A criticality index for air pollution monitors

In this paper, we develop a criticality index for air pollution monitoring networks. The index quantifies the effect on the system of removing an air pollution monitor, which is calculated as a weighted sum of the differences between baseline values interpolated with historical data using the entire monitoring set and values interpolated with each monitor independently removed from the set. The interpolation procedure tests for linear dependencies in the north-south and east-west direction and then universal kriging is selected when linear dependencies exist, otherwise ordinary kriging is applied. The index's cost function is evaluated on a regular grid. The cost function includes a multiplication factor when false-positives or false-negatives above an air quality standard are introduced because of the removal of a monitor. The sensitivity analysis indicated the final index ranks were not sensitive to variations in the multiplication factors. The index captures the effects of wind direction in the study area. We apply the index to an industrial air pollution monitoring network in Hamilton, Ontario, Canada, which consists of eight air pollution monitoring units measuring PM10. We find a large difference in the cost with a minimum of 91,087 to a maximum cost of 1,733,835 or 0.052 and 1.00, respectively, for the proposed index. The removal of the monitor with the highest index value would have a significant effect on the network compared to the impact of removing any of the other seven monitors.