Biological soil crusts (BSCs) are increasingly recognized as common features in arid and semiarid ecosystems and play an important role in the hydrological and ecological functioning of these ecosystems. However, BSCs are very vulnerable to, in particular, human disturbance. This results in a complex spatial pattern of BSCs in various stages of development. Such patterns, to a large extent, determine runoff and erosion processes in arid and semiarid ecosystems. In recent years, visible and near infrared (Vis-NIR) diffuse reflectance spectroscopy has been used for large-scale mapping of the distribution of BSCs. Our goals were (i) to demonstrate the efficiency of Vis-NIR spectroscopy in discriminating vegetation, physical soil crusts, various developmental stages of BSCs, and various types of disturbance on BSCs and (ii) to develop a classification system for these types of ground cover based on Vis-NIR spectroscopy. Spectral measurements were taken of vegetation, physical crusts and various types of BSCs prior to, and following, trampling or removal with a scraper in two semiarid areas in SE Spain. The main spectral differences were: (i) absorption by water at about 1450 nm, more intense in the spectra of vegetation than in those of physical crusts or BSCs, (ii) absorption features at about 500 and 680 nm for the BSCs, which were absent or very weak for physical crusts, (iii) a shallower slope between about 750 and 980 nm for physical crusts and early-successional BSCs than for later-successional BSCs and (iv) a steeper slope between about 680 and 750 nm for the most developed BSCs. A partial least squares regression-linear discriminant analysis of the spectral data resulted in a reliable classification (Kappa coefficients over 0.90) of the various types of ground cover and types of BSC disturbance. The distinctive spectral features of vegetation, physical crusts and the various developmental stages of BSCs were used to develop a classification system. This will be a promising tool for mapping BSCs with hyperspectral remote sensing.