Ecosystem productivity and dynamics issued from multispectral and hyperspectral satellite imagery

摘要

Low resolution images from MODIS multispectral sensor are used for extracting indices correlated with major parameters of productivity, for two deciduous forests (Fagus sylvatica, Quercus sp.) and one shrubland dominated by the semi-deciduous Phlomis fruticosa. Ground ecophysiological measurements were conducted for three growing periods (2005-2007) and are used for indices evaluation as well as input parameters for an ecosystem productivity model. The results of the ground-based productivity model are compared to the 8day MODIS GPP product, showing that MODIS algorithm underestimates productivity and does not closely follow ecosystem dynamics. In an attempt for a more precise productivity product a new light-use efficiency model based on satellite and meteorological data is designed and presented. Moreover, hyperspectral images from CHRIS/PROBA are used for a more detailed study of the semi-decidual Phlomis fruticosa ecosystem. Ground ecophysiological measurements from two growing periods (2006-2007) are used for evaluation purposes. Images are geometrically corrected and atmospherically adjusted. The reflectance spectra obtained are used for extracting indices related to numerous plant physiological parameters. Fast responsive plant processes, such as the function of the photosynthetic apparatus, the photoprotective response to stress factors (low or high temperature, lack of precipitation) and the detailed pigment content of leaves (chlorophyll a, chlorophyll b, carotenoids) may well be followed by such indices issued from hyperspetral data, offering great advantage over multispectral images for ecosystem remote sensing.