Objective: The wildland-urban interface (WUI) is a transitional area between human habitation and natural ecosystems, such as forests, characterized by complex fuel distributions and diverse combustion properties. With the increasing frequency and scale of WUI fires resulting from urbanization and climate change, the need for accurate fire prediction has become critical. The existing models for predicting fire spread are often inadequate because they either analogize urban buildings to vegetation or do not consider differences in fire spread mechanisms between the two. This study addresses these limitations by proposing an extended multimodel combination method for predicting fire spread in WUIs, particularly by focusing on the interaction between vegetation and urban buildings. Methods: This study combines vegetation-building and building-vegetation fire models based on the interaction of vegetation and building fires to accurately predict fire spread in WUI areas. A cellular automata approach is used to divide the study area into grid cells, allowing for the modeling of fire spread across different cell types with unique state and combustion rules. Two specific models are built: a vegetation-building fire model, which evaluates the likelihood of buildings being ignited by adjacent burning vegetation on the basis of thermal radiation, and a building-vegetation fire model, which assesses whether surrounding vegetation could be ignited by fires originating in buildings. The study also integrates well-established models, such as the Rothermel model, for vegetation fire spread and heat radiation calculations for the ignition of buildings. The model is validated using a real-world case study of the Getty Fire in California, USA, which occurred in 2019. The results are compared with actual fire spread data and simulations from the FlamMap6 software to evaluate the model's performance. Results: The Getty Fire case study shows that the proposed multimodel combination method more accurately predicts fire spread than conventional single-model methods. The combination model effectively captures vegetation-building interactions, which are often ignored by traditional models. The burned area and fire perimeter are estimated with higher accuracy than FlamMap6, particularly in areas with mixed fuel types. The incorporation of thermal radiation calculations enhances ignition predictions, particularly in mixed vegetation and building areas, demonstrating the importance of modeling these interactions for higher accuracy. The model successfully predicts ignition timing for vegetation and buildings and dynamic changes in fire spread over time. Compared with FlamMap6, which uses lower grid resolution and lacks interaction modeling, the proposed combination model more precisely predicts fire behavior. FlamMap6 tends to overestimate fire spread in several areas, whereas the proposed method more accurately differentiates fire risks on the basis of fuel type. Conclusions: The proposed extended multimodel combination method addresses the limitations of the existing fire spread models by incorporating distinct models for vegetation and buildings and accounting for their interactions in WUIs. This will improve our understanding of fire spread in complex environments. The results of this case study indicate that this method can provide critical support for emergency management through more accurate and timely prediction of fire spread. Future work should include further optimization, integrating firefighting interventions and more diverse fuel types, to enhance the model's applicability and efficiency in large-scale fire scenarios.