Ensemble prediction of air quality using the WRF/CMAQ model system for health effect studies in China

Jianlin Hu, Xun Li, Lin Huang, Qi Ying, Qiang Zhang, Bin Zhao, Shuxiao Wang, Hongliang ZhangAtmospheric 2017 · DOI

Abstract

Accurate exposure estimates are required for health effect analyses of severe air pollution in China. Chemical transport models (CTMs) are widely used to provide spatial distribution, chemical composition, particle size fractions, and source origins of air pollutants. The accuracy of air quality predictions in China is greatly affected by the uncertainties of emission inventories. The Community Multiscale Air Quality (CMAQ) model with meteorological inputs from the Weather Research and Forecasting (WRF) model were used in this study to simulate air pollutants in China in 2013. Four simulations were conducted with four different anthropogenic emission inventories: MEIC, SOE, EDGAR, and REAS2. Model performance was evaluated against observation data from 422 sites in 60 cities. Ensemble predictions were calculated by linearly combining results from different inventories to minimize squared errors. The ensemble concentrations show improved agreement with observations, with annual $PM_{2.5}$ MFB improving to -0.11 and $O_3$ MNB to 0.03. This study demonstrates that ensemble predictions can significantly improve the accuracy of predicted air pollutant distributions for health effect studies.

Introduction

To study how air pollution affects human health, researchers need accurate maps of pollutant exposure. Chemical Transport Models (CTMs) are the standard tool for generating these maps, but their accuracy is heavily dependent on emission inventories—databases that estimate pollutant output from various sectors. In China, these inventories are notoriously uncertain, with different databases often disagreeing on the amount and location of emissions. This paper presents the first ensemble air quality modeling effort for China, leveraging multiple inventories to create a prediction superior to any single model run.

Methods and Datasets

The researchers utilized the Weather Research and Forecasting (WRF) model for meteorology and the Community Multiscale Air Quality (CMAQ) model for atmospheric chemistry.

Map of China showing the eight study regions and the 422 observation sites used for model evaluation.

The study employed a Linear Optimization ensemble approach:

  1. Four Simulations: The WRF/CMAQ system was run four times for 2013, varying only the anthropogenic emission inventory (MEIC, SOE, EDGAR, and REAS2).
  2. Observational Grounding: Hourly measurements of $PM_{2.5}$, $PM_{10}$, $O_3$, $CO$, $SO_2$, and $NO_2$ from 422 sites in 60 Chinese cities (provided by CNEMC) served as the benchmark.
  3. Weighting: Optimal weighting factors ($w_m$) for each inventory were determined to minimize the squared error between the ensemble output and real-world observations.

Key Insights: The Variability of Inventories

A critical finding was that no single inventory performed best across all regions or pollutants. For instance, while MEIC and SOE were generally more accurate for $SO_2$, REAS2 showed better performance for $CO$.

Performance metrics (MNB and MNE) for $O_3$, $CO$, $NO_2$, and $SO_2$ across the four emission inventories and different regions, illustrating the regional and chemical variability in inventory accuracy.
  • The Power of the Ensemble: By "borrowing" strengths from different inventories, the ensemble smoothed out regional biases and consistently outperformed individual models.
  • Temporal Sensitivity: Optimal weights changed depending on the time scale (daily vs. annual), suggesting some inventories are better at capturing total volume while others excel at capturing the timing of pollution events.
  • Systemic Underprediction: All four inventories underpredicted $CO$, $NO_2$, and $SO_2$, indicating a field-wide need for improved bottom-up emission reporting.

Key Results

The ensemble approach led to marked improvements in statistical performance metrics:

  • $PM_{2.5}$ Accuracy: The Mean Fractional Bias (MFB) for annual $PM_{2.5}$ improved to -0.11, compared to a range of -0.25 to -0.16 for individual models.
  • Ozone ($O_3$) Accuracy: For the daily maximum 1-hour $O_3$, the Mean Normalized Bias (MNB) dropped to just 0.03, significantly better than the individual model range of 0.06 to 0.19.
  • Regional Performance: The ensemble was particularly effective in complex regions like the Sichuan Basin and the North China Plain.
  • Exposure Estimates: The study estimated a population-weighted annual $PM_{2.5}$ concentration of 59.5 $\mu g/m^3$ for China in 2013, providing a vital metric for calculating pollution-related mortality.

Limitations and Future Work

While successful, the study was limited by missing observation data for early 2013 and the use of fixed weights for large regions. Future research could explore spatially varying, pixel-by-pixel weights to further refine local accuracy. The authors have made these ensemble results publicly available to support future health effect studies.

Summary updated

Abstract

Accurate exposure estimates are required for health effect analyses of severe air pollution in China. Chemical transport models (CTMs) are widely used to provide spatial distribution, chemical composition, particle size fractions, and source origins of air pollutants. The accuracy of air quality predictions in China is greatly affected by the uncertainties of emission inventories. The Community Multiscale Air Quality (CMAQ) model with meteorological inputs from the Weather Research and Forecasting (WRF) model were used in this study to simulate air pollutants in China in 2013. Four simulations were conducted with four different anthropogenic emission inventories, including the Multi-resolution Emission Inventory for China (MEIC), the Emission Inventory for China by School of Environment at Tsinghua University (SOE), the Emissions Database for Global Atmospheric Research (EDGAR), and the Regional Emission inventory in Asia version 2 (REAS2). Model performance of each simulation was evaluated against available observation data from 422 sites in 60 cities across China. Model predictions of O3 and PM2.5 generally meet the model performance criteria, but performance differences exist in different regions, for different pollutants, and among inventories. Ensemble predictions were calculated by linearly combining the results from different inventories to minimize the sum of the squared errors between the ensemble results and the observations in all cities. The ensemble concentrations show improved agreement with observations in most cities. The mean fractional bias (MFB) and mean fractional errors (MFEs) of the ensemble annual PM2.5 in the 60 cities are −0.11 and 0.24, respectively, which are better than the MFB (−0.25 to −0.16) and MFE (0.26–0.31) of individual simulations. The ensemble annual daily maximum 1 h O3 (O3-1h) concentrations are also improved, with mean normalized bias (MNB) of 0.03 and mean normalized errors (MNE) of 0.14, compared to MNB of 0.06–0.19 and MNE of 0.16–0.22 of the individual predictions. The ensemble predictions agree better with observations with daily, monthly, and annual averaging times in all regions of China for both PM2.5 and O3-1h. The study demonstrates that ensemble predictions from combining predictions from individual emission inventories can improve the accuracy of predicted temporal and spatial distributions of air pollutants. This study is the first ensemble model study in China using multiple emission inventories, and the results are publicly available for future health effect studies.
BibTeX
@inproceedings{hu2017ensemble,
  author = {Jianlin Hu and Xun Li and Lin Huang and Qi Ying and Qiang Zhang and Bin Zhao and Shuxiao Wang and Hongliang Zhang},
  title = {{Ensemble prediction of air quality using the WRF/CMAQ model system for health effect studies in China}},
  booktitle = {Atmospheric},
  year = {2017},
  doi = {10.5194/acp-17-13103-2017},
}
Figure