The present paper selected Qingyundian town and Weishanzhuang town in DaXing District, and C, aolingying town in Shunyi District as test areas, using MODIS data and ASTER data in different scales. The feasibility of winter wheat LAI inversion by PROSAIL physical model, especially the stability of remote sensing data in different scales, was discussed, and the results from experience model inversion were compared with that from statistical methods. The values of all samples LAI inversion from experience model are close in a region, which means experience model is a reflection of general growing trend, ignoring spatial heterogeneity of the regional leaf area index. But the value of LAI inversion from physical model can be truer in reflecting spatial heterogeneity of the regional leaf area index. The value of LAI inversion from physical model is more real, compared with experience model. With the method of linear weighing, the scale conversion was accomplished, and the LAI inversion results from different remote sensing scale data were compared, and were found similar. The result shows that in the process of largescale regional LAI inversion, physical model inversion is more valid.