Patterns of herbivore browse at small scales, such as the rate of leaf consumption or plant preferences, drive the impact of herbivores on whole-plant processes, such as growth or survival, and subsequent changes in plant population structure. However, herbivore impacts are often non-linear, highly variable and context-dependent. Understanding the effect of herbivores on plant populations therefore requires a detailed understanding of the relationships that drive small-scale processes, and how these interact to generate dynamics at larger scales. We derive a mathematical model to predict annual rates of browse-induced tree mortality. We model individual plant mortality as a result of rates of foliage production, turnover and herbivore intake, and extend the model to the population scale by allowing for between-tree variation in levels of herbivore browse. The model is configurable for any broadleaved tree species subject to vertebrate or invertebrate browse, and is designed to be parameterized from field data typically collected as part of browse damage assessments. We parameterized and tested the model using data on foliage cover and browse damage recorded on kamahi trees (Weinmannia racemosa) browsed by possums (Trichosurus vulpecula) in New Zealand forests. The model replicated observed patterns of tree mortality at 12 independent validation sites with a wide range of herbivore densities and browse damage. The model reveals two key thresholds; in plant foliar cover, indicating when individual trees may be at high risk from browse-induced mortality, and in herbivore intake, leading to high rates of mortality across the whole population.