In critical situations like those experienced recently in Japan, but also in less critical ones like those associated to fires in large industrial plants, the use of swarms of Unmanned Aerial Systems (UAS) operating in a cooperative fashion can be highly beneficial in reducing reaction times and danger. Falling prizes, increased on-board computational power and increased availability on the market allow this transition from strictly military operations to civil emergency and disaster management. One of these applications is the measurement of the concentation of toxic gases after an incident in a chemical plant to estimate the release rate and the danger zone. The optimal distribution of the agents of the swarm equipped with pollution sensors in the area to be scanned is a critical challenge. The communication between the individual agents as well as with the ground station is limited. Furthermore, a simple function allows a weighting of a distinguished part of the area to be scanned. This paper presents technical basics and a simple, decentralized but yet effective on-board implementation as well as results in simulation and from a field test.