ECMWF

ECMWF

Datamodel for the tables in the ECMWF schema.

Data is created by the ECMWF and retrieved from the KNMI ftp site as GRIB files. The original GRIB files can also be retrieved from the flysafe server from :

ECMWF Data Archive

More information about the parameters in these GRIB files can be retrieved from the ECMWF website.

INVARIABLE

The INVARIABLE table contains data that is invariant over time.

LATITUDE	degrees * 100 to make this a integer
LONGITUDE	degrees * 100 to make this a integer
Z	129 Z Geopotential surface/model orography
LSM	172 LSM Land sea mask
LOCATION	Latitude/longitude as GIS location

The GIS LOCATION is only added to this table. This is because always the same values for latitude and longitude are used and then if the GIS location is required for the ANALYSIS or FORECAST tables it can be retrieved by joining to this table on the LATITUDE, LONGUTUDE fields.

FORECAST

The FORECAST table contains predicted values in steps of 3 hours till 72 hours ahead. The MODELTS is the timestamp for which this prediction applies. Two times a day new Forecast GRIB files will arrive. The newer FORECAST data that arrives for the same timestamp will overwrite the older data.

The FORECAST table itself is a master table that does not contain any data. The inherited tables FORECAST<YEAR><MONTH> contain the data, but you the master FORECAST table can be used to SELECT the data.

MODELTS	Time for which this forecast applies = model run + step hours
LATITUDE	degrees * 100 to make this a integer
LONGITUDE	degrees * 100 to make this a integer
STEP	model step (smaller is more recent forecast)
SSRD	169 SSRD surface solar radiation downwards accumulated
TP	228 TP total precipitation accumulated
SP	134 SP surface pressure
BLH	159 BLH boundary layer height
TCC	164 TCC total cloud coverage
U10	165 10U 10 meters wind U-component
V10	166 10V 10 meters wind V-component
T2	167 2T 2 metres air temperature
D2	168 2D 2 metres dew point temperature
LCC	186 LCC low cloud coverage
Pressure level 1000
T_1000	130 T temperature
U_1000	131 U U-wind component
V_1000	132 V V-wind component
GH_1000	156 GH geopotential height
R_1000	157 R relative humidity
Pressure level 925
T_925 real	130 T temperature
U_925 real	131 U U-wind component
V_925 real	132 V V-wind component
GH_925 real	156 GH geopotential height
R_925 real	157 R relative humidity
Pressure level 850
T_850	130 T temperature
U_850 real	131 U U-wind component
V_850 real	132 V V-wind component
GH_850 real	156 GH geopotential height
R_850 real	157 R relative humidity
Pressure level 700
T_700 real	130 T temperature
U_700 real	131 U U-wind component
V_700 real	132 V V-wind component
GH_700 real	156 GH geopotential height
R_700 real	157 R relative humidity

ANALYSIS

The ANALYSIS table contains values from the Analysis GRIB files at 18, 00, 06 and 12 UTC

MODELTS	Time for which this forecast applies = model run
LATITUDE	degrees * 100 to make this a integer
LONGITUDE	degrees * 100 to make this a integer
SP	134 SP surface pressure
TCC	164 TCC total cloud coverage
U10	165 10U 10 meters wind U-component
V10	166 10V 10 meters wind V-component
T2	167 2T 2 metres air temperature
D2	168 2D 2 metres dew point temperature
LCC	186 LCC low cloud coverage
Pressure level 1000
T_1000	130 T temperature
U_1000	131 U U-wind component
V_1000	132 V V-wind component
GH_1000	156 GH geopotential height
R_1000	157 R relative humidity
Pressure level 925
T_925	130 T temperature
U_925	131 U U-wind component
V_925	132 V V-wind component
GH_925	156 GH geopotential height
R_925	157 R relative humidity
Pressure level 850
T_850	130 T temperature
U_850	131 U U-wind component
V_850	132 V V-wind component
GH_850	156 GH geopotential height
R_850	157 R relative humidity
Pressure level 700
T_700	130 T temperature
U_700	131 U U-wind component
V_700	132 V V-wind component
GH_700	156 GH geopotential height
R_700	157 R relative humidity

COMBINED

If you want to use Analysis data if available for a timestimp but otherwise Forecast, the COMBINED VIEW can be used. If a ANALYSIS value is available for a timestamp and latitude/longitude the ANALYSIS value will be used, otherwise the FORECAST value is given.

Historical data

In addition ongoing process of adding ECMWF data since November 11 2007, a historical data set has been loaded. That historical dataset is stored in additional inherited tables FORECASTHISTORY and ANALYSISHISTORY which inherit from FORECAST and ANALYSIS respectively.

The FORECASTHISTORY does not always contain TP (228 TP total precipitation accumulated). Instead of this it contains the following values :

LSP	142 LSP Large scale precipitation
CP	143 CP Convective precipitation

The historical dat sets span the following periods and locations

Start	End	Area	Grid
2007-01-01	2007-11-15	62degN 10degW 335degN 20degE	0.25x0.25deg
2003-02-15	2003-06-15	62degN 10degW 335degN 20degE	0.5x0.5deg
2005-01-01	2006-12-31	62degN 10degW 335degN 20degE	0.5x0.5deg

Helpful Functions

The following functions have been defined in Postgres to select the nearest gridpoint in spacetime for the ECMWF analysis and forecast tables.

These are ecmwf.forecast_nearest and ecmwf.analysis_nearest and they take a latitude, longitude and timestamp as parameter.

They can be used in the following way:

 select * from ecmwf.forecast_nearest( 52.35983, 5.615, timestamp '2008-05-19 13:34:00');
  
 select * from ecmwf.analysis_nearest( 52.35983, 5.615, timestamp '2008-05-19 13:34:00');

These function can for example be used in a join as follows :

 SELECT g.ptt_id
      , g.date_time
      , g.latitude
      , g.longitude
      , g.speed
      , g.course
      , g.altitude
      , (ecmwf.forecast_nearest( g.latitude, g.longitude,g.date_time)).*
 FROM gps.gps_tracking_data g
    , gps.individual i
    , gps.track_session t
 WHERE g.ptt_id = t.ptt_id 
   AND g.date_time >= t.start_date 
   AND i.ring_number = t.ring_number 
   AND g.parser_qc = 1 
   AND g.ptt_id = 41745 
   AND g.latitude <> 0
 ORDER BY g.date_time;

Although these functions were converted to PLPgSQL to speed them up, a faste ways to retrieve the same data is with (thanks to Floris Sluiters):

 SELECT ptt_id, date_time, latitude, longitude, speed, course, altitude, (x::ecmwf.forecast).*
 FROM (SELECT 
      g.*
      ,((SELECT f 
           FROM ecmwf.forecast f
          WHERE f.latitude = round(g.latitude * 100 / 25)::integer * 25
            AND f.longitude =  round(g.longitude * 100 / 25)::integer * 25
            AND f.modelts = TIMESTAMP 'epoch' + round(EXTRACT(EPOCH FROM g.date_time) 
                / (60*60*3))::integer * INTERVAL '3 hours' limit 1)::ecmwf.forecast) as x 
       FROM   gps.gps_tracking_data g
            , gps.individual i
            , gps.track_session t
       WHERE g.ptt_id = t.ptt_id 
         AND g.date_time >= t.start_date 
         AND i.ring_number = t.ring_number 
         AND g.parser_qc = 1 
         AND g.ptt_id = 41745
         AND g.latitude <> 0
         ORDER BY g.date_time) t