Bushfires are complex processes, making it difficult to accurately predict their rate of spread. We present an integrated software system for bushfire spread prediction, SPARK, which was developed with the functionality to model some of these complexities. SPARK uses a level set method governed by a user-defined algebraic spread rate to model fire propagation. The model is run within a modular workflow-based software environment. Implementation of SPARK is demonstrated for two cases: a small-scale experimental fire and a complex bushfire scenario. In the second case, the complexity of environmental non-homogeneity is explored through the inclusion of local variation in fuel and wind. Simulations over multiple runs with this fuel and wind variation are aggregated to produce a probability map of fire spread over a given time period. The model output has potential to be used operationally for real-time fire spread modeling, or by decision makers predicting risk from bushfire events.