Based on a stochastic approach, drainage in a layered soil profile wit
h macropores was predicted using a one-dimensional numerical flow mode
l. Soil hydraulic functions, theta(psi) and K(psi), were obtained from
laboratory measurements on undisturbed 100 cm(3) soil cores. The core
s were collected in 60 different locations for each of three soil hori
zons along a 31 m transect on a sandy loam soil. The effect of a stoch
astic description of theta(psi) and K(psi) on predicted fluxes and wat
er content profiles was evaluated with the aid of Monte-Carlo simulati
ons. Initially simulations were carried out with the actual field data
. Alternatively, 500 sets of the multi-normal parameter vector P = {th
eta(s), alpha, n, K-s} were generated and used in a Monte-Carlo simula
tion approach. In addition, deterministic calculations were carried ou
t using (1) arithmetic mean values of untransformed hydraulic paramete
rs, (2) arithmetic mean values obtained after backtransformation of Ga
ussian distributed parameters, and (3) alternatively using a set of me
an bimodal retention and conductivity curves. Predictions obtained wit
h the two stochastic methods and the deterministic cases are compared
with measured cumulative outflow and moisture content profiles from fi
fteen 1 m long undisturbed soil columns collected along the same trans
ect. All simulations based on unimodal hydraulic functions underestima
ted the mean observed drainage. If actual field data are used, the mea
n outflow is 74% of the mean measured cumulative outflow, compared to
70% for the Monte-Carlo simulation. The range mean +/- two standard de
viations of the simulated outflow for both methods is considerably sma
ller compared to the measured range of variation. Deterministic predic
tions of the total drainage amounted to 83, 83, and 127% of the observ
ed drainage for the cases 1, 2, and 3, respectively. Observed outflow
curves displaying macropore flow could not be predicted with the unimo
dal Richards' flow equation. The results indicate that the field-scale
unsaturated flow behaviour in soils with macropores cannot be accurat
ely predicted assuming a unimodal retention characteristic. This is tr
ue for both the mean drainage but even more so for the extremes. The u
se of bimodal hydraulic functions considerably increased the total dra
inage, resulting in an overprediction of the mean measured drainage. P
redicted soil water contents in the 1 m deep soil profile are mostly h
igher than the measurements. Contrary to the drainage, the range of va
riation of the predicted water contents is always larger than the meas
ured variability, except for the surface values.