Measurements of vegetation and soil dynamics used to anticipate (or reverse) catastrophic transitions in arid and semiarid rangelands are often difficult to interpret. This situation is due, in part, to a lack of empirically-based conceptual models that incorporate the effects of multiple processes, scale, spatio-temporal pattern, and soils. Using observations of multitemporal data from the Chihuahuan Desert, we describe a new approach to classifying vegetation dynamics based on multiple scales of vegetation and soil pattern as well as cross-scale interactions. We propose the existence of six types of mechanisms driving vegetation change including 1) stability, 2) size oscillation of plants, 3) loss and reestablishment of plants within functional groups, 4) loss of one plant functional group and replacement by another, 5) spatial reorganization of vegetation patches, and 6) cascading transitions that spread from small to broad scales. We provide evidence for the existence of these mechanisms, the species involved, and the geomorphic components on which they are observed in the Chihuahuan Desert. These mechanisms highlight the kinds of multi-scale observations that are needed to detect or interpret change and emphasize the importance of soil-surface properties for interpreting vegetation change. The classification is potentially general across arid and semiarid ecosystems and links spatial and temporal patterns in vegetation with ecological and geomorphic processes, monitoring, and restoration strategies.