Accurate determination of surface roughness of rock joints at the large-sca
le is essential for proper rock mass characterization. Surface roughness of
rock joints is commonly characterized using small samples. However, since
roughness parameters of rock joints are scale-dependent and their descripto
rs change with scale. a systematic investigation has been carried out to un
derstand the effect of scale on the surface roughness of rock joints. A sil
icon rubber replica. 1000 mm x 1000 mm in size, was Moulded in-situ from a
natural rock joint surface and its surface was digitized in the laboratory
using a 3-D laser scanner having high accuracy and resolution. The fractal
parameters, i.e. the fractal dimension D and amplitude parameter A describi
ng surface roughness of the replica, were calculated on the basis of the Ro
ughness-Length Method. To investigate the scale-dependency of surface rough
ness of rock joints, ten sampling windows ranging in size from 100 mm x 100
mm to 1000 mm x 1000 mm were selected from the central part of the replica
and their fractal parameters were calculated. The results show that both D
and A are scale-dependent and their values decrease with increasing size o
f the sampling windows. This scale-dependency is limited to a certain size,
defined as the stationarity threshold. and for sampling windows larger tha
n the stationarity threshold, the estimated parameters remain almost consta
nt. It is concluded that. for surface roughness to be accurately characteri
zed on a laboratory scale or in the field, samples need to be equal to or l
arger than the stationarity limit. In this paper we have indicated the meth
odology for establishing the value for the stationarity limit for rock join
ts. (C) 2001 Elsevier Science Ltd. All rights reserved.