Gaussian processes and sensor network calibration

Sensor networks provide a new measurement modality addressing the measurement requirements of societal challenges relating to climate change and environment, for example. In this paper, we first discuss issues with applying the GUM uncertainty framework to uncertainty quantification for sensor network measurements. We then discuss the calibration of the sensors in situ by exploiting the spatio-temporal correlation in the signal being measured. In particular, we look at collocation experiments in which a number of reference and test sensors are located in proximity so that the test sensors can be calibrated against the reference sensors.