Page 1 COST/ESF School: UTLS, Cargese, 3 – 15 October 2005 DA 11: Research satellites and data assimilation Author: W.A. Lahoz Data Assimilation Research.

Page 1COST/ESF School: UTLS, Cargese, 3 – 15 October 2005

DA 11: Research satellites and data assimilation

Author: W.A. Lahoz

Data Assimilation Research Centre, University of Reading RG6 6BB, UK


•Characteristics of research satellite data:

Viewing geometry, coverage and observation types (e.g. chemical species)

•Motivation for assimilation of research satellite data:

High vertical resolution, global coverage, novel data types and benefits of assimilating a novel data type such as ozone; synergies

•Efforts, with examples, at assimilating research satellite data:

NWP centres (GCM and CTM models), research into data assimilation (GCM and CTM models)

•Use of data assimilation techniques to evaluate observations and models

Topics:


Characteristics of research satellite data


1. Resolution: temporal & spatial2. Frequency: temporal & spatial3. Wavelength of measurement: what region of EM spectrum4. Radiometric noise: signal/noise (S/N) ratio5. Coverage: global/local6. Geometry: nadir/limb/occultation7. Level of data: 0: photons; 1: radiances; 2: geophysical parameters8. Errors: random, systematic – biases, “representativeness”9. Platform: sondes, aircraft, satellites – resolution10. Influences on time/space evolution: dynamics: temperature, winds, ozone; chemistry: ozone, ClO.

Features of observations


Observations are our representation of the “Truth”

Photons (L0 data)

Radiances (L1 data)

Geophysical parameters (L2 data)

algorithms

e.g. profiles (ozone), total column (ozone)

What instrument receives

Scientists normally work with L2 dataNWP Centres work a lot with L1 data


Random: Assumed Gaussian; it is reduced by taking averages: Is it Gaussian? How do we check? What could be non-

Gaussian? -> bi-modal distributions, e.g., precipitation

Systematic (bias): Can vary temporally & spatially. If fixed and known, it should be removed

Representativeness: Occurs when information is represented at a scale different from the source of the information (e.g. representation of sonde data in a GCM grid). More important for small-scale observations.

X X?

Observation errors


Real resolution of observations could be coarser than that implied by the apparent resolution/frequency of the observations

-> correlations (horizontal/vertical) between observations

Correlations taken into account in the observation errors covariance:

characterizes observation errors

-> data assimilation (DA): - diagonal errors (e.g. for obs): no correlations - non-diagonal errors (e.g. for model, “background”):

correlations between variables (could be different from obs)

- estimation of “background” error an important issue in DA

Stratosphere: horizontal correlations ~larger than for troposphere:

Flow dominated by smaller wavenumbers in stratosphere

Resolution of observations


GOES-8: ~1 kmHurricane Erin

09/09/01 ~1530 Z

Courtesy James Purdom


Hurricane Erin09/09/01 ~1530 Z

MODIS: ~250m

Courtesy James Purdom


L2: Easier to assimilate than L1 -> historically L2 data has tended to be assimilated before

L1 Recent ideas from Rodgers (“information content”) -> alleviate problems associated with L2 data: a priori

information

L1: less “contaminated” (e.g. by a priori information) Errors less correlated than for L2 data Tendency to assimilate radiances: - nadir radiances already assimilated by met agencies; - limb radiances are much harder to assimilate Improvements in analyses & forecast skill at NWP centres

Level of data L1/L2:


Three ways:

Passive technologies: sense LW radiation emitted by atmosphere, SW reflected by atmosphere.

- imaging: optically thin -> information on Earth surface - sounding: optically thick -> information on atmosphere

Active technologies: emit radiation & measure how much scattered/reflected back

GPS: measure phase delay of signal as it is refracted in atmosphere

BUT: other ways of sounding the atmosphere…

Sounding the atmosphere using satellites


Conventional: surface, sondes (local coverage; high spatial & temporal resolution) – temperature, humidity, winds – “Synoptic”

Aircraft: local coverage; high spatial & temporal resolution – temperature, humidity, winds – “Asynoptic”

Satellites: “Asynoptic”Operational satellites: ATOVS, Satwinds, SSMI (nadir; global

coverage; low spatial & temporal resolution) – temperature, humidity, winds

AND: Recent interest in research satellites (nadir & limb): e.g. SCIAMACHY total ozone assimilated at ECMWF

Research satellites now part of Global Observing System

Observation types used by the Met Office


Operational:Used by NWP centres – objective is to improve weather forecast

-> measure dynamical quantities (temperature & humidity)Continuity/heritage (NOAA-17…)Near-real-time (NRT): typically within 3 hours

Research:Used for research – objective is to study key issues (ozone hole;

climate change…) -> measure a range of quantities, both dynamical & chemical (ozone, …)

Often one-off (continuity for Envisat?)Not NRT: fastest delivery 1-2 days (useless for NWP)

Operational vs Research satellites


ATOVS Global coverage

©Crown copyright, Met Office 16/10/02:

Aircraft Local coverage

Observations used by Met Office


Types of satellite observations: orbits

1. Geostationary (fixed point over the equator): 60N-60S Only one orbit: 35,800 km; ¼ Earth’s surface

2. Polar: quasi-global (e.g. 600 km Hubble, 225-250 km Shuttle )

3. Sun-synchronous (fixed equator crossing time)

4. Non sunsynchronous (variable equator crossing time)

Satellite orbits


NOAA-15 NOAA-16 NOAA-17

Goes-W Goes-W Met-7 Goes-W GMS(Goes-9)

Polar orbitersLEOs

Geostationary satellites

GEOs

Courtesy J-N ThépautECMWF


1. Sun-synchronous satellites (e.g. Envisat, Eos Aura):

Instruments look away from the sun: no manoeuvre to prevent the sun damaging the

instruments Cannot observe the diurnal cycle at a particular place: e.g. diurnal cycle of NO, NO2

2. Non sunsynchronous satellites (e.g. UARS):

Can observe the diurnal cycle at a particular place Have to do manoeuvres to prevent the sun damaging

the instruments -> North look/South look for UARS MLS

Diurnal cycle & orbit


ECMWF satellite dataCourtesy J.-N. Thépaut

Now 29Cyr2 (since Jun 2005)






Motivation for assimilation of research satellite data


NASAUARS (CLAES, HALOE, HRDI, ISAMS, MLS): met data (temp, winds, water vapour); chemical species

(ozone, ClO). Science: JAS 1994, Cal-val: JGR 1996

EP TOMS: ozone column

EOS Terra: land, water & ice (ASTER); radiation (CERES); radiation & biosphere parameters (MISR); biological & physical processes on land & ocean (MODIS); CO & CH4 in troposphere (MOPITT)

EOS Aqua: clouds, radiation & precipitation (AMSR/E); clouds, radiation, aerosol & biosphere parameters (MODIS); temp & humidity (AMSU, AIRS, HSB); radiation (CERES)

EOS Aura: atmospheric chemistry (Eos MLS); temp & constituents (HIRDLS); tropospheric ozone (TES); ozone column & UV (OMI)

What do research satellites measure (observe)?


ESA:ERS-2: ozone column & ozone profiles (GOME)

Envisat: temperature, ozone, water vapour & other constituents using limb, nadir & occultation geometries (MIPAS, SCIAMACHY, GOMOS); aerosol (AATSR, MERIS), sea surface temperature (AATSR); ocean colour (MERIS); land & ocean imagery (ASAR); land, ice & ocean monitoring (RA-2); water vapour column & land parameters (MWR); cryosphere & land parameters (DORIS); RA-2 calibration (LRR)

Synergy: combine nadir (SCIAMACHY) with limb (MIPAS, SCIAMACHY)

-> Use of DA to evaluate Envisat & study chemical distributions

ESA/CSA: ODIN (OSIRIS & SMR): ozone & NO2 (columns & profiles)


NASDA-JAXA:ADEOS: ADEOS-TOMS (ozone column), ILAS (temperature, ozone,

water vapour & other constituents)

TRMM (with NASA): tropical rain -> global change and environmental policy

ADEOS-II: water column, precipitation & ocean and ice parameters (AMSR), land, ice and ocean parameters (GLI), ocean winds (SeaWINDS), radiation parameters (POLDER), temperature, ozone & other constituents (ILAS-II)


Research satellite observations by frequency:Different parts of spectrum sensitive to different species &

different atmospheric levels - can sample the depth of the atmosphere

1. Infrared (IR): ISAMS (UARS), MIPAS (Envisat), HIRDLS (Eos Aura)

2. Visible (Vis): GOME (ERS-2), SCIAMACHY (Envisat)

3. Ultraviolet (UV): GOME (ERS-2), GOMOS & SCIAMACHY (Envisat)

4. Microwave: MLS (UARS), Eos MLS (Eos Aura)

Variety-> opportunity to evaluate observations (e.g. UARS, Envisat)

EM spectrum properties: e.g. microwaves less affected by clouds

EM spectrum & observations


Novel data types: ozone (for NWP) -> radiative transfer (nadir radiances); wind information; UV forecasts

Limb/nadir synergy -> information on tropospheric ozone ; relatively high vertical resolution of limb sounders

Information to help make chemical forecasts: tropospheric ozone -> air quality

Research satellite of today -> operational satellite of tomorrow

Interest in research satellites by the NWP agencies make them more attractive to the EO community

Differences between research and operational satellites becoming blurred: e.g. use of Envisat data

Benefits from research satellites


Is the Earth’s ozone layer recovering?Is air quality getting worse?How is Earth’s climate changing?

Questions that Eos Aura & Envisat could help answer


Examples


GCM: dynamics with “simple” chemistry (ozone sources & sinks; Cariolle + “cold tracer”)

interaction between dynamics/chemistry/radiation; use of operational observations DARC/MetO (3d-var), ECMWF (4d-var)

Used for:

•Ozone forecasts & re-analyses (ECMWF)

•Ozone studies (DARC/MetO)

•Observation & model evaluation (ECMWF, DARC/MetO)

What models to use in DA?


The SH polar vortex split of Sep 2002: ozone at one level

MIPAS ozone

DARC analyses

Blue: low ozone; Red: high ozone; 10 hPa

CourtesyAlan Geer


CTM: range of chemical models (Cariolle + “cold tracer” -> sophisticated photochemistry) forced by off-line winds/temperature

No interaction between dynamics/chemistry/radiation Cariolle: KNMI (KF), GMAO (PSAS) Sophisticated photochemistry: NCAR (KF), BIRA-IASB (4d-var)

Used for:

Testing methodologies (NCAR)

Ozone forecasts (KNMI, GMAO)

Studies of chemical distributions (BIRA-IASB): unobserved species from observed species; information: data-rich -> data-poor regions (reconstruct diurnal cycle, e.g., Lary)

Observation/model evaluation (KNMI, GMAO, BIRA-IASB)


Ozone total column, GOME KNMI/ESA

http://www.knmi.nl/gome_fd

Ozone hole area wrt 220 DU in SHKNMI

http://www.temis.nl

2002

20032005


Assimilated datasets from Envisat

24 Sep 12UTC:

Ozone distribution

BIRA-IASB MIPAS analyses:

Red (high), blue (low)

Ozone evolution 31.5 hPa;22 Sep -> 25 Sep

ACRI GOMOS analysesRed (high), blue (low)


Coupled GCM/CTM: The aim is to have the advantages of GCM & CTM approaches. How to couple?

Météo-France/CERFACS, MSC/BIRA-IASB

Still at an early stage but potentially promising

See also EC GEMS proposal:

http://www.ecmwf.int/research/EU_projects/GEMS/index.html


Use of data assimilation: evaluate models and observations


O-A time series: stability of analysis & “spin-up”

Ozone analysis vs MLS ozone

Operational data + ozone & temperature profiles (MLS) +

ozone column (GOME)

Struthers et al. JGR 2002

46.4 hPa

21.6 hPa

10 hPa

4.64 hPa

“Spin-up”

Stability (desirable property)

12

/04

/97

30

/04

/97

Cal-val: MLS+GOME


Stats for 14-28 Sep 2002

O-F

O: MIPAS ozone obs

From: Geer et al., QJ 2005

Accepted

Rejected

StDev

Bias

Skewness

Kurtosis - 3

Cal-val MIPAS


Motivation: excessive vertical transport in the “old dynamics”

MetO Model

Forecast without chem.- analysis

No chem, no ozone assim, but dynamical DA- analysis

Difference in ozone between experiment and analyses after 7 days, 24th Sept 2002 /ppmm

Forecast with chem.- analysis

Model evaluation

Experiment – Analysis: E-A

Dynamics -> relatively low ozone bias


TOMS column ozone example – 12Z 8th Nov 2003

• TOMS L3 daily gridded product• Analyses interpolated to TOMS grid• 12Z analyses only -> +/- 12 hour colocation time window• Above 5hPa no MOCAGE; BASCOE profile used above.

Independent data

Courtesy Alan Geer


Global mean total column differences

DARC spinup

Courtesy Alan Geer


Summary of all HALOE comparisons – mean Jul-Nov 2003

Courtesy Alan Geer


Intercomparison summary: Analyses; Jul-Nov 2003

Overall, comparison vs independent data indicates that in the stratosphere outside region of the vortex, analyses are broadly consistent & agree with independent data to within 10-20%.

Analyses show largest inconsistencies in following regions:• Mesosphere & stratopause• Region of the “ozone hole”• Tropical tropopause• Troposphere

Problem areas being investigated:http://darc.nerc.ac.uk/asset

Issues: quality of Envisat data;transport scheme; chemistry scheme


• Satellite data have been v. successfully exploited by new DA schemes (4d-var at ECMWF). DA schemes -> introducing additional satellite data that is well characterized improves system.

• Combined availability of new & accurate satellite observations & improvements in models -> improved extraction of information content from these new observations using DA techniques.

• Proliferation of new satellite instruments -> choices on what datasets to use will have to be made (also data selection/thinning).

• Massive investment in data handling (metadata, data management, efficient data dissemination) & monitoring (data evaluation) needed.

• Important that a dialogue is maintained between the data suppliers (space agencies & NWP agencies) & end-users. End-to-end approach.

Use of satellite data: operational/research


1. Operational use of research satellite data (limb geometry -> better vertical resolution; global information on ozone) -> operational use of total column ozone from SCIAMACHY by ECMWF; stratospheric H2O under study

2. Assimilation of novel species: aerosols -> many from research satellites

3. Assimilation of limb radiances by operational centres and research institutions -> fast and accurate radiative models (progress more advanced in the IR) -> research satellites with limb geometry

4. Chemical forecasts -> tropospheric pollution (which model?) -> research satellites to penetrate into troposphere (IASI?); combine with in situ data (observation networks)

The future

Page 1 COST/ESF School: UTLS, Cargese, 3 – 15 October 2005 DA 11: Research satellites and data assimilation Author: W.A. Lahoz Data Assimilation Research.

Documents