






It
is the same as the scalar formula for T, but now x
is a



model
state vector, with 10^{68}d.o.f., and y_{o} is the
set of


observations,
with 10^{59} d.o.f.



R is the observational error covariance,



B the forecast error covariance.



In
3DVar B is constant: it does not include
“errors of the


day”



