부산대학교 도서관

Time series of vegetation indices derived from remotely sensed images have proven great potential in tracking and monitoring green land cover dynamics. Several techniques and tools have been proposed and developed to extract useful information from these time series. However, the statistical properties of these data are scarcely investigated and mostly ignored. In this present paper, particular attention is directed to the probability distribution of vegetation indices time series, more specifically, NDVI time series. NDVI time series of various plant types and geographic locations of Tunisia are investigated to determine the appropriate statistical distribution. Interestingly, the results show that the normal distribution is not a good fit to all our NDVI time series except one case that belongs to the lower Saharian bioclimatic region. Moreover, other distributions provide a good approximation. The best fit depends on the plant type and also on the bioclimatic region.