Proceedings

Find matching any: Reset

Add filter to result:

Crop Rotation Impacts ‘Temporal Sampling’ Needed For Landscape-defined Management Zones

¹E. M. Pena-Yewtukhiw, ²J. Grove

1. West Virginia University
2. University of Kentucky

Yield and landscape position are used to delineate management zones, but this approach is confounded by yield’s weather dependence, causing yield to evidence temporal variability/lack of yield stability. Management options (e.g. crop rotation) also influence yield stability. Our objective was to build a model that would describe the influence of crop rotation on the temporal yield stability of landscape defined management zones. Corn (Zea mays L.) yield data for two rotations, monoculture corn (C-C) and corn alternating yearly (W/S-C) with winter wheat (Triticum aestivum L.)/double-crop soybean (Glycine max L. Merr.), taken in four landscape (shoulder, upper backslope, lower backslope, footslope) management zones, were obtained from a 21-year rotation study established near Lexington, Kentucky. This model would indicate the number of yield maps (year samples) necessary to spatially and temporally characterize the site. Yields were evaluated for spatial and temporal stability by Spearman rank correlation, and time series analysis (Box-Jenkins methodology). The 21-year average yields were lower for C-C (8.3±2.6 Mg ha^-1) than for W/S-C (9.6±2.7 Mg ha^-1). A plot of yield versus time by landscape zones exhibited slightly positive linear trend (0.13 Mg ha^-1 yr^-1) for both rotations. The rank correlations among landscape positions were above 0.80, however rotation choice determined similarities between positions. After removing linear yield trend, Box –Jenkins time series analysis found that C-C yield exhibited greater temporal lag (less yield stability) than W/S-C yield, though this trend was not equally true across the four zones. The lag predicting C-C yield was greatest (5 years) in the upper backslope, while upper backslope and footslope W/S-C yields exhibited maximal autocorrelation at lags of one to three years in length. These 21-year time series models indicate that similar management zones would require a minimum of three prior years of yield data in order to forecast their yield response behavior.

Keyword: Temporal sampling, ARMA models, Spearman rank correlation, corn rotations

Download paper