Abstract: The arid desert meadow and alpine meadow are two main grassland types of Gansu province. From 2005 to 2007, spectral reflectance were measured in Anxi grassland and Maqu grassland. The spectral reflectance characteristics and differences of two main grassland types were analyzed and compared. The results showed that the canopy spectral reflectance of desert meadow was bigger than that of alpine meadow in the visible waves, while in the near infrared waves, that of desert meadow was smaller than that of alpine meadow. The differences of spectral reflectance between different vegetations that belong to the same type were significant in the near infrared waves. The canopy spectral reflectance of desert meadow was clearly smaller than the leaf spectral reflectance, while the canopy spectral reflectance of alpine meadow was bigger than the leaf spectral reflectance in the visible waves and the near infrared waves. The growing status of the same vegetation had great impact on their canopy spectral reflectance. The value of red edge parameters of alpine meadow was bigger than those of desert meadow. In conclusion, hyperspectral remote sensing can be effectively applied to classification and growth monitoring of grassland.