Jensen, S.S., Ketzel, M., Im, U., Løfstrøm, P., Poulsen, M.B., Monies, C., Ellermann, T. (2016): Validering af luftkvalitetsmodellen OML-Highway. Aarhus Universitet, DCE – Nationalt Center for Miljø og Energi, 49 s. - Videnskabelig rapport fra DCE - Nationalt Center for Miljø og Energi nr. 240. http://dce2.au.dk/pub/SR240.pdf
The Danish Road Directorate conducted a mapping of air quality along all state roads in Denmark with the OML-Highway model in 2015. DCE carried out the mapping, and the model results represented air quality in 2012 (Jensen et al., 2015). OML-Highway is an air quality model, which is specifically designed to calculate the air quality along roads in open terrain.
In order to implement the national mapping a number of simplifications was made. These simplifications addressed how detailed roads and traffic are described, as well as how the background concentrations and meteorology are handled, and how nitrogen dioxide (NO2) is calculated.
The aim of this report is to make an assessment of the impact of these simplifications for the quality of the national mapping. A secondary objective is to investigate why modelled concentrations in the national mapping do not decrease as fast as expected with distance from the road; such expectations are based on previous measurement campaigns. To support this analysis, a new measurement campaign at the motorway close to Svogerslev in 2016 was conducted One of the purposes was to assess how air pollution decreases farther away from the highway. Furthermore, a comparison of OML-Highway calculations performed with and without the simplifications in mapping method was conducted. Additional, a comparison of OML-Highway calculations and the measurements from the new measurement campaign at Svogerslev in 2016 was carried out, as well as new OML-Highway calculations compared with the previous measurement campaigns done at the motorway at Køge Bugt in 2003 and the motorway at Taastrup in 2008.
In the following the mapping method is assessed first for the Svogerslev measurement campaign in 2016, as this is the most detailed analysis. Afterwards, the results of the analyses of the campaigns for Køge Bugt in 2003 and Taastrup in 2008.
The simplifications of the mapping method are briefly outlined below.
The road network is based on vejman.dk, which includes all state roads in Denmark. This road network is single digitized e.g. a highway is represented by a single centre line. Road width is defined according to this centre line (excl. emergency lane). OML-Highway allows for a more detailed description of the exact location of road lanes etc., which are not used in mapping method.
The background concentrations are calculated with a regional air quality model, the Danish Eulerian Hemispheric Model (DEHM) and a local scale background model, the Urban Background Model (UBM) for all 1 x 1 km2 grid cells in Denmark along state roads. A simplifying assumption for background concentrations is that a calculation point along a road is represented by the nearest centre point for background concentrations in a 1 km x 1 km grid cell, which should introduce only a marginal difference. A further simplifying assumption (a side effect of a simple way to avoid double counting) is that emissions of municipal roads located in the same 1 km x 1 km grid cell as a state road are not included when calculating background concentrations. This may allow for a minor underestimation. However, roads municipal roads are however included in the calculation of background concentrations for grid cells with no state. State roads are not included in the calculation of background concentrations but contribute to the concentration for the calculations points along the roads.
OML-Highway calculates NO2 concentrations hour by hour, by taking account of photochemical reactions in the atmosphere including NO (nitrogen oxide), NO2 (nitrogen dioxide), and O3 (ozone). There is not a linear relationship between NOx and NO2. It would be very resource demanding to carry out calculations in this way in the context of mapping along the entire State road network. Therefore, instead of calculations, where the background concentration varies for every hour, the mapping method is based on annual mean background concentrations of NO2 on a spatial resolution of 1 km x 1 km. Such concentrations were previously calculated in another project. NO2 at calculation points is calculated based on a method developed by Düring et al. (2011). It is a simplified chemistry model, which is based on annual mean NOx concentrations for the calculation point and the associated background concentrations, as well as, NO2 and ozone for background concentration at the site as well as information on the percentage of directly emitted NO2 emissions of NOx emissions from traffic.
Hence, background concentrations calculated with DEHM/UBM is an integrated part of the methodology for NO2 estimation.
Denmark was divided into 12 regions, representing 5 different meteorological classes based on the average wind speed in order to avoid generation of a meteorological file for every single 1 km x 1 km grid cell along the state road network. The meteorological model MM5 vas used in the mapping, but DCE has subsequently changed model to WRF (Weather Research and Forcasting). A comparison of MM5 and WRF shows no large differences, although the WRF is slightly better than MM5.
The calculations are carried out with a modified version of OML-Highway compared to the version that was used in the mapping. In the modified version contributions from emission sources from state roads further away than 20 km are not included. Furthermore, the boundary-layer height in the meteorological input data has a minimum height of 150 m, as WRF may estimate unrealistically low values in some cases. Both of these factors contribute to lower concentrations than in the mapping, in particular, further away from the road. These modifications reduced concentrations by about 3-5 µg/m3. The previously used version of OML-Highway in the mapping has therefore in particular overestimated concentrations further away from highways, and independently of the simplifications made in order to implement the mapping. It is the modified version that is used for comparisons with the measurements in all three campaigns (Svogerslev 2016, Køge Bugt 2003 and Taastrup 2008).
A new measurement campaign with passive NO2 samplers was carried out at the Holbæk motorway at Svogerslev located west of Roskilde. The measurement period lasted 6weeks from 10.05.2016 2 pm to 21.06.2016 11 am, that is week 20 to week 25. The Japanese produced Advantec filter badges for passive absorption of NO2 were used.
To test the accuracy of the passive NO2 samplers simultaneous measurements at the permanent measuring stations on Jagtvej in Copenhagen (Street station) and at Risø (rural station) under the National Air Quality Monitoring Programme were carried out. These comparisons show high correlation between the two types of measurements, but the passive measurements are slightly lower than the active. The relationship between the measurements is used to calibrate the values from the passive NO2 samplers.
As expected the measurements show maximum concentrations close to the motorway and rapid decline in concentrations with distance to the highway. At a distance of around 400 m the concentrations have almost reached background levels.
OML-Highway calculations show generally higher concentrations than measurements except for the measuring point closest to the motorway. The overestimation is relatively greater further away from the motorway.
In the following, a summary of the impacts of the simplifications of the mapping method is given based on calculations and measurements of the Svogerslev campaign.
A road is single digitized, and the digitalization of the centre line is generally of very high quality, which has been demonstrated by comparing selected samples with aerial photos. A displacement of the centre line of e.g. 1 m will have an impact on the calculated concentrations very close to the road, but would be negligible farther away from the road, and the effect is therefore not quantified.
The mapping method with its simplified NO2 estimation generally leads to lower concentrations than hourly based OML-Highway calculations, and it has less span between lowest (far from motorway) and highest (close to motorway) calculated values. This is a major reason why in the national mapping there is unexpectedly small difference between modelled air quality levels, respectively, close to and further away from the state roads, and that the calculated concentrations in the mapping do not decline as quickly as expected.
Compared to measurements from Svogerslev the mapping method underestimates the highest concentrations close to the motorway and overestimates farther away from the road. This also illustrates that the calculated concentrations farther away from the highway do not decline as quickly as expected.
The simplified NO2 estimation has contributed to the underestimation of the calculated concentrations in general, while the version of OML-Highway applied in the mapping has resulted in a tendency to overestimate the concentrations, especially farther from the road. Thus, these two factors have partly offset each other; however, they have led to the net result that the concentrations especially farther from the road have been overestimated in the mapping.
There is also uncertainty in the level of the annual mean concentrations calculated with DEHM/UBM, but the absolute differences are small compared to the measurements, and cannot explain why the mapping showed relatively high concentrations farther away from the road.
In order to assess the impact on concentrations of using wind speed classes in the mapping method, a sensitivity analysis was conducted for the Svogerslev campaign with different assumptions for meteorological data. The difference between the use of the lower and upper bound of wind speeds in the specific wind speed class showed a difference of up to 10%, and concentrations at all distances of the motorway were affected. Application of wind speed classes doesn’t explain why the mapping has relatively high concentrations farther away from the road.
The OML-Highway underestimates NO2 concentrations close to the motorway, while all other points farther away are overestimated.
Comparison between OML-Highway calculations conducted with, respectively, WRF modelled meteorological data and measured Sonic data showed minor differences in concentrations, which is an indication that the modelled meteorological data from the WRF are not very different from measured meteorological data for longer time averaging periods.
The decline in concentrations with distance from the road was evaluated for different distances up to 10 km from the highway. Calculated concentrations decline quickly the first 1000 m, but the background level is not reached until around 8 km. If measurements of the Svogerslev campaign are used as guidance, this seems to be way too far, since the background level at Svogerslev was almost reached at a distance of about 400 m. This indicates that OML-Highway calculates too high concentrations farther away from the road.
There has not previously been carried out validation studies for the measurement campaign at Taastrup, since the measuring campaign was not designed for this. Its objective was to assess vehicle emissions, especially particles for high speed conditions. The conditions at the measurement site are not suited for a validation study, since they are very complex as there is an embankment and trees along the highway.
Sonic meteorological measurements were carried out during the campaign. These are used to compare the meteorological model calculations with WRF and Sonic measurements of hourly values for the two important parameters wind speed (as Ustar, u*) and wind direction. Ustar is a dispersion model parameter, which is used to describe the wind-induced turbulence. Ustar is proportional to the wind speed – that is, the higher the wind speed, the higher the Ustar. The analysis showed good agreement between the model and measurements for these two important parameters.
There is room for improvement to the OML-Highway model, since the model overestimates concentrations, especially at some distance away from the highway. In April 2017 DCE funded a project to address this issue. The project is expected to be completed at the end of 2017.
