Process Analysis Evaluation of Global and Regional Ozone Models Identifying Sources of Uncertainty Barron H. Henderson; Joseph P . Pinto; Chris Emery University of Florida 2013-10-30 Barron H. Henderson - University of Florida CMAS2013 1/25
Summary As Oreskes (1994) and later Beck (2002) have demonstrated, atmospheric models are “open systems” that have “essentially unknowable” inputs Can have a wide variety of inputs Generated by different groups Minimum level of detail Come from models with their own uncertainty Easily suffer from compensating errors Getting the “right answer” for the “wrong reasons” Model Performance Evaluations look at the result Process Analysis examines the processes that are typically lost useful in identifying important processes useful to constraining development Barron H. Henderson - University of Florida CMAS2013 2/25
Air pollution kills people Data from the World Health 1.4# WHO$Global$Annual$Deaths$(Millions)$ Organization 1.2# Internationally: Easy to accept – developing countries have more 1.0# pollution. 0.8# Developed world too – “Four times more people die in the San 0.6# Joaquin Valley from air pollution 0.4# than they do from traffic fatalities.” – Jared Blumenfeld, 0.2# EPA Regional Administrator 0.0# IARC classifies air as a Outdoor# carcinogen Air#Pollu3on# Traffic# Barron H. Henderson - University of Florida CMAS2013 3/25
We simplify the real world for solution Barron H. Henderson - University of Florida CMAS2013 4/25
We simplify the real world for solution Barron H. Henderson - University of Florida CMAS2013 4/25
We simplify the real world for solution Several modeling systems simulated the globe or northern hemisphere (Zhang et al. 2011; Emery et al. 2012; Mathur unpublished.) Barron H. Henderson - University of Florida CMAS2013 5/25
We simplify the real world for solution Several modeling systems simulated the globe or northern hemisphere (Zhang et al. 2011; Emery et al. 2012; Mathur unpublished.) Evaluation looked only at the continental United States Barron H. Henderson - University of Florida CMAS2013 5/25
Models generally agreed Figure 1 : Time paired predictions vs Figure 2 : Rank paired predictions vs observations observations Barron H. Henderson - University of Florida CMAS2013 6/25
Models generally agreed mostly Figure 1 : Time paired predictions vs Figure 2 : Rank paired predictions vs observations observations Barron H. Henderson - University of Florida CMAS2013 6/25
Where did it come frome? Ozone&Sources& 100%( 90%( Domes.c( 80%( 70%( Domes.c( 60%( 50%( 40%( Intl/Nat( 30%( 20%( Intl/Nat( 10%( 0%( Northeast( Mtn(West( Figure 4 : Conceptual anthropogenic (red) and Figure 3 : Conceptual biogenic (green) contributions: in-phase contributions of contributions (dashed); out-of-phase (solid). “background” air to total. Conceptually we can ask, do the models agree on how much biogenic emissions contribute to total ozone? Mostly, Henderson et al., 2012 Barron H. Henderson - University of Florida CMAS2013 7/25
What differences are there? Problems: Compensating isoprene nitrate issue led to Better rank-paired performance for GEOS-Chem compared to 1 CAMx in the East Less correlation between background and total 2 Still other differences. 3 Two options: Wait for results to disagree and diagnose the problem then? 1 Systematically compare processes 2 Other inter-comparisons show differences and call for details AQMEII Phase II calls for process-based comparisons 1 HTAP shows model differences 2 Barron H. Henderson - University of Florida CMAS2013 8/25
Process Analysis Overview -./00/120# 345206147# 89:./074;# 812<# &$# %"# %$# Δ C = Δ C Emis + Δ C Transport + Δ C Chemistry "# $# !"# '# (# %$# %%# %&# %)# %*# %"# %+# %,# %'# Figure 5 : Conceptual photochemical day Barron H. Henderson - University of Florida CMAS2013 9/25
Process Analysis Overview -./00/120# 345206147# 89:./074;# 812<# &$# %"# %$# Aha:%mostly% "# chemistry!% $# !"# '# (# %$# %%# %&# %)# %*# %"# %+# %,# %'# Figure 5 : Conceptual photochemical day Barron H. Henderson - University of Florida CMAS2013 9/25
Process Analysis Overview -./00/120# 345206147# 89:./074;# 812<# &$# %"# 5.67%x%R01:%A%+%X%. k1 .>%2.0%Y% %$# 1.48%x%R07:%C%+%D%. k7 .>%0.6%X% ....% 9.12%x%R58:%C%+B%. k58 .>%3.0%X% "# ....................................% Produc@on%%%%%%%%%%%%%%%%%+7.8%X% Loss%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%.4.3%X% $# NET%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%+3.5%X% !"# '# (# %$# %%# %&# %)# %*# %"# %+# %,# %'# Figure 5 : Conceptual photochemical day Barron H. Henderson - University of Florida CMAS2013 9/25
Implementation: Integrated Process Rates Implemented: Transport Convective Mixing Wet deposition Emissions a Dry deposition a Heterogeneous Chemistry Gas-phase Chemistry a Gas-phase emissions and dry deposition are solved either in the chemical solver or in asymmetric convection routines. Separation within convection has not yet been implemented. Barron H. Henderson - University of Florida CMAS2013 10/25
Implementation: Integrated Process Rates Implemented: Transport Convective Mixing Wet deposition Emissions a Dry deposition a Heterogeneous Chemistry Gas-phase Chemistry Figure 6 : Illustrative processes summation check where p is a Gas-phase emissions and dry chemistry, transport, deposition, etc. deposition are solved either in the Formaldehyde process sum chemical solver or in asymmetric compared to instantaneous species convection routines. Separation within change in moles. convection has not yet been implemented. Barron H. Henderson - University of Florida CMAS2013 10/25
Implementation: Integrated Reaction Rates Sparse Matrix Vector Gear: available Kinetic Pre-Processor Rosenbrock (coming soon) LSODES (available) Barron H. Henderson - University of Florida CMAS2013 11/25
Implementation: Integrated Reaction Rates Sparse Matrix Vector Gear: available Kinetic Pre-Processor Rosenbrock (coming soon) LSODES (available) Figure 7 : Illustrative reaction sum check. Hydrogen peroxide sum of reactions vs chemistry process. 0 1 X Y ∆ q chem = [ Rct ] j @ α i k i A i j (1) Barron H. Henderson - University of Florida CMAS2013 11/25
Implementation: Integrated Reaction Rates Sparse Matrix Vector Gear: available Kinetic Pre-Processor Rosenbrock (coming soon) LSODES (available) To do: incorporate species specific error correction from SMV-Gear or switch to Figure 7 : Illustrative reaction sum Rosenbrock check. Hydrogen peroxide sum of reactions vs chemistry process. 0 1 X Y ∆ q chem = [ Rct ] j @ α i k i A i j (1) Barron H. Henderson - University of Florida CMAS2013 11/25
Allows for development of chemical indicators Sillman Ratio (eq 3): Ratio of radical losses ( L ) via non-nitrogen pathways to nitrogen pathways In approximation, greater than 0.35 is NO x limited − 1 = L HO x ≈ P ( H 2 O 2 ) L (2) P ( HNO 3 ) L NO x L NO x We can use processes or these chemical metrics to identify regions of interest for further study. Barron H. Henderson - University of Florida CMAS2013 12/25
Isoprene emissions GEOS-Chem CAMx Figure 8 : GEOS-Chem (left) and CAMx (right) isoprene emissions integrated throughout their planetary boundary layers. GEOS-Chem’s isoprene emissions are higher than CAMx’s in the east Consistent with Carlton and Baker ES&T 2011 GEOS-Chem uses MEGAN which emits more isoprene than BEIS, which was used by CAMx Barron H. Henderson - University of Florida CMAS2013 13/25
Sillman Ratio GEOS-Chem CAMx Figure 9 : GEOS-Chem (left) and CAMx (right) SILLMAN integrated throughout their planetary boundary layers. Many of the same features as seen in OPE Along US northern boundary, differences are more clear Recall that >0.35 is NO x sensitive: the differences here are shades of NO x limited Barron H. Henderson - University of Florida CMAS2013 14/25
Focus Area Focus area selected for: Isoprene emission discrepancy Previous findings that western bias may be attributable to BC Large populations exposed on the western seaboard Figure 10 : Region for further analysis. Barron H. Henderson - University of Florida CMAS2013 15/25
Planetary Boundary Layer Height (focus) Note the earlier rise and lower peak for GEOS-Chem The CAMx PBL is diagnosed from vertical diffusivity using ENVIRON’s vertavg algorithm. Figure 11 : GEOS-Chem (green) and CAMx (blue) planetary boundary layer height averaged within the focus area. Barron H. Henderson - University of Florida CMAS2013 16/25
ISOP time series GEOS-Chem CAMx Figure 12 : Time series ISOP plots for GEOS-Chem (left) and CAMx (right) on 2006-07-01 for the focus area. Note: The time step associated with averaged met/biogenics. Barron H. Henderson - University of Florida CMAS2013 17/25
NO x time series GEOS-Chem CAMx Figure 13 : Time series NO x plots for GEOS-Chem (left) and CAMx (right) on 2006-07-01 for the focus area. Huh? Barron H. Henderson - University of Florida CMAS2013 18/25
Recommend
More recommend
