Chemical Box Models

Markus RexAlfred Wegener Institute

PotsdamGermany

(1) Basic concepts, simplified systems (Sunday)(2) The Ox, NOy/NOx, HOx, Cly/ClOx systems (Monday)(3) Application for polar ozone loss studies (Thursday)

bromine cycle (a) bromine cycle (b)

M + M

Dominating ozone loss cycles for polar winter chemistry

"ClO dimer cycle"

M + M

Mneed sunlight

shuts down during nightdue to a lack of ClO

Red: "rate limiting step" - the reaction with the smallest rate or the "bottleneck" of the cycle. Caution: that does not tell us much about the dynamics of the cycle. E.g. under twilight conditions the ClO dimer cycle is surprisingly insensitive to kClO+ClO, but very sensitive on JCl2O2

All cycles depend on[ClOx] and sunlight

Polar ozone loss

ClO + NO2 -> ClONO2

Cl + CH4 -> HCl + CH3

ClO + OH -> HCl + O3

ClONO2 + h -> ClO + NO2

HCl + OH -> Cl + H2O

Polar ozone loss

ClO + NO2 -> ClONO2

Cl + CH4 -> HCl + CH3

ClO + OH -> HCl + O3

HCl + ClONO2 -> Cl2 + HNO3

ClONO2 + H2O -> HOCl + HNO3

HNO3

cold aerosol

Air mass trajectory (day/night)

Lidar stationOzonsonde station

Match project

e

Match animation

15 Jan – 10 Feb 1995470-500 K potential temperature

RegressionOzone loss rate: -5.5 +/- 0.7 ppbv / sunlit hour

Sunlit time [ hours ]

Ozo

ne c

hang

e [

ppbv

]

Rex et al., 1999

•change of ozone only in sunlight•no change in darkness=> no significant dynamical bias

Rex et al., GRL, 2003

Daytime loss vs. nightime loss

O3 = Ls.ts + Ld

.td

Loss rate during sunlittimes

sunlit time

Bivariate regression analysis:

Rate of change during darkness time in

darkness

Schulz, et al.PhD work

=> Ozone loss occurs only in air masses that encountered PSC conditions during the past ten days.

February:Lifetime of ClOx ~10 days

Filters build into the approach

Rex et al., 1999

• Divergence of trajectory cluster small

- avoids shear zones that tend to have larger mixing

- selects dynamical situations where trajectories are more reliable

• PV change along trajectory small- avoids wave breaking events and unreliable trajectories

• Vertical gradient in ozone profiles small

- avoids lamina structures that indicate wave breaking and mixing

- makes results less sensitive on uncertainties in the calculates radiative cooling rates

Effect of the filters

Gross et al., 2003

Results of a virtual Match campaign within the CLAMS model

=> Filters eliminate the bias due to dynamical effects and reduce the statistical uncertainty (broadness of the distribution)

(Ozone loss rate derived from Match - real ozone loss rate in the model)

Match results 1992-2003=475 K

2002Warm winter,no campaign

Ozo

n lo

ss r

ate

[ p

pbv

/ da

y ]

Date [ day of the year ]

Area

of potential P

SC

form

ation [ 10

6 km2 ]

2003

-30

Rex, 1993; von der Gathen, et al., Nature, 1995; Rex et al. Nature, 1997; Rex et al., JGR, 1998; Rex et al., JAC, 1999; Rex et al., JGR, 2002; Schulz, et al., GRL, 2000; Schulz et al., JGR, 2001, Streibel et al., submitted.

ClO

[ p

pbv

]P

oten

tial t

empe

ratu

re [

K ]

Ozo

ne c

olum

nlo

ss r

ate

[ D

U/

sunl

it ho

ur ]

Potential tem

perature [ K ]

ozone column

loss rate

[ DU

/day ]

(c) Ozone loss rate[ ppbv/day ]

(b) Ozone loss rate[ ppbv/sunlit hour ]

Ozone loss rates in Arctic winter 1999/2000

Date [ day of the year 2000 ] Date [ day of the year 2000 ]

Rex et al., 2002

Pot

entia

l tem

pera

ture

[ K

]

Acc

umul

ated

ozon

e co

lum

nlo

ss [

DU

]

Date [ day of the year 2000 ]

Spring equivalent

potential temperature [ K

]

Accumulated ozone loss [ ppmv ]

Accumulated ozone loss [ ppmv ]

Accumulated ozone loss in Arctic winter 1999/2000

Rex et al., 2002

Ozone loss rate[ ppbv / sunlit hour ]

Denitrification in Arctic winter 1995/1996P

oten

tielle

Tem

per

atur

[ K

]O

zonv

erlu

stra

te[

ppb

v /

Ta

g ]

[ ppb

v / S

onne

n-st

unde

]

Datum [ Tag des Jahres 1996 ]

Modelwithout denitrification

80% denitrification

80% denitrificationin 50% of the air masses

Rex et al., Nature, 1997

=> denitrification plays a significant role for severe Arctic ozone losses

Ozone loss versus VPSC

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

) YearO

zone

col

umn

loss

[ D

U ]

(14-

25 k

m,

mid

-Jan

to

late

Mar

ch) Year

Rex et al., GRL, 2004

~ 15 DU additional ozone loss per Kelvin cooling of the Arctic stratosphere

5-6 K temperature change

80 DUozone loss

Comparison with SLIMCAT – 2004 version

With this version the sensitivity of Arctic ozone loss on climate changewould be underestimated by a factor of three

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

) Year

SLIMCAT

Rex et al., GRL, 2004

Box model based on ClOx, BrOx, Ox chemistry, run along Match trajectories to calculate ClOx that is required to explain the observed loss rates.

January ozone loss – model

=> During cold Arctic Januaries ozone loss is consistently faster than can be explained with standard (JPL 2002) reaction kinetics

max. available Cly

ClOx required toexplain loss rate

max. explainable loss rate

observed loss rate

Rex et al., GRL, 2003

Variation of ozone loss ratewith sza (model) [ relative units ]

Fractio

n of tim

e spen

t persza in

terva

l [% / d

eg ]F

raction o

f time

spent per

sza inte

rval [%

/ deg ]

Fra

ctio

n o

f lo

ss p

ersz

a in

terv

al [

% /

deg

]

sza [ deg ]

Fra

ctio

n o

f lo

ss p

ersz

a in

terv

al [

% /

deg

]

Rex et al., 1999

Distribution of ozoneloss vs. sza

•sensitivity to weak photolysis of Cl2O2 in visible light.

•Not inconsistent with lab data.

=> Large effect on January ozone loss rates, weak effect in March

Model uncertainties

Monte Carlo simulations of model uncertaintieshundreds of model runs distributed according to the stated uncertainties in JPL2002e.g.:+/- a factor of 3 for Cl2O2 in the relevant wavelength range+/- a factor of 8.6 for keq ClO/Cl2O2 at 185 K...

Day of the year

ozon

e lo

ss r

ate

for

com

ple

te a

ctiv

atio

n [p

pb/s

unlit

h]

JPL2002median +/- 34 % of the distribution

=> Factor of ~3 uncertainty (one ) of the calculated ozone lossjust due to uncertainty in the gas phase kinetic data. Frieler et al., PhD work

Sample same air mass at different sza during sunset => better constrain keq ClO/Cl2O2

• No measurement of Cl2O2 needed (=> independent from Cl2O2 uncertainties)• No assumption about equilibrium

Self-Match aircraft flight pattern

Flight track 30 January 2003

outbound flight:before sunset

inbound flight:after sunset

airmasses probed during outbound leg

calculatedmatchradius

COPAS (arbitrary units)

Calculated matchradius + COPAS aerosol

contrailencounters

Results from aircraft self Match

30 January 2003

MATCHES

Equilibrium constant smallerthan in JPL2002

• ClOx calculated with box model from measured ClO

• Lifetime of ClOx long => simple model of only the ClOx family

• ClO/Cl2O2 not in equilibrium ! => Calculations along trajectories

von Hobe et al., ACP, 2004

k eq C

l2O

2/C

lO

Keq ClO/Cl2O2 derived from late night measurements close to equilibrium

SOLVE: Daytime Model Results

JPL 2002

Huder & DeMore 1995

Burkholder 1990

Stimpfle et al., 2004

=> Measurements by Burkholder (extrapolated to 450 nm) are more consistent with atmospheric observations of ClO and Cl2O2 than current JPL recommendations

Constraints on JCl2O2 from combining atmospheric measurements of ClO and Cl2O2 with box model calculations

Ratio = [ ClO model ClO model ] / Cl2O2 model

[ ClO meas ClO meas ] / Cl2O2 meas

(J / kProd) model

(J / kProd) actual

REPROBUS~ WMO 2003

DOAS18 February 2000

JPL02,11% BrCl yield

Bromine

• DOAS measurements of BrO (Pfeilsticker et al.) suggest more BrOx than can be explained by long lived source gases

• Canty et al.: Low OClO measurements during night suggest that the branching ratio of ClO + BrO -> BrCl + O2 is ~11% (in JPL02: ~7%)=> BrOx derived from measured BrO would further increase

Canty et al.

Box model based on ClOx, BrOx, Ox chemistry, run along Match trajectories to calculate ClOx that is required to explain the observed loss rates.

January ozone loss – model

During cold Arctic Januaries ozone loss is consistently faster than can be explained with standard (JPL 2002) reaction kinetics

max. available Cly

ClOx required toexplain loss rate

max. explainable loss rate

observed loss rate

With these changes the January ozone loss problem would be largely resolved.

• Kinetic data that is more consistent with recent field measurements of ClO and Cl2O2

• BrOx based on Pfeilsticker et al.

January ozone loss - update

Frieler et al., PhD work

JPL 2002

changed kinetics

not reprocessed yet

not reprocessed yet

Calculated ClOx vs. measured ClOx during SOLVE

JPL 2002, standard bromine„new kinetic“, standard bromine

„new kinetic“, high bromine

ER-2 measurements

Frieler et al., PhD work

Frieler et al., PhD work

Left: JPL02 kineticMid-left: „new“ kineticMid-right: JPL02 kinetic + „new“ BrOx

Right: „new“ kinetic + „new“ BrOx

ClO+ClO

ClO+O

ClO+BrO

Fraction of ozone loss by individual loss cycles

Comparison with SLIMCAT – 2004 version

With this version the sensitivity of Arctic ozone loss on climate changewould be underestimated by a factor of three

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

) Year

SLIMCAT old

Rex et al., GRL, 2004

Comparison with SLIMCAT - 2005 version

New SLIMCAT version reproduces the slope (and degree of scatter !) reasonably well.

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

) Year

Chipperfield et al, GRL, in press

Model uncertainties

Day of the year

ozon

e lo

ss r

ate

for

com

ple

te a

ctiv

atio

n [p

pb/s

unlit

h]

JPL2002median +/- 34 % of the distribution

with atmospheric ClO and Cl2O2 measurements as additional constraintmedian +/- 34% of the distribution

=> Significant reduction in the model uncertainty if information from atmospheric measurements is used

Frieler et al., PhD work

Monte Carlo simulations of model uncertaintieshundreds of model runs distributed according to the stated uncertainties in JPL2002e.g.:+/- a factor of 3 for Cl2O2 in the relevant wavelength range+/- a factor of 8.6 for keq ClO/Cl2O2 at 185 K...

Uncertainty in JClOOCl with and without considering constraints by atmospheric measurements

Normalized reaction constant (JClOOCl/ a priori median of JClOOCl)

Cum

ulat

ive

prob

abili

ty

a posteriori

a piori

• ... the median increases by ~55%

• ... the uncertainty drops to ~35% of the a priori uncertainty

Frieler et al., PhD work

When considering constraints by atmospheric measurements ...

“necessary ClOx” ozone loss, 2ppb ClOx

Main sources of uncertainties

Relative contribution to uncertainty of model results [%]

based on JPL2002

+ constraints by atmospheric measurements

0 20 40 60 80 100 0 20 40 60 80 100

Conatraints by atmospheric measurements strongly reduce the uncertainty of dimer photolysis to the total uncertainty

In case of low chlorine activation the BrO + ClO -> BrO + ClOO reaction becomes the dominant source of uncertainty

Frieler et al., PhD work

Evolution of ANAT compared to previous years

= 380 K

= 400 K

= 475 K

= 550 K

• Maximum loss in 1999/2000 at about 460 K, Ioss in 2004/2005 peaked lower down at ~420 K

• At all levels below 440 K: loss in 2004/2005 was larger than in 1999/2000

Ozone VMR loss profile 2005 vs. 2000

Ozone loss [ ppmv ]

0 1 2 3

pote

ntia

l tem

pera

ture

[ K

]550

500

450

400

350

1999/2000

2004/2005

1998/1999

Ozone loss estimatesvery sensitive to cooling rates and mixing issues

This region was excluded from previouscolumn loss estimates

Ozone loss [ 1012 molecules cm-3 ]

0 2 4 6

Alti

tude

[ km

]24

22

20

18

16

14

12

Ozone concentration loss profile 2005 vs. 2000

1999/2000

2004/20051998/1999

• In terms of concentration: ozone loss in 2004/2005 larger than the previous record from 1999/2000.

• Column loss in 2004/2005 also larger than in 1999/2000.

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

)

2004

2002

2000

1998

1996

1994

1992

VPSC [ 106 km3 ]

Year

Ozone loss (14-25 km) vs. VPSC (400-550 K)

Ozo

ne c

olum

n lo

ss [

DU

](1

4-25

km

, m

id-J

an t

o la

te M

arch

)

2004

2002

2000

1998

1996

1994

1992

VPSC [ 106 km3 ]

Year

2005(preliminary !)

Ozone loss (14-25 km) vs. VPSC (400-550 K)

Ozo

ne c

olum

n lo

ss [

DU

](3

60-5

50 K

, m

id-J

an t

o la

te M

arch

)

2004

2002

2000

1998

1996

1994

1992

VPSC [ 106 km3 ]

Year

Ozone loss (360-550 K) vs. VPSC (360-550 K)

2004

2002

2000

1998

1996

1994

1992

VPSC [ 106 km3 ]

Year

2005(preliminary !

largeuncertainties !!)

Ozo

ne c

olum

n lo

ss [

DU

](3

60-5

50 K

, m

id-J

an t

o la

te M

arch

)

Ozone loss (360-550 K) vs. VPSC (360-550 K)

Long term evolution of VPSC

100

50

Ozone loss [ D

U ]

Cold winters are getting significantly colder ! Reason ??

FU-Berlin dataECMWF ERA15 data

Year

VP

SC [

106

km3

]