Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Australie → 2006 → Using high frequency satellite imagery to estimate forage vegetation in a semi-arid environment

Australian National University (2006)

Using high frequency satellite imagery to estimate forage vegetation in a semi-arid environment

Hume, Iain H.

Titre : Using high frequency satellite imagery to estimate forage vegetation in a semi-arid environment

Auteur : Hume, Iain H.

Université de soutenance : Australian National University

Grade : Doctor of Philosophy (PhD) 2006

Résumé partiel
Large areas of Queensland have 15% tree cover with a grassy understorey. The grasses of this understorey provide the basis for an extensive cattle grazing industry. It is important to monitor this resource base to prevent its degradation and promote its optimal use. Satellite based remote sensing has been used to monitor vegetation at continental to global scales but the difficulty in separating the grass from the trees complicate its application in woodland situations. The spectral difference between tree and grass leaves was measured in laboratory experiments. Variation in the spectral reflectance among tree leaves and among grass leaves was greater than the difference between the spectral reflectance of tree leaves and of grass leaves. They were then spectrally indistinguishable. A laboratory experiment measured the light balance of tree leaves and grass leaves under direct and diffuse irradiance. Tree and grass leaves had similar light balances under beam irradiance ; however, both types of leaf absorbed two times more diffuse irradiance than they did beam. A set of daily NOAA A VHRR observations spanning five years was acquired ; it encompassed a lage proportion of Queensland’s productive woodlands. Maximum value NDVI compositing, the current standard, biased the viewing geometry of the composite image which resulted in maximising the amount of shadow in the image. This effect varied seasonally, spatially and with landcover type. A new, single, measure of image geometry was developed, the ’ray angle’. It was used as the basis of a compositing technique that produced composites with more consistent image geometry enabling more confident ecological interpretation. The new compositing method was compared with the current standard by comparing time series of the Normalised Difference Vegetation Index (NDVI). There were differences between compositing methods and these differences were confounded by correcting for atmospheric and bidirectional reflectance function (BRDF) effects which elevated the NDVI. These correction effects were more pronounced in forests. NDVI time series were assumed to be a mixture of the evergreen (tree) and the raingreen (grass) signal. Classical time series decomposition was shown to be limited in these woodlands and a method capable of unsupervised unmixing of these components was developed and applied. As expected, evergreen NDVI time series of different compositing methods were more similar than their raingreen NDVI time series. The new, ray angle, compositing method gave the raingreen estimates with the most ecologically rational results both at a point and regionally.


Version intégrale (416 Mb)

Page publiée le 28 janvier 2021