Crop residue helps to moderate soil temperature and increase water use efficiency in the short term, while providing improvement in soil quality, increasing soil organic carbon, and facilitating biodegradation of pollutants for long-term sustainability. Since good management of crop residue can also increase irrigation efficiency and reduce erosion, remote sensing-based techniques are receiving increased attention for monitoring crop residue coverage. Indices based on differences and ratios of hyperspectral bands are considered state-of-the-art for operational applications, but are limited because of low signal-to-noise-ratio (SNR) image data from pushbroom sensors such as Hyperion. This study aims to investigate spectral unmixing as an alternative approach to effectively estimate and monitor crop residue cover with airborne and space-based hyperspectral sensors. The secondary aim is to compare traditional linear unmixing to manifold learning-based unmixing approaches to capture nonlinearities inherent in hyperspectral data. For the data in this case study, manifold learning approaches provide more robust estimates than either the cellulose absorption index (CAI) or linear unmixing of airborne and Hyperion hyperspectral data.