Low-cost sensors and microscale land use regression: Data fusion to resolve air quality variations with high spatial and temporal resolution

The strong temporal and spatial gradients in NO2 concentrations frequently observed in urban microenvironments are very difficult to measure and model accurately. Recent developments in low-cost air quality instruments have led to improvement in the spatial coverage of time-resolved measurement, however interpolation is still needed to map pollutant concentrations and connect time-as well as space-dependent variations to urban design features. Here we propose a novel approach that uses a previously-described microscale land use regression (LUR) model to spatially interpolate data from a well-calibrated network of low-cost air quality instruments. We use a semiconducting oxide-based ozone sensor to provide a robust correction of the output of an electrochemical NO2 sensor for ozone interference. We characterise signal noise probably associated with meniscus fluctuations as a significant error source, that can be handled with appropriate signal averaging. The LUR model is used to provide high spatial resolution in the data set, whilst correlation with sensor measurements provides a time-dependent estimate associated with different land use types. Observations from the network of instruments showed marked variability in NO2 concentrations over short distances (on the scale of 100 m), with highest concentrations reached near bus stops, intersections and under shop awnings. This approach connects the complex time- and space-dependent variations to urban design features and is a promising way forward as a basis for objective spatial mapping of time-dependent mean concentration fields and local population exposure estimates.