Transcript [09] Model Assimilation

Model Assimilation
Dr Mark Cresswell
69EG6517 – Impacts & Models of Climate Change
Lecture Topics
• What is assimilation?
•
•
•
•
•
•
Reanalysis – what is it and why is it?
Meteorological stations
Ship and buoy instruments
Radiosonde, dropsonde and Aireps
Climate model assimilation
Spin-up
What is Assimilation?
Assimilation is a process which involves the collation and
analysis of global meteorological observations into a digital
georeferenced format
Once global weather data is stored electronically it is used to
“teach” a climate model the current state of the world ocean,
atmosphere and land conditions
Once the model “learns” the current global state of the
weather it can calculate future change based on these initial
conditions
Reanalysis: form and function
• In order to efficiently calculate forecast
fields for each vertical level and global
gridpoint
• Climate models don’t see individual points
(like meteorological stations) – they see
grids of regularly spaced squares
• Point data (typically derived from
meteorological stations) must be
transformed into a smooth gridded surface
Reanalysis: form and function
• Once all observations are known (station,
ship, buoy, aircraft, satellite, balloon and
radiosonde) the data are blended to generate
a smooth grid
• This observational grid is known as
reanalysis
• Reanalysis represents the most objective
record of what the atmosphere and oceans
were like for a specific date and time
Reanalysis: form and function
Reanalysis fields are
generated for different
pressure levels…from
surface to 31 or so levels up
to the top of the atmosphere
Reanalysis: form and function
• Reanalysis may be used as a “gold
standard” against which model hindcasts
can be compared (to assess model skill,
reliability and bias)
• Reanalysis may also be used to spin-up
global climate change models
• European modellers make use of the
ECMWF reanalysis covering 15 years
(1979-1993) known as ERA-15. We now
have ERA-40 (covering the last 40 years)
Meteorological Stations
• Around the world are a network of stations
where standard meteorological observations
are made – coded as SYNOP
• Observations are recorded at main synoptic
hours only for minor stations or main and
intermediate hours for major stations
• Major stations are manned
• Minor stations may make use of automatic
instruments – recording to magnetic tape or
relaying data via radio
Meteorological Stations
00
03
06
09
12
15
18
21
MAIN synoptic hours shown in RED
INTERMEDIATE synoptic hours shown in BLACK
Data is relayed to the World Meteorological
Organization in Geneva, Switzerland via the GTS:
Global
Telecommunication
System
Currently there are about 10,000 stations globally
(WMO, 2002)
Meteorological Stations
00
03
06
09
12
15
18
21
Data recorded are:
•Air pressure
•Temperature
•Relative humidity
•Precipitation
•Visibility
•Cloud parameters
Meteorological Stations
Meteorological Stations
Typical Meteorological Station Layout
Ship and Buoy
As well as land-based observations, the GTS collects measurements
of meteorological conditions over the sea – coded as SHIP
These data are collected by:
•Ships
•Moored buoys
•Drifting buoys
Measurements made by ships recruited under the WMO Voluntary
Observing Ship Programme. Drifter data is coded as DRIFTER
Ship and Buoy
ABOVE: moored buoy
LEFT: drifting buoy
Upper Air Measurements
As well as land and oceanic surface measurements the GTS collates
upper air data.
Radiosondes – instruments attached to ascent balloons are used to
generate vertical profiles of the atmosphere
Aircraft observations are reported from aircraft as well as specific
pilot reports. Data from aerodromes and airfields comprise the
METAR encoded data
Rockets and radar are also used for vertical data collection
Upper Air Measurements
LEFT:
radiosonde
LEFT: launch
of radiosonde
balloon
RIGHT:
sounding rocket
Satellite Data
Climate Model Assimilation
Following the collation of ALL available data into a uniform gridded
reanalysis dataset, this is “read in” by the climate model
The computer’s memory is used to store the values for each field
(temperature, pressure, windspeed etc) for each level (surface,
850hPa, 500hPa etc)
When ALL of this data is stored in the computer’s memory and the
model initialises all of the variables it uses for each forecast field,
the data is said to have been assimilated.
Climate Model Assimilation
Specific assimilation schemes may be based on a single “snapshot”
of conditions – known as 3D-Variational Assimilation…or it may be
based on several days of reanalysis so uses a 4D-Variational
Assimilation scheme (the 4th dimension being time!)
4D-Var. is a standard data assimilation method and is essentially a
means of estimating the initial conditions of a model by optimising
the fit between real observations and predicted 'observations' found
from a projection of the model forward in time
After assimilation is complete, the model can be run forward in time
to generate the actual forecast
Spin-up
When global weather observations have been successfully
assimilated into a climate model – and it is initialised, the start of the
forecast run (the initial period of time being forecast) may produce
erratic results
The initial period of a forecast run (say the first few days of a 1
month forecast run) is known as the spin-up period. This period
varies according to the model used and the period being forecast
More specifically, Spin-up is the time taken for a model to reach a
state of statistical equilibrium under the applied forcing
Spin-up
Cold start: Usually occurs when a model is first initialised and needs
to be spun up. For example, if a model is configured in a new
domain, it would need to be started in this manner. A cold start could
be from climatology rather than reanalysis. The model is then run
until a statistical equilibrium is achieved.
Warm start: A warm start is a restart of a model, which is used to
eliminate or reduce the model spin up time. The saved fields from a
recent forecast of the same model can be used to initialise a new
simulation, or continue the previous simulation. The saved fields
may be used as a first guess for an analysis including new data, and
then that field is used to initialise the new forecast.