基于机理模型的不同空间分辨率影像提取湖面面积的误差研究

本文提出了一种湖面面积误差机理模型。首先,以椭圆形(或圆形)、三角形、六边形的面状地物为例,推算不同空间分辨率遥感影像提取面积间的关系,据此推测不同空间分辨率影像提取的湖面面积之间存在着较为固定的函数关系,且函数的系数与湖面的几何形状有关。然后,构建了不同空间分辨率影像的湖面面积间的误差模型,并用多组不同影像提取了艾比湖、阿雅克库木湖面积数据验证其有效性和适应性。结果表明:误差机理模型成功模拟了低空间分辨率影像提取的湖面面积与高空间分辨率影像提取的湖面面积之间的误差关系,可在一定程度上估计误差的具体数值并予以校正。因此,该模型具有校正低分辨率影像提取的湖面面积的能力,为在大区域范围准确地提取湖面面积创造条件。

A Mechanism Error Model for the Lake Water Area Estimated through Remote Sensing Images with Different Spatial Resolution

A mechanism error model for the lake surface area derived from the different spatial resolution data is presented. The model is based on the assumption that the differences of lake surface areas estimated by remote sensing images with different spatial resolutions are mainly determined by the mixed pixels on the water-land boundary but not the central part of the lake water surface. The elliptical （including circular）, triangular, and hex- agonal areal ground features are analyzed to illustrate the relationships between areas extracted from images with different spatial resolutions. It is found that for these three kinds of areal ground features, the calculated relation- ships accord with the same law. It is proved that this assumption works for the areas of many internal homoge- neous features derived from images with different spatial resolutions. Lake surface is a certain internal heteroge- neous feature, thus the surface areas derived from the different spatial resolution accord with the similar relation- ship. Taking the lake surface area data derived from images with high spatial resolution as reference, a mecha- nism model for the error of the lake surface area estimated through low spatial resolution data is established. In order to validate this model, it is applied to the data of Ebinur Lake, including lake surface areas based on SPOT VGT and ETM＋. The relationship between areas extracted from VGT and ETM＋ is analyzed, and it is found that the relationship could be described by the error model. To calculate the parameters of the model, the least square method is applied based on the area data derived from VGT and ETM＋. According to this model, the calculation values are obtained based on the area data extracted from VGT, and compared with the area derived from VGT. It is shown that this mechanism model is applicable to Ebinur Lake, and its principle is correct. The model simu- lates the error of the lake surface area extracted from low spatial resolution data, and estimates the value of error successfully. Therefore, it is capable to calibrate the lake surface area derived from low spatial resolution data with the high spatial resolution data. Three sets of data are employed for demonstrating the correction capability of the model, and the result shows that the precision of the lake surface area derived from VGT are improved sig- nificantly after correction. For further evaluation, the areas of Ebinur lake and Yake Kumu Lake estimated by MODIS are calibrated with the areas estimated by TM/ETM＋, and the similar results are obtained. With the grad- ual accumulation of the sample data, the correction effect of the model can be improved. It is helpful for estimat- ing the lake water area more accurately in a large region.