Lena Weissert

Use of a dense monitoring network of low-cost instruments to observe local changes in the diurnal ozone cycles as marine air passes over a geographically isolated urban centre

Ozone (O3) concentrations in urban areas are spatially and temporally variable, influenced by chemical production, depletion through deposition and chemical titration processes and dispersion. To date, analysis of intra-urban variability of O3 concentrations, and the influence of local controls on production and depletion rates, has been limited due to the low spatial and/or temporal resolution of measurements. We demonstrate that measurements made using a carefully managed multi-sensor network of low-cost gas-sensitive semiconductor instruments are sufficiently precise to resolve subtle but significant variations in ozone concentration across a region. Ozone was measured at 12 sites in the isolated subtropical city of Auckland, New Zealand. Overall O3 concentrations in the Auckland region were low (annual mean: 19ppb) across all seasons, with a minimum in summer. Higher O3 concentrations (max. 57ppb) were observed when wind speeds were >5ms-1 and from the W/SW, and were associated with maritime air masses. Ozone formation in the Auckland region is low, which is attributed to a combination of the low O3 background concentrations, the negligible contribution of long-range transport and the effect of NOx titration. Intra-urban variability showed that the lowest O3 concentrations were measured at the residential sites, particularly at night and during rush hours. Ozone depletion from reaction with traffic-generated NO explains the rush-hour minima but did not fully account for the low night-time values. The results suggest that night-time depletion may result from other processes such as the reaction of ozone with nitrite on surfaces such as concrete, pointing towards the need for further studies concerning the rate and mechanism of dry deposition at night in urban areas.

Development of a microscale land use regression model for predicting NO 2 concentrations at a heavy trafficked suburban area in Auckland, NZ

Land use regression (LUR) analysis has become a key method to explain air pollutant concentrations at unmeasured sites at city or country scales, but little is known about the applicability of LUR at microscales. We present a microscale LUR model developed for a heavy trafficked section of road in Auckland, New Zealand. We also test the within-city transferability of LUR models developed at different spatial scales (local scale and city scale). Nitrogen dioxide (NO2) was measured during summer at 40 sites and a LUR model was developed based on standard criteria. The results showed that LUR models are able to capture the microscale variability with the model explaining 66% of the variability in NO2 concentrations. Predictor variables identified at this scale were street width, distance to major road, presence of awnings and number of bus stops, with the latter three also being important determinants at the local scale. This highlights the importance of street and building configurations for individual exposure at the street level. However, within-city transferability was limited with the number of bus stops being the only significant predictor variable at all spatial scales and locations tested, indicating the strong influence of diesel emissions related to bus traffic. These findings show that air quality monitoring is necessary at a high spatial density within cities in capturing small-scale variability in NO2 concentrations at the street level and assessing individual exposure to traffic related air pollutants.

Automated data scanning for dense networks of low-cost air quality instruments: Detection and differentiation of instrumental error and local to regional scale environmental abnormalities

Abstract Recent improvements in low-cost air quality instrumentation make deployment of dense networks of sensors possible. However, the shear volume of data from these networks means that traditional methods for data quality control and data analysis are no longer viable. We propose a real-time data scanning routine that detects local and regional variability within the data sets. This can be used to differentiate errors resulting from instrument malfunction or calibration drifts from natural (environmentally driven) regional changes in ambient concentrations. Our case study considered hourly-averaged ozone data from Texas and from two networks in Vancouver. We used 7 and 28 days of data for the algorithm initialisation with simulated and real instrumental changes. The algorithm output can be used as part of a limited resource maintenance schedule for sensor networks, and to improve understanding of air quality processes and their relation to environmental and public health data.