Due to global biodiversity loss, the estimation of biodiversity using spectral technology has become a hot topic for ecologists and remote sensing scientists. There are many studies on alpha diversity but few studies on beta diversity. There are still some problems worth exploring. To explore the best spectral index and image spatial resolution for estimating plant beta diversity using remote sensing technology, this paper took meadow grassland as the research area. It calculated six beta diversity estimation indices from three aspects: spectral distance, spectral angle and biodiversity concept based on UAV hyperspectral remote sensing images. We developed four indices, and two are existing indices. Mantel tests and correlation coefficients were used to select the best spectral index. Then, the selected index was applied to images with different spatial resolutions to obtain the best observation scale. In addition, to improve the estimation ability of the index, this paper compared two spectral transformation methods, the first derivative transform and Savitzky-Golay filter, and three feature band selection methods: correlation coefficient, successive projections algorithm and the competitive adaptive reweighted sampling. The results showed that in both subscale observation (pixel size