Using a stand-alone 1-dim soil and atmospheric boundary layer model, the fe
asibility of off-line variational soil moisture analysis by assimilation of
near-surface atmospheric observations is investigated. The experiments are
performed using the soil module and boundary layer formulation taken from
the regional forecast model of the German Weather Service. Assimilation exp
eriments indicate that atmospheric information allows to initialize soil hu
midity to realistic values. Sensitivity studies using an adjoint code show
that proper modeling of the soil moisture signal propagation into the atmos
phere is crucial for the algorithm. An assessment of errors corrupting the
fit resulting from erroneous cloud-cover specification is performed. The er
rors are found to be small for the case considered thus confirming the usef
ulness of assimilation of soil moisture for short-term weather forecasts ev
en under non-perfect conditions. (C) 1998 Elsevier Science B.V. All rights
reserved.