基于多源、多时相遥感影像的高原湖泊提取及其不确定性——以西藏羊卓雍错流域为例

Extracting Alpine Lake Information Based on Multi-Source and MultiTemporal Satellite Images and Its Uncertainty Analysis——A Case Study in Yamzhog Yumco Basin,South Tibet

基于野外实地考察,以西藏中南部羊卓雍错流域为例,探讨了基于多源、多时相遥感影像的高原湖泊提取.目前计算机辅助分类方法在高原湖泊提取上存在很大的局限性,目视解译方法虽然繁琐,但能结合丰富的经验信息,故而成为当今高原湖泊提取的较为现实与可行的方法.为此,针对各季湖面特征的不同相态,结合图像增强技术、DEM等辅助信息和野外实地验证,建立了高原湖泊不同时相的遥感解译标志.根据1972—2009年Landsat、CBERS和HJ-1A/1B等3种不同数据源的37幅遥感影像提取了流域内主要湖泊,并分析了湖泊面积的年内与年际变化.在此基础上,讨论了高原湖泊提取过程的不确定性,包括遥感影像甄选、几何校正误差、数字化容差和目视解译经验等给结果带来的不确定性.结果表明:引入高分辨率、多源遥感影像以及不受天气状况影响的微波影像有助于高原湖泊变化监测能力与解译精度的提高;而综合考虑地形信息、湖泊不同时相特征,适用于多种数据源的计算机自动解译算法研发,将提高长时间序列高原湖泊提取的效率.

The research on the lake extraction from satellite images and the extracted lake size variation can provide reliable method and indispensable information to understand the variation of alpine lakes with the accelerating warming.As a case study in Yamzhog Yumco Basin in southern Tibet,the method of extracting alpine lakes from multi-source and multi-temporal satellite images was explored with field survey experience.Analysis showed that automatic classification was hard to satisfy the accuracy for the investigation on the variation of alpine lake size.Although the visual interpretation took more load and time,it had better results by adopting more experience and consequently was a practical way to delineate the boundary of alpine lakes.Thus,the signals of image interpretation were built with the consideration of the characteristics of lake surface in different seasons and topographic information.Accordingly,the boundary of five major lakes in Yamzhog Yumco Basin were digitalized based on 37 scenes of satellite images from Landsat,CBERS and HJ-1A/1B during the period of 1972-2009.Then,the lake surface area was calculated and analyzed at seasonal and annual time scale.The uncertainty of lake extraction was discussed from four respective including images selection,geometric correction,tolerance in digitalization and misjudgment.In the future,higher resolution and microwave multi-temporal satellite images are expected to enhance the ability to monitor lake variation in detail at both of temporal and spatial scale.More advanced automatic interpretation arithmetic is urgent to be developed with the consideration of different feature of alpine lake in different seasons and multi-source images,which would help to deepen the understanding of long-term alpine lake variation.