1. OpenFOAM for Air Quality Ernst Meijer and Ivo Kalkman First Dutch OpenFOAM Seminar Delft, 4 november 2010

2. 4 November 2010First Dutch OpenFOAM Seminar2 Outline Introduction to air quality Application of CFD to air quality problems Example case study OpenFoam versus Fluent OpenFoam 2D test case for urban wind profiles Discussion and conclusions

3. 4 November 2010First Dutch OpenFOAM Seminar3 Air Quality Issues

4. 4 November 2010First Dutch OpenFOAM Seminar4 European guidelines for air quality SpeciesExceedence level Nitrogendioxide (NO 2 ) Annual average40 μg/m³ Hourly average max. 18 time/yr > 200 μg/m³ Particulate Matter (PM10) Annual average40 μg/m³ Diurnal averagemax. 35 times/yr > 50 μg/m³ Primary concern are health effects However allowed PM10 levels are still ~ 10 4 times too high In Netherlands air quality is connected to new building plans

5. 4 November 2010First Dutch OpenFOAM Seminar5 Local Air Quality and Climate Field experiments Wind tunnel Models Meeting European guidelines (NO2, PM10, PM2.5) Evaluation of measures Health assessment; black carbon aerosol Urban Heat Island Integrated assessment on environmental impacts (noise, heat, safety, …)

6. 4 November 2010First Dutch OpenFOAM Seminar6 Application of CFD to AQ Open field: gaussianUrban: wind tunnel Gaussian approach not suitable for urban environment Windtunnel has ‘real turbulence’, but limited capacity Windtunnel gives limited number of information (‘scaled’ field exp) CFD offers capacity CFD gives full 3D, t information CFD allows for chemistry, depositon, multi-phase, heat exchange, …

7. 4 November 2010First Dutch OpenFOAM Seminar7 Example study: air quality near a tunnel exit Establishing annual mean NO2 and PM10 concentrations (2015) Evaluating measures to reduce concentration

8. 4 November 2010First Dutch OpenFOAM Seminar8 Set up calculations Ansys Fluent RANS simulations with k-ε RNG Computational domain 500m x 300m x 90m Logarithmic wind/turbulence profiles with z 0 = 2m Traffic induced momentum 4 tunnel ventilations (0.1 m/s, 1.25 m/s, 3.0 m/s, 4.0 m/s) Stationary flow calculations for 12 wind directions Tracer dispersion calculations per source (tunnel exit, streets) Post processing to annual mean concentrations, based on: Wind statistics (KNMI) Background concentrations (RIVM) Traffic data (#vehicles, emission factors) Calibrating the CFD results Passive NO2 observations for a 8 weeks period Adjust tunnel ventilations speed for best fit with measurements

9. 4 November 2010First Dutch OpenFOAM Seminar9

10. 4 November 2010First Dutch OpenFOAM Seminar10 observations ‘raw’ CFD results calibrated CFD results

11. 4 November 2010First Dutch OpenFOAM Seminar11 From Fluent to OpenFoam Practical Costs AQ require large domains and many computions (48 in example) Specific for atmospheric flows and AQ Surface layer is important (concentrations at 1.5 m) Non-neutral conditions, i.e. stratification, thermal inversions, convective ABL Tool development Data assimilation Coupling of regional, urban, street scale models

12. 4 November 2010First Dutch OpenFOAM Seminar12 Test 1: Comparison Fluent & OpenFOAM After Blocken et al. (2007) RANS standard k-ε model 2D domain, 500 m high, 10 km long Hexagonal grid, cell density graded towards ground. Smallest cells 50 cm high & 10 m long 2 nd order discretization & interpolation schemes Logarithmic ABL velocity profile at inlet (airspeed of 18.5 m/s at top of domain) Ground roughness height 0.012 m

13. 4 November 2010First Dutch OpenFOAM Seminar13 Velocity 0 m 1000 m 10000 m

14. 4 November 2010First Dutch OpenFOAM Seminar14 Turbulent Kinetic Energy 0 m 1000 m 10000 m

15. 4 November 2010First Dutch OpenFOAM Seminar15 Turbulent Dissipation Rate 0 m 1000 m 10000 m

16. 4 November 2010First Dutch OpenFOAM Seminar16 Test 2: airflow in a street canyon RANS standard k-ε model 2D domain, 500 m high, hexagonal grid, 0.5 x 0.5 m sized cells near ground Periodic boundary conditions 2 nd order discretization & interpolation schemes Building reference geometry: 15 m high, 10 m wide, 30 m separation Average airspeed of 5 m/s over inlet Building & ground roughness height 0.01 m Actual velocity profile known from measurements: → Determine z 0, d and u * ABL for different geometries

17. 4 November 2010First Dutch OpenFOAM Seminar17 Wind speed independence

18. 4 November 2010First Dutch OpenFOAM Seminar18 Effect of separation 5 meters 10 meters 15 meters 30 meters 50 meters 100 meters 30 meters separation: d = 14,2 m, z 0 = 0,20 m, u * ABL =0,74 m/s Separation [m] d [m]z 0 [m]u * ABL [m/s] 514,40,060,63 1014,20,100,67 1514,30,070,64 3014,20,20,74 508,013,01,81 10010,020,02,10

19. 4 November 2010First Dutch OpenFOAM Seminar19 Effect of height 15 meters 100 meters

20. 4 November 2010First Dutch OpenFOAM Seminar20 Effect of width

21. 4 November 2010First Dutch OpenFOAM Seminar21 Limitations Solving on a coarse grid and mapping solution onto a finer grid often necessary Test 2: Excessively large number of iterations needed; typically 600,000 Spurious problems with numerical stability, even after optimization of stability parameters Possibly connected with the average speed BC on inlet

22. 4 November 2010First Dutch OpenFOAM Seminar22 Conclusions Test 1: Good match between OpenFOAM and Fluent results! Test 2: Calculated wind speed profiles match known velocity profiles Values of derived parameters mainly depend on the presence of large-scale recirculation zones between the buildings (present when height/separation >≈ 0,3 Velocity at ground level highest when height/separation ≈ 1 Results are in agreement with findings of other studies OpenFOAM is applicable for AQ and has many advantages Still lots to be done… Unstable/stable atmospheric boundary layers Tracer dispersion (OpenFOAM mesh and volume sources?) Moving from RANS to LES

23. 4 November 2010First Dutch OpenFOAM Seminar23 Thank you for your attention ! Dutch OpenFOAM User Group