Low cost sensors and crowd-sourced data to map air pollution in urban areas

Low cost sensors and crowd-sourcing data could potentially revolutionise the way air pollution measurements are collected providing high density geolocated data. In fact, so far data have been collected mostly using dedicated fixed position monitoring stations. These latter rely on high quality instrumentation, well established practices and well trained personnel, which means that, due to its costs, this paradigm entails limitations in the resolution and extension of geographic sampling of an area.

The combination of low costs sensors and volunteer-based or opportunistic acquisition of data can, instead, possibly turn the cost issue into an advantage. This approach, however, introduces other limitations since low cost sensors provide less reliable data and crowd source acquisition are subjects to data gaps in space and time.

In order to overcome these issues redundant data from multiple platforms have to be made available. On one hand this allows statistics to be applied to identify and remove anomalous values, and on the other hand when multiple platforms are used, the chances to have a better coverage and more reliable data  increases.

To implement this approach OGS developed the full suite of tools that has been named COCAL that allow to follow the full path from the acquisition, transmission, storage, integration and real time visualization of the crowdsourced data.

Low cost sensors for the detection of suspended particulate matter size 2.5 and 10 µm, together with atmospheric pressure, humidity and temperature, have been combined with GPS positioning and transmission (being able to opt for GSM, WiFi or LoRaWAN transmission) unit in a black box that can be attached to any moving vehicle travelling in an area. This way large areas can be sampled with high geographic resolution.

Atmospheric data are collected in an InfluxDB database, which allows easy integration with TheThingsNetwork for LoRaWAN network management and directly with GSM and WiFi connections. Public users are provided with a real-time web interface based on OpenLayers for map visualization. Server based processing and conversion scripts generate both filtered data and aggregate data, by computing averages on a spatial and temporal grid.. Finally, automatic interpolation techniques like Inverse Distance Weighting or Natural Neighbours may provide detailed online maps with contouring and boundary definition. All products are available in near real-time through OGC compliant web services, suited for an easy integration with other repositories and services.