Tejon Ranch is a spectacular landscape valued for its biological diversity, livestock production, and cultural heritage. Situated at the convergence of three of California’s major biogeographic zones, it is the largest, contiguous private property in the state. Since 2008, the ranch has operated under a conservation agreement that established adaptive livestock management for conservation and economic outcomes. We developed and applied ecological site concepts to adaptive conservation-focused grazing on 40,000 ha of grasslands on the ranch. Environmental heterogeneity, non-equilibrium system controls, and unpredictable rainfall required novel ecological site descriptions (ESDs) and state-and-transition models (STMs) to understand system history and dynamics, and to predict responses to management. Each ecological site is based on a “geologic landform” class:  a unique combination of biogeographic region, geologic material and age, dominant formative geomorphology, elevation, and slope that supports a distinct set of soils and vegetation dynamics. We randomly established 57 permanent plots within geologic landform classes. Sampling these plots over several years allowed us to evaluate the spatial and temporal variation within and among geologic landform classes and to build data-driven, quantitative ESDs and STMs. Using these models, we developed hypotheses about suitable management practices for conservation goals. We discuss in detail the Holocene Flats ESD, which we identified as having the highest potential for enhancement of conservation values through altered grazing practices predicted to favor several taxa, including the San Joaquin kit fox, blunt-nosed leopard lizard, and native annual forbs. The predicted responses are currently being tested and monitored.