The degradation of rangeland grass is currently one of the most serious environmental problems in South Africa. Increaser and decreaser grass species have been used as indicators to evaluate rangeland condition. Therefore, classifying these species and monitoring their relative abundance is an important step for sustainable rangelands management. Traditional methods (e.g. wheel point technique) have been used in classifying increaser and decreaser species over small geographic areas. These methods are regarded as being costly and time-consuming, because grasslands usually cover large expanses that are situated in isolated and inaccessible areas. In this regard, remote sensing techniques offer a practical and economical means for quantifying rangeland degradation over large areas. Remote sensing is capable of providing rapid, relatively inexpensive, and near-real-time data that could be used for classifying and monitoring species. This study advocates the development of techniques based on remote sensing to classify four dominant increaser species associated with rangeland degradation namely: Hyparrhenia hirta, Eragrostis curvula, Sporobolus africanus and Aristida diffusa in Okhombe communal rangeland, KwaZulu-Natal, South Africa. To our knowledge, no attempt has yet been made to discriminate and characterize the landscape using these species as indicators of the different levels of rangeland degradation using remote sensing. The first part of the thesis reviewed the problem of rangeland degradation in South Africa, the use of remote sensing (multispectral and hyperspectral) and their challenges and opportunities in mapping rangeland degradation using different indicators. The concept of decreaser and increaser species and how it can be used to map rangeland degradation was discussed. The second part of this study focused on exploring the relationship between vegetation species (increaser and decreaser species) and different levels of rangeland degradation. Results showed that, there is significant relationship between the abundance and distribution of different vegetation species and rangeland condition.
The third part of the study aimed to investigate the potential use of hyperspectral remote sensing in discriminating between four increaser species using the raw field spectroscopy data and discriminant analysis as a classifier. The results indicate that the spectroscopic approach used in this study has a strong potential to discriminate among increaser species. These positive results prompted the need to scale up the method to airborne remote sensing data characteristics for the purpose of possible mapping of rangeland species as indicators of degradation. We investigated whether canopy reflectance spectra resampled to AISA Eagle resolution and random forest as a classification algorithm could discriminate between four increaser species. Results showed that hyperspectral data assessed with the random forest algorithm has the potential to accurately discriminate species with best overall accuracy. Knowledge on reduced key wavelength regions and spectral band combinations for successful discrimination of increaser species was obtained. These wavelengths were evaluated using the new WorldView imagery containing unique and strategically positioned band settings. The study demonstrated the potential of WorldView-2 bands in classifying grass at species level with an overall accuracy of 82% which is only 5% less than an overall accuracy achieved by AISA Eagle hyperspectral data. Overall, the study has demonstrated the potential of remote sensing techniques to classify different increaser species representing levels of rangeland degradation. In this regard, we expect that the results of this study can be used to support up-to-date monitoring system for sustainable rangeland management. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2011.
Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:ukzn/oai:http://researchspace.ukzn.ac.za:10413/8583 |
Date | January 2011 |
Creators | Manssour, Khalid Manssour Yousif. |
Contributors | Everson, Terry M., Mutanga, Onisimo. |
Source Sets | South African National ETD Portal |
Language | en_ZA |
Detected Language | English |
Type | Thesis |
Page generated in 0.0024 seconds