A yield trial ends with a number per plot and little to explain the outliers. A gap in the stand, lodging before harvest or a combine error all end up in the dataset as genotype effects, unless someone remembers what happened on that plot.
PhenoScale imagery through the season, and especially at the end of it, builds a plot-truth record: Plant Count, Plant Gaps Percentage, Plant Lodging and the Plot Quality and Plot Quality Score indices. Cloverfield flags every plot whose development is inconsistent with its measured yield.
Compromised plots are corrected or excluded on documented grounds, cleaning the yield dataset and raising the statistical power of the whole trial, with evidence that holds up when a plot exclusion is challenged. Applicable to wheat, barley and canola yield trials.
