Watertable heights and total moisture content were measured in a sand colum
n where the piezometric head at the base ("the driving head") varied as a s
imple harmonic with periods in the range from 14.5 min to 6.5 h. The watert
able height h(t) responded very closely to the driving head compared with t
he predictions of previous analytical and numerical models. The total moist
ure quantified as an equivalent, saturated height h(tot)(t) varied very lit
tle compared with the watertable height. Neither h(t) nor h(tot)(t) deviate
d significantly from simple harmonics when the driving head was simple harm
onic. This indicates that non-linear effects are weak and hence that analys
is based on linear solutions have fairly broad applicability. When h(t) and
h(tot)(t) are simple harmonic, the ratio n(d) = [dh(tot)/dt]/[dh/dt] is a
constant in the complex formalism. Its magnitude \n(d)\ is the usual effect
ive porosity while its argument accounts for the phase shift which is alway
s observed between h(t) and h(tot)(t). Within the current range of experime
nts this dynamic, effective porosity nd appears to be almost independent of
the forcing frequency, i.e., it is a function of the soil and its compacti
on only. Introducing the complex nd enables analytical solution for the wat
ertable height in the column which is simpler and more consistently accurat
e over a range of frequencies than previous models including Richard's equa
tion with van Genuchten parameters corresponding to the measured water rete
ntion curve. The complex lid can be immediately adopted into linear waterta
ble problems in 1 or 2 horizontal dimensions. Compared with "no fringe solu
tions", this leads to modification of the watertable behaviour which is in
agreement with experiments and previous models. The use of a complex nd to
account for the capillary fringe in watertable models has the advantage, co
mpared with previous models, e.g., [Parlange J-Y, Brutsaert W. A capillary
correction for free surface flow of groundwater. Water Resour Res 1987; 23(
5):805-8.] that the order of the differential equations is lower. For examp
le, the linearised Boussinesq equation with complex nd is still of second o
rder while the Parlange and Brutsaert equation is of third order. The extra
work of calculating the imaginary part of the initially complex solution i
s insignificant compared to dealing with higher order equations. On the bas
is of the presently available data it also seems that the "complex n(d) app
roach" is more accurate. This is to be expected since the complex nd accoun
ts implicitly for hysteresis while the Green-Ampt model does not. With resp
ect to linear watertable waves, accounting for the capillary fringe through
the complex n(d) is a very simple extension since the determination of wav
e numbers already involves complex numbers. (C) 2000 Elsevier Science Ltd.
All rights reserved.