Development of land-use regression models to estimate particle mass and number concentrations in Taichung, Taiwan

Highlights

• Both particle mass and number concentrations are measured over one year.

• The model explained variance (R²) is 0.53 for PM_2.5 and 0.51 for PM₁.

• The magnitude of R² ranged from 0.31 to 0.50 for particle number concentrations.

• The built model has the highest R² for particle number concentrations at 0.5–1 μm.

• Models with moderate performance are developed to estimate particle concentrations.

Abstract

Land-use regression (LUR) models have been used to estimate particle mass concentration (PMC), but few studies apply it to predict particle number concentration (PNC) at different sizes. This study aimed to determine both PMC and PNC throughout one year to establish predictive models in Taichung, Taiwan. The annual averages of PM₁₀, PM_2.5, and PM₁ were 71 ± 46 μg/m³, 44 ± 35 μg/m³, and 32 ± 28 μg/m³, respectively. The PNC at size ranges of <0.5 μm, 0.5–1 μm, 1–2.5 μm, 2.5–10 μm, and ≥10 μm were 715098 ± 664879 counts/L, 29053 ± 30615 counts/L, 1009 ± 659 counts/L, 647 ± 347 counts/L, and 3 ± 3 counts/L, respectively. The model-explained variance (R²) values of PM₁₀, PM_2.5, and PM₁ were 0.42, 0.53, and 0.51, respectively. The magnitude of the R² values ranged from 0.31 to 0.50 for the PNC with the highest R² between 0.5 and 1 μm. The differences between the model R² and the leave-one-out cross-validation R² ranged from 4% to 8% for PMC and from 3% to 10% for PNC. This study developed LUR models with moderate performance to estimate PMC and PNC at different sizes in an Asian metropolis. The built LUR models may be improved by combining with other open data to increase the predictive capacity.

Introduction

Particulate matter (PM) emitted from traffic vehicles, industrial operations, power generation, fuel combustion, and other sources has caused ambient air pollution locally and globally. Many epidemiological studies have demonstrated the association between prolonged exposure to PM and adverse health effects in the public, including increased mortality (Beelen et al., 2014; Schwartz et al., 2015), reduced lung function (Hwang et al., 2015), elevated blood pressure (Chang et al., 2015; Chen et al., 2015), diabetes mellitus (Chen et al., 2013), atherosclerosis (Su et al., 2015), cerebrovascular events (Stafoggia et al., 2014), myocardial infarction (Bai et al., 2019), and lung cancer (Raaschou-Nielsen et al., 2013). The International Agency for Research on Cancer (IARC) has classified PM with an aerodynamic diameter less than 2.5 μm (PM_2.5) in outdoor air pollution as carcinogenic to humans (Group 1), based on sufficient evidence of carcinogenicity in humans and experimental animals and strong support from mechanistic studies (IARC, 2015; Loomis et al., 2013). In 2010, exposure to ambient PM_2.5 was estimated to contribute to 3.2 million premature deaths worldwide, due largely to cardiovascular disease, and 223 000 deaths (approximately 15%) from lung cancer. According to these estimates, more than half of the lung cancer deaths attributable to ambient PM_2.5 have been reported in China and other East Asian countries (Lim et al., 2012).

Land-use regression (LUR) has been widely used as an approach to estimate the ambient PM annual average for exposure assessment (Brauer et al., 2003; Eeftens et al., 2012; Hoek et al., 2011; Lee et al., 2015; Moore et al., 2007; Wu et al., 2014). The established LUR models are applied to investigate the association between PM exposure and adverse health effects in the general population (Beelen et al., 2014; Chen et al., 2015; Raaschou-Nielsen et al., 2013; Stafoggia et al., 2014; Su et al., 2015). Unlike other mechanistic models based on comprehensive data collection and physical phenomenon, LUR models empirically rely on statistical methods to combine air pollution measurements at multiple locations with surrounding geospatial features (predictor variables) (Hoek et al., 2008). LUR models also reveal the feasibility of explaining spatial concentration variations in particle mass concentration (PMC) and particle number concentration (PNC) using surrounding land-use features. Most of these LUR models were developed to estimate PMC, such as PM with an aerodynamic diameter less than 10 μm (PM₁₀), coarse particles (PM_2.5-10), and PM_2.5 (Brauer et al., 2003; Eeftens et al., 2012; Hoek et al., 2011; Lee et al., 2015; Moore et al., 2007; Wu et al., 2014). Some studies have established LUR models for PNC, but they are limited to ultrafine particles (aerodynamic diameter less than 0.1 μm) (Eeftens et al., 2016; Ghassoun et al., 2015; Hoek et al., 2011; Patton et al., 2015; Wolf et al., 2017) due to measured data from the available instrument. To the best of our knowledge, no studies have applied LUR to predict PNC at different sizes.

Because PM may undergo complex physicochemical changes related to mass- and number-based concentrations upon internalization by cells leading to a biological impact and toxicity (Warheit and Brown, 2019), the spatial distribution and sensitivity to primary particle emission are noticeably different between PNC and PMC (Chen et al., 2018), and the adverse health effects caused by PM exposure were found to be different between PMC and PNC (Aguilera et al., 2016; Hennig et al., 2018). It is important to elucidate and estimate PMC and PNC at different sizes for future epidemiological studies. By using the portable and newly developed device to measure PMC and PNC at different sizes simultaneously, this study can provide the exposure estimates to build the LUR models for PMC and PNC. This study aimed to develop LUR models of PMC and PNC at different sizes by using 24-h average measurements of ambient levels encompassing one year for inhabitants living in Central Taiwan.