Peat wildfires are the largest fires on Earth involving both flaming and smouldering combustion, with one leading to the other. A common ignition source of smouldering fires in tropical peatlands are intentional flaming fires used to clear surface vegetation. To capture the behaviour of these fires, it is necessary to consider the interaction between flaming vegetation and smouldering peat. However, doing so is infeasible with the state-of-the-art wildfire models, as they do not consider the transition from flaming to smouldering and are computationally too expensive at the field-scale hundreds of hectares. In this work, we overcome these limitations and model both flaming and smouldering at the field-scale using cellular automata: that is a discrete mathematical model that uses simple rules to capture complex behaviour while remaining computationally light. The model was calibrated against existing experiments in the literature and used to predict the effect of peat moisture content on the behaviour of peatland wildfires. The model shows how flaming creates smouldering hotspots and how these hotspots merge – flaming spreads rapidly, consuming surface vegetation, leaving behind hotspots of smouldering peat which consumes most of the peat. The model was then applied to study a real prescribed fire of 573 ha peatland in Borneo in 2015, observed by drone footage. The model captured the spread patterns of the fire and predicted that 2.9 ha of peatland burnt after 3 months with 70% peat moisture content (dry-based). This ioutcome could have been reduced to 0.02 ha if the peat moisture content had been above 100%. This work improves the fundamental understanding of how peat wildfires spread at the field scale which has received little attention until now.