- Title
- Assessing land degradation and the effectiveness of calcrete bontveld rehabilitation in a grassridge PPC cement mining area using multi-sensor remotely sensed data and machine learning techniques
- Creator
- Mpisane, Khanyisa
- Subject
- Land degradation -- South Africa
- Subject
- Environmental degradation
- Subject
- Mines and mineral resources -- South Africa
- Date Issued
- 2023-12
- Date
- 2023-12
- Type
- Master's theses
- Type
- text
- Identifier
- http://hdl.handle.net/10948/62527
- Identifier
- vital:72821
- Description
- This study uses multi-temporal remote sensing data to spatially visualise and quantify land degradation due to mining as well as Calcrete Bontveld rehabilitation at the Grassridge PPC Cement mine, Gqeberha formerly known as Port Elizabeth in the Eastern Cape, South Africa. Botanical field data is also used to verify the suasses of rehabilitation in the area. SPOT-6 and Landsat multi-spectral images were compared, and Support Vector Machine and Random Forest algorithms were used for classification in order to determine which yields more accurate results for a limestone mine. Support Vector Machine analysis using SPOT-6 images yielded the best results. This was due to the high spatial resolution of SPOT-6 compared to Landsat and Support Vector Machine classifier was able to classify images with fewer training points compared to Random Forest. The spatio-temporal land cover change at the mine was then determined between the years 2000, 2015 and 2020. Land cover classification is useful for monitoring land degradation and, in this case, was able to show the extent of rehabilitation success. For the year 2020, a 17% area was rehabilitated; however, the algorithm could not distinguish between unmined Calcrete Bontveld matrix and rehabilitation sites that were older than five years. The performed change detection also showed that 29.50% of unmined Calcrete Bontveld matrix had changed to “mature rehabilitation” (rehabilitation sites older than five years). Again, for this percentage in some areas the algorithm could not distinguish between the unmined Calcrete Bontveld matrix and rehabilitation sites that were older than five years due to high similarities between the two land cover types. Area changes of the different land cover classes could also be used to demonstrate how rehabilitation areas have matured over time and lead to the conclusion that most of the Calcrete Bontveld which was mined, has over the years been successfully rehabilitated. Vegetation analysis was conducted to further validate the rehabilitation success of Calcrete Bontveld matrix. Multivariant Detrended Correspondent Analysis showed that rehabilitation sites which were younger than five years (2–year-old rehabilitation plots that were sampled) had great dissimilarity to the natural unmined Calcrete Bontveld matrix and that rehabilitation sites older than five years, in this case 16–years older, had high similarity and resemblance to natural unmined Calcrete Bontveld matrix and therefore could be considered as being mature. This was a more definitive assessment as it considers all aspects of the vegetation. Species cover and species richness also showed that Calcrete Bontveld matrix rehabilitation sites which have been rehabilitated for more than 5 years had greater similarity to natural unmined vegetation compared to areas that have been rehabilitated for less than five years. This study, therefore, demonstrates that due to the high similarity between mature rehabilitation sites and unmined Calcrete Bontveld, rehabilitation has been successful.
- Description
- Thesis (MSc) -- Faculty of Science, School of Environmental Sciences, 2023
- Format
- computer
- Format
- online resource
- Format
- application/pdf
- Format
- 1 online resource (x,93 pages)
- Format
- Publisher
- Nelson Mandela University
- Publisher
- Faculty of Science
- Language
- English
- Rights
- Nelson Mandela University
- Rights
- All Rights Reserved
- Rights
- Open Access
- Hits: 59
- Visitors: 59
- Downloads: 4
Collections
NMU School of Environmental Sciences
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | Mpisane, K.pdf | 4 MB | Adobe Acrobat PDF | View Details Download |