1. Atmospheric modelling of HMs Sensitivity study
Oleg Travnikov
EMEP/MSC-E
TFMM 6th Meeting, Zagreb 2005

2. Outline Overview of the model formulation
Analysis of the model sensitivity and uncertainty
TFMM 6th Meeting, Zagreb 2005

3. Computation domain Regional model (MSCE-HM) Coverage – EMEP region
Resolution – 50×50 km2
EMEP region
EMEP region
Hemispheric model (MSCE-HM-Hem)
Coverage – Northern Hemisphere
Resolution – 2.5°×2.5°
TFMM 6th Meeting, Zagreb 2005

4. Computation domain Vertical structure: (s-p) coordinate system
15 layers up to 15 km
First layer height ~ 70 m
TFMM 6th Meeting, Zagreb 2005

5. Model scheme MSCE-HM Input Output Boundary conditions
Atmospheric transport
Atmospheric chemistry
Dry and wet deposition
MSCE-HM
Pb, Cd, … Hg
TFMM 6th Meeting, Zagreb 2005

6. Atmospheric transport
Species continuity equation
Horizontal transport:
Monotone Bott scheme (fourth-order area-preserving polynomials)
Vertical transport:
Bott scheme adapted to irregular vertical grid
Vertical eddy diffusion:
Second-order implicit finite-difference scheme
TFMM 6th Meeting, Zagreb 2005

7. Mass consistency Air continuity equation
Calculation of vertical transport velocities from the air continuity equation:
Step I: Calculation of the air horizontal transport using the Bott advection scheme
Step II: Calculation of vertical velocities using analytical inversion of the Bott scheme
TFMM 6th Meeting, Zagreb 2005

8. Atmospheric transport
Basic tests
“Rotational flow”
Results:
Low numerical diffusion
Monotonicity
Insignificant distortions
Stability in strong deformational flows
Low time-splitting error
TFMM 6th Meeting, Zagreb 2005

9. Dry deposition Main characteristics: Ecosystem-dependent scheme
Resistance analogy approach
Size-segregated deposition velocities of particles
Deposited species:
Particles (Pb, Cd, Hgpart, …)
Gases (Hg2+gas, Hg0)
Aqueous Hg species (fog)
TFMM 6th Meeting, Zagreb 2005

10. Dry deposition Particles deposition to vegetation:
Theoretical formulation [Slinn, 1982]
Empirically derived parameters [Ruijgrok et al., 1997; Wesely et. al., 1985]
Ruijgrok
Wesely
TFMM 6th Meeting, Zagreb 2005

11. Dry deposition Particles deposition to water surface:
Based on the approach from [Williams, 1982]
Effects of wave breaking and aerosol washout by seawater spray
TFMM 6th Meeting, Zagreb 2005

12. Aerosol scavenging efficiency
Wet deposition
Main characteristics:
In-cloud and below-cloud scavenging
Empirically derived approach:
Aerosol scavenging efficiency
[Kasper et al., 1998]
Rp – precipitation rate
A, B – empirical constants
In-cloud scavenging efficiency of aerosol:
Cw – cloud liquid water content
TFMM 6th Meeting, Zagreb 2005

13. Atmospheric chemistry
(Hg)
Gas-phase chemistry:
Oxidation by ozone, chlorine and hydroxyl radical
Aqueous-phase chemistry:
Oxidation by dissolved ozone, chlorine and hydroxyl radical
Reduction to elemental form
Formation of chloride complexes
Adsorption by soot particles
TFMM 6th Meeting, Zagreb 2005

14. Boundary conditions Mercury
Monthly mean concentrations at the domain boundaries from hemispheric modelling
Lead and cadmium
Prescribed air concentrations based on measurements data
TFMM 6th Meeting, Zagreb 2005

15. Sensitivity and uncertainty analysis
Model sensitivity to input parameters and processes
Influence of meteorological variability
The model uncertainty
TFMM 6th Meeting, Zagreb 2005

16. Model sensitivity Metals: Lead and mercury Model input parameters:
Anthropogenic emission
Pb, Hg
Speciation of anthropogenic emission Hg
Natural emission and re-emission
Wet deposition coefficient
Dry deposition velocity
Boundary concentration
Eddy diffusion coefficient
Liquid water content
Oxidation rate by O3
Oxidation rate by OH
Oxidation rate by Cl2
Reduction rate in aqueous phase
Hg ion-chloride equilibrium constant
Adsorption equilibrium constant
pH of cloud water
Chloride ion concentration
Aerosol solubility
Henry’s low constants
Metals:
Lead and mercury
Model input parameters:
(Pb – 7, Hg – 25 parameters)
Model outputs:
Concentration in air
Concentration in precipitation
Total deposition flux
TFMM 6th Meeting, Zagreb 2005

17. Relative deviation of Pb air concentration
Model sensitivity
Relative deviation of Pb air concentration
Vd x 2
Analysis procedure:
Variation of input parameters:
X = Xbase · K
Relative deviation of the model output:
TFMM 6th Meeting, Zagreb 2005

18. Relative deviation of Pb total deposition
Model sensitivity
Relative deviation of Pb total deposition
Vd x 2
Analysis procedure:
Variation of input parameters:
X = Xbase · K
Relative deviation of the model output:
Sensitivity coefficient:
TFMM 6th Meeting, Zagreb 2005

19. Sensitivity coefficient
Pb air concentration
Pb total deposition
Intervals – 90% confidence range
TFMM 6th Meeting, Zagreb 2005

20. Sensitivity coefficient
Hg air concentration
Hg total deposition
TFMM 6th Meeting, Zagreb 2005

21. Meteorological variabilityPb Hg
Analysis procedure:
Calculation period:
The same emission data
Relative deviation:
Mean square relative error:
TFMM 6th Meeting, Zagreb 2005

22. Meteorological variability
Variation of modelling results
Intervals – 90% confidence range
Pb: ±25-30% for all parameters
Hg: ±10% for air concentration, ±20-30% for others
TFMM 6th Meeting, Zagreb 2005

23. Model uncertainty Expert estimates of input uncertainties: (20%-90%)
Input parameters ex
Anthropogenic emissions
50%
Speciation of anthropogenic emission
40%
Natural emission and re-emission
90%
Wet deposition coefficient
75%
Dry deposition velocity
Boundary concentration GEM
20%
Boundary concentration of Pb
Eddy diffusion coefficient
Liquid water content
Oxidation rate by O3
Oxidation rate by OH
Oxidation rate by Cl2
Reduction rate in aqueous phase
Hg ion-chloride equilibrium constant
Solution-adsorption equilibrium constant
pH of cloud water
Chloride ion concentration
Aerosol solubility
Expert estimates of input uncertainties:
(20%-90%)
50%
TFMM 6th Meeting, Zagreb 2005

24. Model uncertainty Expert estimates of input uncertainties: (20%-90%)
Hg air concentration
Expert estimates of input uncertainties:
(20%-90%)
Uncertainty due to individual parameters:
Overall model uncertainty:
TFMM 6th Meeting, Zagreb 2005

25. Model uncertainty Lead Mercury
Model intrinsic uncertainty (without emissions)
Overall model uncertainty (with emissions)
Lead
Mercury
TFMM 6th Meeting, Zagreb 2005

26. Conclusions
Detailed model description was prepared and distributed for the model review
Sensitivity analysis has shown that the modelling results for particle bound metals (Pb, Cd, etc.) are the most sensitive to emission data
Mercury modelling results are very sensitive to boundary conditions
The model intrinsic uncertainty does not exceed 50%
Evaluation of the overall model uncertainty requires realistic estimates of emission data uncertainty
TFMM 6th Meeting, Zagreb 2005