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

Accueil du site → Doctorat → Royaume-Uni → 1989 → Applications of remote sensing to arid grasslands : experimental and Nigerian case studies

University College of North Wales (1989)

Applications of remote sensing to arid grasslands : experimental and Nigerian case studies

Hassan, Bukar

Titre : Applications of remote sensing to arid grasslands : experimental and Nigerian case studies.

Auteur : Hassan, Bukar

Université de soutenance : University College of North Wales

Grade : Doctor of Philosophy (PhD) 1989

Résumé
Available from UMI in association with The British Library. This thesis contains the results of radiometric studies of glasshouse and field vegetation canopies, radiometric and physiochemical analysis of a Welsh sand and 11 soil samples from N.E. Nigeria, a small comparison of ground and satellite estimates of biomass on a sand dune system and a land use classification of N.E. Nigeria. Emphasis was placed on vegetation indices and separating senescent vegetation from bright sandy dry soils. Milton multiband radiometer (MMR) data were collected from glasshouse grown canopies of Lolium perenne cv. S23 L. ryegrass subjected to experimental regulation of water, fertilizer, cutting treatments and sands applications. Outside weather fluctuations in brightness affected readings, but this was corrected by introducing artificial lighting with two 500-watt fluorescent tubes. Near-infrared (NIR) was found to be more significantly related to biomass than the red. Some of the derived vegetation index values, the vegetation ratio (VR), the normalized difference vegetation index (NDVI), the green leaf area index (GLAI), the brown leaf area index (BLAI) and the perpendicular vegetation index (PVI) were related to harvested biomass. The VR, NDVI and GLAI were well correlated (r-value = 0.57 $<$ 0.86) with biomass. But the BLAI did not relate to yield (r-value = 0.32). The soils were 50-90% sands with base saturation of about 60% and the heavy mineral analysis show that they were derived from metamorphic rocks. Low iron oxides contents resulted in pale tones with very high red and NIR reflectance values largely to colour, low hygroscopic water and organic contents. There was little agreement between the Spectron ’Model SE590’ radiometer and MMR used in the spectral analysis. The same MMR radiometer was used in the field experiments. There were weaker relationships between vegetation variables and spectral data. Weather brightness, sun angle, background and vegetation conditions and topography were the major factors that affected radiometric results. But despite all these, the NDVI and the PVI were found to be significantly (r-value = 0.58 and r-value = 0.48, respectively) correlated to dune vegetation biomass of Aberffraw. A Landsat image of N.E. Nigeria was classified into eight unsupervised and seven supervised broad cover classes. The unsupervised classification was less useful as it could not identify some major land features given the environmental conditions. The similarity in infrared reflectance from drying vegetation and the sandy soil presents an inherent problem in Landsat studies of such semi-arid areas. As a result, it was impossible to separate lands under different agricultural practices in spite of the knowledge gained in the simulated glasshouse experiments, and some knowledge of the field situation.

Mots clés : Vegetation mapping in Nigeria Geography Pattern recognition systems Pattern perception Image processing Botany Geography Pattern recognition systems Botany

Présentation : EThOS (UK)

Proquest Dissertations & Theses

Page publiée le 25 mai 2009, mise à jour le 15 octobre 2018