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

Accueil du site → Master → Afrique du Sud → Analysis of land cover changes of the Hartebeesthoek Radio Astronomy Observatory environment over five decades

University of Pretoria (2018)

Analysis of land cover changes of the Hartebeesthoek Radio Astronomy Observatory environment over five decades

Mbwia, Lisa Nyadzua

Titre : Analysis of land cover changes of the Hartebeesthoek Radio Astronomy Observatory environment over five decades

Auteur : Mbwia, Lisa Nyadzua

Université de soutenance : University of Pretoria

Grade : MASTERS OF SCIENCE IN GEOINFORMATICS 2018

Résumé partiel
The focus of this research was to determine the multi-temporal land cover changes (LCC) of the Hartebeesthoek Radio Astronomy Observatory (HartRAO) environment over 5 decades. The HartRAO site is a strategic point in the mapping system of South Africa and acts as fundamental reference node in the International Terrestrial Reference Frame (ITRF). In this study, the conducted field assessments determined the land cover type of the study area and its environment. The use of aerial photographs, Google Earth images, satellite images and climate data were platforms to assess the LCC of the area and surroundings. The ENVI 5.1 software package was used to determine the LCC of the area from the Landsat TM and Landsat OLI satellites using image analysis. In addition, ArcGIS 10.2 was used to determine the hydrology of the area from the SRTM DEM file, provide a platform to view aerial photographs, and map out LC images determined from ENVI 5.1. Field assessments and demarcation of the study area were conducted using Google Earth images. The e-Water TREND software proved useful in determining the statistical significance of climate data over 5 the decade period. Microsoft Excel software was used to tabulate, generate graphs and charts from satellite image and climate data analysis. According to the first objective the delineation of land cover types in the area was done using aerial photographs and field assessments of the study provided pictorial information to interpret land cover changes over the years of the study. Analysis of satellite data for the last 19 years also showed the changes in the land cover type and this was used to delineate land cover types. This was followed by satellite image processing and analysis over the last 19 years involving : image ratioing, image classification, change detection and accuracy assessment. Trends in rainfall and temperature of the area over period in the study area were determined using climatic data. Combining primary and secondary data to provide visual interpretation on the changes in land cover type and the seasonal variability were tools identified for the study. Hence there is a note that there is an increase in woody vegetation within HartRAO and changes in land cover and land use activity which does affect the changes in climate and land cover type. Image processing : image ratioing of the Landsat TM and Landsat OLI images, the EVI, NDVI, NDBI and NDBaI spectral indices were used to give an overview of the land cover type and coverage in the area using the ENVI 5.1 software. A supervised image classification technique (pixel-based) was the ideal method for the Landsat images in this study this is because of the variability of the land cover type and when analysing will depend on spectral information. The second and third objective focuses on remote sensing techniques, where there is an importance in removing biasness when comparing of classification algorithms to determine the most suitable algorithm for image classification. Therefore in the study maximum-likelihood (parametric) and support vector machine (nonparametric) classification algorithms were used to determine the extent of the land cover type in the area. Change detection in the study was used to determine the level of changes within and between the land cover type in the area of study. The aim of accuracy assessment is to determine the performance of the classification algorithms based on the land cover type. This proved that both performed well but SVM had a slightly higher accuracy with most land cover types and as part of the study, classification algorithms can be used individually while assessing specific land cover types.

Présentation

Version intégrale (6,8 Mb)

Page publiée le 13 janvier 2019