Abstract Citrus greening is a serious disease affecting citrus production in Florida and different parts of the world. This disease is spread by an insect vector and the trees are killed several years after infection. There is no known treatment for the disease. Disease detection and removal of infected trees is a critical part of citrus greening disease management efforts. This paper reports the evaluation of spectral features extracted from visible-near infrared spectroradiometer spectra for their potential to detect citrus greening disease. The extraction of spectral features is an effort to lower the cost of the optical sensor while maintaining their performance. Spectral features: (i) spectral reflectance bands and (ii) vegetation indices (VIs) were derived from 350–2,500 nm spectral reflectance data using two feature extraction methods: stepwise discriminant analysis and stepwise regression analysis. Following the selection of spectral features, the features were assessed using two classifiers, quadratic discriminant analysis (QDA) and soft independent modeling of classification analogies (SIMCA) to determine the overall and individual class classification accuracies. The classification results indicated that both the spectral features (spectral bands and VIs) yielded good overall (higher than 80%) and healthy class (higher than 85%) classification accuracies using the QDA-based algorithm. The SIMCA-based algorithm yielded good average citrus greening class classification accuracy (higher than 83%) using selected spectral features. Thus, the present study demonstrates the applicability of utilizing spectral features for detection of greening in citrus.