Answers are grounded in this paper and cite their sources.
Purpose. Sugarcane technologies are often evaluated in commercial strip trials with a few crop cuts and limited replication. We asked what Sentinel-2 imagery and georeferenced sampling add to such a trial, and how much independent evidence their dense sampling contains. Methods. A third-ratoon trial under natural nematode infestation (nine strips, three treatments in a repeated sequence) was reconstructed from archived materials: a one-year cloud-screened Sentinel-2 baseline of the normalised difference vegetation index (NDVI), 27 NDVI-stratified crop cuts and 90 nematode points sampled twice. Analyses combined NDVI to yield calibration, residual variograms, a random-strip mixed model, exact permutation and an independently built pre-treatment covariate. Results. A field-scale trend ran along the strip sequence, aliased with treatment position. With an intraclass correlation of 0.25, the 90 points carried the information of 27.7. For the main nematode contrast, p was 0.004 with points as independent, 0.162 in the mixed model and 0.046 to 0.095 by permutation; five defensible strip-level specifications gave 0.023 to 0.61. More points per strip could cut its standard error by at most 12%; 80% power needs about ten strips per treatment. Root Pratylenchus fell by 63 to 67% under both bionematicide treatments. Two pre-treatment covariates, correlated at 0.65, moved a crop-cut contrast from p = 0.176 to 0.048. Conclusion. The digital layer makes a strip trial auditable when the strip is the unit. It densifies sampling but adds no experimental units: with nine strips, the choice of model and covariate carries more of the p-value than the data do.
No takes yet. Share an insight, caveat, or question.
Otoboni et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: