DELINEATING POTENTIAL MANAGEMENT ZONES FOR COTTON BASED ON YIELDS AND SOIL PROPERTIES.

Procedures to delineate management zones as a basis for making variable rate applications of fertilizers within a field have not been well defined. The objective of this study was to evaluate a framework of procedures for delineating potential management zones with data collected from 1-ha grids on an irrigated cotton (Gossypium hirsutum L.) field in Texas from 1998 through 2000. Selected measurements included lint yield, soil pH, exchangeable Ca^2+ and Mg^2+, K saturation, sand and clay content, depth to free carbonate layer, depth to caliche, NO3^-N, available P, elevation, and slope. Data were processed with k-means cluster analysis, multivariate analysis of variance (MANOVA), and discriminant analysis. Cluster analysis allowed grouping lint yield and soil properties into high and low yielding classes by their data structure. Using the most significant linear combination in terms of distinguishing two yield classes from MANOVA, soil pH, extractable Ca^2+ and Mg^2+, K saturation, clay content, and soil N to P ratio were identified as variables that were most related to cotton yield classification and resulted in two potential management zones. The lowest misclassification rate (27%) appeared in the two classes developed from the six influential variables and misclassification mostly located near the pivot boundary and near high to low or low to high yield transition areas, possibly due to changes in soil properties coupled with differences in weather-year patterns in the 3 years. Combined with other information, the delineated high and low yield classes can serve as potential management zones for making detailed management prescriptions for irrigated cotton on these soils.