Toward targeted observations of the meteorological initial states for improving the PM2.5 forecasts in the Beijing-Tianjin-Hebei region

The Beijing-Tianjin-Hebei region (BTH region) has suffered from successive heavy haze events during the past several decades. The PM2.5 has been dominated as one of the main air pollutants during the haze events, which poses a higher health risk and even a higher death rate. An accurate prediction of the PM2.5 concentration is critical for providing early warnings to residents and helping governments take timely actions. However, due to the complex chemical and physical processes, and uncertainties of meteorological and emission conditions, the numerical chemical transport models (CTMs) are not able to provide presice forecast results, especially on heavy haze days. Here, to reduce the uncertainties caused by meteorological initial conditions and improve the air quality forecasting skills in the CTMs, an advanced approach of conditional nonlinear optimal perturbation (CNOP) was adopted to identify the sensitive area for targeted observations of meteorological fields associated with PM2.5 concentration forecasts. By considering one heavy haze episode that occurred in 2017 in the Beijing-Tianjin-Hebei (BTH) region, we first explored the effect of possible targeted meteorological observations on improving the PM2.5 forecasting skills during both the accumulation and dissipation periods of the haze event. Since assimilating the additional observations on the sensitive areas is verified to be effective on improving the PM2.5 forecasts, we then investigate how to refine the ground meteorological observation network to greatly improve the PM2.5 concentration forecasts by identifying sensitive areas for targeted observations. Our study will provide a new strategy to reduce the uncertainties of air quality forecasts, and finally help improve the air quality forecast skills.