A principal goal of seed germination modelling for wild species is to predict germination timing under fluctuating field conditions. We coupled our previously developed hydrothermal time, thermal and hydrothermal afterripening time, and hydration–dehydration models for dormancy loss and germination with field seed zone temperature and water potential measurements from early summer through autumn to develop predictions of germination timing for Bromus tectorum at a semi-arid site in north-central Utah, USA. Model predictions were tested with a validation dataset based on concomitant seed retrieval experiments in 2 years. Predictions were generally in agreement with observed field germination time courses, even though integration across multiple precipitation events was necessary. Success of the modelling effort hinged on two factors. First, we used a soil capacitance sensor that measured seed zone (5 mm soil depth) water content accurately over a wide range. Second, simulations were built using physiologically based threshold models that can incorporate differences in germination timing for multiple germination fractions and for multiple stages of dormancy loss. Our results suggest that simulation models using hydrothermal time concepts can predict field germination phenology accurately. Seeds in this study integrated their experiences in a widely fluctuating environment in a manner consistent with the assumptions of hydrothermal time. Such threshold-based models also have the advantage of generality, as these concepts can be applied to many different species, environments and weather scenarios.