Assumptions:
o
The
observation errors are spatially and temporally uncorrelated. This allows
to handle each observation separately;
o
The
horizontal and vertical forecast error covariances are independent and
the vertical forecast error covariance is neglected;
o
Only
the time evolution of the diagonal elements (the variances) of the forecast
error covariance matrix are computed;
o
The
covariances are computed by assuming that the horizontal correlation of
the forecast error field may be represented by a simple function of the
distance;
o
The
operator H is a bilinear interpolation acting on the four points of the
model surrounding the location of the observation.