Wireless sensor networks consist of hundreds or even thousands of tiny sensor nodes, which communicate with each other via radio, and gather information about, for example, temperature, pressure, or humidity. This proposal aims to develop useful models of the data traffic in such networks, describing in a statistical manner how information travels through the network of nodes. These models will be based in some cases, upon access to traffic measurements from real test-beds and network deployments and, in others, upon calculations from first principles. Besides using results from the extensive literature of simulation studies, this project will leverage data from projects that are already in progress, namely the DTI-funded project SECOAS, and the EPSRC-funded projects PROSEN and MC-DIAS. All three academic partners in the project are already working on these projects. Based upon these traffic models, we will assess, optimise and improve specific existing protocols, to make them suitable for our selected applications. Wherever possible, our findings will be validated in field trials.Protocols for wireless sensor networks are already well under investigation; however modelling of the traffic they generate is an important and virtually untouched topic, which will facilitate a much more informed approach to protocol design, modelling, modification and deployment. For example, models for wireless sensor network traffic will differ fundamentally from those for Internet traffic. Some of our new traffic models may be amenable to analytical solutions, if suitable assumptions are made.There will be two distinct sources of information for these traffic models:1. We will gather traces of wireless sensor traffic from existing field trials and test-beds. These will be made available to us from other projects that the project partners are involved in.2. We will work from first principles, by producing a statistical description of how much data a sensor node generates in a given application, which will be referred to as traffic source statistics throughout. By making assumptions about, for example, network layout and aggregation strategy, we will be able to produce models of traffic behaviour, which can then be compared with the measurements described above.The models we produce will lead to the development of new ways of emulating sensor network traffic. For example, using our traffic models, a sensor node could be programmed to mimic traffic from a large group of nodes. In this way, large sensor networks could be emulated using only a modest number of nodes, many of which are emulating part of a much larger network. It will hence be possible to obtain useful performance information more quickly than otherwise, using less equipment. This concept will be investigated and developed as part of our work. Moreover, there is the potential to make the emulation software we develop as part of this activity available to other groups that are researching this topic.We will also integrate the software development from earlier in the project, and deploy it in real-world trials. The proposers' involvement with SECOAS, PROSEN and MC-DIAS represents access to experimental scenarios that will enable studies of the protocol performance and analysis of results in the applications areas of the environment, power, water and telecommunications (see letter of support form British Telecommunications). Relationships will be established with these programmes and trials which will allow the integration of new protocol strategies with their extended programmes of activities e.g. wind farm test site at TUV NEL Ltd. The key aim will be to demonstrate the benefits of understanding the application and the resultant traffic profiles and the appropriate analysis of the protocol that supports that application.