Summary of data deficiencies in the operational archive

Date	Time	Parameter	Level	Notes
23 April 1985	06/12	all	surface	missing data
24 April 1985	06/12	all	surface	missing data
29 April 1985	18	all	surface	missing data
30 April 1985	00	all	surface	missing data
30 April 1985	18	all	surface & pressure levels	data archived incorrectly
June, July 1985	00/06/12/18	Temperature	surface	analysed sea surface temperature was constant for this period (corrected 31 July 1985); forecasts may also have been affected
13 June 1985	18	all	surface & pressure levels	missing data
14 June 1985	00/06/12	all	surface	missing data
14 June 1985	06	all	pressure levels	missing data
8 November 1985	06	all	surface	no data archived
1-2 January 1986	00/06/12/18	all	pressure levels	due to an error in the earth location of soundings at NOAA/NESDIS, Washington, bad data was used in the analysis and this could affect the quality of the data
1-4 December 1991	00/06/12/18	albedo	surface	all values set to zero
12 October 1999 - 10 April 2000	00/06/12/18	Humidity & Temperature	lower stratosphere	the humidity fields in the lower stratosphere both in the analysis and forecast were seriously affected by erroneous moisture. Read more
15 August 1993 - current day	00/06/12/18	The size of waves of low magnitude (less than 1 meter) and the corresponding mean periods	sfc	The size of waves of low magnitude (less than 1 meter) may be too high and the corresponding mean periods too high. This problem is largely confined to enclosed or semi-enclosed seas. Read more
21 November 2000 - 26 November 2002.	00/06/12/18	Orography	sfc and 1st model level	The spectral orography on model level 1 and the Gaussian grid point orography at the surface are not in agreement. Read more

Table 1 Summary of miscellaneous data deficiencies in the archives

Total cloud cover
All total cloud cover fields archived as analysis data are incorrect.Cloud cover at ECMWF is deduced during the post-processing of various direct model output fields. Due to some of these fields not having been initialised with respect to analysis data, values archived with the analysis archive are of no practical use.

Precipitation
All precipitation fields archived as analysis data are incorrect.
Research works at ECMWF are developing schemes to use reported precipitation data from observations. To assist this process. precipitation data are contained within the analysis archive; these data are for internal ECMWF use only and are of no practical use to external users.
Forecast precipitation fields before 1991 are of doubtful quality because of limitations in the model physics resulting in "spin-up" problems during the early stages of the forecast.

U- and V-Wind Components at the Poles
In 1991 it was discovered that, on a regular latitude/longitude grid, the ECMWF u- and v- components of wind were incorrect at the poles. The problem was that the horizontal components of wind gave inconsistent polar values of wind magnitude and direction. Changes have been made to the interpolation routines used to create the ECMWF/TOGA Basic Data Sets and to extract data from the ECMWF/TOGA Advanced Data Sets and the Supplementary Fields Data Set. These changes have had the following effects on u- and v-wind fields at the poles:

(a) Surface data. The grid points at each of the poles will contain horizontal wind components from the nearest neighbouring Gaussian latitude circle interpolated to the required resolution. For the T213 model the nearest latitude circle is ± 89.578132.

(b) Upper air data. The grid points at each of the poles contain the correct horizontal wind components derived from the spherical harmonics coefficients, i.e. the values of the wind magnitude derived from the horizontal wind components will be constant, while the u- and v- components oscillate with a wave number 1 pattern around the poles.

From 1 January 1992, all u- and v-wind components at the poles supplied from the ECMWF/TOGA Advanced Data Sets will be in the format described in (a) and (b) above.

(N.B.) Prior to August 1998, data from the ECMWF/TOGA Basic Data Sets for periods between 1985 and June 1991 will contour incorrect u and v components at the pole as described in (a) and (b) above.
Since August 1998, the whole of the Basic Data Set archives contain corrected values of u and v at the poles.

The following paragraph contains a method for calculating wind components at the poles using the values at a neighbouring latitude circle. This is only one method that can be used. The user may wish to use another method for calculating polar winds, for example by using a polar stereographic projection.

Calculation of wind at the pole from values at a neighbouring latitude circle
Given winds Ui, Vi , i =1, 2, 3, ..., NLON along a latitude circle close to the pole, compute

Where

approximate wind at the pole for longitude i=l based on the wave number 1 components of U and V along the latitude.
Winds at other longitudes are given by



Example
Using T213 wind data at 500 hPa (from September 199l):

Distance of latitude circle from pole	2.5°	1.5°	0.421868°*
NLON	144	240	640
Relative error in wind strength	2.9%	2.2%	0.1%
Absolute error in wind direction	13°	7°	0.7°

* Gaussian latitude nearest to pole

Humidity for the period 12/10/1999-10/04/2000

ECMWF has issued the following warning about upper air data between 12 October 1999 and 10 April 2000.

During the operation of IFS Cycle 21r4 of the ECMWF model from 12 October 1999 to 10 April 2000 the humidity fields in the lower stratosphere both in the analysis and forecast were seriously affected by erroneous moisture. These fields should not be used for research and it is recommended that the model results from the operational archive are not used to provide initial conditions for numerical experiments during this period.

The error is up to 1 to 2 orders of magnitude in a shallow layer just above the tropopause.

This may have a slight affect on the temperature in some regions.

Alternative analysis (but not forecast) data can be supplied for this period from research experiments which were run with the error corrected. However the pressure level data from these experiments does not include specific humidity

WAM - Size of waves
There has been a problem with the wave analysis since the introduction of ERS data on 15 August 1993. The size of waves of low magnitude (less than 1 meter) may be too high and the corresponding mean periods too high.

This problem is largely confined to enclosed or semi-enclosed seas such as the Mediterranean Sea, Black Sea, Baltic Sea, the Gulf of Mexico and the East China Sea.

A partial correction for this problem was introduced on 11 April 2000 and further changes will be introduced shortly which should eliminate the problem.

Orography

The spectral orography (parameter 129, archived in MARS on model level 1) and the Gaussian grid point orography (parameter 129, archived in MARS as a surface field) of the high resolution model T511 were not in agreement for the period 21 November 2000 - 26 November 2002.

The discrepency between the two fields was introduced on 21 November 2000 with the introduction
of the T511 forecasting system.

Differences between the two orography fields can be up to a couple of hundred metres in particular on the slopes of steep orography such as the Alps or the Pyrenees. The spectral orography field on model level 1 is the correct field. The surface Gaussian grid point orography is in error through the application of a wrong filter.