Abstract

Agronomic researchers are increasingly accountable for research programs and outcomes relevant to producers. Participatory research—where farmers assume leadership roles in identifying, designing, and managing on-farm field-scale research—addresses this directive. However, replication is often unfeasible at this level of scale, underscoring a need for alternative methods to estimate experimental error. We compared mean square errors to evaluate: (i) within-field variability for estimating experimental error (in lieu of replication) and (ii) classified within-field variability, using apparent electrical conductivity (EC_a), for estimating plot-scale experimental error. Eight 31-ha fields, within a contiguous section of farmland (250 ha), were managed as two replicates of each phase of a no-till winter wheat (Triticum aestivum L.)–corn (Zea mays L.)–millet (Panicum miliaceum L.)–fallow rotation. The section was EC_a–mapped (approximately 0- to 30-cm depth) and separated into four classes (ranges of EC_a). Georeferenced sites (n = 96) were selected within classes, sampled, and assayed for multiple soil parameters (0- to 7.5- and 0- to 30-cm depths) and residue mass. Within-field variance effectively estimated experimental error variance for several evaluated parameters, supporting its potential application as a surrogate for replication. Comparison of data from the field-scale experimental site to that from a nearby plot-scale experiment revealed that EC_a–classified within-field variance approximates plot-scale experimental error. We propose using this relationship for a systems approach to research wherein treatment differences and their standard errors, derived from EC_a–classified field-scale experiments, are used to roughly evaluate treatments and identify research questions for further study at the plot scale.

Copyright © 2003. American Society of Agronomy. Published in Agron. J.95:602–613.