Atrial tachycardia (AT), atrial flutter (AFL) and atrial fibrillation (AF) are among the most common cardiac arrhythmias and are driven by localized sources (ectopic focus in AT, macro-reentrant circuit in AFL and rotors in AF) that can be targeted for ablation. We aimed to characterize such mechanisms from the non-invasive perspective of body surface potential mapping (BSPM), using realistic computer models. Dominant frequency (DF) maps were studied to estimate the frequency of the driving mechanism (fdrive) and to analyze its spatio-temporal distribution. Phase singularity points (SPs) were detected in phase maps and their distribution and rotational patterns were compared between arrhythmias. The fdrives were more expressed on the anterior portion of the torso for the right atrium and on the posterior portion for the left atrium. AF models presented distinct fdrive regions, with more regions occupying a significantly smaller portion of the torso. Rotational activity was detected in all arrhythmias, with increasing levels of spatial-temporal stability (AF, AT and AFL, respectively, with significant differences between AFL and AF). These results can be used to identify the driving mechanisms and, in the future, to locate them in the atria.