Setup Model Evaluation CLIWANET

The objective of Cliwanet Workpackage WP4100 includes a comprehensive evaluation of output from atmospheric models with observations retrieved from the CLIWA-network. The focus will be on cloud parameters with emphasis on cloud liquid water path. Observed quantities are inferred from ground-based measurements (see WP2000) during three 2-month observational campaigns, i.e. CNN1 (Aug/Sep 2000), CNN2 (Apr/May 2001), and BBC (Aug/Sep 2001). Next to this, 2-dimensional fields of precipitation (3h/12h accumulated values) produced by the BALTRAD (the Baltic Radar Centre) and 2-dimensional fields of cloud liquid water inferred from NOAA/AVHRR satellite overpasses (see WP3200) will be made available for model evaluation.

Crucial for a meaningful evaluation is the extraction of an adequate and useful combination of model quantities from the forecast runs. During the Kick-off meeting this item was discussed at length and the results are summarized hereafter in a prescriptive format.

Institutes and Models


DWD (Lokalmodell), ECMWF, Rossby Center (HIRLAM/SWECLIM) and KNMI (RACMO/HIRHAM)

Horizontal and vertical resolution of models

Specification of Model Time Series Output per Station

Temporal aspects of output

Spatial coverage of subdomains

about 50x50 km2 around each station
Avoid the central point to be a sea point (except for Gotland).

Export model output averaged over the subdomain. For ECMWF this is trivially exporting the direct model output for each of the specified grid points. For DWD/MIUB, SMHI and KNMI, this requires some intermediate processing likely to be done inside the model. Besides the averaged values, they should also export minimum values, maximum values, and direct model values of the grid point nearest to the station coordinates. An alternative for DWD/MIUB, SMHI and KNMI, is exporting model output of all gridpoints involved in each of the subdomains and do the processing at a later stage. For DWD/MIUB the latter implies the export of 12*27=324 sets of model output which might be a bit overdone ...

General Information

Single-level parameters

Surface pressure is assumed instantaneous. All other values should be averaged (fluxes, cloud cover) or accumulated (precip) over the past temporal interval.

  1. Verifying date [yyyymmdd]
  2. Verifying time [hhmn]
  3. Surface Pressure [Pa]
  4. Sensible heat flux at surface [W/m2]
  5. Latent heat flux at surface [W/m2]
  6. Momentum flux at surface [Pa] (rho <u'w'>)
  7. Downward SW-flux at surface [W/m2]
  8. Upward SW-flux at surface [W/m2]
  9. Downward LW-flux at surface [W/m2]
  10. Upward LW-flux at surface [W/m2]
  11. Downward SW-flux at TOA [W/m2]
  12. Upward SW-flux at TOA [W/m2]
  13. Upward LW-flux at TOA [W/m2]
  14. Precipitation Convective [m/s]
  15. Precipitation Large Scale [m/s]
  16. Precipitative Fraction in GridBox [0..1]
  17. Total Cloud Cover [0..1]

Multi-level parameters

All multi-level values are assumed instantaneous.

  1. Verifying date [yyyymmdd]
  2. Verifying time [hhmn]
  3. Model layer value
  4. Pressure [Pa]
  5. Temperature [K]
  6. Zonal wind component [m/s]
  7. Meridional wind component [m/s]
  8. Vertical wind speed [Pa/s]
  9. Turbulent Kinetic energy [m2/s2]
  10. Specific Humidity [kg/kg]
  11. Specific Liquid Water [kg/kg]
  12. Specific Ice Content [kg/kg]
  13. Cloud fraction [0..1]
  14. Short Wave In-Cloud Optical Thickness [..]
  15. Long Wave In-Cloud Emissivity [0..1]
  16. Liquid Precipitative Flux at lower interface [kg/m2/s]
  17. Solid Precipitative Flux at lower interface [kg/m2/s]

File Structure
All time series are formatted as ASCII. The principle rule is that all information within one record refers to a single verifying time (apart from the general information). To avoid (too) long records there are more records (typically nlev+1) referring to the same verifying time.
All files must start with the following header lines.
If any of the header information changes, repeat the entire header before adding new information to the same file. 

 Headerlines
    line 1: '#',A16      ... name institute
    line 2: '#',A16      ... name participant
    line 3: '#',A16      ... name model
    line 4: '#',A16      ... name experiment
    line 5: '#',I10,I6.4 ... initial date[yyyymmdd]/time[hhmn] model run 
    line 6: '#',A16      ... name or acronym CLIWANET station
    line 7: '#',2F8.2    ... longitude/latitude receptorpoint 
    line 8: '#',E16.5    ... surface geopotential height [m2/s2] 
    line 9: '#',E16.5    ... land sea mask gridbox       [] 

FORMAT FILE WITH SINGLE-LEVEL FIELD INFO:

(start with header line: next 4 lines form one record ...)
    line 10: '#   vdate vhrm'
	   ,'        PSURF         SENF         LATF         MOMF         SWSD'
           ,'         SWSU         LWSD         LWSU         SW0D         SW0U'
           ,'         LW0U     PRECCONV       PRECLS     PRECFRAC       TOTCOV'
 Loop over time-index 
    line 11 ... : data with FORMAT: I9,I5.4,E14.6,14E13.5

FORMAT FILE WITH MULTI LEVEL FIELD INFO:

(next 2 lines form one record ...)
    line 10: '#   vdate vhrm lev'
	   ,'         PRES         TEMP           UU           VV        WMEGA'
           ,'          TKE         QVAP         QLIQ         QICE       CLFRAC'
           ,'      SWCLTAU     LWCLEMIS      PRFLLIQ      PRFLICE'
 OUTER Loop over time-index 
 INNER Loop over level-index
    line 11 ... : data with FORMAT: I9,I5.4,I4,14E13.5

File Naming
Naming of files is as follows. Files with single-level (multi-level) parameters are named SLEV_dtgin_XX (MLEV_dtgin_XX), where XX is substituted by the 2-character acronym of the actual CLIWANET station (see first column table). If information of more or all stations is contained in one file, leave XX as XX. The suffix dtgin refers to the initial datum time group of the model run in yyyymmddhh format.

Co-ordinates and altitudes

CLIWANET Stations
acro
nym		 station operator LAT LON ALT CNN1 CNN2 BBC
BE Bern UNIBE 46.95 N 7.73 E 575 m x x -
CA Cabauw KNMI 51.97 N 4.93 E -1 m x x x
CH Chilbolton CLRC 51.14 N 1.44 W 84 m x x -
GE Geeshacht GKSS 53.41 N 10.42 E 58 m x x -
GO Gotland SMHI 57.24 N 18.39 E 56 m - x -
HE Helsinki SMHI 60.18 N 24.83 E 14 m x x -
KI Kiruna Chalmers 67.87 N 21.06 E 492 m x x -
LI Lindenberg DWD 52.22 N 14.12 E 112 m x x -
ON Onsala Chalmers 57.40 N 11.93 E 11 m x x -
PA Palaiseau CETP 48.80 N 2.30 E 175 m x - -
PO Potsdam DWD 52.38 N 13.05 E 100 m x x -
SP St. Petersburg IREE/MGO 59.95 N 30.70 E 72 m x x -

Specification of Model Field Output

Precipitation

Within CLIWANET (WP2200/2300/2400) high quality precipitation fields (3-hour and 12-hour accumulated fields) will be inferred from the BALTRAD network intended for use in model evaluation (WP4100). These products will be made available off-line on a monthly basis. To utilize this product atmospheric models must archive accumulated precipitation, preferably at 3-hour intervals, for the CLIWANET network domain.

Archiving of Accumulated Precipitation Fields


Cloud Liquid Water Path

A significant task formulated within WP3200 includes the retrieval of LWP (liquid water path) from satellite observations (over sea from AMSU data, over sea and land from AVHRR data). As the NOAA/AVHRR overpasses are to a certain extent irregularly distributed in time, it is not possible to define the instants of overpass beforehand. But it is probably sufficient to have model fields of cloud cover, liquid water column, and cloud top temperature at 6-hour intervals (if feasible 3-hour frequency would be preferable).

Archiving of Cloud Parameter Fields