Use of Satellite Earth Observations in Operational Meteorology 1. The state of EO use in NWP 2. Challenges and current priorities 3. Suggestions for the

Use of Satellite Earth Observations in Operational Meteorology

1. The state of EO use in NWP

2. Challenges and current priorities

3. Suggestions for the next 5-10 years

Globmodel workshop, 12 Sept 2007 - F. Bouttier

1.a. Who are the users ?

the big ones: a few 'large' numerical weather prediction centres, with own satellite assimilation teams & good contact with data providers (e.g. ECMWF)

many smaller institutes, with regional models = growing modelling capability, but tiny R&D teams : de facto excluded from complex organisations.

(some universities & research centres increasingly work like small NWP centres)


The big ones: global forecast scores

time

2002 2003 2004 20062005

1.b. What are their objectives ?

to improve the overall forecast quality: cf. scores to improve forecasts of high impact events to provide numerical input (or coupled model systems) for

non-meteorological environmental models: ocean, chemistry, hydrology, etc. 'Earth model'

Subject to unpredictable trends in institutional strategies, fashion and funding opportunitites...


1.c. How is EO data used in NWP ?

as direct input to data assimilation (DA) as ancillary data for verification & validation for research


EO use in direct assimilation

Most visible application, but not the only one - and it's not limited to global models.

a major factor of improvement over the past 20 years - in synergy with increased CPU power & new algorithms

current DA algorithms (3D-Var and 4D-Var) were primarily developed for EO data assimilation

impact of 'old' radiances (AMSU, HIRS): large, and still growing in quality & coverage

DA of new instruments (GPS, scatterometres, spectrometres...) will take several years to develop fully.


Example of beneficial satellitedata assimilation: SEVIRI radiances into ALADIN

regional model

forecast range (hours)

precipitationforecast errors

(top: rms,bottom:bias)

without SEVIRI

with SEVIRI

Regional impact of EO is severely limited by their coverage (which is global-model oriented)

EO use as ancillary NWP data these issues are neither 'pure' NWP nor adequately solved

by other communities (yet) they often suffer from chronic understaffing and lack of

structuration (with lots of uncoordinated initiatives)

= orphan problems at many NWP centres: sea surface, sea ice, sea roughness lake temperature/ice snow on ground sources of sand/dust soil moisture, soil temperature/ice, soil/vegetation properties surface properties for radiation computations...


SSM/I, 19V-19H:SSM/I, 19V-19H:

November 2006November 2006

Mapping of microwave surface emissivityfor radiance data assimilation

Importance of soil moisture:T forecast from 2 models with slightly different soil

analyses

Time (hours over one day)

truth

T(C)

EO use for verification & validation

needed in quasi real time (for application to current NWP system versions)

verification: bias correction, quality control, score computation, e.g.: rarity of good precipitation data

validation:

to check & tune some non-assimilable model aspects e.g. radiation balance, cloud cover

as « ground truth » in case studies, e.g. imagery


EO use for NWP research

mostly to improve key physical parametrisations: subgrid convection & turbulence, radiation, microphysics, surface processes.

useful datasets:

instrumented sites (ARM) field experiments (IHOP, COPS, CSIP, RICO, IPY...) research satellites (EOS, Aquatrain, Envisat)

use of EO data is still incomplete, but promising.


Challenges for NWP 1/4

Different players may have different priorities: NWP user: How to sustain the current spectacular

improvement in NWP thanks to EO Data provider: How to improve the use of EO in NWP Meteorologist: What are the key needs of NWP Manager: How to improve cooperation between institutes



How can we sustain the current spectacular improvement in NWP thanks to EO ?

Keep flying existing technologies as long as they do not saturate the DA systems:

well known IR and microwave radiances, scatterometre, AMW winds, GPS, etc.

emphasis on improved coverage, spatial resolution, repeat frequency, programme continuity, instrumental quality, stability of data dissemination.

Not 'sexy', but very valuable - such data can be assimilated nearly as soon as an instrument flies. Coverage & repeat frequency requirements are crucial for new mesoscale NWP DA systems.



How can we improve the use of EO in NWP ?

i.e. can we avoid EO data being underused by NWP ?

Main reasons for NOT using EO data include: short instrument lifespan (actual or expected) lack of visibility on successor missions, and increases in

coverage duplication with other instruments in terms of observed

frequencies/regions/orbits etc) lack of added value (low information content, difficulty

to incorporate into NWP DA software) may require excessive CPU or R&D staffing



The key needs: 3D information on T, wind, Hu everywhere, with emphasis

on dynamically sensitive areas and high impact weather. actual dx,dt resolutions of NWP DA systems: global ~50km

x 6h, regional ~5km x 1h. vertical resolution should be about 300m (!) current T coverage is adequate in upper troposphere,

often poor below (under cloud tops; over land) lack of precise (bias-free) humidity observations wind coverage is very poor (except on sea surface) ancillary data: see above, e.g. soil moisture.


Quikscat limitation in heavy rain (useful anyway)

ALADIN model forecast (24/2/2007)

La Réunion

Quikscat 20m/s wind speed observations

Cyclone as seen by model humidity field

example of sensitive forecast structure(COPS campaign thunderstorms)

Cloudy area

Data-poor area

3. Suggestions: DA

For 'hard' DA of EO data, the SAF-NWP (and ITWG) approach may be the best: it is helpful & cost-effective to have an interface team between data providers and NWP institutes.

The core of DA is still done by NWP R&D teams, which is not affordable to smaller institutes: software interoperability (e.g. obs handling, monitoring) should be improved, perhaps by creating ad hoc teams.

More perennial structuration is needed to facilitate access to ancillary data (esp. data fusion from various sources): e.g. GHRSST-PP is a good approach ? Is it cost-effective to tackle permanent problems using short-lived projects ?


3. Suggestions: programmesFact: long-lived instrument concepts are much better used in

NWP than high-tech, isolated ones. Because it is very difficult to enlarge NWP teams to feed

new instrument technologies into DA systems. New versions of older instruments should focus on

backwards compatibility, (orbit) complementarity, and global long-term continuity.

Innovative instruments are welcome if they can be flight-tested quickly and (if ok) made operational with a long-term perspective.

The “prove there's an impact in NWP and then we'll fund it” approach is very slow and unefficient. Weather obs. show a benefit after they are given good coverage, not before...


3. Suggestions: organisation

The SAF concept has worked very well. The European NWP consortia (COSMO, HIRLAM, AALDIN,

MetOffice) and Eumetnet/SRNWP provide convenient opportunities to reach many NWP institutes and users.

for directly assimilated data, the emphasis should be on spreading common expertise & software for the NWP centres: foster networking, training & interoperable solutions. (avoid excessive meetings & paperwork)

for ancillary data, the data merging can/should be done upstream by specific teams and freely distributed. Activities on a single instrument are of limited usefulness.


3. Suggestions 'from the trenches'according to NWP staff: data from new satellites should be accessible earlier to DA

scientists, even if the data is not perfect. It would speed things up.

discussion with the data providers should be easier and more interactive. (the US-led mail forums are more productive)

instruments simultaneously flown on several satellites can be used quickly because it is faster to accumulate statistics about them.

high opinion of SAF-type work organisation no time to work on DA of many interesting instruments

because they are too unusual.


Conclusions: current NWP trends

Emphasis on 'severe' weather events: growing need for data about cloudy & precipitating areas.

Decreasing funding for R&D teams: only a limited variety of EO data types can be used. (but many satellites are ok)

High CPU & data handling costs: informativity may matter more than data quantity. High potential value of currently poorly observed regions and parameters (e.g. polar AMWs).

Multiplication of smaller NWP institutes = more demand for regional, high resolution & high frequency EO data.


Documents

Use of Satellite Earth Observations in Operational Meteorology 1. The state of EO use in NWP 2. Challenges and current priorities 3. Suggestions for the