| The Yellow River delta is a new land. It is formed by the Yellow River epeirogenesis in the last hundred years. The sedimentary environment, climatic conditions and soil parent material determines that the primary saline soil is widely distributed in the region. With the development of the local agricultural economy and heavy irrigation light row tillage, and shallow ground water level with high salinity, soil salinization in the Yellow River delta is serious. Affected by salinization, vegetation habitat and a variety of rare wild animal habitat of the area are threatened, the ecosystem is vulnerability. Soil salinization has become the most important environmental problem is the local ecological system and agricultural sustainable development. Therefore, it is very important to control saline soil and prevent its degradation and realize the sustainable development of agriculture, which is timely, accurate and dynamic to obtain the information of water and salt in saline soil.This study explored hyperspectral field and satellite-based remote sensing of soil salt content. Using Kenli County in the Yellow River Delta as the study area, in situ soil field spectra and satellite-based remote-sensing images were integrated with laboratory measurements of soil sample salinity to study remote sensing-based soil salt estimation and inversion procedures. The main conclusions of the paper were:(1) Through the correlation between the reflectance and the salt content, the study screened out the sensitive band of soil salinity. The results showed that the sensitive bands of soil salinity were mainly in visible light and near infrared. The spectral reflectance of saline soil was relatively low in the visible range, but increased quickly with increasing wavelength.The increase was less pronounced in the near-infrared range. When the water content was less than 30%, the spectral reflectance of saline soil was increased with the increase of salt content.When the water content was greater than 30%, the spectral reflectance of saline soil declined with the increasing of salt content. But the spectral patterns were basically similar and remain parallel. The study used the sensitive bands for constructing new spectral features, whose correlation coefficients were larger than 0.6. This study found that the correlation between therecombined spectral features and soil salinity increased notably compared to.The study used multiple stepwise linear regressions on the sensitive spectral parameters to analyze the soil salt spectrum and construct the field hyperspectral estimation models of soil salinity. And the study chose the best field hyperspectral estimation model of the soil salinity that fit the study area, which was y=2.119–0.015×(M1+M4)/(M1–M4)–3.508×(M4+M9)+0.012×(M1+M3)/(M1–M3)–M2/(M2–M3)–0.021×M5/(M1–M5). Among them, y expressed salt content, M1 expressed the 763 nm band,M2 expressed the 780 nm band, M3 expressed the 885 nm, M4 expressed the 899 nm band, M5 expressed the 995 nm band, and M9 expressed the 1794 nm band. This model had a verification R2 of 0.811. It showed that the model established by this method had the best fit and the most stable, and it was feasible.(2) The paper fitted narrow-band field hyperspectral reflectance to the wide-band reflectance of the satellite Landsat 7. And it used the same method to construct the hyperspectral estimation models of soil salinity based on Landsat 7. In addition the study chose the best estimation model of the soil salinity that fit the study area, which was y=1.345–25.89×g SWIR1–245.440×g Red×(g Red–g NIR)–0.252×(g Red+g NIR)/(g Red–g NIR)–19.563×(g Red–g SWIR1). This model had a verification R2 of 0.867. Similarly, the best estimation model applied for Landsat 8 is y=0.685–11.359×g SWIR1–165.302×g Red×(g Red–g NIR)–0.23×(g Red+g NIR)/(g Red–g NIR)–25.385 ×(g Red–g SWIR1). This model had a verification R2 of 0.848.These models had the best fit and the most stable, which showed the models had good availability.(3)The paper used Landsat remote sensing image in the study area. After the passage of the radiation correction, geometric correction and atmospheric correction pretreatment, the study used a fully constrained least squares mixed-pixel decomposition method to remove the water and coastal beach signals, and used the linear mixed pixel decomposition method to minimize the spectral information of the vegetation. Then the study scale-corrected the remote sensing reflectance and applied the soil salinity model to the corrected satellite remote sensing image by the ratio- average method. Finally the model was applied to the correction of satellite remote sensing image for inversion. And the study got the better results. This study showed that the method of combining the field hyperspectral readings with satellite remotesensing to estimate the soil salt content by inversion was both feasible and practical.(4)Two best models were applied to Landsat 7 and Landsat 8 satellite remote sensing image, and the inversion results were compared and analysis. The study found that the inversion results were consistent, and the results were similar with the measured interpolation.Thus, the model had certain significance for the dynamic monitoring of soil salinity in Kenli County in the Yellow River Delta in spring and was effective for regional satellite-based saline soil investigation. |
PDF Full Text Request