Rangelands are globally being transformed drastically by climate change and anthropogenic activities such as the settlement expansion, crop farming and grazing activities. These changes impact ecosystem services by altering land cover, resource availability, and ecological balance. However, the lack of spatially explicit criterion and frameworks for monitoring these changes and their impacts on ecosystem services necessitates timely accurate and effective methods evaluating transformation and fragmentation. This study utilised the random forest classifier and Landsat data in Google Earth Engine (GEE) to assess the spatio-temporal variability of rangelands KwaZulu Natal communal areas from 2000 to 2020. Fragmentation levels were also assess using geospatial statistically techniques. In Inhlazuka, grasslands classification achieved overall accuracies of 75%, 79% and 83%, while Vulindlela achieved 89%, 85%, and 89% for the years 2000, 2010 and 2020, respectively. Rangelands in Vulindlela decreased at a rate of 37.52 ha per year between 2000 and 2010 and at 76.46 ha from 2010 to 2020. In contrast, Inhlazuka exhibited a decreased of 40 ha per year between the years 2000 and 2010 followed by an increase of 45.28 ha annually from 2010 to 2020 due to a decline in the forest class. Key spectral features for discriminating the grasslands were Bands 5 (NIR) and 7 (SWIR). Rangeland fragmentation increased, with mean patch sizes in Vulindlela shrinking from 32 ha in 2000 to 9 in 2020, because of urban expansion. In Inhlazuka, mean patch sizes decreased from 2 to 1 ha in 2020. These findings underscore the importance of developing spatially explicit monitoring frameworks to guide sustainable communal rangeland management policies, preserving ecosystem services and mitigating land degradation.