| dc.description.abstract | Current air-quality maps in The Netherlands are produced on a 4x4 km² spatial resolution by a interpolation model, called RIO. This model is based on a low resolution measurement network, which misses air-quality on micro-scale. However, deployment of low-cost sensors has helped in producing detailed air-quality maps in the recent years. One of the entities that implemented a low-cost sensor network, through the citizen science project called Snuffelfiets, is the Province of Utrecht. These sensors measure particulate matter with particles smaller than 2.5 micrometer (PM2.5). This research paper investigated the quality of the observations done by low-cost sensors of the Snuffelfiets project and compared these to the estimated PM2.5 concentrations from the RIO model. First, the raw dataset is cleaned to remove unreliable observations. Observations are then aggregated within a 1x1 km² grid cell on a weekly basis between 06:00 and 20:00 on Monday through Friday and the mean PM2.5 concentration is assigned to the corresponding grid cells. The calculated PM2.5 means are then compared to the referenced RIO concentrations using the one-sampled t-test. The outcome is used to produce t-score and p-value maps, which show if there is a statistical difference between the low-cost
sensors and the RIO model. Results did not show any spatial pattern between the weeks of analysis, which was due to the irregularities in temporal and spatial distribution from the Snuffelfiets observations. In addition, the quality of the Snuffelfiets sensors combined has been investigated against the advanced RIVM sensors. The analysis showed no systematic bias between the low-cost sensors and the advanced sensors. | |
