The development of ecological sites as management units has emerged as a highly effective land management framework, but its utility has been limited by spatial ambiguity of ecological site locations in the U.S., lack of ecological site concepts in many other parts of the world, and the inability to accurately assess the state of ecological sites or monitor changes in state through time. Here we present a modeling framework for high-resolution mapping of ecological sites in a semi-arid ecosystem using hyper-temporal remote sensing (i.e., hundreds of images) and support vector machine (SVM) classification. Results from this study show that SVM classification was effective in modeling ecological sites using a 28-year time series of normalized difference vegetation index (NDVI), with a 62% correct classification. Results were compared to Soil Survey Geographic database and expert delineated maps of ecological sites which had a 51 and 89% correct classification, respectively. The hyper-temporal remote sensing framework is effective in modeling the spatial distribution of ecological sites through its ability to characterize the soil-vegetation relationship and its response to climatic variability (i.e., drought or elevated rainfall). Additionally, the proposed hyper-temporal remote sensing technique may provide an objective framework to evaluate and test ecological site concepts through examining differences in vegetation dynamics in response to climatic variability, as well as the ability to assess and monitor changes in ecological state due to other change drivers. Given limited financial and human resources, an improved understanding of ecosystem potential is needed to maximize ecosystem services, promote the recovery of degraded lands, and adapt to and mitigate the impacts of climate change. The hyper-temporal remote sensing approach presented here has potential to greatly improve the efficiency of high-resolution ecological site mapping, and highlights its utility in terms of reduced cost and time investment relative to traditional manual mapping approaches.