This paper investigates the predictability of the soil erodibility fac
tor (K) in the revised universal soil loss equation from soil clay con
tent (C), the Naperian logarithm of the geometric mean particle size (
D-G) and organic matter content (OM), using a global dataset compiled
from published studies. A multiple regression equation accounted for o
nly 41% of the observed variance. Because of the large unexplained var
iance, an alternative procedure was explored to describe the K data in
terms of the lower and upper bounds of the range of variation and the
most probable value in this range. However, this approach did not pro
vide sufficient information about the distribution of the observed K v
alues. A procedure based on fuzzy logic and fuzzy mathematical theorie
s was then developed, using the program FUZKBAS, which describes the f
requency distribution of observed K values for a given soil, character
ised by D-G, C and OM, in terms of membership functions. (C) 1997 Else
vier Science B.V.