Choosing Models for Soil Chronofunctions and

Fitting them to Data

 

R.      J. Schaetzl, L. R. Barrett and J. A. Winkler

 

            The effectiveness and utility of soil chronofunctions is examined in the light of existing pedogenic theory.  Statistical treatments applied in chronofunction research are reviewed, including linear transformations of raw data, which may improve the utility of the chronofunction.  We advocate using a particular statistical model only if it can be justified based on our current understanding of the functioning of the pedologic system.  We emphasize the potential difficulties of using linear chronofunction methods; simple linear and logarithmic functions are not always the best option for chronofunctions.  Hyperbolic, polynomial or nonlinear functions might improve not only the fit of the chronofunction but also advance our understanding of the pedologic system. 

            When the chronofunction explaining the most variance is not best suited to a process-based understanding of pedogenic theory, other functions with slightly smaller r2 values that have been judged suitable by a priori reasoning or theory might better reflect the functioning of the pedologic system.  We justify limited extrapolation of some chronofunctions to time zero and discuss how this can be useful in the identification of pedogenic thresholds and ‘step functions’.  Interpretation of the Y-intercept of chronofunctions can, at times, aid our understanding of the soil system near time zero and be useful in the identification of the existence and timing of thresholds.