The 2016 “Ecological Sites for Landscape Management” special issue of Rangelands recommended an update to our thinking of Ecological Sites, suggesting that in our desire to make Ecological Sites more quantitative, we abandoned consideration of Ecological Sites’ spatial context. In response, Ecological Site Groups (ESGs) and associated general state-and-transition modelwere proposed as a framework for describing landscape-level processes occurring across multiple ecological sites, and thus integrating multiple ecological sites from similar landscapes into common behavioral units. We hypothesized that the spatial distribution of ESGs could be predicted using readily available geospatial data due to the theoretical association between ESGs and landscape biophysical properties. Here we test ESG concepts with a spatial modeling framework using machine learning algorithms, a SSURGO modified NASIS point dataset, and a suite of remote sensing-based spatial covariates (e.g., hyper-temporal remote sensing, terrain attributes, climate data, land-cover, and lithology). Our modeling approach was tested on two Major Land Resource Area (MLRA) study areas within the western U.S., representing 6.1 million ha within MLRA 35 and 7.5 million ha within MLRA 42. Results show our approach was effective in mapping ESGs, with a 64% correct classification based on 1,406 point observations across 8 expertly-defined ESG classes in MLRA 35 and a 75% correct classification based on 2,626 point observations across 9 expertly-defined ESG classes in MLRA 42. National coverage of the training and covariate data used in this pilot study provides opportunities for a consistent national-scale mapping effort of ESGs.