Atrial fibrosis is characterized by expansion of extracellular matrix and increase in the number of fibroblasts which has been associated with the development and maintenance of atrial arrhythmias. However, the mechanisms how the fibrosis contributes to atrial arrhythmia remain incompletely understood. In this study, we used a proposed fibroblast model coupled with the human atrial myocyte to investigate the effects of fibrosis on atrial excitability and repolarization at both cellular and macroscopic levels. The 12-lead electrocardiogram (ECG) was also simulated to explore the index of clinical diagnosis for fibrosis. The simulation results showed that the fibrosis can modify action potential morphology of human atrial myocyte, slow down wave propagation, and have rate adaptation, thus causing the atrial electrical heterogeneity. The fibrosis alone was sufficient to cause arrhythmia, induce reentry wave, and result in low amplitude and wide P waves at normal heart rate and significant prolonged and inverse P waves at high heart rate. All these symptoms aggravated when the level of fibrosis increased. Our simulations demonstrated that fibrosis is the substrate of atrial arrhythmia and thereby may be a potential target in the treatment of atrial arrhythmias.