Revisions during 1986

2 January 1986 Bad data was used in the analysis from 860101 00Z to 860102 12Z inclusive due to an error in the earth location of the soundings at NOAA/NESDIS, Washington. No satellite soundings were used in the analysis until 4 January 1986, when the problem was resolved.

4 March 1986 A modification was made to the initialisation scheme in order to help preserve the tidal waves in the data assimilation. As a result, better use is made of surface pressure observations, in particular in the Tropics. where the tidal fluctuations contribute to a large extent to the observed diurnal pressure variations.

11 March 1986 The humidity analysis was modified. The precipitable water content is now used in the atmosphere as observed from space and reported in the satellite data. Furthermore, the procedures for using bogus humidity data derived from synoptic surface reports - dew point, cloud amounts - were used in a modified way. The changes had modest but consistent positive impact on the temperature forecast in the free atmosphere and results in a more realistic forecast of the precipitation amount.

13 May 1986 Forecast model cycle 26. The 16-level model was replaced by a 19-level model. The three extra levels were introduced in the stratosphere.

15 July 1986 Forecast model cycle 27. A parametrisation scheme for representing the momentum transports due to subgrid gravity wave drag was incorporated into the operational ECMWF forecast model. It counteracts the zonalisation tendency of the large scale flow and prevents the development of too strong westerlies in winter. Errors in large scale condensation and snow melting were corrected.

19 August 1986 The calculation of low, medium and high cloud amounts were corrected in post-processing.

9 September 1986 A new analysis system was implemented which makes better use of observations; data and differences from the first-guess will be used at reported levels. This will effectively increase the actual vertical resolution of the analysis, in particular in the boundary layer and near the tropopause.
Unnecessary vertical interpolation of analysis increments between model and standard pressure levels are eliminated. Analysis increments on the Gaussian grid of the model were evaluation (1.125° resolution).
The data selection criteria in boxes of flexible size depending on data sensity and extending over the depth of the troposphere were improved.
A univariate optimum interpolation scheme replaced the correction scheme for the analysis of humidity.

30 September 1986 Modifications to analysis. Indication of inner data in upper analysis slab was corrected. Negative and zero values of q and q_s were forced to small positive numbers. The check on the analysis error file date was reimplemented.

7 October 1986 Forecast mode cycle 28. The snow melt logic was corrected. The old scheme caused a drop of the surface temperature when snow has to melt on the ground at the expense of surface energy.

15 October 1986 An error in soil moisture analysis was corrected.

11 November 1986 Analysis modified. The u and v departures are checked together. The persistence error calculation and the indication of inner data in upper slab were corrected. Serious data problems with satellite soundings (SATEMs and TOVS) at the beginning of November. The performance of the ECMWF analysis and forecast especially over the Southern Hemisphere was adversely affected by these data deficiencies.