Mapping tea plantation health in proportion to invasive alien plants and climate change in Tshivhase Tea Estate, South Africa: A remote sensing approach

Master of Earth Science in Hydrology and Water Resources

Department of Earth Sciences

Invasive alien plants (IAPs) pose substantial threats to agriculture, biodiversity, and ecosystem services delivery globally. IAPs decrease species richness and abundance through competition, predation, hybridization, and indirect effects. Tea plantations like any other agricultural entity are threatened by invasive alien species. Tea is important as it contributes significantly to job creation and export earnings in many developing countries as well as its tasteful flavour and health benefits. The control and removal of these invaders require accurate mapping. In recent years, many countries have adopted the use of satellite remote sensing to map tea health at various scales. The increasing effects of climate change threaten the health and production of tea. This study assessed the co-occurrence of IAPs present in the Tshivhase Tea Estate and mapped the dominant species (Solanum mauritianum) using geostatistics techniques (Inverse Distance Weighting (IDW), kriging, and regression splines). The study also explored the potential of SPOT 7 and Sentinel 2 satellite data in mapping the occurrence of S. mauritianum, Lantana camara and the co-occurrence amongst S. mauritianum, L. camara, and Chromoleana odorata in the Tshivhase Tea Estate in Limpopo, South Africa. The Vegetation Condition Index (VCI) has been used to assess tea plantation health and climate data to assess climate change. Tea plantation health was then related to S. mauritianum distribution. The most frequently occurring species was the S. mauritianum with a 53% occurrence proportion and 73% likelihood of occurrence followed by the L. camara at 25% proportion of occurrence. None of the geostatistic methods were significant in mapping the occurrence of alien invasive plants species with IDW yielded R2=0.04; the root mean square error (RMSE) =0.48; the mean absolute deviance (MAD) =0.42; kriging (R2=0.02; RMSE=0.47; MAD=0.45) and regression splines (R2=0.004, RMSE=3.95; MAD=1.81). The presence/absence and remotely sensed environmental data were used to generate a logistic regression model in R for S. mauritianum and L. camara occurrence as well as the observed and conditional co-occurrence probability of S. mauritianum (P1), L. camara (P2) and C. odorata (P3). The logistic regression model coefficients were then mapped on QGIS to produce invasive alien plant species (IAPs)’ occurrence frequency and co-occurrence probability maps. The SPOT 7 model performed the highest receiver-operating characteristic (ROC) in the area under the curve (AUC) of 0.94 in predicting the conditional co-occurrence of S. mauritianum and L. camara whereas the sentinel 2 model performed a ROC AUC of 0.86. This could be because SPOT 7 has a spatial resolution than Sentinel-2. The VCI was high throughout the Tshivhase Tea Estate with a 50-100% range. The S. mauritianum occurred in areas of high VCI thereby affecting the health of tea plantations through the competition of nutrients and other things. The temperature and precipitation showed a positive or significant correlation with VCI in 2015. The study has attempted to map tea health concerning invasive plant species using high-resolution satellite images. The study has concluded that geostatistics methods are not suited for mapping the occurrence of alien invasive plants in homogenous plantations especially when the area of study is small and with scattered invasive alien plant species distribution. On the other hand, remote sensing variables and co-occurrence in combination with logistic regression are well suited for mapping the prediction occurrence of alien invasive plants in the homogeneous plantation, especially when using high and spectral resolution satellite images.

NRF