The presence of process-induced meso-scale pores, typically present in a variety of cast and additively manufactured alloys, typically leads to a low fatigue limit and an important dispersion related to the distribution of pores. On the other hand, a scatter in fatigue life still exists due to inherent microstructural heterogeneity like silicon precipitates in the matrix in between pores. Therefore, a new fatigue model modelling uncertainty on two scales, i.e. due to varying pore distributions and micro-structural heterogeneity is proposed. A stochastic approach for the micro-plasticity based on a separation of scales is chosen to model uncertainty due micro-structural heterogeneity, and a weakest link assumption is applied to extend this theory to incorporate uncertainty due to varying distribution of pores. The finite element method is used to take into account the complex morphology of pores, and a previously implemented Neuber-type method was used for fast approximation of the full-field elasto-plastic stresses in porous structures. The parameters of the multi-mechanism model are obtained via a maximum likelihood estimate using tomography-based randomly generated synthetic porous specimens on experimental fatigue data.