基于高分一号卫星影像的延安市顾屯流域耕地盐渍化定量反演

Quantitative Inversion of Salinization of Cultivated Land in Gutun Watershed of Yan’an City Based on Gaofen-1 Satellite Image

2013年陕西省延安市顾屯流域实施治沟造地工程后,部分新增耕地内出现盐渍化现象。为获取盐渍化空间分布特征,以顾屯流域作为研究区,基于高分一号影像及相应时间段内室内实测数据,通过盐分指数模型与高分影像特征波段建立的支持向量机(support vector machine,SVM)模型对工程后所有耕地进行了盐渍化定量反演。研究表明,选取的盐分指数(salinity index,S3)与盐分相关性最好,反演模型R^(2)为0.742;筛选了3个特征波段组合,利用SVM方法建模,结果显示基于波段1与波段3组合(P_(1)P_(3))的SVM模型表现了最优的反演能力,R^(2)达到0.867;SVM模型反演结果显示研究区正常土壤面积占26.62%,轻度盐渍化占43.52%,中度盐渍化区域为29.82%,重度盐渍化几乎没有。研究对于耕地资源保护及后续工程建设具有积极意义。

After the gully land consolidation(GLC)project was implemented in Gutun Watershed of Yan’an City,Shaanxi Province in 2013,salinization has been found in some of the new farmland.In order to obtain the spatial distribution characteristics of salinization,Gutun Watershed was selected as the study area,and the Gaofen-1 satellite image of the area and soil samples over the same period were collected.Based on the collected data,the salinization of all cultivated land after the project was quantitatively inverted by the salinity index model and the support vector machine(SVM)model established through the feature bands of Gaofen-1 satellite image.Research shows that the selected salinity index S3 is most correlated with salinity,and its inversion model R^(2) is 0.742.The selected three characteristic band combinations used for SVM modeling,and the results showed that the SVM model based on band 1 and band 3 combination(P_(1)P_(3))had the optimal inversion capability,with R^(2) up to 0.867.The inversion results of SVM model show that the normal soil area in the study area accounts for 26.62%,mild salinization accounts for 43.52%,moderate salinization accounts for 29.82%,and severe salinization accounts for almost none.The study has positive significance for protection of cultivated land resources and the follow-up construction.