WMO Workshop of Global Producers of Seasonal to Interannual

advertisement
WORLD METEOROLOGICAL ORGANIZATION
GPC-SIF/Doc. 5.14
(30.I.2003)
____________________
_________
WORKSHOP OF GLOBAL PRODUCERS OF SEASONAL
TO INTERANNUAL FORECASTS
GENEVA, 10-13 FEBRUARY 2003
ITEM: 5.14
Original: ENGLISH
Met Office issued SI products - current status and plans
(Submitted by Dr Mike Davey)
GPC-SIF/Doc. 5.14
Met Office issued SI products - current status and plans
Several types of seasonal-to-interannual (SI) forecast products are currently made
available by the Met Office via the Met Office website at :
http://www.metoffice.com/research/seasonal/
Most of our SI products are openly available to the public. From 1998, some
password-protected webpages with specialised information have been specifically
available to NMSs and organisations such as FAO. (The password is provided on
request.) See : http://www.metoffice.com/research/seasonal/nms/
The global SI products are based on dynamical model ensemble forecasts. Up
until March 2003, these products are based on weekly 9-ensemble integrations of a
2-tier AGCM; from March 2003 the products will be based on monthly 40-ensemble
integrations of a coupled ocean-atmosphere GCM.
Current state (January 2003):
Public seasonal global-gridded products (provided since May 2002):
Graphical maps (gif format) of probabilities for above normal temperature (T2m)
and precipitation (Pr) anomalies for 3-month periods for months 1-3 and 2-4 from the
forecast start time. Updated monthly, based on the most recent 3 weeks of weekly 9ensemble AGCM forecasts.
NMS-only seasonal global-gridded products:
 Graphical probability maps for above normal T850 and Pr anomalies for months 13 and 2-4, updated weekly, based on the most recent 9-ensemble AGCM forecast.
 Additional maps of deterministic (ensemble-mean) T850 and Pr anomalies for
months 1-3 and 2-4 ahead are also provided.
Spatial resolution:
Atmospheric GCM grid (2.5 latitude x 3.75 longitude)
Domain:
Global. For ease of use, regional sub-maps are also provided at the same
resolution.
Definition of anomaly:
Relative to model climatology, based on retrospective forecast data (currently
1987-1998).
Calibration:
Products are based on the forecast ensemble only (e.g. retrospective forecasts
and model output statistics are not used to adjust forecast probabilities).
Issuance time:
For NMS webpages:
weekly, as soon as the products have been generated
(monthly, mid-month when we switch to the CGCM in March 2003).
For public webpages:
in the first week of each month (mid-month from March 2003)
GPC-SIF/Doc. 5.14, p. 2
Skill information:
For NMS webpages:
Optional skill masks are provided. For probability graphs, the mask threshold is
a ROC score of 0.6 at each gridpoint (using 21 years of retrospective forecasts).
For deterministic graphs, the mask threshold is based on significance of anomaly
correlations from Monte Carlo estimates: the threshold roughly corresponds to a
correlation of 0.4.
For public webpages:
For public use, skill information is provided in the form of maps of Heidke score
for each product, based on retrospective forecasts. This score was selected as it
is related to 'percent correct' and is more readily understandable to non-experts
than other scores. (See the website for details.)
Extensive skill information from the same forecast system (but with ERA40
re-analysis atmospheric initial conditions) is available on the DEMETER
project website, for both our AGCM and CGCM systems.
(http://www.ecmwf.int/research/demeter/)
Confidence:
No specific confidence information is provided. (For a separate commercial
1-month range UK product, confidence information based on bin spread is provided.)
Dissemination:
Only via the Met Office website. (A fax service to NMSs has been discontinued as
it is no longer required.) There are links to various product and information pages.
For the global products, pull-down menus provide access to various forecast dates,
ranges, variables, and geographic regions.
Public webpages are freely available; NMS webpages are password protected (to
provide NMS access to additional forecast information, and to avoid misinterpretation of technical information by non-expert users)
Verification:
Past issued forecast products (back to 1998) are available on the website, but no
verification of individual forecasts is provided.
Operational capability:
Currently the weekly AGCM 9-ensemble and the monthly CGCM 40-ensemble are
non-operational. From March 2003, the CGCM system and the processing to
generate web-page products are expected to have operational status.
Disclaimer:
On the public and NMS webpages we state 'These forecasts are considered
experimental. The Met Office accepts no responsibility for actions taken on the basis
of these forecasts.'
Conditions of use/access:
There are currently no specific conditions of use regarding the public or NMS
webpages. A password is needed to access the NMS webpages. General conditions
for the Met Office website (copyright etc.) are at :
http://www.metoffice.com/corporate/legal/
Additional information:
Descriptions of models and products are provided on the website.
GPC-SIF/Doc. 5.14, p. 3
Other seasonal forecast products:
More detailed forecasts for some specific regions and seasons (e.g. Sahel
rainfall); winter NAO index; global-mean annual-mean T
(see http://www.metoffice.com/research/seasonal/) Some regional products are
actively disseminated by email to NMSs in the relevant regions. Most are published
in the Experimental Long Lead Forecast Bulletin (see http://www.iges.org/ellfb/).
There are also commercial products that are not public: e.g. to monthly range for the
UK, and select site-specific forecasts.
Foreseen public SI products:




T2m, Pr, SST above/below normal probabilities for months 2-4, 3-5, 4-6 ahead;
Nino3, Nino3.4, Nino4 SST anomaly ensemble plumes to 6 months ahead;
single and multi-model (combined Met Office + ECMWF) products;
possibly also:
tercile probabilities as well as above/below for T2m and Pr and SST; tropical
cyclone seasonal activity; probabilities of T2m and Pr 'extremes'; (content will
depend to some extent on what other centres make publically available)
Foreseen NMS-only products:
As far as technically possible, and provided our products have better-than-chance
skill, we will aim to provide the main agreed common products based on the draft list
prepared by the ET ILRF (see doc3 appendix 1). We will also work with our multimodel partners (currently ECMWF) to develop multi-model versions of the products.
The main fields in the list are:
monthly and 3-month-average gridded global fields of T2m, SST, Pr, Z500, T850,
MSLP ensemble mean and a spread measure (e.g. standard deviation? outer and
tercile boundaries?) for anomalies from model climate using an agreed reference
period
monthly and 3-month-average tercile probabilities for T2m, SST, Pr, Z500, T850,
MSLP
Nino3, Nino3.4, Nino4 ensemble plumes of monthly SST anomalies to 6-month lead
SOI and NAO index ensemble plumes to 6-month lead, using agreed definitions of
the indices.
NB products to be provided in numerical form as well as graphical. We note that the
recommended format for numerical data is GRIB2 for fields.
Other requirements in the ET ILRF list include forecasts of degree-days, frequencies
of events (e.g. spells, frosts), monsoon onset, tropical cyclone activity. Of these, we
will first aim to provide tropical cyclone activity, with others potentially following as
demand is determined and prioritised in more detail.
Note: gridded individual ensemble fields, and daily data, will not be provided
except where some particular future applications require such information, such as
for forecasts of regional monsoon onset.
Verification: to follow ET SVS guidelines for SI products, noting the possible use of
further verification measures according to demand.
GPC-SIF/Doc. 5.14, p. 4
Issues
Several issues were noted by ET ILRF regarding the common provision to NMSs
by global producers. Clearly, agreement is needed on various technical details to
facilitate the usefulness and practicality of the proposed SI forecast provision, but
there will be limitations on how 'common' such provision can be in practice. Some
remarks:
 a common grid is feasible
 common forecast periods (calendar months and 3-month averages) should be
used
 the actual start time and issue time of diverse forecasts may not matter, as long as
the forecast product period is common and forecasts are issued promptly.
 the use of common reference climatologies is very desirable (normals, terce
categories etc.), but may be difficult because different providers may have
different model reference climatologies for model forecast anomalies.
- would a recent 15-year climatology be adequate? (e.g. 1986-2000) (there are
problems with category definitions if the reference period is too short)
- there may be advantages in using different reference climatologies for
different regions
To help reduce this problem, should forecasters provide actuals (as well as
anomalies) for forecasts and retrospective forecast datasets, to enable NMSs to
construct their own tailored products for particular reference periods?
 common skill measures: ET SVS recommendations provide a starting point, but
this aspect requires further consideration. Again, provision of retrospective
forecast data would allow extension and customisation of skill measures.
 common format: are all participating LRF providers and receivers able to use
GRIB2?
Should NetCDF fields also be provided? It would also be convenient to agree a
common ASCII file format for plumes.
 graphical data: it would be very useful to have common graphical formats, e.g.
colour conventions, intervals, file type?
Should there be a lead centre (cf verification) for a common display of some key
products?
 it would be useful to provide common retrospective forecast data where possible
to facilitate development of products and applications.
 should there be a common point-of-access website with links and information
about forecasts and contributing centres?
 there should be a process to obtain feedback from customer NMSs regarding
content, accessibility, usefulness, other requirements etc.
It may be advantageous to arrange a trial 'exchange' of agreed SI products among
global
producers
before
general
release
to
NMSs,
to
test
feasability/content/practicality etc.
It would be advantageous to consider holding further workshop(s) to review
progress, requirements, and possible extensions of the 'common' product list.
Download