The spatial distribution of soil moisture defines preferential flow pa
ths in the unsaturated zone. Hence, three dimensional (3D) estimates o
f soil moisture are of great importance to understand transport of con
taminants as well as remediation processes in the unsaturated zone. In
this study 3D estimates conditioned on spatially frequent observation
s of soil moisture, have been obtained by kriging. The observations we
re divided into subdomains consistent with the local stratigraphy, and
directional semivariogram analysis was applied It was found difficult
to clearly identify a 3D semivariogram function in this case, but fro
m a georadar survey two semivariogram functions were derived. describi
ng two different sedimentological units, By conditioning the estimates
of soil moisture on the sedimentological architecture computed by ind
icator kriging, more accurate estimates were achieved. These improveme
nts were quantified by a 'jackknife' cross validation procedure. Besid
es the practical aspects of finding the most important flow paths, est
imates of soil moisture are valuable when validating unsaturated flow
models.