The grassland biome, which is classified as a terrestrial ecosystem, contributes significantly to carbon sequestration. About one third of this ecosystem covers the land surface in south Africa and faces the danger of being eradicated. Much of the pressure is attributed to by synthetic initiatives that seeks to expand the economy of the country thereby meeting the demands of cumulative population. Mining, agriculture, and human settlement are the main characters. Given the paucity of research on the threatened ecosystem, the support vector machine learning algorithm (SVM) is employed to investigate fragmentation from 2016 to 2023. We used Sentinel-2A/B satellite images to learn more about spatial and temporal patterns, as well as the distribution of fragmentation in the grassland biome, using the Google Earth Engine platform. The findings revealed that grassland occupied 66% of the area in 2016, decreased to 52% in 2019, and then increased to 59% by 2023. The inconsistency in the pattern or trend of the grassland class is likely attributable to the expansion of the other classes. The SVM model indicated an overall classification accuracy of 97.62%, 97.66% and 98.58% in 2016, 2019, and 2023, respectively. In contrast, the models developed to relate LAI to NDVI, MSAVI2, OSAVI, and NDRE produced R2 values of 0.6396, 0.6325, 0.6387, and 0.6344, respectively. An in-depth understanding of the fragmentation patterns observed in grasslands yields valuable information for the formulation of conservation strategies, sustainable land-use planning, and climate-resilient management approaches aimed at safeguarding South Africa's distinctive grassland ecosystems.
Read full abstract