Two short stories on atmospheric composition by xqg16657


									       Two short stories on
     atmospheric composition

• Re-evaluation on the role of N2O5

• A ‘cheeky’ bottom-up evaluation of global
  mean OH
NO + HO2  O3 + OH

           Last IPCC report
Sources of oxides of nitrogen

                      Last IPCC report
  Sinks of oxides of nitrogen
• Lots of inter-conversions between
  different species
    NO, NO2, NO3, N2O5, HONO, HO2NO2, HNO3, PAN
     , PPN, MPAN, ‘other PANs’, organic nitrates

• Loss mechanisms dominated by wet
  and dry deposition of HNO3
  How do we make HNO3 1
1. OH + NO2  HNO3
   Looks simple but has caused at least a decade of
         OH+NO2  HOONO  HNO3
         OH+NO2  HNO3

   Missing chemistry – ‘magic aerosol’ reactions
     How do we make HNO3 2
•   NO + O3  NO2
•   NO2 + hn  NO + O3
•   NO2 + O3  NO3
•   NO3 + hn  NO2 + O3
•   NO3 + NO2  N2O5
•   N2O5 + hn / T  NO3 + NO2
•   N2O5  2 HNO3 (on aerosols)
How do we parameterize the


   How do we parameterize the
• How many molecules hit the surface of
  the aerosol per second?
  – Gas kinetics – mean free path etc etc etc
• What fraction of the molecules that hit
  the surface of the aerosol react
        How do we find the g
•   Lab studies
•   For N2O5 important in the stratosphere
•   Lots of work done
•   g = 0.1 ish
    What impact does this have on
    the tropospheric composition?
    OF NOX, O3, AND OH, Dentener FJ, CRUTZEN PJ, J.G.R. 1993
•   Abstract:
    Using a three-dimensional global model of the troposphere, we show that the heterogeneous reactions of
    NO3 and N2O5 on aerosol particles have a substantial influence on the concentrations of NO(x), O3, and
    OH. Due to these reactions, the modeled yearly average global NO(x) burden decreases by 50% (80% in
    winter and 20% in summer). The heterogeneous removal of NO(x) in the northern hemisphere (NH) is
    dominated by reactions on aerosols; in the tropics and southern hemisphere (SH), with substantial smaller
    aerosol concentrations, liquid water clouds can provide an additional sink for N2O5 and NO3. During spring
    in the NH subtropics and at mid-latitudes, O3-concentrations are lowered by 25%. In winter and spring in
    the subtropics of the NH calculated OH concentrations decreased by up to 30%. Global tropospheric
    average O3 and OH burden (the latter weighted with the amount of methane reacting with OH) can drop by
    about 9% each. By including reactions on aerosols, we are better able to simulate observed nitrate wet
    deposition patterns in North America and Europe. 03 concentrations in springtime smog situations are
    shown to be affected by heterogeneous reactions, indicating the great importance of chemical interactions
    resulting from NO(x) and SO2 emissions. However, a preliminary analysis shows that under present
    conditions a change in aerosol concentrations due to limited SO2 emission control strategies
    (e.g., reductions by a factor of 2 in industrial areas) will have only a relatively minor influence on O3
    concentrations. Much larger reductions in SO2 emissions may cause larger increases in surface O3
    concentrations, up to a maximum of 15%, if they are not accompanied by a reduction in NO(x) or
    hydrocarbon emission.
•   Cited 234 times
            End of story?
• Well g were derived for stratospheric
• Cold
• Pure sulfuric acid
• Troposphere though is
  – Warm and sulfuric acid aerosol is
    Rumblings of discontent
• Tie et al., [2003] found gN2O5<0.04 gave
  a better simulation of NOx
  concentrations during TOPSE
• Photochemical box model analyses of
  observed NOx/HNO3 ratios in the upper
  troposphere suggested that gN2O5 is
  much less than 0.1 [McKeen et
  al., 1997; Schultz et al., 2000]
Jacob, Atmos. Env., 2000

Ozone is produced in the troposphere by gas-phase oxidation of hydrocarbons and CO catalyzed by hydrogen
oxide radicals (HOx º OH + H + peroxy radicals) and nitrogen oxide radicals (NOx º NO+NO2). Heterogeneous
chemistry involving reactions in aerosol particles and cloud droplets may affect O3 concentrations in a number of
ways including production and loss of HOx and NOx, direct loss of O3, and production of halogen radicals.
Current knowledge and hypotheses regarding these processes are reviewed. It is recommended that standard
O3 models include in their chemical mechanisms the following reaction probability parameterizations for reactive
uptake of gases by aqueous aerosols and clouds: gHO2 = 0.2 (range 0.1-1) for HO2 ® 0.5 H2O2, gNO2 = 10-4
(10-6-10-3) for NO2 ® 0.5 HONO + 0.5 HNO3, gNO3 = 10-3 (2x10-4-10-2) for NO3 ® HNO3, and gN2O5 = 0.1
(0.01-1) for N2O5 2 HNO3. Current knowledge does not appear to warrant a more detailed approach.
Hypotheses regarding fast O3 loss on soot or in clouds, fast reduction of HNO3 to NOx in aerosols, or
heterogeneous loss of CH2O are not supported by evidence. Halogen radical chemistry could possibly be
significant in the marine boundary layer but more evidence is needed. Recommendations for future research are
presented. They include among others (1) improved characterization of the phase and composition of
atmospheric aerosols, in particular the organic component; (2) aircraft and ship studies of marine boundary layer
chemistry; (3) measurements of HONO vertical profiles in urban boundary layers, and of the resulting HOx
source at sunrise; (4) laboratory studies of the mechanisms for reactions of peroxy radicals, NO2, and HNO3 on
surfaces representative of atmospheric aerosol; and (4) laboratory studies of O3 reactivity on organic aerosols
and mineral dust.
              2003 / 2004
•   Working on TRACE-P
•   NOx too low in the model
•   What could we do about it?
•   Normally look at sources
•   But also look at sinks
•   New literature
          New lab studies
• Warmer temperatures
• More humid conditions
• Tropospherically applicable aerosols
            New literature
• Kane et al., 2001 - Sulfate – RH
  – JPL
• Hallquist et al., 2003 - Sulfate - temp
  – Tony Cox’s group in Cambridge
• Thornton et al., 2003 - Organics - RH
  – Jon Abbatt’s group at U Torontio
    Parameterization based on best
         available literature
Aerosol type     Reaction probabilityb             Reference

Sulfatea         g = a(RH)10b(T)                  [Kane et al., 2001]
                 a = 2.7910-4 +                   [Hallquist et al., 2003]c
                        1.310-4  RH -
                        3.4310-6  RH2 +
                       7.5210-8  RH3
                 b = 410-2(T-294) (T ≥ 282K)
                 b = -0.48           (T < 282K)

Organic Carbon   g = RH  5.210-4 (RH < 57%)      [Thornton et al., 2003]d
                 g = 0.03            (RH ≥ 57%)

Black Carbon     g = 0.005                         [Sander et al., 2003]

Sea-salt         g = 0.005          (RH < 62%)     [Sander et al., 2003]e
                 g = 0.03           (RH ≥ 62%)

Dust             g = 0.01                          [Bauer et al., 2004]f
What gs do we
 • Much lower than 0.1
 • Dry low values
 • Higher at the surface
What is the impact
on composition?

Lower gN2O5 = higher NOx
Higher NOx = Higher O3
Higer NOx = Higher OH
Does this make the model better?
• Complexity in model isn’t automatically
  a good thing
• More complex models are not
  necessarily better at simulating the
• Complex models take longer to run and
  confuse the issue
     Compare with observations

Emmons et al. [2000] climatology of NOx

Mass weighted model bias changes from
      –14.0 pptv to –7.9 pptv
Mean ratio changes from
      0.77 to 0.86
Middle troposphere (3-10km) changes from
      0.79 to 0.91
     Compare with observations
Logan [1998] Ozonesonde climatology
Mass weighted model bias
      -2.9 ppbv to -1.4 ppbv
Mean ratio changes from
      0.94 to 0.99.
Ox (odd oxygen) budget
      Chemical production increases 7%
      3900 Tg O3 yr-1 to 4180 Tg O3 yr-1
   Compare with observations
Global annual mean tropospheric OH
 0.99106 cm-3 to 1.08106 cm-3
 8% increase.

Both values are consistent with the current constraints
on global mean OH concentrations based on methyl-
chloroform observations:
1.07 (+0.09 -0.17)  106 cm-3 [Krol et al., 1998]
1.16  0.17         106 cm-3 [Spivakovsky et al., 2000]
0.94  0.13         106 cm-3 [Prinn et al., 2001]
• Aerosol reaction of N2O5 is very
  important for the atmosphere
• Previous estimates have been too high
• New laboratory data allows a better
• Sorting out old problems although not
  ‘sexy’ is important
    A ‘cheeky’ bottom-up
evaluation of global mean OH
NO + HO2  O3 + OH

           Last IPCC report
Global mean OH
  How do they calculate global
          mean OH
• Methyl chloroform – solvent in Tippex
• Made by a few large chemical
• Sources are known (nearly)
• Can measure concentrations across the
    Measured across the world
•   (1) Ireland, Mace Head
•   (2) USA Cape Meares, California
•   (3) Barbados
•   (4) American Samoa
•   (5) Tasmania
• Have emissions in your Chemistry –
• Emit the MCF
• Blow it around
• Compare to the observations
• Optimize the OH concentration to get
  best possible fit
       Top down approach
• Don’t directly observe the OH
• Indirect method
• Can directly observe OH
• But lifetime of OH is ~ 1s
• So measurements at one site don’t tell
  you much about global concentrations
• Is this true?
• OH measured by the FAGE group in
• Time series of OH
• Can we use this to provide information
  about global OH
• ‘Couple’ global atmospheric chemistry
  model and the observations
                      Observed vs Modelled OH
Mace Head - Ireland

                                Observations dominated by day
      What have we learnt?
• In general high OH at solar noon
     Model 
• Low OH at night
     Model 
• Daily variability
     Model ?
More useful comparison
        Measured mean is 1.8 × 106 cm-
        3, Modelled mean is 2.3 × 106 cm-3

        Ratio of 1.56 ± 1.62.
        The statistical distribution of the ratio is
        not normal and so more appropriate
        metrics such as the median (1.13) or the
        geometric mean (1.13 +1.44 -0.64 ),
        The model simulates 30% of the linear
        variability of OH (as defined by the R2).
        The uncertainty in the observations (13%)
        suggests that the model systematically
        overestimates the measured OH
Other HOx components
 Sampled for the
             Over a year
EASE ‘97 campaign

                                Sampled for the
                            Smoothed mean OH from
                              NAMBLEX campaign

           Observed Campaign means
Why the annual variability?

All data over a year   All data over a year
R=0.92                 Smoothed R=0.98
Is this only at the surface?
Cape Grim - Australia

                    Other places
   So what have we learnt?
• Mace Head we tend to over estimate
• Cape Grim doesn’t seem so bad
• Can we combine this information and
  the model to get a global number?
• Very Cheeky!
              What do we get?
       A Priori   Compare   A Posteri   Prinn et al.
       (Model)    FAGE
NH     1.12       -19%      0.91        0.90 ± 0.20

SH     1.02       +1%       1.03        0.99 ± 0.20

Global 1.07       -9 %      0.97        0.95
     What does this mean
• Very, very lucky!!!!
• The FAGE OH and the MCF inversions
  seem consistent
• Model transfer seems to work
• Uncertainties suggest it could have
  gone the other way
     Can we do this better?
• Include more data
  – Aircraft campaign
  – Surface sites
  – Ships
• How do we incorporate this information?
             One approach
•   Have the modelled concentrations
•   But want to reduce the dimensionality
•   Principal Components Analysis
•   Takes concentrations from model
•   Comes up with different way of thinking
    about the model gridboxes
Component 1
Component 2
Component 3
Component 4
     How might we use this?
• Compare OH modelled with OH measured
• For each point workout the fraction of that box
  represented by each component
• R (Box Model / Measured) = Σ Cstrength Rcomponent
• Find the Rs
• Reapply to the model OH field
• Calculate a global OH
• CTM comparison with OH looks pretty
• We can use this information to constrain
  the model OH and this gives a
  reasonable result
• To take this further requires a bit more

To top