Abstract
Sclerocarya birrea is an extensively used and distributed plant in southern Africa. Almost all components of the tree are utilized by local people throughout its native habitat. In addition, the fruit has long been regarded as a valuable item with significant cultural, social, and economic value in many rural communities across the continent. The rapid decline of the S. birrea population in both private and public areas in South Africa have been a subject of concern in literature for an extended period. This necessitates efficient and urgent monitoring tools. So, modelling the occurrence of S. birrea is a first step toward conservation efforts. This research examined the accuracy of Landsat-8 and Sentinel-2 in predicting S. birrea presence based on a logistic regression model. It further evaluated the importance of Sentinel-2 bands, vegetation indices, and environmental variables in predicting S. birrea presence. The results presented Sentinel-2 as a better tool for predicting S. birrea because of the overall accuracy of 76.92% as opposed to the 62.82% obtained by Landsat-8. Moreover, the Sentinel-2-derived model obtained an Area Under Curve of 0.881 whilst Landsat-8 obtained 0.728. The success of sentinel-2 is attributed to the red edge bands and for Landsat-8, the shortwave infrared bands were significantly influential in the results they obtained (p>0.05). The analysis of the second objective indicated that only three vegetative indices, ARVI, GEMI, and S2REP, were significant in predicting S. birrea presence at a 95% confidence level. Furthermore, the aspect was the most significant environmental factor (p>0.05). The results of this study demonstrated and endorsed the use of Sentinel-2 and Landsat-8 for monitoring regional vegetation. These findings may assist local communities that rely on S. birrea for a living since S. birrea hotspots may be economic growth hotspots, which may benefit stakeholders. Many S. birrea-based enterprises, such as Amarula liquor and Portia M cosmetics, may provide employment for locals while generating income. Similarly, local authorities can use this study for the development of efficient conservation efforts.
Keywords: Remote sensing, Marula trees, Sclerocarya birrea, Logistic regression, Sentinel-2, Vegetation Indices