NDVI obtained from NOAA AVHRR is suitable for vegetation monitoring at global scale. However, NDVI values are affected not only by seasonal changes of vegetation but also by other factors, such as anisotropy of ground, cloud, and atmospheric effect. In this paper, we propose a new method of correcting reflectance to remove undesirable variations from NDVI. This method is based on statistical time series analysis. Simple BRDF model is modified to state space model, which is a kind of time series model, and Kalman filter algorithm is applied for the estimation of surface reflectance. The result of sample points shows the possibility of correcting the ground reflectance. And some are discussed for the further improvement.
抄録全体を表示