Role of export industries on ozone pollution and its precursors in China

This study seeks to estimate how global supply chain relocates emissions of tropospheric ozone precursors and its impacts in shaping ozone formation. Here we show that goods produced in China for foreign markets lead to an increase of domestic non-methane volatile organic compounds (NMVOCs) emissions by 3.5 million tons in 2013; about 13% of the national total or, equivalent to half of emissions from European Union. Production for export increases concentration of NMVOCs (including some carcinogenic species) and peak ozone levels by 20–30% and 6–15% respectively, in the coastal areas. It contributes to an estimated 16,889 (3,839–30,663, 95% CI) premature deaths annually combining the effects of NMVOCs and ozone, but could be reduced by nearly 40% by closing the technology gap between China and EU. Export demand also alters the emission ratios between NMVOCs and nitrogen oxides and hence the ozone chemistry in the east and south coast.


Supplementary Note 1 Case setting
To study the impact of export industries and the pathways to mitigate its footprint, this study sets up a few cases in the validated modeling platform. A base case and the other 3 cases were constructed. The differences between different bases were emission inputs.
For the base case, air pollutant emissions of NOx, NMVOCs and CO for the year of 2013 were adopted, which represented the 'true' emissions (emissions in reality under the best knowledge) in 2013. Case 1 to 3 used reconstructed emissions to either study the impact of export or the effectiveness of the proposed manners.
Case 1 was set up to study the impact of export-driven emissions on BTEX and O 3 concentration. In Case 1, NOx, NMVOCs, and CO emissions relevant to export demands were excluded. By comparing the modelling results from base case and Case 1, the contribution of export-driven emissions on O 3 and BTEX concentration can be revealed.
Case 2 was set up to test the effectiveness of the cleaner production manners (in line with the proven and affordable technologies in EU) in the export capacities. By comparing the sectoral emission intensities in China and EU, we estimated that an 1,165 kt of NMVOCs can be reduced from the export-relevant industrial capacities. Given that NOx have been reduced aggressively and the persistent growth of NMVOCs from industries, we focused on the reduction of NMVOCs in this study and did not extend the discussion to the reductions of other precursors. Therefore, an 1,165 kt of NMVOCs were excluded from the emissions inputs of Case 2, while the inputs of NOx and CO remained the same as that of base case.
Please note that the reductions of NMVOCs were done by sectors. In other words, sectoral specific reduction was made instead of an even reduction of all sectors. By comparing the results of Case 2 and base case, we can estimate the effectiveness of the cleaner production manners in the export capacities.
Case 3 was developed to understand the wider impact of cleaner production manners. Case 2 assumed that cleaner production manners were only applied to export production lines.
However, it might not be true in reality. If a shoe-making factory is decided to upgrade its technology and management, for example, it is more plausibly done for the whole production line rather than only for the shoes for export. We estimated that the proposed manners in Case 2 would lead to a reduction of 4,437 kt of NMVOCs if they targeted the whole production capacities. Therefore, Case 3 was set up by reducing the NMVOCs from industries by 4,437 kt, while the inputs of NOx and CO remained the same as that of base case. Results of Case 3 were compared with base case to fathom the effectiveness of an industry-wide NMVOCs reduction effort.
The study domain of this work is mainland China, with a spatial resolution of 27km × 27km. The air quality modeling platform coupled the Weather Research and Forecast (WRF) model [1], SparseMatrix Operator Kernel Emissions (SMOKE) model [2], and CMAQ model [3]. The Weather Research and Forecast (WRF) model v3.9 was used to provide meteorological data. The CMAQ v5.0.2 with the CB-05 gas-phase chemical mechanism was used to simulate the ambient O 3 mixing ratios under different precursor emissions scenario. The SMOKE provided model-ready emission data by allocating the annual emissions at province level into hourly interval and grid cell. Species allocations were also involved. Take Case 1 as an example, the annual bulk emission inventory for the base year of 2013 is first developed. Given that the input-output table is also in an annual basis, the bulk emission inventory is used as an input for the environmentally-extended input-output (EEIO) analysis to produce the consumption-based emission inventory. The consumption-based emission inventory reveals that how many emissions in each source category are associated with the demand of export. By excluding the emissions driven by export, a new bulk emission inventory is generated for Case 1, which is the case for us to study the impact of export. The new bulk emission inventory is processed with the temporal and spatial surrogates and emission processing systems to have the monthly, daily and even hourly emission inputs for the air quality modelling system and other analysis. The model-ready meteorological and emission data was then fed into air quality model. The model was spin-up for 3 days in each month to eliminate the impact of initial conditions. Detailed model configurations of CMAQ and WRF are shown in Table 2.
Ground-level O 3 measurements were used to validate the modeling platform. The locations of ambient ozone monitoring sites are shown in Supplementary Fig.4. In China, ambient O 3 mixing ratios were not regularly measured nation-wide until 2013. The records of ambient O 3 from China's national air quality monitoring network were adopted for the reference year 2013. Specifically, the performance of modeling platform in July and October 2013 were evaluated. These two months represented two typical O 3 seasons in China. Normalized mean bias (NMB), normalized mean error (NME), and correlation coefficient (R) were used as indicators of model performance. According to recommended benchmarks for photochemical model performance statistics, the NMB for the 1 hour average or maximum daily 8 hour average ozone should be no larger than 15%, and the R should be higher than 0.50 [4]. The model performances of this work (See Supplementary Table 3) were within the above suggested range. The NME of this study was similar to those of previous studies in China [5]- [7]. For example, the NME for the 1 hour average O 3 over the eastern China in July was around 58.8~62.7% [7]. The modeling system can reproduce the O 3 mixing ratio reliably. In case study, this study mainly refers to the maximum 8 hour average since it was reproduced well in the model and it is more relevant to the health impact.