An objective in mapping air quality attributes such as concentrations of airborne particles (particulate matter – PM) is to determine those areas which can be considered as hotspots and determine factors that contribute to their formation. Classical kriging has been applied extensively in mapping air quality variables, with most applications focusing on average pollutant concentrations. Generalized linear spatial process models are being applied as an alternative to classical kriging. In this paper we compare ordinary and regression kriging models to the Poisson log-linear spatial model (Diggle et al. 1998, Diggle et al. 2007) with and without covariate information in mapping annual average exceedance frequencies of the South African PM10 air quality standard of 120 µg/m3 (RSA Govt. Gazette 2009, 2012). We use daily PM10 data from 36 air quality monitoring sites in the Highveld (Gauteng and western Mpumalanga provinces) for the 48 months period from September 2009 to August 2012. Higher concentrations are observed in high density residential areas, with high proportion of informal and mixed types of dwellings. Therefore, significance of household energy use, number of households and settlement type as explanatory variables in mapping the yearly exceedance rates are explored.
Reference:
Khuluse, S. 2013. Mapping the annual exceedance frequencies of the PM10 air quality standard - Comparing kriging to a generalized linear spatial model. In: South African Statistical Association Conference, Polokwane, South Africa, 4-8 November 2013
Khuluse, S. (2013). Mapping the annual exceedance frequencies of the PM10 air quality standard - Comparing kriging to a generalized linear spatial model. SASA Conference 2014. http://hdl.handle.net/10204/7780
Khuluse, S. "Mapping the annual exceedance frequencies of the PM10 air quality standard - Comparing kriging to a generalized linear spatial model." (2013): http://hdl.handle.net/10204/7780
Khuluse S, Mapping the annual exceedance frequencies of the PM10 air quality standard - Comparing kriging to a generalized linear spatial model; SASA Conference 2014; 2013. http://hdl.handle.net/10204/7780 .