Procedures are presented and evaluated for developing probability dist
ribution functions for rill numbers (density) and rill flow rates that
can be used to represent the stochasticity of rill networks in recent
erosion models such as PRORIL. Subsoil and topsoil data sets, includi
ng photographs, collected at the University of Kentucky were used in t
he evaluation. Photographic images were corrected for optical distorti
on and visually analyzed to develop the rill networks. A digital terra
in model (DTM) that allowed combining of channels, but not flow splitt
ing, was also utilized to develop a flow network and compared to the p
hotographically determined network. The DTM generated network did not
provide a good fit to the photographically determined network, likely
because of problems with interpolation and with the inability to predi
ct rill splitting. The DTM generated networks were utilized to develop
probability density functions (PDFs) for rill numbers and conditional
PDFs for rill flow rates given a number of rills. The binomial distri
bution provided a good fit to rill number distributions as defined by
the Kolmogorov-Smirnov test. The Weibull distribution provided the bes
t fit to the conditional PDF for flow rates, but the goodness of fit w
as poor. This lack of fit, likely due to inadequacies of the DTM, shou
ld improve with improved DTMs.