MULTIVARIATE CROP PRODUCTIVITY ZONES IN THE ALABAMA COASTAL PLAIN

Various data are used to develop management zones for site-specific crop production. Most evidence indicates that the technique used for zone development is crop and management dependent. The objective of this research is to evaluate which field-scale data are most appropriate for developing management zones for characterizing crop productivity and variability over multiple growing seasons and managements. Specifically, we are evaluating: 1) if field-scale crop yield variability is better described in zones derived from temporal or static data, and 2) the relationships between zone development approach and soil management system. This study was conducted on a field-scale (20 acre) experiment evaluating the interaction of soil management systems (conventional versus conservation) with soil landscapes on a site in the Alabama Coastal Plain. Six replications in a cotton (Gossypium hirsutum L.)- corn (Zea mays L.) rotation traverse the landscape. Soil landscapes range from Typic Paleudults on well drained uplands to imperfectly drained Oxyaquic Paleudults in drainageways. Field-scale data include satellite remote sensing imagery, terrain attributes generated from a LiDAR derived digital elevation model (DEM), field-scale electrical conductivity, and a first-order soil survey (1:5000). Management zones were developed using fuzzy k-means clustering, and geo-referenced crop yield data have been collected (2001-2006). Our results indicate that all evaluated data were generally suitable for characterizing crop productivity and variability using a clustering approach. As expected, satellite remote sensing data collected in season were more highly related to yield compared to terrain and soil variables. The relative effectiveness of these data for describing yield variability is most dependent on crop, and somewhat dependent on management