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 hexagonal 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 relationships accord with the same law. It is proved that this assumption works for the areas of many internal homogeneous features derived from images with different spatial resolutions. Lake surface is a certain internal heterogeneous feature, thus the surface areas derived from the different spatial resolution accord with the similar relationship. Taking the lake surface area data derived from images with high spatial resolution as reference, a mechanism 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 simulates 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 significantly 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 gradual accumulation of the sample data, the correction effect of the model can be improved. It is helpful for estimating the lake water area more accurately in a large region.