BASICS Flashcards
Q 3.. Type of observations
- GTS data (Conventional observations) : SYNOP, SHIP, BUOY, SYNOR, TEMP, PILOT, AIREP, AMDAR, ACAR, SATEM, SATOB).
- Non Conventional Data (Data not transmitted through GTS, and private in general) : RADAR, TOVS, SSMI, …etc
- Old forecast valid at the time of this initial situation (used to compensate data in regions devoid of observations).
Post processing
- post processing calculates a large number of diagnostic fields
- post-processing applications are very heavy but mostly serial
- (the sun grid: most of the servers are dedicated for post processing and graphics creation)
- Post-processing fields is a part of the models operational suites
- Post processing techniques such as model output statistics (MOS) have been developed to improve the handling of errors in numerical predictions
Model operational suit in NWP
- WRF/U.A.E operational suit uses mainly GRADS and NCL to produce graphics
- The operational suite of a model is a set of programs developed to automate all its execution steps from the gathering of data until the production of outputs.
- Will perform all tasks needed to produce the forecast with no further manual intervention. There are, however, facilities to manually override tasks should any problems occur. The operational suite embraces all the individual tasks that are required to produce a forecast. Most of the software within the operational suite has been written in-house. The suite itself is controlled by what is known as the suite control system (SCS). The SCS can be used to select which tasks are run, how they are run and when they are run.
- It is based on Unix/Linux/Perl Command Languages.
- It requires an HA (High Availability) and HPC (High Performance Computing) computing machine.
- It must be permanently monitored by specialized operators.
Models operational suites:
- Gathering and processing observations.
- Getting LBC data
- Getting / Creating INIT data.
- Integrating the forecast model in time.
- Post-processing fields.
- Creating graphics and portable files.
- Dispatching data to different users
- Checking the quality of the forecasts
Weather forecast range from
24 hours to 7 days
Numerical Weather Prediction (NWP)
ois the part of Meteorology Science which is dealing with modeling of the atmospheric conditions and trying to predict these conditions in the near future.
oIt is among the most expensive technologies in term of CPU consumption.
oAt GHQ Meteorological Department, It is based on three specific models:
- WRF (Weather Research & Forecasting) model for weather parameters prediction
- WAM (Wave Model) for sea state forecast
- WRF-CHEM for pollutants and dust transport
NWP consists in :
- Subdividing a chosen geographic 3D area in thousands (or millions) of little cubes (boxes)
- Gathering all current (and past) actual information about atmosphere and ocean : all types of observational data.
- Affecting one value for the main meteorological parameters (Pressure, Temperature, Wind, Humidity) in each cube.
- Calculating through complex equations (momentum, thermodynamics, physics) the modifications affecting these meteorological parameters in time.
- Presenting the predicted parameters values in a comprehensible standard format (charts, meteograms, soundings, …etc).
Horizontal resolution
- ◦When resolution is fine enough (< 10km), small scale phenomena (thunder storms, convective cells, sea breeze, local sand storms… etc) could be well depicted.
- ◦Otherwise, only large scale phenomena (systems movement, large scale precipitations, Jet Stream evolving, … etc) could be well simulated.
Vertical resolution.
◦Fine near the surface, weak upward.
◦When the number of vertical levels is large enough, convective circulation is better taken into account by the models.
Increasing horizontal resolution by a factor of two (2) increases
the number of “cubes”, and implies computation machines 8 times faster.
The horizontal resolution of an NWP model is directly related to
the size of the weather feature it can simulate.
The horizontal resolution of an NWP model is directly related to the size of the weather feature it can simulate.
The resolution is related to
either the spacing between grid points for grid-point models or
the number of waves used to represent weather data for spectral models.
Models Grids Vertical Resolution
- Vertical resolution must be quite fine (on the order of a few millibars) near the earth’s surface.
- An increase in resolution is necessary near and below the tropo-pause to predict the jet stream accurately.
- Different numerical models use a variety of vertical coordinates types to represent atmospheric layers.
Hydrostatic approach
assume hydrostatic equilibrium, in which the downward weight of the atmosphere balances the upward-directed pressure gradient.
Non-hydrostatic
processes and their effect become important when the length of a feature is approximately equal to its height.
◦Convective storms.
◦Gust fronts.
◦Gravity waves (mountain waves and turbulence)
Necessity of physics parameterization
Some atmospheric phenomena need to be parameterized in order to take into account their impact, because:
- Computers are not yet powerful enough to directly treat them.
- They are often not understood well enough to be represented by an equation.
- Their effects profoundly impact model fields and are crucial to creating realistic forecasts
Global models
- resolve atmospheric equations on the whole glob.
- They can not use very fine resolution because of computers limitations.
- Because of their weak resolution, they can not detect small scale phenomena.
The most popular global models are :
- ◦ECMWF/IFS (partially public and received on MDD) : http://www.ecmwf.int.
- ◦NCEP/GFS (completely public) : http://www.ncep.noaa.gov.
- ◦Météo-France/ARPEGE (not available on the net).
- ◦German DWD global model.
- ◦METOFFICE/UKMO Unified Model.
- ◦Japan Meteorological Agency JMA Global Model.
Global models are used to
forecast general synoptic circulation and to provide Initial and Lateral Boundary Data for Limited Area models.
Limited Area Models (LAM).
- They resolve the atmospheric equations on regional or local limited area domains.
- They can use very high resolution (100m to 50km) and more vertical levels. They can catch very small phenomena.
Limited Area Models (LAM).
They resolve the atmospheric equations on regional or local limited area domains.
They can use very high resolution (100m to 50km) and more vertical levels. They can catch very small phenomena.
They are adequate to incorporate complex physics parameterizations (especially microphysics, deep/shallow convection and boundary layer turbulence).
They can run on small to medium computers (normal PCs, workstations, Servers, Clusters)
Limited Area Models (LAM).
They are obliged to get
LBC and Initial data from global models
The most popular LAMs are:
- ◦WRF / NMM (Weather Research & Forecasting) developed by an international community and maintained at NCAR (USA)
- ◦ALADIN (Private Consortium guided by Meteo-France : 15 European and north African countries)
- ◦COSMO / LM (COnsortium for Small scale MOdeling guided by DWD).
- ◦HIRLAM (Private Consortium : Scandinavian countries and Spain).
- ◦COAMPS, RAMS, RUC, …etc.
- ◦ETA (used in more than 50 Centers and Universities).
- ◦MM5 (AFWA and more than 20 centers over the world).
U.A.E. WRF
2km
Initial Data.
The actual situation used by the model to start integrate equations.
initial data is created by techniques called
data assimilation
The information used to create initial data are:
◦GTS data (Conventional observations) : SYNOP, SHIP, BUOY, SYNOR, TEMP, PILOT, AIREP, AMDAR, ACAR, SATEM, SATOB).
◦Non Conventional Data (Data not transmitted through GTS, and private in general) : RADAR, TOVS, SSMI, …etc
◦Old forecast valid at the time of this initial situation (used to compensate data in regions devoid of observations).
The process of initial data creation (analysis and data assimilation) is
more complicate than the forecast model itself, and more consumer in term of CPU time.
Lateral Boundary Conditions data
are used in the LAM models for the following reasons:
- ◦To be able to compute derivatives at the model borders.
- ◦To know what is likely to penetrate the domain of interest from outside.
- ◦To avoid noisy fields at the borders.
LBC data come from
global or regional models including the LAM domain.
WRF/U.A.E. can use LBC either from
NCEP/WAFS or from NCEP/GFS.
The operational suite of a model is a
set of programs developed to automate all its execution steps from the gathering of data until the production of outputs.
The operational suite of a model is a set of programs developed to automate all its execution steps from the gathering of data until the production of outputs.
It is based on
Unix/Linux/Perl Command Languages.
The operational suite of a model is a set of programs developed to automate all its execution steps from the gathering of data until the production of outputs.
It is based on Unix/Linux/Perl Command Languages.
It requires
an HA (High Availability) and HPC (High Performance Computing) computing machine
Output products.
The raw outputs concern only
only the prognostic variables (Pressure, Temperature, Wind, Humidity).
Post-processing
calculates a large number of diagnostic fields.
The direct raw model outputs could be in a format
not understandable by the forecasters (spectral, gridded binary, …etc).
The direct raw model outputs could be in a format not understandable by the forecasters (spectral, gridded binary, …etc).
They must be
transformed to comprehensible standard charts by a graphics procedure.
WMO recommends the use of GRIB format of output products for the following reasons :
- ◦GRIB files are auto documented.
- ◦GRIB files are portable on all kind of platforms.
- ◦GRIB data are coded in compressed binary form, their size is optimized.
- ◦GRIB data are very easy to transform into graphics charts. They can be ingested by most available graphics software.
- ◦GRIB data could be exchanged very easily.
NWP charts
consist in visualizing model prognostic and diagnostic variables in fields form.
One chart can contain more than one field (Height, Temperature, Wind for example).
NWP charts consist in visualizing model prognostic and diagnostic variables in fields form.
One chart can contain more than one field (Height, Temperature, Wind for example).
The charts can be produced for relevant levels:
- ◦Surface (wind, Gust, temperature, pressure, humidity, Dust load)
- ◦850mb (wind, temperature, height)
- ◦700mb (wind, temperature, height, humidity)
- ◦500mb (wind, temperature, height, vorticity)
- ◦250mb (wind, height)
- ◦Level less variables (Clouds, Visibility, Cape, Cin, Mocon, Severe weather indexes)
- ◦Some characteristic fields (jet stream level, tropopause , iso-zero temperature, significant wave height and direction, ..etc)
Meteograms
- They correspond to time series of some relevant surface parameters at a given location.
- They are very useful and give a continuous (in time) overview of the meteorological situation at a given location.
Meteograms
They show generally
10m wind speed and direction, 2m temperature and relative humidity, mean sea level pressure, hourly accumulated rainfall, convection indexes, visibility …etc.
Predicted (T-Phi gram)
- They correspond to the forecasted vertical profile of temperature, dew point temperature and wind above one given location.
- Their quality depends on the quality of the generating model.
- They give an idea about situation of the air mass on the vertical. They are useful in depicting the occurrence of some phenomena like convection, fog, ..etc.
Objective Verification
- To monitor forecast quality - how accurate are the forecasts and are they improving over time?
- To improve forecast quality - the first step toward getting better is discovering what you’re doing wrong.
- To compare the quality of different forecast systems - to what extent does one forecast system give better forecasts than another, and in what ways is that system better?
Driven Models
- Sea/Ocean/Coastal Wave models (WAM, WaveWatch, Swan, ..etc).
- Dust and Aerosol Transport models (DREAM, CARMA-DUST, CHIMERE-DUST, …etc).
- Oil spill models.
- Transport models (pollution, volcanic ash, nuclear and biological pollutants transport, …etc)
WRF model HPC Characteristics
two components:
- oData assimilation to prepare an initial state for the forecast
- oPrediction model
oIt is run four times a day at:
- 00:00 and 12:00 to 168 hours forecasts
- 06:00 and 18:00 to provide 48 hours forecasts.
The operational configuration is covering
othree nested domains with increasing horizontal resolution
oThe grid of the model is constituted of:
o4752000 grid points in domain d01
o8694000 grid points in domain d02
o3937860 grid points in domain d03
As a result, a model run for 7 days forecast takes
takes 2 hours (all optimizations are ON, and the most consuming physics chosen)
The WRF model is running over
three two-way nested domains d01, d02 and d03 with respectively increasing horizontal resolution.
1.The WRF model is running over three two-way nested domains d01, d02 and d03 with respectively increasing horizontal resolution.
In d03 the horizontal resolution is about
2 km
The U.A.E/WRF is performing its own
data assimilation using variational techniques.
High Performance Computing (HPC)
- workstations and servers: low to medium performance (cheap prices)
- clusters: medium to high performance (acceptatble prices)
- mainframes: high performance (expensive)