Downscaling Winter Daily PM2.5 Concentrations From Large-Scale Meteorological Dataset -a Case Study for a City in North China

Abstract

Winter air pollution in North China becomes a serious environmental problem in recent years, which has aroused a widespread concern. Estimating PM2.5 concentration is necessary for the government to take actions in leading times, or to reproduce the historical values. In this study, we attempt to construct statistical downscaling (SD) models based on large-scale meteorological variables, to estimate the PM2.5 in Jiaozuo, a city in the heavy-pollution area of North China. Predictors were screened from large-scale meteorological variables, by selecting the grid boxes with the values highly correlated to the PM2.5 in Jiaozuo. Correlation maps show that PM2.5 is usually related to comparatively low pressure and high relative humidity at the lower atmosphere and comparatively high pressure at the upper atmosphere, high air temperature in all the troposphere and weak winter monsoonal winds. After training the SD models with the winter samples during 2014-2017, the daily PM2.5 is simulated with correlation coefficients of 0.607 (mean) and 0.548 (maximum) to the observation, by the logarithmic transformed PM2.5 values. The SD models can roughly reflect the daily variation of PM2.5. Compared to the PM2.5 model based on the local meteorological data, larger correlations are obtained (0.57 versus 0.51-0.53, without logarithmic transformation). In the days of “APEC Blue Sky” and the “SCO Blue Sky” with emissions largely reduced by strict controls on industry and traffic in surrounding areas, low PM2.5 is indeed observed, however, the SD models without considering any change of emissions also produced low PM2.5 simulations, implying that the large-scale circulation plays a main role in the daily variation of air pollution. Similar effects were obtained for the traffic-control month and the Chinese Spring festival with massive fireworks burning. Basing on the large-scale reanalysis variables and assuming the same emission level as that in the model training period, the downscaled winter PM2.5 has a significant decreasing trend during 1979-2017.

Downscaling daily fine Particulate Matter (PM2.5) concentration maps for Ho Chi Minh City

“…The results showed a signi cant improvement in the resolution of the current PM 2.5 product, increasing it by more than 30 times from 0.1° (approximately 10 km) to 0.003°( approximately 300m) while maintaining high accuracy. Simultaneously, Liu et al, (2021) [5] applied statistical downscaling models in North China using hourly PM Their study downscaled short-range wind forecasts from ERA5 from a horizontal resolution of 31 km to mimic high-resolution (HRES) (deterministic) short-range forecasts at 9 km resolution. These studies demonstrate the effectiveness of downscaling methods in improving the resolution and accuracy of various environmental datasets.…”