vegetation Coverage Area

The soil is one of the world’s most important resource which is closely related to the survival and development of organisms. In recent years, due to the impact of human activities and natural environmental changes, the soil salinization is worsening, and it has become one of the main factors of soil degradation. Saline soil is one of the main low-yielding soil type in our country. The productivity of Saline Soil is restricted by the condition of soil texture. In this paper, the Yellow River Delta is chosen as our study area because of its vast offshore area. Otherwise, in this area the increasing salinization and the constantly amplification of saline soil area, greatly restricted the sustainable development of agriculture, forestry and animal husbandry. In this paper, we research the quantitative inversion and spatial distribution of saline soil about the Yellow River Delta for providing the decision basis for the soil salinization governance.In this paper the spectral features of saline soil, spectral modeling based on partial least-squares regression (PLSR) and spectral mixture analysis (SMA) are researched depend on the hyperspectral data which measured in the laboratory and field, the hyperion image and the physical and chemical analysis data. Finally, the inversion mapping of soil salinity in the study area (including vegetation coverage area) is implemented. The main works and conclusions in the thesis are as follows:(1) The spectral features of saline soil are analysed. The otherness between field and laboratory spectra is discussed deeply. Then, the otherness of spectral reflectance between these two spectra, the otherness of continuum removal transformation and the otherness of correlation between these two spectra and the soil salt content were generated. Finally, the reasons resulting to these differences are analysed from the point of saline-alkali soil.(2)The soil salinity inversion models are set up based on PLSR. Partial least squares regression (PLSR) is employed to build three salinity inversion models. These models are model I (RMS=0.627g/kg, R2=0.890) built by laboratory spectra with 63 bands, model Ⅱ(RMS=0.783 g/kg, R2=0.766) built by field spectra with 143 bands and model Ⅲ(RMS=0.824 g/kg, R2=0.741) built by field spectra with 88 bands. All the models have high precision.(3) Spectral mixture analysis (SMA) is operated by image and field endmembers. The influence of the other spectra especially vegetation spectra are removed from saline soil spectra. Finally, the image of residual soil spectra based on image and field endmembers were obtained and named IR and FR.(4) The inversion mapping of soil salinity in the study area is implemented. Soil salinity inversion of the study area (including vegetation coverage area) is implemented as the salinity inversion models that established based on PLSR are applied to the image of residual soil spectra. The inversion result and precision were analysed. Results indicate that FR-MODELⅡ and FR-MODELIII have better inversion precision, the RMS is 0.930 g/kg and the R2 is 0.700 of the fitting of inversion value and measured value in FR-MODEL II, the RMS is 1.244 g/kg and the R2 is 0.594 of the fitting of inversion value and measured value in FR-MODELⅢ.(5) The inversion model is optimized in order to improve the precision of soil salinity. An optimized salinity inversion model MODELIV is applied to the image of residual soil spectra based on field endmembers. Results indicate that the inversion precision of FR-MODELIV improves compared with before, the RMS is 0.823 g/kg and the R2 is 0.736 of the fitting of inversion value and measured value. To sum up,the approach proposed in this study provides a new direction for inversion soil salt content for vegetation coverage area.