Degree Type


Date of Award


Degree Name

Doctor of Philosophy




Objectives of this dissertation research were: (1) to develop multiple regression equations to predict available (Bray-1) P levels at various subsoil depths, using soil horizon, soil profile, parent material, location, and climatic variables; (2) to develop prediction equations for subsoil P without the soil horizon variables; and (3) to compare location (legal township and range numbers) and climatic variables (mean annual precipitation and temperature) for predicting subsoil P. Input data were from 3913 soil horizons from 696 soil profiles from 22 Iowa counties representing the most soil association areas;The depth to each horizon variable was included in all regressions. Soil horizons variables included pH, soil test P (dependent variable), soil test K, clay, organic carbon, and bulk density. Parent material variables included loess, pedisediment materials above till, till and paleosol, colluvium in loess areas, alluvium in loess and till areas, and eolian sands. Profile variables included genetic horizon, site slope, thickness of A horizon, minimum pH, depth to minimum pH, drainage class, biosequence, and depth to maximum clay horizon;Variables were initially selected in alternate models including linear functions of parent material variables, cubic function of depth, and quadratic functions of all others. Climatic variables, more important than location variables, were retained. Significant interactions between linear, quadratic, and other variables and then interactions between other variables were then selected. Three final prediction models were selected from (1) all variable groups (88 variates, R('2) = 0.776), (2) all except horizon variables (81 variates, R('2) = 0.745), and (3) all except horizon and genetic horizon variables (75 variates, R('2) = 0.719). Appropriate final model depends on availability of horizon data and soil profile descriptions;The final all-variable prediction model (MODEL M-6) for subsoil P (STP) included 18 linear, 7 quadratic, 1 cubic, and 62 interaction variates (15 DEPTH*X(,i), 9 DEPTH('2)*X(,i), 2 DEPTH('3)*X(,i), and 36 linear*linear interactions between other variables. Effects of the variables on STP were examined in MODEL M-6 using partial derivatives of STP with respect to the X(,i) variables and by computing predicted STP values for depth and combinations of two other variables from simplified regression equations. The dominant effects of biosequence modified by interactions with the cubic function of depth and many other variables and the joint effects of pH and bulk density with depth on STP showed that MODEL M-6 can predict widely varying STP distributions. Effects of other variables on STP distributions were also shown.



Digital Repository @ Iowa State University,

Copyright Owner

Hammed Mohammad Salih



Proquest ID


File Format


File Size

248 pages