Little is known about the spatial structure of yield across fields, no
r of the temporal stability of this structure. We determined the spati
al structures of grain yield for 6 yr of data and examined the stabili
ty of these structures over time. Yields were measured each year at 22
4 locations on an 8 x 24 grid within a 16-ha field, using plot combine
s. The field was planted either to corn (Zea mays L.) or soybean [Glyc
ine max (L.) Merr.] in an alternate-year rotation. Soybean yields aver
aged 3.16, 3.14, and 3.12 Mg ha(-1) in 1990, 1992, and 1994. Corn yiel
ds averaged 9.30, 9.34, and 5.02 Mg ha(-1) in 1989, 1991, and 1993 wit
h the lower yield in 1993 a result of excessive rainfall that gear. Th
e large-scale deterministic structure of the spatial yield data was co
nstructed for each year by median polishing. The trend surfaces were d
ifferent for each gear, indicating a lack of temporal stability in lar
ge-scale structure, although the trend surfaces shared some similar fe
atures. Overall, the deterministic structure accounted for about 25% o
f the overall yield variance. The small-scale stochastic spatial struc
ture was determined by computing variograms of the yield residuals aft
er subtracting the trends. Variograms showed strong spatial structure
of the yield residuals, with zero nugget and correlation ranges approa
ching 150 m. Variograms differed from year to year and were not relate
d to the crop being grown. The range was significantly correlated to p
recipitation, Increasing with increasing total growing-season rainfall
. Long-term monitoring of yield will be necessary to fully characteriz
e its spatial-temporal pattern.