DB1

1
Weather Prediction and the use of weather
data The European Centre forMedium-Range
Weather Forecasts ECMWF
David Burridge with help from A. Simmons, G.
Kelly, J-N Thepaut, D. Marbouty, A. Thorpe, A.
Lorenc, T. Palmer
2
Annual means of the forecast range at which
the anomaly correlation of 500 hPa forecasts
first reaches The 60, 655, 70, 80, 85 and
the 95 levels for the Northern hemisphere (left
panel) and the Southern hemisphere (right panel).
3
Cevennes floods 9 September 2002

• Peaks beyond 500mm/day (670mm at Anduze)
• A large area with more than 200mm in 24h
• Early warnings were available

4
(No Transcript)
5
NOAA16 9 Sept. 2002 1220UTC
6
Le Gardon à Collias
11 Sept
9 Sept 2002
7
Forecast consistencyfive consecutive T511
forecasts valid at the same time
8
T511 precipitation forecastscumulated over 24h
and valid for 20020908 18z to 20020909 18z
9
EPS precipitation probabilitiesvalid for TP
cumulated over 24h 20020908 18z to 20020909 18z
Sa 7/9
Fr 6/9
Th 5/9
We 4/9
Tu 3/9
CTRL
Prob. RRRgt50mm
Prob. (area) RRRgt50mm
10
S-France flood Summary

• A stationary convective system generates huge
  amounts of precipitation in a few hours ( 600
  mm/24h in some places) over a limited area.
• The medium-range deterministic forecast is very
  consistent. The synoptic pattern is correctly
  forecasted already 5 days in advance. Good
  ability of the model to simulate the basic
  ingredients for severe convection.
• EPS supports the deterministic forecast showing
  unusual high probability values. EFI index gives
  early warning but spread a large area.

11
Precipitation accumulated over the Po' river
catchment area predicted by the 51 EPS members
started on 2 November 1994 and run for 10 days
(grey lines, cyan for the control) and the 6
nested LEPS members (violet lines, green for the
control) started on 3 November and run for 3
days. The red line depicts the area-average over
the river-catchment computed from synoptic
observations.
12
Global Weather Prediction A Triumph for science
and computing Today we have global operational
forecasts with useful skill varying between 7 and
8 days
During the satellite data era there has been a
gain of more than 2 days and 4 day gain since
Miyakodas experimental forecasts which he
carried out in the late 1960s (published in
1972) the hemispheric skill that can be
achieved with the 1950s observing system is
around 6 days truly a triumph for science and
computing
13
Number of used observational data per 12 UTC
cycle in ECMWF's operational assimilation system,
1997-2002.
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
A
F1
Four-day forecasts for Sunday 29 December 2002 A
analysis F1 full system F2 - without
satellite data
F2
18
Real-time data from new research satellites

ENVISAT from ESA, flying a range of instruments
to measure ocean waves and ozone - LAUNCHED
AQUA from NASA, flying AIRS the first of a new
generation of high-resolution infrared
instruments - LAUNCHED
19
Major enhancements of the operational satellite
observing system
SEVIRI on MSG, the new generation of EUMETSAT
geostationary satellites - 2003
IASI and other instruments on the first
generation of EUMETSAT polar-orbiting
satellites - LATE 2005
20
(No Transcript)
21
(No Transcript)
22
THORPEX?
A combination of inaccurate initial conditions
and errors in model formulation contribute to

• significant failures in forecasting high-impact
  weather
• the inability to extend the range of skillful
  predictions beyond 7 days
• poor prediction of tropical-extratropical
  interactions
• inadequate skill in predicting specific aspects
  of mesoscale weather, e.g., precipitation

23
Rossby Wave Dispersion 6-28 November
2002 Time/Long. Diagram 250-mb Meridional Wind
35-60 N
6 Nov 9 Nov 12 Nov 15 Nov 18 Nov 21
Nov 24 Nov 27 Nov
Cyclogenesis
India/T.C.
Tornadoes
Oil Tanker
Cyclogenesis
Alps Flood Foehn Wind
Snow/Ice Storm
Cyclogenesis
Cold-Air
Cyclogenesis

Moroccan Flood
Cold-Air
Alps flood




UK
Japan
UK
Cal.
24
THORPEX Objectives

• advance basic knowledge of global-to-regional
  influences on the predictability of high-impact
  weather
• contribute to the development of a
  dynamically-controlled interactive forecast
  system
• consider short-range (0-3 days), medium-range
  (3-7 days) and week-two forecast problems the
  middle ground between weather and climate
• develop methods for assessing the economic and
  societal value of improved weather forecast skill

25
Interactive Forecast System

• optimal design of fixed and adaptive observing
  system
• use of targeting techniques for observing and
  data assimilation
• two-way flow of information between model and
  observing system

26
Driftsonde System
Hourly data at flight level
High-resolution vertical profiles of Temperature,
Wind, Moisture, Pressure
Ground Station
27
An example of possible Driftsonde deployment from
Japan
28
Numerical Weather Prediction

• Errors in initial conditions and forecast model
  dominate.
• NWP centres concentrate on
• sophisticated data assimilation schemes,
• built round high-resolution models.
• Error growth is important
• Sensitivity to initial conditions,
• Chaotic error growth - limits to predictability,
• THORpex international research programme.

29
Assimilation of cloud precipitation

• Weather systems (In order of priority
  possibility?)
• 1. Resolved, dynamically-forced, vertical motion
  (eg depressions)
• 2. Horizontally (but not vertically) resolved
  forcings and feedback on parametrized
  boundary-layer processes (eg Strato-Cu)
• 3. Resolved forcings and feedback on
  parametrized deep convection (eg MCS, embedded
  convection in fronts)
• 4. Scattered convection.

30
Sources of errors in model predictions

• Initial conditions
• Forecast model
• External forcing
• Errors in all these accumulate,
• and in some circumstances grow,
• to limit the predictability.

31
Seasonal Prediction

• Chaotic growth ? Detailed initial conditions of
  the atmosphere are not important.
• Coupled systems have components without this
  rapid growth, e.g. ocean, land surface.
• Seasonal forecasts do depend on initial
  conditions for these components.
• Data assimilation for coupled models.

32
Niño-3 SST hindcasts
ACC 1987 1994 (32 start dates)
DEMETER multi-model ECMWF CNRM UKMO
LODYC Persistence
33
Real-time dynamical multi-model seasonal forecast
Risk of wet / dry winter 2002/03
34
Climate Prediction

• Changes in external forcing dominate.
• Model validation and development is crucial
• Data assimilation is important for this.

35
Coupled Data Assimilation

• Skill in seasonal predictions comes from slowly
  evolving sea and land surface.
• We do not have enough observations for these the
  atmosphere is relatively better known.
• Need atmosphere in coupled Data Assimilation,
  even though atmosphere Initial Conditions are not
  important.
• But dont need fully coupled Data Assimilation
• the insertion of obs into each model is
  independent
• with coupling, via fluxes, in the forecast steps.

36
Bias

• If a model has bias, it is impossible for Data
  Assimilation to give consistent bias-free fields
  and fluxes.
• Exacerbated by biased obs Data Assimilation
  methods.
• Coupled Data Assimilation must give higher
  priority to coping with bias than does NWP Data
  Assimilation.

37
Summary

• Non-NWP Data Assimilation has different
  priorities
• Biases are unavoidable, and must be allowed for
  in coupled Data Assimilation
• Re-analyses useful to validate seasonal
  prediction systems

38
Prospects for NWP and Seasonal Forecasting

• Comprehensive Earth-System forecast facilities
• Comprehensive Earth-System assimilation
  facilities
• Strong Satellite Capabilities
• Improving ground-based observing systems with
  experimental
• systems to provide guidance
• Robust and efficient Numerical methods
• Thoroughly Validated Parametrizations
• Extensive Ensemble Capabilities
• Strong Computing
• We may confidently expect
• Increasing forecast quality
• Further spin-offs and new products

39
The Weather Community can advance Global
monitoring of the environment in

• SYSTEMS FOR RISK ASSESSMENT
• delivering operational support to weather-related
  risk management (early warning, impact assessment
  and reaction) in sensitive areas for floods
  forest fires oil spills and support for
  humanitarian aid
• GLOBAL ATMOSPHERE MONITORING
• delivering regular assessments of state of the
  atmosphere with particular attention to aerosols,
  ozone, UV radiation and specific pollutants
• GLOBAL OCEAN MONITORING
• in support of seasonal weather predictions,
  global change research, commercial oceanography
  and defence.
• GLOBAL VEGETATION MONITORING
• to assess carbon fluxes to/from the biosphere.

40
Areas for GMES Collaboration Global
Monitoring / Forecasting of Greenhouse Gases

• 2.1 Map the seasonal variations of total
  column amounts of Greenhouse Gases
• 2.2 Model and assimilate ocean colour data, to
  estimate ocean carbon uptake.

2.3 Model and assimilate global aerosol
information (to improve weather forecasts the
use of ocean colour data)
2.4 Model and assimilate information on the
Land Biosphere and carbon cycle.
41
Global Monitoring / Forecasting of Reactive
Gases The Chemical Weather Forecast

• Current operational ozone monitoring capability
  is a good basis for developing a global
  capability to monitor reactive gases and
  associated aerosols
• 3.1 Integrate chemical modules with weather
  models, to provide global assimilation
  forecasts of the distributions of
• ozone and its precursors
• sulphate aerosol
• other aerosol
• The global models can drive regional chemistry
  / air quality models.

Ozone CO
42
Environmental Predictions

• Environmental Predictions can be very effective
  if environmental models are driven by
• Global Earth-System Assimilations
• Global Deterministic Ensemble Forecasts
• Inverse Model for Carbon attribution
• Atmosphere Regional
  Weather Model
• Chemical Aerosol Transport Model
• Trajectory Model
• Hydrological Model
• Land Crop Model
• Fire Model
• Disease Model
• Ocean Oil-spill Model
• Storm Surge Model
• Coastal Zone Model
• Regional Ocean Model

43
(No Transcript)
44
(No Transcript)