Description of the ECMWF data assimilation system in 1987

The ECMWF data assimilation system in 1987 consisted of a multivariate optimal interpolation analysis, a non-linear normal model initialisation and a high resolution spectral model which produced a first-guess forecast for the subsequent analysis. Data were assimilated every 6 hours.

Domain	Global
Data assimilation frequency	6-hr (± 3-hr observation time window)
First guess	6-hr forecast
Dependent variables	Ps,F,u,v,RH
Vertical coordinate	Hybrid, Pk+1/2=Ak+1/2+Bk+1/2 Ps, details as above
Horizontal grid	160 x 320 points on a quasi-regular (1.125°) "Gaussian" grid
Analysis method	Mass and wind:	3-dimensional multivariate statistical interpolation
	Relative humidity:	3-dimensional uni-variate statistical interpolation up to 250 hPa
	Surface:	Sea surface temperature from NMC analysis
		Soil water content using rainfall observations,
		estimated evaporation
		Snow depth using snow depth and snowfall observations
Initialisation method	Non-linear normal mode, 5 vertical modes, non-adiabatic