Training Data assimilation and Satellite Data – Johannes Flemming Modelling and Assimilation of Atmospheric Chemistry [email protected]

Training Data assimilation and Satellite Data – Johannes Flemming

Modelling and Assimilation of Atmospheric Chemistry

[email protected]

[email protected]


Overview

Motivation

Basic concepts of atmospheric chemistry modelling

Data assimilation of trace gases

Observations

Chemical data assimilation at ECMWF (GEMS & MACC)

Examples for O3 and SO2

Summary


Why Atmospheric Chemistry at NWP centres ?- or in a NWP Training Course?

Environmental concern

Air pollution

Ozone hole

Climate change

Expertise in data assimilation of satellite, profile and surface obs.

Best meteorological data for chemical transport modelling

Interaction between trace gases & aerosol and NWP

radiation triggered heating and cooling

precipitation (condensation nuclei)

Satellite data retrievals improved with information on aerosol

Hydrocarbon (Methane) oxidation is water vapour source


Change in Aerosol Optical Thickness

ClimatologiesNew: reduction in

Saharan sand dust

& increased sand dust over Horn of Africa

Old aerosol dominated by Saharan sand dust

26r3: New aerosol (June) Tegen et. al 1997 997):

26r1: Old aerosol (Tanre et al. 84 annually fixed)

Thickness at 550nm

Example: Impact of Aerosol Climatology on SW Radiation

J.-J. Morcrette A. Tompkins


Surface Sensible heat flux differences

20 W m-2 ~ 20-30%

Boundary layer height increases >1km

Example: Impact of Aerosol Climatology on SW Radiation

old

new

New-old


Improved Predictability with improved Aerosol Climatology

Published in Quart. J. Roy. Meteorol. Soc., 134, 1479.1497 (2008)

Rodwell and Jung

Figure 3: Average anomaly correlation coefficients (see main text for details) for forecasts of meridional wind variations at 700 hPa with the `old' (solid) and the `new' (dashed) aerosol climatology for (a) the African easterly jet region (15oW.35oE, 5oN.20oN) and (b) the eastern tropical Atlantic (40oW.15oW, 5oN.20oN). Forecast lead-times for which the scorewith the `new' aerosol is signicantly better (at the 5% level) are marked with circles. Results are based on the weather forecasts (see main text for details) started at 12 UTC on each day between 26 June to 26 July 2004.


An other motivation …


Modelling atmospheric composition Gas phase

Transport

Source an Sinks

Chemical conversion

Emissions

Deposition


N2

O2

H2OArgon

20%

78%

1%

N2O 310

H2

CO

Ozone

500

100

30

ppb1:109

CO2

CH4 (1.8)

ppm1:106

380

Ne

18He (5)

HCHO 300

Ethane

SO2

NOx

500

200100

ppt1:1012

NH3 400

CH3OOH 700

H2O2 500

HNO3 300

others

Atmospheric Composition

•The small concentrations do matter because•chemical conversion is non-linear•small concentrations could mean high turn-over, i.e. high reactivity


Emissions

ChemicalReactions

AtmosphericReservoir

wet & dryDeposition

Transport TransportcatalyticalCycles

Dr. Martin Schultz - Max-Planck-Institut für Meteorologie, Hamburg

Atmospheric Chemistry

Photolysis


Modelling atmospheric composition

Mass balance equation for chemical species ( up to 150 in state-of-the-art

Chemical Transport Models)

,

.

concentration of species i

( ) ... Emission

( , , , ...) ... Chemical conversion

... Deposition

i ih h i c i Z

i

i i

i i j k m

i Dep i

c cc w c K E R D

t z z z

c

E f c

R f c c c c

D l c

V

Source and SinksTransport


Examples of Global Mass Budget

Global transport contribution is zero if model conserves mass


Nitrogen Oxides - sources and sinks

Total ColumnsConcentrations

Surface Emissions

Chemical Production and Loss & LightningVincent Huijnen, KNMI

MOZART-3 CTM2003070500

Note: High Loss is related to high concentrations


Some very general remarks about gas phase chemistry in the Atmosphere …

Under atmospheric conditions (p and T) but no sunlight atmospheric chemistry of the gas phase would be slow

Sun radiation (UV) splits (photolysis) even very stable molecules such as O2 (but also O3 or NO2) in to very reactive molecules

These fast reacting molecules are called radicals and the most prominent examples are

O mainly in stratosphere and above, but also in troposphere

OH (Hydroxyl radical) and HO2 (peroxy radical) in troposphere

Reaction with OH is the most important loss mechanism in the troposphere for very common species such as CO , NO2, O3 and Hydrocarbons

We need to quantify the concentration change due to chemistry


Chemical Kinetics (I)

Gas-phase reactions

A + B C + D

A + B C

C A + B

A + B + M C + D + M

Photolytic reactions if < limit

A + h B + C

heterogeneous (aerosol-, liquid-phase) reactions

surface reactions

often A+B C


Chemical Kinetics (II)

Reaction speed (= concentration change per time) is proportional

to product of concentration of reacting species

Example: A + B C

Example: Photolysis A + h B + C

Chemical loss of A is proportional to concentration of A

1

1

[A] [B] [C][A][B]

A concentration of A

exp chemical rate constant

d d dk

dt dt dt

Ek

kT

[A][A]

photolysis constant

b

dj

dtj


Detour …Sub-grid scale chemistry parameterisation??

The non-linear nature of the chemical Kineticis leads to a potential

influence of the spatially unresolved sub-scale variability on the

model resolved scale (“turbulent” reynolds averaging)

There is no solution for sub-grid scale variability of chemistry yet

No solution to this problem yet

1 1

A = A + A

[A][A][B [A ][B ]]

dk

dtk

Well mixed – reaction occurs Not mixed – no reaction


◄ into stratosphere

No transport modelled

Chemical Lifetime vs. Spatial Scale


Ozone Production in Stratosphere (Chapman 1932)

O2

b

O3

d

j2

k2 3

j3 2

k3 2

a) O O + O

b) O + O + M

+h (240 )

+h (900

O + M

c )) O O + O

d) O O 2O

nm

nm

2O ~

[ ]0 quasi steady state assumption (QSSA)

p

d O

dt

The shape of the ozone profile can be qualitatively explained by the derived ozone production

(a) is most efficient in upper atmosphere (more UV input)

(b) is most efficient in lower atmosphere (higher density)

32 O3 3 3

O2 2 2 O3 3 3

OO O j O O O

[ ]j O O O j O O O

b d

b d

dk k

dt

d Ok k

dt

Chemical Equations Kinetics equations

Assumptions


Ozone profile and life time predicted with Chapmann Cycle

theory

observed

Catalytic reactions with NOx, HOx,ClOx and BrOx


Data assimilation of atmospheric composition

Observations

Assimilation System

Examples


Special Characteristics of Atmospheric Chemistry data (vs NWP) assimilation

Quality of NWP depends predominantly on Initial conditions whereas Atmospheric Chemistry modelling depends on initial state (lifetime) and emissions

Emissions data are uncertain and difficult to measure and biased

Chemical atmospheric fields have strong horizontal and vertical gradients for atmospheric composition

small scale emission variability

Heterogonous reactions on surface

Observation

Limited representativeness

Sparse

Poor near real time availability


Air quality observations at surface (… biased toward polluted areas )

NO2 annual mean in Berlin

Regional model (25 km) vs.Air quality observations

Large variability of observationsWithin GRID box

Is the model result “good” ?

Could data assimilation improve model result?


Profile Observations ( … far to few)

Ozone sondes- GAW stations

MOZAIC flight observations


Satellite observations Assimilation of retrievals vs. radiances

RetrievedPartial/total

columns

3D Concentration Field

Analysis

Total Columns (averaged)

Retrieval Algorithm (A priori atmospheric

state)

Complex ObservationOperator

(Radiative Tranfer)CO2

Assimilating Model(model atmopsheric

state)

ObservationOperator

GRG & AERGEMSsystem

Radiance observation

Evaluation

retrievals analyses

Radiance assimilation


DOAS analysis

Total Slant Column

Tropospheric Slant Column

Tropospheric Vertical Column

SCIATRAN RTM (airmass

factor)

A priori information needed:

aerosol loading and profile

vertical profile of the observed species

surface reflectance at time of measurement

Obtained from climatologies or Models

Problems:

spatial resolution

loss of independence of measurement and model

clouds

NO2 retrievals from satellite observations

Joana Leitao. Uni Bremen

Retrieval of trop. NO2 from SCIAMACHY measurements


Special Characteristics of Atmospheric Chemistry satellite observations

Total or partial column retrieved from radiation measurements

No or only low vertical resolution

Ozone (and NO2) dominated by stratosphere

Weak or no signal from planetary boundary layer

Global coverage in a couple of days (LEO)

Limited to cloud free conditions

Fixed overpass time (LEO) – no daily maximum

Retrieval algorithms are ongoing research


MOPITT CO (TC) Data countExample April 2003

•Very few observations in tropical regions (clouds)•Only Land Points were assimilated•Data have been thinned to 1°x1° grid


CO total column retrievals from different instruments/ retrievals

MOPITT- retrieval

IASI - retrieval A IASI – retrieval B

27.-31.08.2008

Different retrievals tend to differ …


Trace gas assimilation system at ECMWF

Stratospheric Ozone with linearized ozone chemistry since 1999

GEMS-project (2004-2009) / MACC-project (2009-2011)

Ozone in troposphere and stratosphere

CO, SO2, Formaldehyde, NOx, Aerosol and CO and CH4

Full Chemistry (CTM MOZART-3)


GEMS / MACC Global Production

2003 -2008


The IFS is coupled to a CTM for data assimilation of atmospheric composition

CTM

0 3 6 9 12 …..

IFSM

IFSC

C

IC

P L

OA

SIS

4OASIS4

Meteorology

Tracer Concentrations

Tracer ConcentrationsProduction/Loss

Initial Conditions

Concentrationfeedback


ECMWF 4D-VAR Data Assimilation Scheme Assimilation of Reactive Gases and Aerosol

transport +

“chemistry”

advection only

transport +

“chemistry”


Reactive gases assimilation – approach in GEMS and MACC

Include NOx, SO2, O3, CO and HCHO species in IFS (Transport and

Assimilation)

Introduce source and sinks by coupling with Chemical Transport Models

MOZART (MPI-Hamburg), TM5 (KNMI), Mocage (Meteo France)

1. Assimilate species in IFS with existing 4D-VAR implementation developed for meteorological fields

Apply coupled system in out loops (forward trajectory run) only

NMC method (i.e. differences to different meteorological forecasts) to obtain background error statistic

Implement and test feedback mechanism to coupled CTM

Implement diagnostic NO2 (fast chemistry, observed) to NOX (slow chemistry, modelled) observation operator


Assimilation of Ozone

Dominated by the stratosphere

Assimilated Ozone retrievals

Total Columns from UB-VIS instruments (OMI, SCHIMACHY and SBUV)

Low-resolution stratospheric profiles from Microwave-Limb-Sounder (MLS)


Ozone Total Columns – Inter-annual variability in GEMS re-analysis


Ozone Hole Development

Winter – no sun light:

Cold stable Vortex -> formation of Polar Stratospheric Clouds

Accumulation of Chlorine/Bromine compounds on PSC surface

Spring – gradually more sun-light

Rapid release of CLO on PSC surfaces

Quick catalytic destruction of ozone –> ozone hole

Vortex becomes more permeable – closure of ozone hole


Temperature and O3 over South Pole – Sonde observations

Temp

O3

Ozone Hole Closure by transport PSC Formation Chlorine Activation


Antarctic Ozone Hole 2008

Different Modelling Schemes for the Chemistry

Operational ECMWF ozone with O3 chemistry parameterisation

NRT coupled-system IFS-MOZART with full stratospheric chemistry:

coupled-system IFS-TM5 with stratospheric O3 climatology

Each scheme was run with (AN) and without (FC) data assimilation of O3 satellite observations at 0 UTC to provide initial ozone conditions

Assimilated Observations

Total columns (OMI, SCHIAMACHY, SBUV)

Stratospheric partial Columns (MLS)

Questions:

Inter-instrument biases

Impact of different chemistry schemes


Instument - Biases over Antarctica

MLS can observe during polar night

Biases small

•MLS can observe during polar night•Large differences due to different sampling•Actual Biases are small (2-3%)


Total Columns vs. Ozone Sondes

Without assimilation With assimilation


Ozone hole size with different CTMs and assimilation

No assimilation

Assimilation

Forecast initialised by analyses every 15 days


Vertical Profiles at Neumayer Station

IFSLinearChemistry

MOZARTFull Chemistry

TM5Climatology

No MLS


Assimilation of Volcanic SO2


Volcanic eruptions are a major

natural SO2 source

Volcanic eruptions can penetrate the

tropopause

SO2 is a aerosol precursor (SO4)

SO2/SO4 is long-lived in

stratosphere

Volcano ash emission are an

aviation hazard

Volcanic Eruptions


Volcanic SO2 assimilation experiment

Questions:

What is the Volcano SO2 / ash flux

What is injection height and plume height

Can assimilation of satellite data pick up plume or do we need to model the dispersion of the SO2 / ash emissions in the assimilating model

Case study of Nyamuragira eruption 27/11/06- 4/12/06

Observations:

SCIAMACHY SO2 total column (BIRA) - Assimilated

OMI SO2 total column (plots) for comparison – Not Assimilated


Iceland Volcano Plume ForecastInjection height 3, 5 and 10 km


28.11

6.12

1.12

3.12

OMI SO2 columns [DU] from http://so2.umbc.edu/omi/

SO2 total column observations

SCIAMACHY SO2 columns 1.12

OMI plots to estimate Volcano flux and injection height

Increase in total column – loss = flux estimate

Test runs with variable injection height to match with observed plume

Background error profile constructed according to estimated plume height

SO2 background error stdv

profile


SO2 forecast assimilation: Total column SO2

1 Dec

Dobson Units

3 Dec 5 Dec 7 Dec

Control run – no assimilation)

Tracer run injection height 14 km – resembles OMI SO2

SO2 Forecast

SO2 tracer emission estimated from OMI SO2 day to day total column change


SO2 assimilation: Total column SO2

1 Dec, 0z

Dobson Units

3 Dec, 12z 5 Dec, 12z 7 Dec, 12z

Assimilation SO2 (no source)

Assimilation with SO2 volcano emission

SO2 Assimilation

Assimilation without SO2 volcano emissions fluxes provides reasonable plume forecast if injection height is correctly guess (Background error statistics)


Summary

Atmospheric composition and weather interact

Chemical lifetimes determine to what extent species are distributed in the

atmosphere

Emissions are often uncertain but greatly determine concentrations

Chemical data assimilation has to deal with very heterogeneous fields

Chemical data assimilation should also help to improve emissions

MACC forecast system produces useful forecast and robust data assimilation

products

Re-analysis of Ozone, CO and Aerosol (2003-2008) and NRT forecast up to

present are available at http://www.gmes-atmosphere.eu/


Thank You !

Documents

Training Data assimilation and Satellite Data – Johannes Flemming Modelling and Assimilation of Atmospheric Chemistry [email protected]