The TROPOMI instrument aboard Sentinel-5P is a relatively new, high-resolution source of information about atmosphere composition. One of the primary atmospheric trace gases that we can observe is nitrogen dioxide. Thanks to TROPOMI capabilities (high resolution and short revisit time), one can describe regional and seasonal $N{O}_2$ concentration patterns. Thus far, such patterns have been analysed by either ground measurements (which have been limited to specific locations and only to the near-surface troposphere layer) or numerical models. This paper compares the TROPOMI and GEM-AQ derived vertical column densities (VCD) over Poland, focusing on large point sources. Although well established in atmospheric science, the GEM-AQ simulations are always based on emission data, which in the case of the energy sector were reported by stack operators. In addition, we checked how cloudy conditions influence TROPOMI results. Finally, we tried to link the $N{O}_2$ column number densities with surface concentration using boundary layer height as an additional explanatory variable. Our results showed a general underestimation of $N{O}_2$ tropospheric column number density by the GEM-AQ model (compared to the TROPOMI). However, for the locations of the most significant point sources, we noticed a systematic overestimation by the GEM-AQ model (excluding spring and summer months when TROPOMI presents larger $N{O}_2$ VCDs than GEM-AQ). For the winter months, we have found TROPOMI $N{O}_2$ VCD results highly dependent on the choice of qa_value threshold.

High $P{M}_{10}$ concentrations are still a significant problem in many parts of the world. In many countries, including Poland, 50 $\mathsf{\mu }$g/m${}^3$ is the permissible threshold for a daily average $P{M}_{10}$ concentration. The number of people affected by this threshold’s exceedance is challenging to estimate and requires high-resolution concentration maps. This paper presents an application of random forests for downscaling regional model air quality results. As policymakers and other end users are eager to receive detailed-resolution $P{M}_{10}$ concentration maps, we propose a technique that utilizes the results of a regional CTM (GEM-AQ, with 2.5 km resolution) and a local Gaussian plume model. As a result, we receive a detailed, 250 m resolution $P{M}_{10}$ distribution, which represents the complex emission pattern in a foothill area in southern Poland. The random forest results are highly consistent with the GEM-AQ and observed concentrations. We also discuss different strategies of training random forest on data using additional features and selecting target variables.

TRPOMI instrument aboard Sentinel-5P is a relatively new, high-resolution source of information about atmosphere composition. One of the primary atmospheric trace gases that we can observe through it is nitrogen dioxide. By now, we were using the chemical weather model (GEM-AQ) as a mean for estimating nitrogen dioxide concentration on a regional scale. Although well established in atmospheric science, the GEM-AQ simulations were always based on emission data, which in the case of the energy sector were reported by stack owners. In this paper, we attempted to compare the TROPOMI and GEM-AQ derived VCDs over Poland with a particular focus on large point emitters. We also checked how cloudy conditions influence TROPOMI results. Finally, we tried to link the NO2 column number densities with surface concentration using boundary layer height as an additional explanatory variable

High PM10 concentrations are still a significant problem in many parts of the world. In many countries, including Poland, 50μg/m3 is the permissible threshold for a daily averaged PM10 concentration. The number of people affected by this threshold’s exceedance is challenging to estimate and requires high-resolution concentration maps. This paper presents an application of random forests for downscaling regional model air quality results. As policymakers and other end users are eager to receive a detailed resolution PM10 concentration maps, we propose a technique which utilizes the results of regional CTM (GEM-AQ, with 2.5km resolution) and local Gaussian plume model. As a result, we receive a detailed, 250-meter resolution PM10 distribution, which resembles the complex emission pattern in a foothill area in southern Poland. The random forest results are highly consistent with the GEM-AQ and observed concentration. We also discuss different strategies of data training random forest - using additional features and selecting target variables.