Turbulence-permitting air pollution simulation for the 1 Stuttgart metropolitan area 2

Air pollution is one of the major challenges in urban areas. It can have a major impact on human health 10 and society and is currently a subject of several litigations at European courts. Information on the level of air 11 pollution is based on near surface measurements, which are often irregularly distributed along the main traffic 12 roads and provide almost no information about the residential areas and office districts in the cities. To further 13 enhance the process understanding and give scientific support to decision makers, we developed a prototype for 14 an air quality forecasting system (AQFS) within the EU demonstration project “Open Forecast”. 15 For AQFS, the Weather Research and Forecasting model together with its coupled chemistry component (WRF16 Chem) is applied for the Stuttgart metropolitan area in Germany. Three model domains from 1.25 km down to a 17 turbulence permitting resolution of 50 m were used and a single layer urban canopy model was active in all 18 domains. As demonstration case study the 21 January 2019 was selected which was a heavy polluted day with 19 observed PM10 concentrations exceeding 50 μg m. 20 Our results show that the model is capable to reasonably simulate the diurnal cycle of surface fluxes and 2-m 21 temperatures as well as evolution of the stable and shallow boundary layer typically occurring in wintertime in 22 Stuttgart. The simulated fields of particulates with a diameter of less than 10 μm (PM10) and Nitrogen dioxide 23 (NO2) allow a clear statement about the most heavily polluted areas apart from the irregularly distributed 24 measurement sites. Together with information about the vertical distribution of PM10 and NO2 from the model, 25 AQFS will serve as a valuable tool for air quality forecast and has the potential of being applied to other cities 26 around the world. 27


28
Currently more than 50 % of the global population live in cities whereas the United Nations (UN) expect a further 29 increase by about 10 % in 2030 (UN, 2018). The UN also expect that in 2030 34% of the world population will 30 reside in cities with more than 500 000 inhabitants.

31
To protect human life, the World Health Organization (WHO) proposed maximum permittable pollution levels 32 (Maynard et al., 2017 and references therein). E.g. for particulate matter with particle diameters less than 10 µm 33 (PM10), the critical value is an annual mean concentration of 20 µg m -3 or a daily mean value of 50 µg m -3 (WHO, and chemistry models to predict air quality. Regional and global atmospheric models like the Weather Research      description of the selected case study. Section 4 shows the results including a discussion, sect. 5 summarizes our 112 work and gives an outlook on potential future enhancements of the AQFS prototype.

121
As seen from Fig. 1b, the Stuttgart metropolitan area is characterized by an elevation variation of more than 300 122 m. The lowest elevation is approx. 220 m in the basin while the highest elevation reaches up to 570 m. As the main 123 traffic roads are in the basin, especially during wintertime this often leads to a worsening of the air quality as the 124 surrounding prevents an air mass exchange due to the stationary temperature inversion.

125
For the WRF model system land cover and soil texture fields are not available at resolutions higher than 500m. As suggested by the WRF user guide, we applied the sub-grid turbulent stress option for momentum 150 (Kosovic, 1997) in domains two and three. The complete namelist settings are provided in the supplement.

151
As the finest resolution applied for the AQFS is 50 m, the more sophisticated Building Effect Parameterization

156
Atmospheric chemistry is parametrized by the Regional Acid Deposition Model 2nd generation (RADM2) model 157 (Stockwell et al., 1990). RADM2 features 63 chemical species including photolysis and more than reactions.

167
Model output is available in 5 min intervals for the innermost model domain.

168
Our single day case study on the turbulence permitting (TP) scale is designed to serve as a test bed to set up an air 169 quality forecasting system prototype for the Stuttgart metropolitan area. For process studies, the model chain itself

170
can be applied to other areas over the globe as long as 1) detailed land cover and soil texture data are available,

175
The meteorological initial and boundary conditions were provided by the operational ECMWF integrated

196
The third emission data set (BW-EMISS) deployed in our study was obtained from the Baden-Württemberg State

197
Institute for the Environment (LUBW). This data set contains annual mean emissions from different sectors 198 following the GNFR classification and is currently available only until 2014 and has a horizontal resolution of 500 199 m. Unfortunately, more recent quality-controlled data sets were not available when our study was performed. It is

209
Due to its much higher horizontal resolution, the BW-EMISS data set (Fig. 3b) shows much more detailed 210 structures for the NO2 emissions which are mainly caused by road traffic.

211
In addition, the following adjustments have been performed: 1) NOx emissions from forest grid cells have been

215
The WRF-Chem model only ingests one emission data set per species, hence emissions from the different GNFR 216 categories have been accumulated to a single emission data set before performing the simulation. Figure 4 217 summarizes all necessary steps and the complete data and workflow of the AQFS prototype.

233
For our study, we selected 21 January 2019. This day was characterized as "fine dust alarm" situation (Stuttgart  A sufficient criterion is a higher PM10 concentration following (1). If (1) is not fulfilled, then (2) and (3) together 244 with either (4) and/or (5) must be fulfilled. If only (4) or (5) is fulfilled, then (6) must be considered. For our case 245 study, the criteria 1-5 were fulfilled.

253
This day was a typical winter weather situation. Central Europe was located at the east flank of a blocking high 254 pressure system located over the East Atlantic together with moderate to low horizontal geopotential gradients and 255 resulting weak winds at 500 hPa in southwestern Germany (Fig. 6a).

256
Near surface temperatures are below freezing level, between 1000 and 850 hPa very light easterly winds 257 characterize the flow, and a dry layer is present around 925 hPa (Fig. 6b). Above 850 hPa, the wind direction The inversion between the two air masses inhibits vertical mixing leading to higher concentrations of aerosols in 260 the lowest few hundred meters above ground (AGL) and preventing air mass exchange aloft. This inversion is 261 further enhanced by the special orography of Stuttgart city (see later Fig. 15).     For the airport station, the model stays too warm with a positive bias of almost 2 K between 05 and 09 UTC.

296
During the further course of the day, the bias reduces to 1 K at noon while after sunset it turns into a negative bias 297 of 1 K.

298
A possible reason for the larger differences at the airport and IPM before (after) sun rise (sun set) is the occurrence 299 of low stratus or fog. At the beginning of the simulation, cloud coverage were reported by 5-7 octas (broken 300 clouds) over Schnarrenberg and the airport at approx. 500 m AGL (not shown) while after 04 UTC the low level 301 clouds started to diminish at Schnarrenberg first leading to a strong cooling until the early morning which is seen 302 as a temperature drop in the observations shown in Fig. 9. This temperature drop at Schnarrenberg and IPM is also

307
Although no measurements of sensible heat and ground heat fluxes are available, diurnal cycles of the fluxes at 308 the three locations IPM, Schnarrenberg, and airport were investigated. Figure 10 shows the simulated surface 309 sensible heat and ground heat flux at the three different meteorological measurement sites.

310
The sensible heat flux (Fig. 10a) shows a typical diurnal cycle with fluxes around zero before (after) sunrise  intensity residential (category 31) with an urban fraction of 0.5 and the UCM is applied here, energy is mainly 325 stored in the urban canopy layer instead of being transferred into the soil.

326
As this day was characterized by a shallow PBL and a temperature inversion, it is worth to investigate the PBL 327 evolution during the day. Figure 11 shows time-height cross sections of potential temperature at IPM (top) and    (Fig. 12b), when turbulence is fully evolved (Fig. 11), the simulated NO2 concentrations are less than 30 µg m -3 350 on average apparently due to vertical mixing of NO2 (see next section). In the evening (Fig. 12c)     During the morning traffic (Fig. 13a), PM10 accumulates in the Stuttgart basin as this is an area with heavy traffic 364 during the morning and an atmospheric inversion is present (Fig. 7). Interestingly, the high NO2 concentrations 365 along the motorway (Fig. 12a) do not lead to very high PM10 concentrations potentially due to chemical transitions 366 caused by low temperatures.

367
During daytime when turbulence is fully evolved, the concentration of PM10 decreases to less than 20 µg m -3 due 368 to vertical mixing and horizontal transport (see next section). After sunset (Fig. 13c)

Vertical distribution of NO2 and PM10
373 In addition to the horizontal distribution of near surface NO2 and PM10, TP simulations with a fine vertical 374 resolution also enable qualitative insights into the vertical distribution of pollutants. Figure 14 shows West-East 375 cross sections at Neckartor (Fig. 1b) during the morning rush hour and at noon time. Neckartor is one of the 376 heaviest traffic locations in the Stuttgart city area.

377
The NO2 concentration during the morning rush hour shows an accumulation along the motorway (red arrow in 378 Fig. 14a) and in the region around Neckartor (white arrow in Fig. 14a) with concentrations exceeding 100 µg m -3 379 as the atmospheric inversion prevents exchange with the layers above (Fig. 7). The vertical extent of concentrations 380 higher than 30 µg m -3 is about 200 m AGL with a strong reduction above.

381
During noon time (Fig. 14b), the simulated NO2 concentration is much lower (less than 30 µg m -3 ) as turbulence 382 leads to a stronger mixing throughout the boundary layer up to 400 m AGL which is in accordance with the 383 simulated potential temperature timeseries shown in Fig. 11.

430
The simulation also shows that pollutants can be advected from the motorway A81 towards Stuttgart, depending 431 on the wind situation potentially leading to an increase of the NO2 and PM10 concentrations in the Stuttgart basin.

432
As can be seen from Figs

441
II. Improving the chemical background e.g. by applying higher resolution products from the CAMS

442
European Air quality project (Marécal et al., 2015). This will help to have a more detailed structure of

454
Although air quality modeling on the TP scale is a very challenging and computationally expensive task, we are 455 convinced that the AQFS will have a great potential to further improve process understanding and will certainly 456 help politicians to make decisions on a more scientifically valid basis.

457
Code and data availability to belong to an ECMWF member state to benefit from these data sets. Due to restrictions on the input data sets for 462 this simulation, the data can only be made available upon special request from the corresponding author.

469
The authors declare that they have no conflict of interest.

471
This study has been performed within the EU-funded project Open Forecast (Action number 2017-DE-IA-0170).

472
We acknowledge ECMWF for providing analysis data from the operational IFS and CAMS reanalysis. The