Sensor networks allow data collection on virtually continuous scales.
The primary limiting factor associated with such networks is available energy. In this regard, communication is the dominant consumer. So, decisions must be made in terms of when to transmit from a sensor to the gateway; we refer to this as a suppression strategy. Also, there are issues with respect to transmission failure leading to strategies involving possibly redundant transmission from the sensors and appropriate record-keeping at the gateway. In this talk we describe some initial investigations into the impact of such mechanisms on global model fitting and data reconstruction in the context of an incomplete data likelihood reflecting these mechanisms. We examine information loss and reconstructive performance under fully Bayesian model fitting using simulated data under simple network and model specifications. Our broader objective is to implement such strategies for sensor networks collecting data to learn about soil moisture and carbon uptake models.