Toward targeted observations of the meteorological initial state for improving the PM_2.5forecast of a heavy haze event that occurred in the Beijing–Tianjin–Hebei region

Abstract

Abstract. An advanced approach of conditional non-linear optimal perturbation (CNOP) was adopted to identify the sensitive area for targeted observations of meteorological fields associated with PM2.5 concentration forecasts of a heavy haze event that occurred in the Beijing–Tianjin–Hebei (BTH) region, China, from 30 November to 4 December 2017. The results show that a few specific regions in the southern and northwestern directions close to the BTH region represent the sensitive areas. Numerically, when predetermined artificial observing arrays (i.e. possible “targeted observations”) in the sensitive areas were assimilated, the forecast errors of PM2.5 during the accumulation and dissipation processes were aggressively reduced; specifically, these assimilations, compared with those in other areas that have been thought of as being important for the PM2.5 forecasts in the BTH region in previous studies, exhibited a more obvious decrease in the forecast errors of PM2.5. Physically, the reason why these possible targeted observations can significantly improve the forecasting skill of PM2.5 was interpreted by comparing relevant meteorological fields before and after assimilation. Therefore, we conclude that preferentially deploying additional observations in the sensitive areas identified by the CNOP approach can greatly improve the forecasting skill of PM2.5, which provides, beyond all doubt, theoretical guidance for practical field observations of meteorological fields associated with PM2.5 forecasts.

An approach to refining the ground meteorological observation stations for improving PM2.5 forecasts in Beijing-Tianjin-Hebei region

“…In our study, since we focused on the uncertainties of meteorological initial condition associated with the PM2.5 forecast, following Yang et al, (2022), the state vector 𝒙 consists of zonal and meridional wind (U and V, respectively), temperature (T), water vapor mixing ratio (Q) and pressure (P) components, 185 which are considered as important meteorological fields on PM2.5 forecasts over the BTH region (see the review paper of Chen et al, 2020). The perturbations 𝛿𝒙 0 are superimposed on the ground meteorological field 𝒙 0 of interest.…”

“…Therefore, when one computes the CNOP-type initial perturbation superimposed on the better simulation initialized by ERA5 (i.e. "truth run"), it can be regarded as an approximation to the most sensitive initial error that disturbs the meteorology forecast of sensitive area of the meteorological field (see next subsection) and preferentially assimilating additional observations in the sensitive area of the control forecast will make the updated forecast (hereafter as "assimilation run") approach to the truth run (see Yang et al, 2022). Such idea is a kind of observation system simulation experiment (OSSE, Masutani et al, 2020).…”

“…In terms of the effect of meteorological initial conditions, lots of studies have shown that small uncertainties in meteorological initial fields will result in large uncertainties in PM2.5 forecasts (Gilliam et al, 2015;Bei et al, 2017). Recently, it has been recognized that a bad meteorological initial 45 condition may even affect the forecast of the accumulation or dissipation processes of the PM2.5 event and could result in a false alarm of the heavy haze event (Yang et al, 2022). Therefore, an accurate meteorological initial condition is also crucial for the regional PM2.5 forecasts, besides the initial chemical concentrations and emission.…”

“…Among the various meteorological observation sources, the observations from the ground meteorological stations are often assimilated to predict the meteorology fields (Hu et al, 2019;Devers et al, 2020;Yao et al, 2021). Yang et al (2022) studied the uncertainties of meteorological initial fields to PM2.5 forecasts The dominant meteorological stations to be identified, as mentioned above, would provide the meteorological observations that will have the largest impact on the PM2.5 forecasts of the concerned region. This idea belongs to the new observational strategy of "targeted observation", that is assimilating additional observations at the target time t1 in some key areas (i.e.…”

“…The CNOP has been adopted to identify the sensitive areas in the studies of tropical cyclones, El Nino-Southern 105 Oscillation events, oceanic mesoscale eddies, and marine environments and has successfully improve the forecasting skills (see the review of Duan et al, 2022). Especially, Yang et al (2022) applied the CNOP to determine the sensitive areas for targeted observation of a heavy hazy event which was not warned in time by the monitoring center and demonstrated that assimilating additional observations in such sensitive area leads to successful forecast of the PM2.5 concentrations with much higher skill. Then in 110 this study, we would use the CNOP to recognize the dominant ground meteorological stations applicable for PM2.5 forecasts by investigating the sensitive areas of eight winter heavy hazy events over the BTH region during years of 2016-2018, consequently providing an idea to refine the current ground meteorological stations for improving the PM2.5 forecasts in the BTH.…”