Interpretation of yield data from a long-time soil fertility experiment

Yield data for ear corn and wheat grain are reported from a longtime soil fertility experiment on Tilsit silt loam in Muhlenberg County, Kentucky. The 31-year period covered is from 1914 to 1944, inclusive. The experiment includes plots receiving nothing or manure in combination with superphosphate, rock phosphate, ground limestone, and muriate of potash. Yields were obtained in a 4-year rotation of corn, wheat, and 2 years of mixed legume and grass hay during the last half of the period and in a 4-year rotation of corn, soybeans, wheat, and mixed hay during the first half. Various methods of treatment of the data are discussed, as well as the effect of treatment on yield trends. These include regression and analysis of variance. Graphs showing yield trends are fitted by the method of regression. All treatments gave significant positive regression coefficients except for plots receiving nothing or manure and ground limestone. Plots receiving nothing showed a downward trend in corn yields, showing a decided decline in productivity over the experimental period. Most yields showed a linear trend throughout. Phosphate fertilizer was found to be the most important factor affecting the yield trends. By means of analysis of variance, yield variation and differences were broken down into their component parts. A very significant difference between treatments, between years, and for the reduction in variance due to the trends in yields was found. Because of lack of replication of treatments on each crop in any one year, considerable information was lost. Replication of each treatment for each year in the rotation however, took the place of replication in any one year to a considerable extent. Variability of yield data for different years was evaluated by calculating coefficients of variability for the various treatments. Low-yielding treatments were considerably more variable than high-yielding ones. Calculations of correlation coefficients of corn yields with weather gave some measure of the effect of precipitation and temperature on yield and variability. Significant correlations of corn yields with weather data for critical periods were found.