中文摘要：

选取位于中国北方干旱区的湖泊乌梁素海为研究区,基于2002-2004年的LandsatTM数据以及同期实测水深数据,根据水深遥感原理,通过1次,2次,3次实测水深数据与遥感影像反射率及其组合之间的相关关系,选取五变量,三变量,二变量及一变量的非对数及对数模型进行湖泊水深估值,继而通过误差分析及模型对比,最终确定出3次实测数据五变量非对数模型作为乌梁素海的最优估值模型。该模型利用多光谱组合确定水深的估值模式,克服了单光谱信息难以综合反映沉水植物、芦苇、水色及底质空间变化的困难,符合干旱区浅水湖泊的特性,结果可为干旱区其它浅水湖泊水深反演提供新的思路和参考。

英文摘要：

In the paper, Wuliangsuhai lake, located in the arid area of North China, is taken as a case, in consideration of the characteristics of the shallow depth in the arid area. Through the relative analysis of one, two and three times＇ real-measured depth data and the reflectance of satellite images, the water depth is estimated based on the logarithm or non-logarithm model of five-, three-, two- and one-variable using the Landsat TM/ETM data and the real-measured depth data from 2002 to 2004 according to the principle of water depth remote sensing. By model comparison and error analysis, the non-logarithm model of five-variable with three-time mesured data is selected as the optimal depth model. The model uses the multi-spectral composition method to estimate water depth, solving the problem that the single spectral can not reflect the spatial changes of submerged weeds, reeds, water color and bottom matter. The model accords with the characteristics of shallow lake in the arid area and can provide the new approach and reference for the other lakes of the arid area.