Despite many efforts to develop evapotranspiration (ET) models with improved parametrizations of resistance terms for water vapor transfer into the atmosphere, estimates of ET and its partitioning remain prone to bias. Much of this bias could arise from inadequate representations of physical interactions near nonuniform surfaces from which localized heat and water vapor fluxes emanate. This study aims to provide a mechanistic bridge from land‐surface characteristics to vertical transport predictions, and proposes a new physically based ET model that builds on a recently developed bluff‐rough bare soil evaporation model incorporating coupled soil moisture‐atmospheric controls. The newly developed ET model explicitly accounts for (1) near‐surface turbulent interactions affecting soil drying and (2) soil‐moisture‐dependent stomatal responses to atmospheric evaporative demand that influence leaf (and canopy) transpiration. Model estimates of ET and its partitioning were in good agreement with available field‐scale data, and highlight hidden processes not accounted for by commonly used ET schemes. One such process, nonlinear vegetation‐induced turbulence (as a function of vegetation stature and cover fraction) significantly influences ET‐soil moisture relationships. Our results are particularly important for water resources and land use planning of semiarid sparsely vegetated ecosystems where soil surface interactions are known to play a critical role in land‐climate interactions. This study potentially facilitates a mathematically tractable description of the strength (i.e., the slope) of the ET‐soil moisture relationship, which is a core component of models that seek to predict land‐atmosphere coupling and its feedback to the climate system in a changing climate.