Classification of Crop Yield Variability in Irrigated Production Fields

yield map represent grain mixed from a certain area, and some uncertainty is associated with the exact size Crop yield maps reflect stable yield patterns and annual random and geographical location of this area as well as meayield variation. Procedures for classifying a sequence of yield maps surement error. For the same location, this uncertainty to delineate yield zones were evaluated in two irrigated maize (Zea mays L.) fields. Yield classes were created using empirically defined is likely to vary from year to year because of different yield categories or through hierarchical or nonhierarchical cluster combine travel paths. Therefore, a single-year yield map analysis techniques. Cluster analysis was conducted using average is useful for interpretation of possible causes of yield yield (MY), average yield and its standard deviation (MS), or all variation but may be of limited value for more strategic individual years (AY) as input variables. All methods were compared SSCM decisions over mediumto long-term periods. based on the average yield variability accounted for (RVc). Methods Procedures must be developed to correct or eliminate in which yield was empirically classified into three or four classes recognizable errors of yield monitor measurement and accounted for less than 54% of the yield variability observed and integrate multiyear sequences of yield maps. Here, we failed to delineate high-yielding areas. Six to seven yield classes established by cluster analysis of MY accounted for 60 to 66% of the yield assume that a sequence of corrected and interpolated variability. Differences among cluster analysis methods were small yield maps, which need to be classified to delineate areas for MY as data source. However, fuzzy-k-means clustering had lower with different yield expectation within a field, has been RVc than other methods if used with the MS or AY data. The spatial obtained. Such classification will result in a map of past fragmentation of yield class maps increased in the order MY MS yield performance. With multiple years of georeferAY. Univariate cluster analysis of mean relative yield measured for enced yield data, repeating patterns and their more staat least 5 yr should be used for yield classification in irrigated fields ble natural causes may be separated from random variawhere six to seven classes appear to provide sufficient resolution of the yield variability observed. More research should be conducted to tion in each year, providing a basis for spatially varying develop methods that result in spatially coherent yield zones and to yield goals and other SSCM decisions. understand differences between rainfed and irrigated environments Interpretation and classification of multiple-year in the importance of mapping yield goals for crop management. yield maps has often involved empirical criteria or decisions on how many yield classes should be formed. Blackmore (2000) proposed an empirical classification G on-the-go yield mapping using comin which the sample mean and the coefficient of variabine-mounted yield monitors has become one of the tion (CV) were used to classify yield into groups such most widely used precision-farming tools. Yield monias high yielding and stable, low yielding and stable, and tors generate spatially dense data at relatively low cost, unstable. Pringle et al. (2003) proposed an “Opportunity potentially allowing characterization of the spatial and Index” for identifying fields with the greatest overall temporal yield variability. However, the analysis and potential for SSCM, which they calculated from the interpretation of yield map data has lagged behind yield magnitude of yield variation, its spatial structure, and monitor adoption by farmers. As more yield monitors empirical “thresholds” for both. Lark and Stafford are used and multiple-year yield data accumulated, (1997, 1998) used fuzzy-k-means clustering for pattern there is an increasing concern about how to process and recognition in multiple-year yield maps. Taylor et al. interpret these data for site-specific crop management (2001) attempted to create yield goal maps by aggregat(SSCM). ing 3 to 7 yr of maize yield data into larger cells, calculatOn a field average basis, grain yield measured by yield ing average past yields for different periods, and commonitors and certified electronic scales agrees within paring the different yield goals with the actual yields 2 to 5% (Doerge, 1997). With careful calibration and obtained. They concluded that there was a greater opoperation, yield monitors are sensitive to changes in portunity for classifying consistently high-yielding areas yield although a variable time delay exists and the grain than consistently low-yielding areas based on the mean flow through a combine resembles a diffusive process relative yield and temporal standard deviation (SD). (Arslan and Colvin, 2002). Individual data points on a These as well as other classification methods have not been evaluated using uniform data sets and statistical criteria that express how well spatial and temporal yield A. Dobermann, J.L. Ping, G.C. Simbahan, and R.B. Ferguson, Dep. of Agron. and Hortic., Univ. of Nebraska, P.O. Box 830915, Lincoln, NE 68583-0915; and V.I. Adamchuk, Dep. of Biol. Syst. Eng., Univ. Abbreviations: AY, yields in all individual years; CV, coefficient of of Nebraska, P.O. Box 830726, Lincoln, NE 68583-0726. Contribution variation; Dv, fractal dimension; ISODATA, Iterative Self-Organizing of the Nebraska Agric. Exp. Stn. Scientific J. Ser. Paper no. 14009. Data Analysis; KME, k-means cluster analysis; MS, mean and stanReceived 21 Jan. 2003. *Corresponding author (adobermann2@ dard deviation of yield; MY, mean yield; RVc, average yield variability unl.edu). across years accounted for by the classification; RVj, proportion of yield variability in one year accounted for by the classification; SD, Published in Agron. J. 95:1105–1120 (2003).  American Society of Agronomy standard deviation; SSCM, site-specific crop management; WAR, hierarchical cluster analysis using Ward’s method. 677 S. Segoe Rd., Madison, WI 53711 USA