Selection of management variables for regressing Iowa corn yields on management, climatic, and soil variables

Major objectives of this dissertation research were: (1) to determine the most important management variables for predicting corn yields and (2) to develop a multiple regression prediction equation of Iowa corn yields on management, climatic, and soil variables. Corn yields and related data were from 2657 observations from 1957 to 1970 at 712 sites in 15 Iowa counties representing most soil association areas;From the first series of quadratic models, 25 of 50 management and 2 climatic variables were retained in final MODEL A-35 (R(\u272) = 0.603). The relative importance of variable groups were: climatic indexes \u3e tillage and planting \u3e plow-layer soil tests \u3e environmental (lodging, insect damage, weeds) \u3e fertility management \u3e crop rotation;The final yield prediction MODEL J-10 (R(\u272) = 0.681) was derived from models testing soil variables and interaction variates. It had 33 linear, 14 squared, and 33 interactions involving 20 management, 2 climatic, and 11 soil variables. Effects of the moisture stress index were modified by 8 interactions, N fertilizer had 6 interactions, 6 variables had no interaction, and the rest were involved in 1 to 4 interactions on yield. Effects of the variables and their interactions on yield in MODEL J-10 were determined from the partial derivatives of yield with respect to the variables and by computing (DELTA)YIELD values for selected combinations and levels of 2 or 3 variables;Yield responses to both plant density and N fertilizer and their levels associated with maximum yield increased greatly as moisture stress decreased. Responses to N were also markedly affected by interactions with crop rotation and soil test N level. Soil variables of drainage class, site slope, and thickness of A horizon affected yield directly and through interactions with the N-related variables. Planting data associated with maximum yield varied from about May 10 in southern Iowa to April 26 in northern Iowa. Yield losses from corn rootworm and corn borer were influenced by several interactions. Negative effect of second-brood corn borer increased as plant density and N fertilizer levels increased;Plow-layer pH, depth to carbonates, site slope, and subsoil P had interrelated effects on yield; all should be considered in liming recommendations. Other variables that affected yield directly and through interactions included S-N and E-W directions in Iowa, clay content of plow layer, biosequence, and plant available water capacity;Interaction effects between variables should be considered for predicting corn yield and for recommendations of management practices. Final prediction MODEL J-10 can be used to predict corn yields of most soil mapping units in Iowa