The article presents stochastic wildfire model based on Monte-Carlo method and remote sensing data integration into wildfire propagation model. Currently, the accuracy of wildfire models is limited by both error in input data and models themselves. Stochastic approach allows accounting for sources of errors. Monte-Carlo method is used to calculate multiple realisations of wildfire spread accounting for different possible propagation scenarios and evaluate probability of fire spreading to nearby territories. Evaluation of model performance was done for a few hundred wildfires in Siberian larch forests. Performance was analyzed by way of comparing model’s probability estimates and actual frequency of wildfire reaching nearby areas. Analysis showed good agreement between model’s estimates and actual remote sensing data. This model can be used to estimate danger for settlements and infrastructure objects from nearby fires.