Fractional vegetation cover (FVC) is a critical canopy structural variable essential for understanding vegetation dynamics. Estimating FVC in arid environments often remains difficult due to sparse vegetation, weak spectral signals, and high spectral confusion with the background. This study investigates the potential of spaceborne imaging spectroscopy to overcome these limitations by estimating the FVC of Tillandsia landbeckii, a fog-dependent bromeliad endemic to the Atacama Desert. Owing to its low reflectance and lack of chlorophyll absorption features, Tillandsia remains largely undetectable using conventional multispectral sensors. We used hyperspectral data from the EnMAP satellite and applied six semi-supervised sparse spectral unmixing algorithms at both local and regional scales. Unlike traditional supervised approaches, our method does not rely on labeled training data. Instead, it uses a limited set of field- and image-based endmember spectra to perform subpixel unmixing. Validation was conducted using a high-resolution reference dataset combining a UAV orthomosaic (1.7 cm) and Pl & eacute;iades Neo imagery (30 cm). While the best local model reached a mean absolute error (MAE) of 3.1%, restricting the regional regression to confirmed Tillandsia pixels further reduced the MAE to 1.8%. This study presents the first operational demonstration of EnMAP for subpixel mapping of Tillandsia landbeckii cover, highlighting the value of semi-supervised unmixing techniques for vegetation analysis in hyper-arid environments with limited reference data and strong spectral background similarity. The approach establishes a transferable framework for estimating low-signal vegetation in hyper-arid regions, advancing the state of the art in fractional cover mapping.