Electrophysiological and anatomical variability are key to the initiation, maintenance and termination of atrial fibrillation, the most common arrhythmia. Limitations in depth, breadth and consistency of experimental and clinical recordings hamper the identification of the key properties regulating population-level variability in atrial fibrillation and its treatment. In this presentation, we will describe recent collaborative efforts to develop populations of whole-atria models informed by large clinical and experimental datasets to reflect population-level variability in atrial electrophysiology and arrhythmias. We will cover the methodologies required to encapsulate anatomical, electrophysiology and structural variability in the whole atria model populations, with clear strategies towards model calibration and validation based on key properties. We will also demonstrate how the populations of whole-atria models can be used to evaluate the efficacy of pharmacological therapy and to identify biomarkers for patient risk stratification.