| Land desertification is a special form of land degradation, and also is a kind of process in which land surface soil structure and its corresponding vegetation changes, leaded by severe land degradation. It has an serious impact on the development of social economy. The situation of sandification combating works remain stern in visage in our country, there are great needs to take further effective measures to prevent sandy desertification. With the introduction of remote sensing technology, information extraction and detection for dynamic change of sandification has entered a new phase. As a new technology, remote sensing has been characterized as a large amount of information, huge observation range, high precision and speed, real-time and dynamic characteristics, having played an important role on further exploring the theories and methods for extracting desertification information accurately and the national desertification monitoring and evaluating.The breakthrough point of this paper is the changes of spectral reflectance caused by the physical and chemical features of soil surface characteristics in the process of desertification. On the basis of field investigation, index factors for sandy land are retrieved quantitatively using measured spectrum data and other remote sensing image. A quantitative evaluation method has been proposed upon land desertification information extraction and soil characteristic parameters inversion, therefore, it has a guiding effect on evaluating whether the soil has occurred sandy phenomenon and which degree the desertification is at. First of all, analyzed the soil information features parameters, and distinguished their difference between the sandy and unsandy land. Secondly, analyzed the correlation between measured spectrum and characteristic parameters of sandy land, determined the best response band, and established the inversion model. Thirdly, retrieved the soil organic matter (SOM) and soil particle composition (SPC)based on the reflectance of remote sensing image band. Finally, determined the threshold of these index factor and extracted the desertification information based on the inversion results. The main research contents are as follows: (1) The analysis for soil characteristic parameters of the sandy land. Through analyzing the different content of organic matter, total nitrogen, total phosphorus, potassium, particle diameter between the sandy and unsandy land, it was found that the content of organic matter in unsandy land was about three times as much as that in sandy land, and it descended graduately with the soil decaded. The content of TN in sandy land was also different from that of unsandy land greatly, while the content of TP and TK didn’t have the huge difference. In the soil particle composition, silt and clay were the important index factor for the soil sandification, the content of them descended greatly in the sandy land. Happened with the soil degradation, SOM、TN、silt and clay presented a law of diminishing, and they were the most important contribution factor to distinguish the sandy and unsandy land.(2) The quantitative inversion for soil surface parameters based on measured spectrum. Through analyzing the correlation between measured spectrum and soil nutrient, it was known that the correlation coeffecient of SOM and TN was the highest, and their effective spectral wavelength range was about600nm-800nm. In the soil particle composition, clay and silt had the best correlation with the spectrum, the correlation of weighted soil particle and course sand was secondry to it, while the gravel and fine sand had a bad correlation to the measured spectrum. The sensitive wavelenght for silt and clay was1000nm-1900nm, while there was a small peak at550nm in the visible spectrum, and it was the best inversion symbols with visible spectrum band. Building and validating the inversion model for SOM, TN, silt and clay using its best correlated bands, the invertion model R2of SOM and TN was above0.8, comparatively speaking, the R2of SPC inversion model was more lower, it was just about0.5.(3) The inversion of soil characteristic parameter based on the remote sensing image data. Compared with measured spectrum, the modeling effect based on multispectral image is not ideal. But generally speaking, its inversion model R2is about0.5, it can also able to reflect the trend of SOM and SPC change with the land desertification. Although the correlation between TN and multispecral image was high, the R2of SOM invertion model is higher than that of the TN, and its predicted value are more closer to the measured one. Comparatively speaking, the correlation among silt, clay, WSD and multispectral image is lower than that of soil nutrient, but the correlation of silt and clay was more higher than that of the WSD. The inversion precision of ungrowing season image is higher than that of growing season one, to some extent, the soil information in the growing season image has been influenced by vegetation information, it enhanced the background noisy.(4) Method reseach for sandy information extraction based on soil surface parameters. Through the classification of SOM, the results show that the level of organic matter content reflects the soil nutrient characteristics, soil organic matter in sandy land is under1.0%. In addition, testing the threshold through the sample data and artificial interaction adjustment, as a result, the threshold of silt is4.0, clay threshold is7.0and the threshold of weighted soil diameter is40. The classification accuracy based on the clay threshold value is the highest, precision of silt and weighted soil particle are relatively lower. No matter which index factor based on, the classification accuracy for unsandy land is low, while the classification accuracy for the sandy land is higher, above80%. The precision of sandy information extraction based on growing season image is more precise, the total precision is as high as79.45%. |
PDF Full Text Request