Impacts of sectoral emissions in China and the implications: air quality, public health, crop production, and economic costs

This study adapted the chemical transport model, Community Multi-scale Air Quality (CMAQ) Model v4.7.1 (Byun and Schere 2006) to simulate the air quality in China in 2010 and the contributions from different emission sectors. In each case a 7 day spin-up simulation was conducted. Hourly meteorological fields were extracted from the outputs of the weather research and forecasting model v3.7 (Skamarock and Klemp 2008). Note that the original land cover information from USGS was updated by the land use datasets from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC), which has a better representation of land information at a 1 km resolution (Liu et al 2014). Boundary and initial conditions were derived from outputs of a global chemical transport model, GEOS-Chem v8.3.2 (Wang et al 2004, Bey et al 2001). Photolysis rates were calculated by the JPROC module, based on radiative forcing and chemical mechanism scheme. The model performance was evaluated by comparing model results with available observations. The model evaluation is provided in section 1.4 in the supporting information (SI) available at stacks.iop.org/ERL/13/084008/mmedia.

The baseline emission inventory for China in 2010 was compiled by anthropogenic air pollutant emissions from HTAP-V2 dataset (Janssens-Maenhout et al 2012), biogenic emissions from MEGAN-MACC dataset (Sindelarova et al 2014), and fire emissions from FINN v1.5 dataset (Wiedinmyer et al 2014). The anthropogenic emission dataset contained annual emissions from agricultural fertilization, road transport, aviation, shipping, energy generation, industry production, and domestic living. The fire emissions dataset recorded the emissions from both biomass burning and waste burning with the assistance of satellite observations. The dust emissions were estimated by the in-built windblown dust emission module of the CMAQ model. The summation of these emission sources was processed to construct a three-dimensional inventory, covering the whole of China except for the southern part of the Spratly Islands, see figure 1. The model was configured to have a 27 km gridded space with 26 vertical layers. CB05 and AE5 were chosen for gas-phase chemistry and aerosol chemistry mechanism, respectively. The speciation was also conducted in accordance with the chemical mechanism framework of air quality model. The information of emissions in our compiled model-ready emission inventory is provided in table S14 in the SI.

Zoom In Zoom Out Reset image size

The model performance was comprehensively evaluated by satellite-retrieved products and available ground measurements. We employed widely used statistical indicators to quantify the model performance. The statistical indicators included mean bias, correlation coefficient, and index of agreement. In general, the evaluation results show that the model results were robust in this study. The detailed discussion is provided in section 1.4 in the SI. The evaluation results were taken into account in our uncertainty quantifications as discussed later.

We calculated the baseline emission impact due to all emission sectors based on a simulation that included all the emission sources. Six sector scenarios whereby the emissions of each sector were removed in sequence. We note that a direct sum-up of all sector impacts may not be equivalent to the results of baseline simulation due to the nonlinear effect of emission to concentration. The relative contribution to PM_2.5 and O₃ concentration of a sector was calculated by dividing the impacts of the sector by the total impacts summed up of all the sectors. The relative contribution was then multiplied by the baseline emission impact to estimate the relative impact of the respective sector. On the other hand, the contributions to human health impact of a sector was estimated by multiplying the baseline health impact by the corresponding sector's contribution to baseline concentration. This method was widely used in air quality studies, such as Lelieveld et al (2015). The uncertainty of nonlinearity was further addressed and discussed in the uncertainty qualification section.

Different health endpoints were assessed in response to the sectoral contributions to PM_2.5 and O₃ concentration, which have been reported to have a direct link with cardiopulmonary disease and lung cancer (Pope et al 2009). Here we adapted China-specific concentration-response functions (CRF) (Gu and Yim 2016) to estimate the impact of PM_2.5 on mortalities due to chronic obstructive pulmonary disease, ischemic heart disease, lung cancer, and stroke. Note that the toxicity of various PM_2.5 species was assumed to be identical due to the lack of solid corresponding information in current epidemiological research. The mortality numbers due to the exposure of O₃ were quantified by the log-linear form of CRF, specifically adapted for China. The basic form of this CRF for health endpoints (E) was shown as follows:

$$\begin{equation} E = \sum\nolimits_k {({\rm RR}_k - 1)/{\rm RR}_k \cdot P_k } \cdot f_k \end{equation} \tag{ 1 }$$

where RR for PM_2.5 was resolved by:

$$\begin{equation} {\rm RR}_k = 1 + \alpha \cdot \left\{ {1 - \exp \left[ { - \beta (X_k - X_0 )^\delta } \right]} \right\}. \end{equation} \tag{ 2 }$$

RR for O₃ was resolved by:

$$\begin{equation} {\rm RR}_k = \exp \left[ {\gamma \cdot \left( {X_k - X_0 } \right)} \right] \end{equation} \tag{ 3 }$$

where k refers to the index of the domain grid; P refers to the population size based on various population datasets (Fu et al 2014, Columbia University 2005, 2013, Population Census Office 2012) for different area; and f refers to the baseline incident rate above 30 years of age based on the statistic results of Mainland China, Hong Kong, and Taiwan in 2010 (MOH 2011, DGBAS 2011, Census and Statistics Department of the Hong Kong Special Administrative Region (CSDHK 2015). Note that historical records of the regional incident rates in China were classified into urban and rural categories. The urban-rural division was therefore based on the land use types, derived from the Data Centre for RESDC (Liu et al 2014) and the Centre for International Earth and Science Information Network (Center for International Earth Science Information Network (CIESIN)/Columbia University 2013).

RR refers to relative risk. X_k and X₀ are the pollutant concentration (µg m⁻³) in the target scenario and the threshold value, below which no additional risk was assumed to exist, respectively. The parameters α, β, and δ were fitted by nonlinear regression in our previous efforts (Gu and Yim 2016). The parameter γ was fitted by meta-regression based on the previous epidemiological study for China. The detailed information is provided in sections 1.5.1 and 1.5.2 in the SI.

The crop production loss (CPL) of four major crop types: rice, wheat, maize, and soybeans due to O₃ exposure were estimated by the product of annual yield (γ) and yield loss ratio (YLR):

$$\begin{equation} {\rm CPL}_k = \frac{{{\rm YLR}_k }}{{1 - {\rm YLR}_k }} \cdot \gamma _k \end{equation} \tag{ 4 }$$

where annual yield and growing calendar data for four types of crops in China were retrieved from Monfreda et al (2008). YLR was calculated based on the exposure-relative yield (ERY) relationships according to O₃ exposure and specific types of crops (Wang and Mauzerall 2004). Crop exposure to tropospheric O₃ was quantified by five exposure indices, including M7, M12, SUM06, AOT40, W126. The corresponding calculation method was provided in section 1.5.3 in the SI. These indices were initially proposed and mostly applied in the national crop loss assessment network (NCLAN) experimental studies in the US. ERY equations for rice, wheat, maize, and soybeans are provided in table S13 in the SI.

The social cost of the air pollution-attributable public health impact and crop production loss, due to one-year air pollutants emissions in China were estimated through different impact pathways of sectoral emissions.

The cost of public health damages was assessed in consideration of hospital expenditure (HE) and absence work loss (AWL). Hospital expenditure of a single morbidity or mortality case includes outpatient fee and hospitalization fee, which were estimated by a factor of local per capita gross domestic product (GDP). Such factor and residue were derived from linear regression between the statistics of local per capita GDP and local expenditure per mortality and morbidity (Xie et al 2016):

$$\begin{equation} {\rm HE} = \sum\nolimits_{f,k} {E_k \cdot \left( {\alpha _f \cdot {\rm PCGDP}_k + \beta _f } \right)} \end{equation} \tag{ 5 }$$

where PCGDP refers to per capita GDP; f denotes the categories of health expenditure; k refers to the index of the domain grid; E refers to two health endpoints: mortality and morbidity. α and β are the regression coefficients extracted from Xie et al (2016). The spatial distribution of GDP in 2010 was derived from the data center for RESDC (Liu et al 2014).

On the other hand, AWL refers to the incomplete production due to the absence of working. Such loss for morbidities is the fraction of GDP proportional to the average per capita work loss days (WLD) out of annual per capita total working days (WD: 260 days year⁻¹). The rest of work loss comes from the accumulative work loss years for mortalities aged from 15–65 years old. The summation of work loss due to morbidities and mortalities contributes to the total work loss, which is expressed as:

$$\begin{equation} \begin{array}{l@{}l@{}l@{}} {{\rm AWL}} \hfill & = \hfill & {\sum\nolimits_k {\frac{{{\rm WLD}_k }}{{{\rm WD}}} \cdot {\rm GDP}_k + \sum\nolimits_{k,y \leqslant 65} {} } } \hfill \\ {} \hfill & {} \hfill & {E_k \cdot {\rm PCGDP} \cdot \left( {65 - y} \right) \cdot {\rm WD}} \hfill \\ \end{array} \end{equation} \tag{ 6 }$$

where WLD was concluded as linear with annual average PM_2.5 concentration, at the magnitude of 0.0207 (95% CI: 0.0176–0.0238) (Hurley et al 2005); y is mortality age estimated via China health statistics (MOH 2011).

The cost of crop production loss for four major crops was quantified as the product of the loss amount and crop value in the corresponding year. In this study, crop value by unit was calculated as the total crop value over the annual crop productions in 2010 extracted from the Food and Agriculture Organization of United Nations (FAO 2015).

The Monte-Carlo approach was applied to assess the uncertainties of health outcomes and crop production loss. We calculated a central estimation with 95% confidence intervals (95% CI) of overall uncertainty for each impact, which was realized 20000 times randomly assuming a triangular probability distribution T (low, nominal, high). Specifically, model uncertainty was adjusted by the normalized mean bias of model evaluation result: T (0.99, 1.14, 1.31) for PM_2.5 and T (0.76, 0.98, 1.24) for O₃; Spatial varied factors for addressing nonlinearities between sum-up of all sectoral impacts and baseline emission impacts were included for PM_2.5 and O₃, with the confidential range T (0.37, 2.24, 2.46) and T (2.26, 4.16, 15.95), respectively. Uncertainties of health impact estimation were originated to the coefficient uncertainties of concentration-response functions, which were resolved by the variety of nonlinear fitting. The detailed information is described in the section 1.5 in the SI. The results together with work loss days: T (0.0176, 0.0207, 0.0238) constructed the uncertainty in the economic cost of health impacts; in terms of crop production loss, uncertainties were attributable to the different O₃ exposure estimation methods; the uncertainty of climate impacts was estimated by the coefficients uncertainty of response function. Each uncertainty factor was modeled through 20 000 realizations, each of which was picked up randomly to calculate impacts, and forming an approximate probability distribution. The ultimate impact uncertainty was estimated based on the median and 95% confidential interval of impact distribution. We note that the chemical mechanism of PM_2.5 formation has yet to be fully understood, especially in the winter haze periods (Cheng et al 2016). This may induce some uncertainties in our PM_2.5 calculation.

Figure 2 depicts the spatial distributions of annual surface PM_2.5 and O₃ concentration contributed by the various sectors in China. The industrial sector was the largest contributor to the annual surface PM_2.5 in China, accounting for, on average, 32.0% of annual surface PM_2.5 in China. The largest impact occurred in northern China, Sichuan and some metropolises in southern China. The second contributor was the residential and commercial sector, accounting for 24.9% of annual surface PM_2.5 in China. Such impact dominantly came from the space heating and in-house cooking, which were particularly impacted in the winter season. The agriculture and power generation sectors exerted a similar level of contribution to annual surface PM_2.5 in China, ranking the third and the fourth largest contributors, respectively. When combined, the two sectors (the agriculture and the power generation sectors) accounted for 28.4% of the annual surface PM_2.5 in China. We note that the PM_2.5 related to the agriculture sector was dominated by ammonium nitrate (see table S15 in SI). Ammonia (NH₃) is predominantly emitted by cultivation activities due to the volatilization of fertilizers. Fu et al (2017) reported that excessive NH₃ emissions could enhance the formation of PM, especially secondary inorganic PM. This supports our findings, as the agriculture sector had a relatively high level of secondary PM_2.5 (shown in table S15 in SI). The ground transportation sector was estimated to contribute 7.3% of the annual surface PM_2.5 in China. The remaining 7.4% of the annual surface PM_2.5 was attributable to other sectors.

Zoom In Zoom Out Reset image size

The spatial patterns of sectoral contributions to the annual surface of O₃ and PM_2.5 were different. The industrial sector remained the largest contributor of annual surface O₃ in China, accounting for 32.4% of the annual surface O₃ concentration. Despite the significant sectoral impact on the annual surface O₃, a reduction in O₃ due to the industrial emissions occurred in major cities. The reduced production of O₃ arose from urban O₃ titration (Seinfeld and Pandis 2012), which occurred when excessive NO_x was emitted in an environment with a relatively high ratio of nitrogen oxides (NO_x) and volatile organic compounds. The second and third largest contributors were the power generation sector and the resident and commercial sector, which respectively contributed 28.7% and 9.5% of the annual surface O₃. These three sectors together accounted for more than 70.0% of the annual surface O₃ in China. The ground transportation sector contributed marginally (3.1%), whereas the agriculture sector caused a slight reduction in O₃ (−2.0%), as mentioned (see figure 2). Agricultural emissions denote to the excessive NH₃ emission due to the usage of fertilizer. Previous studies have reported the impacts of NH₃ on O₃. Makar et al (2009) and Wang et al (2011) reported that NH₃ could affect aqueous chemistry by providing a weak base condition, thus enhancing the O₃-induced cloud SO₂ oxidation rate because O₃ exhibits a higher oxidation rate under a weak base condition. Therefore, the concentration of O₃ was reduced slightly while the concentration of SO₄²⁻ increased, as shown in our results. Other sectors accounted for the remaining 28.3% of the annual surface O₃ concentration in China.

We calculated the public health impacts of PM_2.5 and O₃ pollution due to emissions of the six studied sectors, which are summarized in figures 3 and 4. The results show that the emissions from the six sectors collectively caused an estimated 1 143 000 (95% CI: 168 000–1 796 000) premature mortalities, accounting for 12.0% of all-cause mortalities in 2010 in China. 76.3% of air pollution-related mortalities were due to PM_2.5 pollution, while the remaining was attributable to O₃ pollution. In general, the spatial distribution pattern of PM_2.5-attributable mortalities and O₃- attributable mortalities are similar, see sub-figure A of figures 3 and 4. The largest health impact attributed to PM_2.5 and O₃ pollution all appeared in the urban areas in eastern China, such as the Yangtze River Delta region and the Pearl River Delta region. Rural areas in Hebei, Henan, and Sichuan provinces (see their locations in figure S8) were also identified to receive noticeable health impact from PM_2.5 and O₃ pollution. By contrast, marginal impacts were found in western China due to its complex topography and smaller emission amount relative to the eastern part (Gu and Yim 2016). In terms of the sectoral contributions to pollution-related mortality, as shown in sub-figure B of figures 3 and 4, industrial sector contributed to the most of prominent PM_2.5-attributable premature mortalities, while other sectors contributed to the most of prominent O₃-attributable premature mortalities. Such difference may be due to the different nature of PM_2.5 and O₃ formation, as mentioned in the section of air quality impact.

Provincial health impact pattern due to PM_2.5 and O₃ pollution were also varied. The industrial and the residential commercial sector were estimated as the dominant sectors to the PM_2.5-attributable mortalities for almost all provinces in China, except for some developed provincial divisions with the ground transportation as the second-largest contributor, including Shanghai, Zhejiang, Taiwan, Hong Kong, and Macau. Shandong, Henan, and Hebei were recognized to receive the top-three PM_2.5-attributable mortalities in China, because the high exposure of PM_2.5 pollution. As for O₃-attributable mortality, the major contributing sectors were typically the industrial, power generation, and other sectors. Note that the industrial and other sectors exerted prominent negative impacts on ozone-attributable mortalities in some provinces, like Henan, Jiangsu, Zhejiang, and Chongqing. Such reductions probably due to the emissions of NH₃ that enhance the oxidation rate of O₃ and urban O₃ titration, as described in the section of air quality impact. Shandong, Henan, and Sichuan were recognized to receive the top-three O₃-attributable mortalities in China.

Our results show that the industrial sector and the residential and commercial sector were identified as major contributors to total health impacts, accounting for 36.2% (∼414 000 (95% CI: 61 000–640 000)) and 20.0% (∼229 000 (95% CI: 34 000–354 000)) of the total values, respectively. In terms of spatial distribution, the premature mortalities caused by these two sectors showed more spatial heterogeneity, compared with the other four sectors. A possible explanation for this is the higher ambient air pollution in urban areas resulting from the aggregation of the residential and commercial sector and the industrial sector near urban areas. Overall, 43.8% of the total health impacts apportioned to the remaining sectors are outlined as follows: power generation (∼183 000 (95% CI: 27 000–288 000)), ground transportation (∼73 000 (95% CI: 11 000–115 000)), agriculture (∼129 000 (95% CI: 19 000–203 000)), and other sectors (∼115 000 (95% CI: 16 000–196 000)).

O₃ pollution affects crop production in various ways, including damaging foliage, decreasing photosynthetic efficiency, reducing nutrient and water uptake, and increasing grain abortion (Wargent and Jordan 2013). Figure 5 and table 1 show the spatial distribution and summary of production losses of four major crop type due to sectoral emissions in China, respectively. Our calculations show that the total crop production loss majorly occurred in northeastern, eastern, and southwestern China. Industrial sector was the dominant sector of this impact for almost all of China, while the power generation and other sectors showed their dominance in the northern China and some sporadic regions in the southwest part of China, respectively, see the lower panel of figure 5. For different crop type, rice production loss happened in the southern part of China, dominantly affected by industrial emissions. Soybean production loss occurred in the northeastern and southwestern part of China, majorly contributed by industrial and other sectors' emissions. Wheat production loss located in the central and western part of China, mainly due to the emissions from industrial and power generation sectors. Maize production loss mainly took place in the northeastern and western part of China, which was attributable to the industrial and power generation emissions. Based on our results the industrial sector (41% of the total impact) and the power generation sector (23% of the total impact) were the two largest contributors. With the exception of agriculture, all the sectors had a negative effect on crop production. Previous section has explained the fact of positive impact of agriculture sector on O₃ concentration. The positive impact of NH₃ emissions on O₃ pollution compensated for the total O₃-induced crop production loss due to sectoral emissions. In summary, the emissions from the six major sectors caused the loss of approximately 20 035 (95% CI: 6776–32 166) Gg crops in China. The industrial sector caused the most damage to crop production (41%), and the corresponding estimates are outlined as follows: wheat (approximately 3087 crops (95% CI: 348–5795)) Gg, rice (approximately 2589 (95% CI: 762–4416)) Gg, and maize (approximately 2056 (95% CI: 1167–2845)) Gg.

Table 1. Total crop production loss (Gg) caused by sectoral emissions in China (95% confidence intervals) of each sector to the impacts. The crop types include maize, rice, wheat, and soybeans. The results were rounded up to the nearest integer. The values in the parentheses are the 2.5th and 97.5th percentiles.

Table 2 shows the economic cost of air pollution-related premature mortality, and crop production loss due to sectoral emissions in China. Our estimation indicates that the sectoral air pollutants emissions in China led to an economic cost of ∼267 (95% CI: 180–360) billion yuan (0.66% annual GDP) in 2010. These findings suggest an urgent need to reinforce efforts to reduce anthropogenic emissions in China.

The industrial sector and residential and commercial sector were the two largest contributors, together accounting for 57% of the total economic cost. More effort should be thus put on controlling the emissions of these two sectors. The breakdowns show that the relative contributions of power generation emissions to air pollution were lower than those of the residential and commercial sector, while power generation was one of the dominant contributors of crop production loss, larger than residential and commercial sector. When combining, the economic cost of power generation sector and residential and commercial sector were similar. Here, we provide the detailed cost table for each breakdown, which is anticipated to serve as a useful reference for enacting cobeneficial emission control strategy with different emphasizing aspects to maximize its effectiveness.

Except for the aforementioned three sectors (the industrial, power generation, residential and commercial sectors), the agriculture and ground transportation sectors also show the comparable impacts on public health and crop production. These results highlight a higher potential to formulate cobenefit emission control policies for these two sectors. Even though agricultural emissions exerted a positive impact (negative number in table 2) on crop production due to the associated O₃ reduction, the agricultural sector shows a larger impact on public health than on crop production loss. More stringent emission control policies on agricultural sector may thus have an important implication on air quality due to its NH₃ emission that is a key precursor of secondary inorganic particulate matter. Emissions of other sectors accounted for merely 6% of the air pollution-related health impact, while their contribution to crop production loss can be as high as 25%, which reflect their considerable impact on O₃ and the corresponding spread spatial pattern.

Table 2. Economic costs (billion RMB, 2010 currency rate) of various impact types due to sectoral emissions in China in 2010 and the relative contributions (RC) (%) of each sector to various impacts. The results are at a 95% confidence interval and rounded up to the nearest integer. The values in the parentheses are the 2.5th and 97.5th percentiles.