基于遥感影像的新疆玛纳斯河流域土壤盐渍化分类

Soil salinization classification in Manas river basin of Xinjiang based on remote sensing images

【目的】研究新疆玛纳斯河流域盐渍化严重程度及分布情况,为治理土壤盐渍化提供参考。【方法】采用2007-08-19的玛纳斯河流域Landsat TM影像图,通过主成分分析选出典型波段,计算归一化植被指数和归一化差异水体指数,并结合近红外波段进行决策树分类,获取不同程度盐渍化土壤的分布情况。【结果】根据归一化植被指数、归一化差异水体指数、近红外波段及第3主成分,构建了玛纳斯河流域土壤盐渍化决策树分类流程,分类结果显示:玛纳斯河流域各个灌区都有不同程度的土壤盐渍化,其中重度盐渍化土壤主要分布在安集海灌区、132团灌区及玛纳斯河流域附近,约占玛纳斯河流域耕地面积的0.5%;轻中度盐渍化土壤主要分布在绿洲与沙漠过渡区、莫索湾灌区、136团灌区、石河子灌区、121团灌区以及重度盐渍化土壤周围,约占耕地面积的28%。地下水位埋深较浅的区域盐渍化程度严重。【结论】利用决策树分类法,通过内陆河小流域的遥感影像进行土壤盐渍化分类具有一定的可行性。

【Objective】 The degree and distribution of soil salinity in Manas river basin of Xinjiang were studied to improve the control of soil salinization.【Method】 Based on the Landsat TM image on August 19,2007 of Manas river basin,the typical band was selected through principal component analysis.The normalized difference vegetation index and the normalized difference water index were calculated.And decision tree classification was conducted by combining with the near infrared band.Finally,the distributions of different levels of saline soils were obtained.【Result】 The flow of decision-tree classification in the Manas river basin of Xinjiang was constructed based on the normalized difference vegetation index,the normalized difference water index,the near infrared band and the third principal image.The result of the classification showed that each irrigated area in the Manas river basin suffered from a certain degree salinization.Severe saline soil mainly distributed in Anjihai irrigation district,132 regiment irrigation district and the areas near Manas river,accounting for 0.5% of the whole cultivated land of Manas river basin.Mild-to-moderate saline soil mainly distributed in the edge of desert and the oasis,Mosuowan irrigation district,136 irrigation regiment district,Shihezi irrigation district,121 regiment district and the areas around the severe saline soildistricts,contributing to 28% of the whole cultivated land of Manas river basin.The salinization degrees were serious in the places where the groundwater tables were shallow.【Conclusion】 Decision tree is feasible to conduct salinity classification based on remote sensing image data in small watershed of continental rivers.