發(fā)布時間：2019-03-22 10:02

【摘要】：土壤鹽漬化作為土地荒漠化和土地退化的主要類型之一,是造成干旱、半干旱地區(qū)土壤流失的重要原因,及時快速發(fā)現(xiàn)和治理土地鹽漬化問題顯得尤為重要。為建立土壤鹽漬化遙感監(jiān)測模型,選取寧夏平羅縣典型土壤鹽漬化發(fā)生區(qū)域作為研究區(qū),以野外原位光譜測量數(shù)據(jù)和實驗室內測得的土壤含鹽量與pH數(shù)據(jù)為基礎,運用高光譜數(shù)據(jù)處理方法,分析不同鹽漬化程度土壤的光譜特征；對實測土壤光譜反射率進行倒數(shù)、對數(shù)、均方根及其一階微分等光譜變換,計算高光譜指數(shù)；與土壤樣本含鹽量進行相關性分析,篩選鹽漬化土壤的光譜特征波段,利用多元線性回歸分析建立土壤鹽漬化監(jiān)測模型。通過本文的研究,得到如下結論：(1)野外實測光譜數(shù)據(jù)與實驗室測得的土壤含鹽量數(shù)據(jù)相結合,將土壤類型主要分為非鹽漬化土壤、輕度鹽漬化土壤、中度鹽漬化土壤、重度鹽漬化土壤四大類型。不同程度鹽漬化地區(qū)的土壤光譜特征曲線在形態(tài)上基本趨于一致；但在可見光波段,不同程度鹽漬化土壤的反射率并不呈現(xiàn)規(guī)律性變化。(2)將實測的土壤及植被光譜反射率數(shù)據(jù)進行倒數(shù)、對數(shù)、對數(shù)倒數(shù)、均方根等四種形式的變換,再對原始光譜反射率以及四種變換形式的光譜反射率數(shù)據(jù)進行一階導數(shù)微分變換以及原始光譜的二階微分導數(shù)變換,共計22種變換形式的反射率數(shù)據(jù)。再同四種不同類型的土壤含鹽量數(shù)據(jù)進行統(tǒng)計分析,計算相關系數(shù),得到相關系數(shù)圖。圖中顯示,各種變換形式可以有效提高光譜反射率和土壤含鹽量兩者之間的相關性。(3)以篩選出的特征波段為自變量與土壤含鹽量進行統(tǒng)計回歸分析,構建土壤鹽分動態(tài)監(jiān)測模型,結果表明構建模型的模擬值與實際測量的土壤含鹽量值之間的相關性較高。尤其是反射率倒數(shù)一階微分模型,相關系數(shù)高達0.81,土壤鹽漬化遙感監(jiān)測模型的預測效果較好。(4)變換后的土壤和植被光譜反射率,其中倒數(shù)對數(shù)一階微分變換的光譜反射率與實驗室測得的土壤含鹽量相關性最好。選擇相關性最好的特征波段450nm、685nm構建鹽分指數(shù)模型及960nm和1094nm構建植被指數(shù)模型,結果表明兩者和土壤含鹽量的相關性較高,因此協(xié)同兩指數(shù)構建區(qū)域的土壤鹽漬化遙感監(jiān)測模型,經驗證,模擬效果很好,可以用來快速提取該區(qū)域的土壤鹽漬化信息,為今后土壤鹽漬化監(jiān)測提供一種新的手段。
[Abstract]:Soil salinization, as one of the main types of land desertification and land degradation, is an important cause of soil erosion in arid and semi-arid areas. It is very important to find and control soil salinization in time and quickly. In order to establish the remote sensing monitoring model of soil salinization, the typical soil salinization area in Pingluo County, Ningxia, was selected as the study area, based on the in-situ spectral data measured in the field and the soil salinity and pH data measured in the laboratory. The spectral characteristics of soils with different degree of salinization were analyzed by using hyperspectral data processing method. The spectral reflectance of the measured soil was calculated by inverse logarithm root mean square and its first order differential isospectral transformation to calculate the hyperspectral index. The spectral characteristic band of salinized soil was selected and the monitoring model of soil salinization was established by multiple linear regression analysis. The results are as follows: (1) the soil types are divided into non-salinized soil, mild salinized soil and moderate salinized soil according to the combination of field measured spectral data and soil salt content data measured in laboratory. Four types of severely salinized soil. The spectral characteristic curves of soil in different degree salinized areas tend to be consistent in morphology. However, in visible light band, the reflectivity of salinized soil in different degrees does not change regularly. (2) the measured spectral reflectance data of soil and vegetation are inversed, logarithmic reciprocal, root mean square and other four forms of transformation. Then the first derivative differential transformation and the second order differential derivative transformation of the original spectral reflectivity and four kinds of spectral reflectivity data are carried out. The reflectivity data of 22 kinds of transformation forms are made up of the first derivative differential transformation and the second order differential derivative transformation of the original spectrum. Then four different types of soil salt content data were statistically analyzed, the correlation coefficient was calculated and the correlation coefficient diagram was obtained. The results show that the correlation between spectral reflectivity and soil salt content can be effectively improved by various transformation forms. (3) Statistical regression analysis is carried out with selected characteristic bands as independent variables and soil salt content as independent variables. The dynamic monitoring model of soil salinity was constructed, and the results showed that the correlation between the simulated value of the model and the measured soil salt content was high. The correlation coefficient is as high as 0.81, and the prediction effect of the remote sensing monitoring model for soil salinization is better. (4) the spectral reflectivity of soil and vegetation after the transformation, the correlation coefficient is 0.81, and the correlation coefficient is 0.81. The spectral reflectivity of the reciprocal logarithm first order differential transformation is the best correlation with the soil salt content measured in the laboratory. The best correlation band was 450 nm, the salt index model was constructed at 685nm, and the vegetation index model was constructed by 960nm and 1094nm. The results showed that the correlation between them and soil salt content was high. Therefore, the remote sensing monitoring model of soil salinization is constructed in collaboration with two indices. The simulation result is very good, which can be used to extract the soil salinization information quickly and provide a new method for the monitoring of soil salinization in the future.
【學位授予單位】：寧夏大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：S156.41

【引證文獻】

相關會議論文 前1條

1 陳建軍;張樹文;;基于MODIS數(shù)據(jù)的東北地區(qū)土地覆蓋分類的精度評價研究[A];第十四屆全國遙感技術學術交流會論文選集[C];2003年

，

本文編號：2445497