Identification of scar and fibrosis from late gadolinium enhanced (LGE) MRI is a key input to patient specific simulations of cardiac arrhythmia and proposed treatment strategies across ventricular, atrial, and whole heart modeling applications. There is therefore a clear need to have a robust and reproducible method of extracting this substrate information from LGE MRI, however, several automated approaches exist and there is not clear consensus on which is best. We compared scar identification across four automated method sand 3 replicates of human guided scar identification using post first time atrial fibrillation ablation LGE MRI scans. Across 50 patients we applied: image intensity ratio (IIR) > 1.6 (either by axial slice or globally), the mean plus 3 standard deviations of the wall intensities (STD wall), and the mean plus 3 standard deviations of the blood pool (STD BP). Human guided scar identification also leveraged the ablation sites applied during catheter ablation. Identified scars were then compared via DICE coefficient, LGE MRI intensity thresholds,and downstream outcomes such as arrhythmia induction via simulation using openCARP. Inter-observer correlation of identified scar volumes(average DICE coefficient over all 50 subjects) was high between human ob-servers (0.70 to 0.81 across three observers). Correlation between human-guided and automated methods was poorer excepting STD BP (IIR slice: 0.39– 0.47, IIR global: 0.36 – 0.46, STD wall: 0.36 – 0.52, STD BP 0.66 – 0.76).Automated methods agreed poorly with each other. Modes of error (over vs underestimating scar) varied. Ongoing evaluation of simulation results suggest differences in simulated arrhythmia induction and possible ablation strategies suggested by each method. We plan to include these simulation results in preparation for the final manuscript.