Modelling intra- and inter-day variability of NO2 concentrations in Portugal
Title: Modelling intra- and inter-day variability of NO2 concentrations in Portugal
Speaker: Dr. Raquel Meneze, Department of Mathematics and Applications,Minho Universit, Azurém, Portugal
Date: Thursday, 28th April 2016
Time: 4pm
Location: Room E0.01, Science East
Abstract:
The air quality is the term usually coined to refer to the degree of pollution in the air that we breathe. The air quality index is calculated based on five different pollutants, including nitrogen dioxide (NO2). This pollutant is toxic by inhalation and there is evidence that long-term exposure to NO2 at high concentrations has adverse health effects, namely in respiratory and cardiovascular systems. Furthermore, NO2 concentration levels closely follow vehicle emissions, in many situations, thus providing a reasonable marker exposure to traffic.
The goal of this study is to characterize the spatial and temporal evolution of NO2 concentration levels, taking into account that environmental data often incorporate distinct recurring patterns in time, imposed by social habits. We aim at capturing the cyclic nature of these environmental indicators, identifying the intra- and inter-day variability. Simultaneously, we aim at modelling the temporal and spatial correlation inherent to this type of data, following a similar approach to that described in Menezes et al. (2015). This study focus on NO2 hourly data collected in Portugal from October 1 to December 31, 2014. These months correspond to the highest mean NO2 concentrations along the year.
An initial exploratory study shows that there are two main seasonal effects in the data set. NO2 levels show two daily peaks, in the morning (8:00 A.M.) and afternoon (6:00 P.M.) which coincide with rush–hour traffic, with the second peak being more significant than the first. Moreover, the mean NO2 concentrations are much lower on weekends (particularly on Sunday) than on weekdays, displaying, also, smaller variations on weekends, which reflect reduced levels of vehicular emissions on non-working days. This analysis indicates that variables such as the type of site (background, industrial or traffic), the type of environment (urban, suburban or rural), and the day of the week (work day or weekend) are possible explanatory variables. Furthermore, the analysis of the correlation between meteorological parameters, as temperature, wind speed and relative humidity and NO2 levels identified significant negative associations among them. Thus, we first model the trend of NO2 data using a Generalized Linear Model (GLM) with the above mentioned six explanatory variables. After characterizing the trend function, geostatistical approaches are applied to the resulting residuals with the aim of characterizing the space-time variability and deriving the predicted values through the kriging tools.
This methodology allows to make use of kriging techniques for prediction, by considering the multiple cycles underlying the environmental data, and reconstruction of the space-time pattern followed by the NO2 concentrations as well as the associated prediction error. Moreover, knowledge of the space-time dependence can be used to complement the current design sampling, where there are districts without monitoring stations or with many missing values.
Series: Statistics
Social Media Links