Burning houses, outbuildings and vehicles emit toxic fumes including benzene, hydrogen cyanide, carbon monoxide and many others. The conflicting needs of fire fighters to prevent fire spread by acting in the burning urban environment on the one hand, and to avoid hazardous exposure to toxins on the other hand, is a risk management problem.  Estimates of exposure from a single burning house are compounded when a number of burning houses in close proximity interact. The winds convey toxic pollution from each house, but the natural correlations of the winds and plumes in the atmospheric boundary layer mean that the peak short term exposure to toxins can be larger than for single-house emissions, and more frequent, depending on how closely spaced the burning elements of the cluster are, and how many houses are in the cluster.  Analysis of the structure of the plume of emissions downwind suggest some simple practical rules on approaching within the hazardous range of the plume: for example, approaching the fringe and not the centreline as a tactic when it is necessary to engage the fire. This requires an estimate of wind direction. Similarly deciding when too many houses in a confined burning cluster are emitting and cannot be approached, even when we expect that single burning house may be approachable. This requires a reliable observation of the number of houses burning and a spatial estimate of separations. More advanced systems of wind and plume-image observations can ultimately provide decision support tools for these risk management problems, but just a greater qualitative knowledge of the peak fluctuation properties for multiple source toxic emissions, which are given by plume dispersion models with correlated source winds, allows better informed judgement for firefighting scenarios in the urban-rural fringe environment.