Open Collections

Spatial and temporal analysis of ambient hourly PM10 in Vancouver

University of British Columbia

Fine particulates (PM10, PM2.5) have attracted many studies from a variety of disciplines. Because of the composition of higher proportion of various toxic metals and acidic sulfur species, fine particulate pollution is of specific concern to public health. They can contribute to the development of multiple adverse health effects. Many communities have put the particulate monitoring, evaluation and emission reduction as the prior issues in their air quality management plans. A current three-year Harvard-UBC study supported by the Environmental Protection Agency of the United States focuses on the development of new statistical methods and models for estimating population-level exposures to PM10 and PM2.5 in the greater Boston and Vancouver areas. That study will among other things require the spatial interpolation of Vancouver PM10 field, for example, down to the Census Tract levels. This thesis reports preliminary findings about the spatial and temporal variations of ambient hourly PM10 collected in the Greater Vancouver Regional District and nearby Fraser Valley. The spatial interpolator to be used will require independent response over time. To ensure such as assumption leads to the removal of temporal autocorrelation and whiten residuals for interpolation. The results from descriptive analysis of the raw data exhibit obvious consistency of monthly, weekly, week-day and day-hour patterns from site-to-site. In the study region motor vehicle emissions are the major source of PM10 pollution and the week-day and day-hour effects capture the traffic pattern. Poor air quality episodes at some stations are associated with anomalous weather conditions, such as coastal fogs, stagnant air conditions. Our space-time model for PM10 includes a deterministic trend and stochastic residual. We introduce a spatially homogeneous trend that comprise the above-mentioned site-to-site consistent temporal effects. For the detrended residuals, we find that the ARMA(2,1) or AR(3) model seems to capture the short-term autocorrelation structure satisfactorily. The coherent and slight spatial variation of the coefficients of ARMA(2,1) or AR(3) models suggests us to employ a single ARMA(2,1) or AR(3) model for all sites to eliminate the time-direction correlation and get a parsimonious description of our data. There remains some daily autocorrelation (lag-24 hours) in ARMA(2,1) or AR(3) residuals which can not be described by linear time series models. Nevertheless this correlation is not more than 0.06, and does not seem to cause concern in spatial interpolation of these residuals. It is suggested to incorporate the meteorological variables into models in future studies that might explain this daily serial correlation. It is noteworthy that there is little spatial correlation in station residuals after removing the trend and the same temporal autocorrelation structure represented by either a single ARMA(2,1) or AR(3) model. This low inter-site correlation may suggest that any given site in the study region picks up most of the variations of PM10 and station residuals contain little useful additional information for the construction of PM10 field. At the same time this analysis may suggest that the current network (with a scale of 7km) may not have sufficiently high spatial resolution to reflect the local important spatial gradients of PM10. For certain purposes, a finer monitoring network (down to the scale, say 2km) may be required to determine the fine spatial structure of PM10.

Thesis/Dissertation

application/pdf

2009-05-26

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/8208

Statistics

University of British Columbia

UBCV

DSpace