Soil apparent conductivity measurements for planning and analysis of agricultural experiments: A case study from Western-Thailand

- The use of ECa for design and analysis of agricultural experiments was investigated.
- ECa measurements were used to explain variability in the height of maize.
- Estimation of treatment effects were more precise using ANCOVA with ECa than ANOVA.
- ECa and plant height were used to simulate three experimental designs.
- Power analysis demonstrated benefits if ECa is used as covariate in the analysis.

In experimental trials, the success or failure of agricultural improvements is commonly evaluated on the agronomic response of crops, using proper experimental designs with sufficient statistical power. Since fine-scale variability of the experimental site can reduce statistical power, efficiency gains in the experimental design can be achieved if this variation is known and used to design blocking, or some proxy variable is used as a covariate. Near-surface geophysical techniques such as electromagnetic induction (EMI), which describes subsurface properties non-invasively by measuring soil apparent conductivity (ECa), may be one source of this information. The motivation of our study was to investigate the effectiveness of EMI-derived ECa measurements for planning and analysis of agricultural experiments. ECa and plant height measurements (the response variable) were taken from an agroforestry experiment in Western Thailand, and their variability was quantified to simulate multiple realizations of ECa and the residuals of the response variable from treatment means. These were combined to produce simulated data from different experimental designs and treatment effects. The simulated data were then used to evaluate the statistical power by detecting three orthogonal contrasts among the treatments in the original experiment. We considered three experimental designs, a simple random design (SR), a complete randomized block design (CRB), and a complete randomized block design with spatially adjusted blocks on plot means of ECa (CRBECa). Using analysis of variance (ANOVA), the smallest effect sizes could be detected with the CRBECa design, which indicates that ECa measurements could be used in the planning phase of an experiment to achieve efficiencies by improved blocking. In contrast, analysis of covariance (ANCOVA) demonstrated that substantial power improvements could be gained when ECa was considered as a covariate in the analysis. We therefore recommend that ECa measurements should be used to characterize subsurface variability of experimental sites and to support the statistical analysis of agricultural experiments.