To monitor the recovery of regenerating rainforest a network of wireless sensor nodes are being us in Queensland’s Springbrook National Park. Located in areas of open grassland, regenerating rainforest and old rainforest, the nodes monitor micro-climate indicators including; temperature, humidity, leaf wetness, soil moisture, wind speed and wind direction. The sensor network is a valuable research platform for the study of land-use change. Measuring the effects of invasive species on biodiversity, the ecological functioning of rainforests and the impacts of climate change
Rainforests are typically characterized by areas with very limited solar energy with adverse and dynamic radio environments. In order to develop the network and energy management protocols required for robust and reliable performance of long-term, rainforest networks, we had to first quantify the performance of current WSN technology under these conditions.
As wireless sensor networks (WSNs) continue to increase in their use across a broad range of applications, another cat- egory of energy consumers becomes increasingly significant — the sensors themselves. WSNs have been seen, from the earliest days of the research community, as a new means to measure and monitor the natural world. In order to quantitatively measure characteristics of any phenomena in the nat- ural world requires a sensor – a device that measures some physical quantity and converts it into an electrical signal. In the case of a single node measuring multiple phenomena (or physical quantities) at a single spatial location, the sensor energy load can begin to increase rapidly.
The issue of management of sensor energy load is an under-explored and important area for the sensor networks community. As nodes include a higher number of sensors, some of which have energy cost (for instance, audio and video), sensing load will certainly increase in significance relative to overall node energy consumption. If future sensor network applications are to provide lifetime guarantees, nodes must also be able to efficiently manage their sensing energy load and still deliver sufficient information about the phenomena being measured in the first place. Proper management of the sensor energy load can enable higher nominal sampling rates for many types of sensors, allowing better detection of anomalies and the ability to better describe a broader range of phenomena.
To address the above questions, we design a new WSN energy architecture which allows nodes to adapt their sampling strategy for each sensor individually in response to changing environmental conditions, in a way which maximises the value of information returned to the user within the tight energy constraints at nodes. While previous work has attempted to maximize data or information flow from WSN’s, our work here aims at maximizing the extraction of interesting information from the network, by basing decisions on sampling strategy on Bayesian surprise in the individual sensor data stream.This work makes the following key contributions: