Results are presented from of experiments to determine the susceptibility of rooftops to embers staying in contact during wildfires. Wind tunnel experiments were run in which the rooftops of model houses were covered with model embers and exposed to the wind. Tests covered a range of wind speeds, wind directions, and roof geometries. Experiments were filmed and image analysis showed regions on the rooftops where embers remained after prolonged exposure to the wind. Rooftop ember removal is controlled by the flow separation over the roof and around the building as well as the building geometry. Counterintuitively, the flow separation means that, for certain wind directions, increasing the roof slope actually results in larger regions of the rooftop where embers are retained. Internal corners, such as around dormers, can be regions of high ember stability. However, these regions of high stability are also dependent on the wind direction. For one building tested, the fraction of the roof that remained covered varied by 46% for the same wind speed but different wind angles. The results show that ember retention on rooftops is highly dependent on the flow over and around the whole building and, therefore, tests should model the entire building.