Temporal changes in field calibration relationships for Aeroqual S500 O3 and NO2 sensor-based monitors
Temporal changes in field calibration relationships for Aeroqual S500 O3 and NO2 sensor-based monitors
复制标题
DOI：
10.1016/j.snb.2018.07.087
复制
发表时间：
2018-11-10
影响因子：
8.4
通讯作者：
Beverland, Iain J.
中科院分区：
文献类型：
--
作者：
Masey, Nicola;Gillespie, Jonathan;Beverland, Iain J.
基于传感器的监测器越来越多地用于测量空气污染物浓度，但需要在实地条件下进行校准。我们进行了间歇性的比较（6次超过6个月的时间）之间的臭氧和二氧化氮浓度测量的气敏半导体（O-3）和电化学（NO2）传感器（每两个）和参考分析仪在英国自动城市和农村网络。将每个部署期分为相等的（n = 48 x 1小时）训练和测试数据集，分别推导和测试校准方程。我们观察到Aeroqual O-3和参比O-3浓度之间存在显著的二元线性关系，Aeroqual NO2与参比NO2和Aeroqual O-3浓度之间存在显著的多元线性关系。监测器响应随时间的变化（包括O-3传感器输出的明显基线漂移，以及2个Aeroqual NO2传感器之间的差异）导致浓度估计相对不准确（参见。参考浓度），来自第一个训练期推导出的校准方程，并应用于随后的测试部署（例如，对于两个监测器之一，所有测试期组合的数据集，NO2 RMSE = 47.2 μ g m（-3）（n = 286））。对）。通过将重复的间歇训练数据组合成单个校准数据集（对于上述相同的测试数据集，NO2 RMSE = 8.5 μ g m（-3）），实现了估计浓度的准确性的实质性改进。建议采用后一种现场校准方法。
Sensor-based monitors are increasingly used to measure air pollutant concentrations, but require calibration under field conditions. We made intermittent comparisons (6 times over a 6-month period) between ozone and nitrogen dioxide concentrations measured by Aeroqual gas-sensitive semiconductor (O-3) and electrochemical (NO2) sensors (two of each) and reference analysers in the UK Automatic Urban and Rural Network. Each deployment period was split into equal (n = 48 x 1-hour) training and test datasets, to derive and test calibration equations respectively. We observed significant bivariate linear relationships between Aeroqual O-3 and Reference O-3 concentrations, and significant multiple linear relationships between Aeroqual NO2 and both Reference NO2 and Aeroqual O-3 concentrations. Changes in monitor responses over time (including apparent baseline drift in O-3 sensor output, and discrepancies between the 2 Aeroqual NO2 sensors) resulted in relatively inaccurate concentrations estimates (cf. reference concentrations) from calibration equations derived in the first training period and applied to subsequent test deployments (e.g. NO2 RMSE = 47.2 mu g m(-3) (n = 286) for a dataset of all test periods combined, for one of the two monitor pairs). Substantial improvements in accuracy of estimated concentrations were achieved by combination of repeated intermittent training data into a single calibration dataset (NO2 RMSE = 8.5 mu g m(-3) for same test dataset described above). This latter approach to field calibration is recommended.