•A novel calibration method of the modified Mohr-Coulomb fracture model is presented.•Localization analyses and two material tests are used in the model calibration.•Experiments on AA6016 specimens covering a range of stress states are performed.•FE simulations predict crack initiation and propagation with good accuracy. This paper presents a novel calibration procedure of the modified Mohr-Coulomb (MMC) fracture model by use of localization analyses and applies it for three tempers of an AA6016 aluminium alloy. The localization analyses employ the imperfection band approach, where metal plasticity is assigned outside the band and porous plasticity is assigned inside the band. Ductile failure is thus assumed to occur when the deformation localizes into a narrow band. The metal plasticity model is calibrated from notch tension tests using inverse finite element modelling. The porous plasticity model is calibrated by use of localization analyses where the deformation histories from finite element simulations of notch and plane-strain tension tests are prescribed as boundary conditions. Subsequently, localization analyses are used to establish the failure locus in stress space for proportional loading conditions and thus to determine the parameters of the MMC fracture model. Finite element simulations of notch tension and in-plane simple shear tests as well as two load cases of the modified Arcan test are used to validate the calibrated fracture model. The predictions by the simulations are in good agreement with the experiments, even though some deviations are seen for each temper. The results demonstrate that localization analyses are a cost-effective and reliable tool for predicting ductile failure, reducing the number of mechanical tests required to calibrate the MMC fracture model compared to the hybrid experimental-numerical approach usually applied. Display omitted