Towards a Better Understanding of the Long-Term Yield Response of Corn to the Repeated and Random Seasonal Effects of Fertilizer Nitrogen and Tillage.
J. H. Grove, Plant and Soil Science Dept, N-122 ASCN, Univ of Kentucky, Lexington, KY 40546-0091 and Eugenia Pena-Yewtukhiw, West Virginia Univ, Davis College of Agriculture, Forestry and Consumer Sciences, Division of Plant and Soil Sciences, 1104 Agricutlural Sciences Building, Evansdale Campus, PO Box 6108, Morgantown, WV 26506.
Natural processes resulting in yield have temporal and spatial frameworks. These frameworks determine characteristics such as magnitude, stability, and seasonal occurrence. In time and space, an action (management practice) has consequence/causes a reaction (an autocorrelation). Long-term experiments are not done because the investigator expects a simple effect of time, but because of the expectation of an interaction between time and one or more treatments/treatment combinations. These time by treatment effects may contain two components. First, the treatments may cause fixed (non-random) effects due to their repeated annual application to the same plot areas. Second, there may be a stochastic effect due to random annual seasonal climate, where treatment differences display dependence upon the general quality of the production year. The literature is replete with multi-year field research reports where the season or year "effect" was significant. This is oft explained as a simple effect of a good (high yielding), or a bad (low yielding), year on crop development. But treatment by season interactions are also observed, most often by plant geneticists reporting genotype by environment interactions. These interactions are often poorly characterized, hindering understanding. Our objective was to better understand annual yield response of corn (Zea mays L.) to fixed/repeated applications of fertilizer nitrogen (N) and tillage in a long-term field experiment. If, as expected, there was a time by treatment interaction, we desired to better understand/describe: a) the impact of random seasonal climate; and b) the role of fixed/repeated application of the treatments; on corn yield response. Corn yield data were taken from a 35-year (1970-2004) trial initiated into a Kentucky bluegrass (Poa pratensis L.) sod underlain by a deep, well-drained Maury silt loam (fine, mixed, mesic Typic Paleudalfs) located near Lexington, Kentucky. Annual mean temperature is 13„a C and average annual rainfall is 1140 mm, 40% of which occurs during the 175-day average growing season (May through September). The experiment was laid out in four split blocks with the four randomly assigned fertilizer N rates (0, 84, 168 and 336 kg N/ha) randomly imposed perpendicularly across the two tillage treatments (no-tillage [NT] versus moldboard plow [MP]). Individual plot size was 12.2 m long by 5.5 m (6 rows) wide. Moldboard plowing to 20-25 cm, followed by tandem disking to 8-10cm, was done one to two weeks prior to planting in late April to early May. Ammonium nitrate was surface broadcast within two weeks of planting. Corn yield was determined after hand harvest of 6.1m of the center two rows of each plot in early October. A winter rye (Secale cereale L.) cover crop was established each year. Yield data were subjected to analysis of variance, both by year and across years, using PROC GLM and PROC MIXED, as appropriate. Time by treatment interaction was found. Both the stochastic and fixed, repeated roles of time on the time by treatment interaction were examined using regression of annual treatment yield against an index of annual seasonal quality, temporal trends in annual treatment yield, and time series analysis with autoregressive moving average (ARMA) models. As random seasonal climate improved, MP corn yield increased more than NT corn yield, reflecting the greater moisture stress in the MP crop. The tillage related difference in responsiveness to seasonal quality was lowest in N fertilized corn and greatest at 0 kg N/ha. A deterministic trend for yield decline, especially at 0 kg N/ha, and more especially for MP corn, was observed, indicating loss of soil quality/productivity relative to the initial sod condition. Detrended yield, when subjected to time series analysis, resulted in "white noise" for yield of NT corn fertilized at 0 and 84 kg N/ha and also for yield of MP corn at 0 kg N/ha. At higher N fertilizer rates, yields of both NT and MP corn were not temporally independent. Rather, yields in these treatment combinations were subject to "carryover", residual effects related to yield magnitudes in prior years. ARMA model coefficients suggested that these carryover effects persisted between 2 and 3 years. We conclude that the time by treatment interaction on corn grain yield has manifested itself in this study in several ways. First, well N fertilized MP corn was most responsive to random seasonal climate, while unfertilized NT corn was least responsive. Second, soil quality was diminished with annual corn cropping without N fertilization, and especially with plowing prior to cropping. Third, N fertilized corn yield, especially MP corn, was more dependent on yields achieved in the previous 2 to 3 years.