G. Dagan et al., STOCHASTIC-ANALYSIS OF CONCENTRATION MEASUREMENTS IN THE TRANSPORT EXPERIMENT AT TWIN-LAKE SITE, Water resources research, 33(4), 1997, pp. 559-567
A procedure to identify the parameters characterizing flow and transpo
rt in heterogeneous aquifers with the aid of concentration measurement
s in tracer field experiments is developed. Unlike previous studies, w
hich employed the measured plume spatial moments at different times an
d their theoretical expressions, we rely here on breakthrough curves a
nd temporal moments in order to analyze the field tests at Chalk River
Site. In these experiments, breakthrough curves of a radioactive trac
er were measured continuously at a large number of points in parallel
control planes. We derive theoretical expressions of the temporal mome
nts of the breakthrough curves by the same Lagrangian approach that wa
s used previously for spatial moment. We assume a stationary random ve
locity field of constant mean and relate it to the axisymmetric log co
nductivity covariance by a first-order approximation in the variance,
with neglect of the effect of pore-scale dispersion. The final theoret
ical results relate the temporal moments to U (the mean velocity), sig
ma(Y)(2) (the log conductivity variance), I-Yh (the horizontal integra
l scale), and b(e) (a function of the anisotropy ratio e = I-Yv/I-Yh).
By assuming ergodicity, we identify the temporal moments at the Chalk
River Site experiment from measured breakthrough curves. With the aid
of the theoretical results and by a best fit we could estimate U, sig
ma(Y)(2), I-Yh, and I-Yh/b(e). An attempt to identify the vertical and
transverse integral scales from temporal-spatial moments in the contr
ol planes was not successful. We took advantage of the dense measureme
nts of breakthrough curves along vertical transects (at intervals of 1
cm) in order to identify the experimental concentration two-point cov
ariance. We derived a simplified theoretical expression for its depend
ence on the log conductivity vertical integral scale I-Yv, which was i
dentified by a best fit with experimental results. This procedure, app
lied for the first time to analysis of field tests, led to the identif
ication of the estimates of sigma(Y)(2), and I-Yh, as well as of their
variances of estimation.