Grass to shrubland conversions have been widely reported in drylands world wide, but remain poorly understood. Knowledge of ecosystem spatial properties and the environmental constraints underlying landscape pattern is crucial to determining why shrublands have replaced grasslands in some areas and not others as well as predicting where such transitions may be most likely to occur in the future. A useful framework for understanding patterns of grassland-to-shrubland transitions consists of three interacting concepts: 1) direct impacts 2) spatial heterogeneity of underlying constraints, and 3) spatial variation in process feedbacks. The application of this framework allows spatial modeling of landscape change to be appropriately constrained at various spatial/temporal scales ranging from those operating at the scale of plants to those operating at the scale of landforms. Mesquite (Prosopis glandulosa) invasion of grasslands in southern New Mexico and the proliferation of juniper (Juniperus monosperma) in pinion-juniper savanna in central New Mexico are two cases where conceptual models of spatial influences on state transitions may be developed and evaluated. Repeat air photography and remotely sensed images from the Jornada LTER and Chupadera Mesa were analyzed to quantify landscape-scale spatial patterns of transition at decadal time-scales. Patterns of change on imagery were related to contemporary plot-scale (20 X 20 m) field measurements of vegetation structure and soil to identify settings that conferred vegetation stability or susceptibility to a type conversion. On the sandy mesquite site, stable grass and grass-mesquite patches were structured by local soil characteristics at the plant-patch scale (e.g. petrocalcic horizons and surface texture) and landscape position at the basin scale. Conversion to shrub-dominated coppice dunes and loss of grass was a function of the density and arrangement of mesquite plants at the plant-patch scale. At the juniper site, persistence of a savanna structure and the conversion to juniper woodland was controlled by thresholds related to juniper density and arrangement that were dictated by landscape scale variation in soils, geomorphology, elevation, and topographic setting. This example illustrates contrasting sites where the processes differ but the outcomes do not (i.e., spatial variation in temporal changes in grass-woody ratios). The relative importances of each element of the conceptual framework provide ecological insight to grassland-to-shrubland transitions as a whole as well as these specific cases which may inform future management.