Slide 1 GlobMODEL workshop, ESRIN, 12-13 Sept 2007 Assimilation of Earth-observation data into Earth-system models Practical considerations in the transition.

GlobMODEL workshop, ESRIN, 12-13 Sept 2007 Slide 1

Assimilation of Earth-observation data into Earth-system models

Practical considerations in the transition from research to operations

Erik Andersson and Adrian SimmonsEuropean Centre for Medium-Range Weather Forecasts


What do we understand by Research?

Research into the use of Earth-Observation data in Earth- System models can include work on:

– Use of new data in assimilation systems

– Data assimilation techniques

– Assimilating models

– Retrieval algorithms

– Radiative transfer modelling

– Design of new instruments and missions

– …


What do we understand by Operations?

• Routine numerical weather prediction

– For ranges from hours to seasons ahead

• Routine environmental monitoring

– GEMS follow-on project envisages operational running of system in both near-real-time and with a 5- to 6-month delay

• Periodic reanalysis

– ERA-15, ERA-40, ERA-??, …



Operational NWP already makes a very prominent use of EO data

Nu

mb

er

of in

stru

men

ts p

rovi

din

g d

ata


Operational NWP already makes a very prominent use of EO data

Am

ou

nt o

f dat

a u

sed

per

da

y (m

illio

ns)


How has this been achieved?

• Through ECMWF dedicating staff to the use of satellite data

• Through consultants at ECMWF funded by the Space Agencies (especially EUMETSAT and ESA)

• Through cooperation among members of the European Meteorological Infrastructure

– Eumetsat SAFs (NWP, GRAS, Ozone, OSI, …)

– KNMI’s near-real-time supply of retrievals (SCIAMACHY, OMI)

– …

• Through cooperation with ESA on the monitoring and exploitation of data from specific missions

• Through cooperation with weather prediction centres and space agencies outside Europe

• Through general scientific interaction


Implications of the already widespread use of Earth-observation data

• Impact studies based on systems assimilating fewer types of observation may overestimate the impact of a new type of observation

• Good bias correction (inter-calibration) is essential for benefit to be derived from a new type of observation

• Transition from research to operations is most efficient if research is carried out close to operational system


Near-real-time data supply and general preparedness for new missions

• Lack of near-real-time data supply is a barrier to the transition from research to operations for NWP

– MIPAS

– MLS

– (and potentially, in the past, SMOS)

• ADM good example of early preparedness


Impact of MLS ozone profiles on ECMWF 4D-Var ozone analyses (Elias Holm, R. Uni)


ILIAD - Impact of P, T and backscatter ratio on Rayleigh Responses of the ADM-Aeolus mission

Line shapeDepends on

P, T and s

Rayleigh(P,T)

Mie

Transmissionthrough the Double FP

Light transmitted through TA and TB Response (function of fD)

Response curveFunction of P, T and s

Dabas Meteo-France


Emerging Earth-system models

• Earth system models are being developed for climate monitoring (and prediction) and for environmental monitoring / air quality

– Atmosphere

– Ocean

– Land and sea surface

– Biosphere

– Cryosphere

• Current available Earth-model systems are incomplete

• Processes captured by the EO measurements may be missing or inadequately described by the model – hampering data/model fusion


An Advanced Surface Analysis System

SMMR,SSM/I,TMI,AMSR,GRACE,AVHRR,MODIS,MERIS,ASCAT,ERS scat.,SEVIRI,GOES Imager,....


The Current Surface Analysis System


Ongoing/recent research projects – surface analysis

Geoland I (EU)- MODIS LAI(Jarlan et al. 2007)

SMOS DA I (ESA)- SMOS TB

H-SAF (EUMETSAT)-ERS scat (demo, re-analyses)-ASCAT (operations)(Scipal et al. 2007)

SMASS (DFG)- MW TB(Wilker et al. 2007)

NWP-SAF (EUMETSAT)- SMOS TB(Holmes et al. 2007,Drusch et al. 2007)

ELDAS (EU)-MW TB(Seuffert et al. 2003,Seuffert et al. 2004)


Expanding the contribution of the academic sector

Opening up the research systems of operational assimilation centres to much wider use by external research groups would:

• Enable external research to be more relevant and readily applicable

• Require support at both the operational and external research centres, as operational systems tend to be sophisticated and complicated, change over time, and are not necessarily

designed to be friendly to outside users

• Require long-term support, as support-team expertise needs to be built up, and lessons need to be carried over from research

to near-real-time operations, and later to reanalysis

• Train young researchers for future roles at operational centres


Impact of MIPAS limb-radiance assimilation on zonal-mean analyses (Niels Bormann, …, ASSET project)

Humidity (%) Ozone (ppmv)

Pre

ssu

re (

hP

a)


Impact of ESA/KNMI SCIAMACHY total column ozone on ECMWF 4D-Var ozone analyses (Rosana Dragani)


Conclusions – transition from research to operations

In many cases, interpretation of EO data is significantly enhanced through assimilation (fusion) into a comprehensive Earth system model, or components thereof

• Research and development of Data assimilation techniques, Assimilating models, Retrieval algorithms, Radiative transfer modelling, … can be conducted in a distributed manner

• New developments can be integrated into the operational system, after thorough testing and evaluation of its impact

• (Components of) the operational system can be made available for experimentation in the research community – provided the required support teams can be established

• The uptake of EO data can be accelerated by making the data available in (quasi) near-real time, as soon as possible after launch – later/reprocessed data are vital for reanalyses

Slide 1 GlobMODEL workshop, ESRIN, 12-13 Sept 2007 Assimilation of Earth-observation data into Earth-system models Practical considerations in the transition.

Documents

data slide

uni slide

globmodel workshop

operational system slide

datamodel fusion slide

use of new data

s dabas meteofrance

earth system models