Spatial and temporal stability of corn grain yields

Year-to-year consistency of crop yields within a farm field is needed to use grain yield monitor data for site-specific management decisions such as yield goals for fertilizer recommendations. A 5-yr study was conducted from 1991 to 1995 to determine whether patterns of corn (Zea mays L.) grain yields are similar over a number of years and whether grain yields from one or more years can be used to predict grain yields for subsequent years. The experimental site was located at the Northern Cornbelt Sand Plain Management Systems Evaluation Area near Princeton, MN. The research area was 4.4 acres with soils mapped as three variants of the Zimmerman fine sand (mixed, frigid, argic, Udipsamment) and a Cantlin loamy fine sand (sandy, mixed, frigid, typic, Udipsamment). Continuous corn was grown from 1990 through 1995 after alfalfa (Medicago sativa L.) from 1981 through 1989. Cultural practices were applied uniformly to the 4.4 acre site each year. The 4.4 acres were divided into 60 grid cells (50 ft. by 60 ft.) and grain yields, corrected to 15.5% moisture, were determined by hand harvesting an area (two rows 20 ft. long) within each of the 60 grid cells. Differences between highest and lowest continuous corn grain yields in the research area were 72 bu/acre in 1991, 44 bulacre in 1992, 45 bu/acre in 1993, 51 bu/acre in 1994, and 57 bu/acre in 1995. Grain yields were not spatially consistent from year to year. Areas with better grain yields were not consistent from year to year, and conversely, poor production areas were not found in similar locations each year. Only 4 to 42% of the grain yield variability for a given year is accounted for by a knowledge of the grain yields from a previous year. The lack of grain yield stability as measured by ranked correlations on a sandy soil raises serious questions for the potential for use of this information. The data indicate that the use of grain yield maps for fertilizer recommendations on a site specific basis may not be possible or may require a much longer term database than the normally recommended 5 yr, unless there is a construct of inputs that explains the grain yield patterns each year.