In general, roughness profiles of rock joints consist of nonstationary
and stationary components. At the simplest level, only one parameter
is necessary to quantify nonstationary joint roughness. To capture the
nonstationary roughness, it is suggested that the average inclination
angle I, along with the direction considered for the joint surface, b
e used. Most natural rock-joint surface profiles do not belong to the
selfsimilar fractal category. However, they may be modeled by self-aff
ine fractals. It is shown that even though the fractal dimension D is
a useful parameter, it alone is insufficient to quantify the stationar
y roughness of non-self-similar profiles. A new strength criterion, wh
ich includes three roughness parameters (D, I, and a spectral coeffici
ent, K-s), is suggested for modeling the anisotropic peak shear streng
th of rock joints at low normal effective stresses. Joint roughness da
ta should be used to estimate the roughness parameters in different di
rections on the joint surface. Parameter D reflects the rate of change
in length of the profile in response to a change in the scale of meas
urement r. Parameter K-s may be used to model the scale effect. A vali
dation study shows that the new equation has good capability to predic
t the anisotropic peak shear strength of joints. In practice, to allow
for modeling uncertainties, the new equation should be used with a fa
ctor of safety of about 1.5. Comparisons are also shown between the pr
edicted shear strengths based on Barton's equation and the measured pe
ak shear strengths.