Browsing by Subject "Sentinel-2"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Open Access
    Exploring the utility of Sentinel-2 imagery in combination with environmental variables for monitoring Acacia mearnsii and Pinus pinaster in the Fynbos biome
    (2026) Masimula, Ntandokazi Sijabulisiwe; Visser, Vernon
    Acacia mearnsii and Pinus pinaster are invasive woody alien plant species threatening the biodiversity of the Fynbos biome, a global biodiversity hotspot in South Africa. Effective monitoring is essential due to their extensive spread and potential for further invasion. This study assessed the use of freely available Sentinel-2 satellite imagery combined with vegetation indices, environmental variables and machine learning for different seasons of peak and non-peak greenness to improve species detection and mapping. Field data from the Langkloof Valley, including GPS invasion points and vegetation cover, were used to train six Random Forest classification models. Predictor variables selected based on a correlation threshold (r ≥ 0.7) included spectral bands, vegetation indices, and environmental factors across peak and non-peak seasonal means. Results highlight the importance of topographic variables, specifically the topographic wetness index, in increasing classification accuracy. The most accurate model, incorporating Sentinel-2 spectral data, vegetation indices, and environmental variables during peak greenness, achieved 94% overall accuracy and a kappa coefficient of 0.79 using 5-fold cross-validation. These findings demonstrate that incorporating spectral data with environmental data offers a cost-effective, repeatable approach for invasive species monitoring in the Fynbos biome. Future research should refine classification models with phenological insights and expand monitoring efforts to improve targeted management interventions.