Defining Yield-Based Management Zones for Corn–Soybean Rotations

of information on spatial variability in yield practically possible but the utility of this information for MZ delinUnsupervised clustering has been proposed for developing geoeation remains largely unexplored. referenced agronomic information into management zones (MZs). Our objectives were to use fuzzy c-means clustering to identify yield-based A decade ago, little was known about the spatiotemMZs, and to compare spatial association and agreement among corn poral variability in corn and soybean yields (Jaynes and (Zea mays L.) yield-based (CYB) MZs, soybean [Glycine max (L.) Colvin, 1997). Numerous, recent studies have used corMerr.] yield-based (SYB) MZs, and published soil survey map units. relation and regression to examine temporal stability of Six years of yield monitor data (three per species) from four fields spatial yield patterns within fields, and a common result were used with the clustering software MZ Analyst. Clustering success has been the lack of stable yield patterns over time. was evaluated with four performance measures. Two measures of variFor example, Lamb et al. (1997) found that corn yields ance reduction and the fuzziness performance index (FPI) indicated observed in 1 yr accounted for between 4 and 42% of clustering optimization with 4 to 6 MZs. In contrast, the normalized the variation in yields observed in subsequent years. classification entropy (NCE) indicated that yield data were optimally Likewise, Jaynes and Colvin (1997) reported rank correorganized with only 2 MZs. On average, the 4-MZ delineation reduced the yield variance to 40% of the whole field variance (corn within lation coefficients of –0.09 to 0.54 and 0.03 to 0.52 for CYB MZs and soybean within SYB MZs); mean relative yields within within field comparison of multiyear corn and soybean MZs were significantly different from each other, ranging from 23% data, respectively. The lack of strong temporal correlabelow to 12% above the whole-field mean. With 4 MZs, CYB and SYB tion among yield data sets collected from a given field MZs were significantly associated in all fields, but weighted agreement has been cited as evidence that management based on between CYB and SYB MZs was only slight (0.06 Kw 0.34), site-specific prediction of yield may not be successful. indicating crop-specificity in MZ delineation. In general, highest yieldHowever, a MZs approach requires that we identify ing MZs were significantly associated with areas mapped as a poorly areas within a field that behave similarly through time. drained, level soil series while lower yielding MZs corresponded to Jaynes et al. (2003) suggest that such MZs may be identimap units for eroded or more slopping soils. However, clustering yields fiable and useful for management even when fields exby soil series reduced yield variance less than unsupervised, yield-based clustering. Routine application of MZ Analyst likely requires more hibit poor temporal stability in relative yield patterns; decision support for identifying clustering success. they advocate cluster analysis instead of correlation and regression for converting long-term yield data into management information. Unsupervised clustering algorithms have been proposed for delineating MZs from A premise of precision agriculture or site-speyield monitor data (Lark and Stafford, 1997 and Stafford cific management is that management can be imet al., 1998). This multivariate clustering approach groups plemented on a spatial scale smaller than that of a whole similar observations into distinct classes but does not field. A major barrier to widespread adoption remains require the user to direct or train the classification algothe lack of simple decision rules and protocols on how rithm with benchmark data, a requirement for superto practically and routinely delineate such management vised clustering. Because they do not require any a priori zones (MZs). Since the beginning of the precision agritraining, unsupervised clustering techniques are simpler culture technology era, patterns of yield variability have to apply than supervised techniques and, therefore, may been considered important for variable rate nutrient be better suited to general, on-farm applications (Fridgen management because productivity expectations or yield et al., 2004). goals influence rate recommendations in much of the In the eastern cornbelt of the USA the dominant Midwest (Lamb et al., 1997). Many university recommenproduction system is corn grown in annual rotation with dations for fertilizer management are based on identisoybean, and recommendations, especially for fertility fying soil productivity potential and providing fertilizer management, are often for the whole rotation and not to compliment the soils native nutrient supply in meeting for the crop within the rotation. Temporal variation the needs of a high yielding crop. Theoretically, consisin rainfall pattern coupled with the varying drainage tent, within-field variation in yield should reflect withinpotential of the major agricultural soils are a governing field variation in soil productivity potential and related, factor for year-to-year variation in whole-field yields as soil-specific input needs. Widespread availability of comwell as the within-field spatial variation in yield producbine-mounted yield monitors has made the collection tivity potential. Numerous reports have documented the extent to which both corn and soybean yields can be Department of Agronomy, 3351 Lilly Hall of Life Sciences, 915 W. State St., Purdue Univ., West Lafayette, IN 47907-2054. Received 21 effected by inadequate drainage (reviewed by Evans and Aug. 2004. *Corresponding author ([email protected]). Fausey, 1999); where needed, artificial drainage enhancePublished in Agron. J. 97:1115–1128 (2005). Site-Specific Management Abbreviations: ANOVA, analysis of variance; CYB, corn yield-based; FPI, fuzziness performance index; MZs, management zones; NCE, doi:10.2134/agronj2004.0220 © American Society of Agronomy normalized classification entropy; SYB, soybean yield-based; YB, yield-based. 677 S. Segoe Rd., Madison, WI 53711 USA 1115 Published online June 17, 2005