There is an emerging need for technologies that can rapidly map the location and concentration of pollutants within a range of environments, particularly in the aftermath of an environmental disaster. The advent of UAVs and similar autonomous agents provide a platform for sampling pollutants within the environment, but what is currently lacking is the autonomous behaviour that determines where best to place the agent at any given moment in time. The ultimate aim of this research is to develop this autonomy using physical models that capture the dynamics of pollutant dispersal upon an underlying environmental flow of specific interest. These models, along with relevant external data-streams, can then be used to solve an optimisation problem to determine the ‘best’ location for placing sensors. However, to achieve this in real-time will necessitate the use of relatively simple models, which sacrifice prediction accuracy for numerical efficiency. The uncertainty induced by simple models can be mitigated through the real-time incorporation of sensor measurements, using sensor fusion techniques. As such, this project is highly interdisciplinary, requiring modelling from fluid mechanics and techniques borrowed from the control systems community.