Continuous classification was applied to variables measuring landform, extr
acted from a digital elevation model (DEM) of a field. It was shown that so
il physical properties (gravimetric water content at two depths and sand/cl
ay content) and soil mineral nitrogen in the subsoil were linearly related
to membership values in these classes (the memberships having been spatiall
y smoothed and transformed from the simplex space to a real space by the lo
g-ratio transform). Adjusted R-2 values between 0.57 and 0.84 were obtained
. Stronger evidence existed for linear relationships of these soil properti
es to the transformed membership values than to the underlying landform var
iables from which the classes were derived (using the Akaike information cr
iterion to compare regressions on different numbers of predictor variables)
. This suggests that the classification procedure significantly enhances th
e information contained in the DEM for purposes of predicting soil properti
es. (C) 1999 Elsevier Science B.V. All rights reserved.