Research Projects

The montane grasslands of the Shola Sky Islands in the Western Ghats are ecologically critical ecosystems that harbor endemic flora and fauna while playing a vital role in maintaining hydrological regimes and water availability for downstream agricultural systems. These grasslands are increasingly under threat due to the invasion of alien plant species such as Acacia mearnsii, Pinus spp., Eucalyptus spp., Scotch broom (Cytisus scoparius), Gorse (Ulex europaeus), and Introduced historically for tannin / timber production and land stabilization, these species have aggressively expanded, transforming native grassland habitats, reducing biodiversity, and altering ecosystem functions. Our previous research made significant contributions to understanding these landscape changes. In a study focused on the Palani Hills, it was found that over two-thirds of native montane grasslands had been lost to plantations and agriculture over four decades (Arasumani et al., 2018). This work was extended to the entire Shola Sky Island landscape, where we quantified ecosystem-wide losses caused by invasive species (Arasumani et al., 2019). In another study, potential restoration zones in the Nilgiris, Anamalai, and Palani hills were identified based on spatial patterns of invasion (Arasumani et al., 2021a). Further, the suitability of hyperspectral (AVIRIS-NG), multispectral (Sentinel-2), and radar (Sentinel-1) imagery for detecting invasive species was evaluated for limited areas in the Nilgiris (Arasumani et al., 2021b). Building on this foundation, the current project aims to conduct a large-scale, species-level mapping of invasive alien species and remnant grasslands across the Shola Sky Islands using PlanetScope next-generation satellite imagery (3 m resolution). A regression-based spectral unmixing approach will be employed to generate fractional cover maps. Synthetic training data will be developed using spectral libraries and field-based reflectance profiles. These will be used to train a Support Vector Regression (SVR) model for pixel-level fractional cover estimation, even in spectrally mixed conditions. This methodology builds on our recent work using spaceborne imaging spectroscopy for mapping vegetation communities (Arasumani et al., 2023). PlanetScope pixels will be subdivided into 1 m grids, where visual interpretation and field-based cover estimation will be conducted. Accuracy metrics such as MAE, RMSE, and R² will be used for validation. To ensure greater accuracy, the predictions will be validated through detailed field sampling. Time-series PlanetScope data will be analysed to quantify the expansion dynamics of different invasive species and their relative impact on grassland loss. The final deliverables will include species-wise invasion maps, prioritized restoration sites, and a mobile GIS application to assist forest officials and restoration teams in locating and managing invasive patches.