Low-cost sensors and crowd-sourced data: Observations of siting impacts on a network of air-quality instruments.

Low-cost sensors offer the possibility of gathering high temporal and spatial resolution crowd-sourced data-sets that have the potential to revolutionize the ways in which we understand individual and population exposure to air pollution. However, one of the challenges associated with crowd-sourced data ('citizen science'), often from low-cost sensors, is that citizens may use sites strongly affected by local conditions, limiting the wider significance of the data. This paper examines results from a low-cost network measuring ground-level ozone to evaluate the impact of siting on data quality. Locations at both reference stations and at private homes or research centers were used, and thought of as a typical 'crowd-sourced' network. Two instruments were co-located at each site to determine intra-site variability and evaluated by standard performance statistics and local-scale activity logs. The wider application of the data for both regional Inter-site variability was evaluated to show-case the wider value and usefulness of crowd-sourced data. Analysis of intra-site variability showed little differences at most sites (<5ppb). Large differences in intra-site variability were detected when sensors were exposed to direct sunlight (causing thermal variations within the instrument) and proximity to large emission sources. Short-term local activities, such as lawn-mowing, were identifiable in the data, but had minimal impact on standard reporting time-scales, and so did not pose as being significant limitations or errors. Inter-site evaluation demonstrated that dense networks of low-cost sensors can add value to existing networks, with minimal impact on the overall data-set quality. Sensors located in crowd-sourced locations nearby to regulatory analyzers were able to capture similar trends and concentrations, supporting their ability to report on wider conditions. Thus crowd-sourced approaches to monitoring (with suitable calibration and data quality control checks) may be an appropriate method for increasing the temporal and spatial resolution of air quality networks.