Review Papers Claim by Claim

Visual summary — main empirical facts (visualize first)

Primary citation

Core dataset and analyses are from the focal paper:

Method validation reference (image software)

Critical evaluation — strengths

• Transparent, practical pipeline: The paper documents camera settings, color card calibration (2016), ImageJ plugin usage, RGB→HCL conversion (Smith 1978 formulas), and thresholding steps, enabling reproducibility at the level of image processing when similar equipment is used Image processing methods.
• Appropriate statistical modeling: Mixed models + ASReml‑R for variance components, separable covariance structure for GE, BLUPs and reliabilities, and testing of storage × genotype interactions — standard, rigorous quantitative genetics workflows for plant breeding Statistical approach.
• High phenotype reliability: Plot-based h2 >0.75 indicates the image phenotypes have low within-plot noise relative to genotype signal under the study conditions, supporting use for selection and mapping within similar germplasm and environments Heritability results.

Critical evaluation — limitations & blindspots

• Limited genetic diversity and sample size: Only 15 selected red clones (effectively 13 per year) — a highly selected breeding panel — reduces ability to generalize the color-collinearity finding and limits power to discover quantitative loci or G×E interactions beyond this program's germplasm Sample composition limitation.
• Single-location field design: All trials at one research station limit environmental breadth (soil, climate, management), so high genetic correlations across the three local environments do not imply stability across broader production areas or under different management Single-location limitation.
• Potential measurement bias from thresholding: Semiautomated color and skinning thresholds were followed by visual inspection — appropriate but subjective decisions (threshold choice, exclusion of defects) can create systematic biases if not standardized or automated across datasets; color card used only in 2016 (not 2015), which can introduce inter-year calibration inconsistencies Calibration & thresholding caveats.
• Limited mechanistic linkage: HCL reflects perceived color; the paper did not measure anthocyanin chemistry (e.g., total anthocyanins, specific glycosides, acylation) which are known to affect color stability and lightness/chroma changes in storage — linking image metrics to pigment chemistry would strengthen causal interpretation (references: anthocyanin stability, acylation effects) Anthocyanin stability review.

Interpretation of the surprising collinearity (Hue, Chroma, Lightness ≈ 1 axis)

The authors report that a single PC explains ~83–86% of color variation. Two plausible, non-exclusive explanations are:

1. Biological constraint: in these red-skinned genotypes, pigment biosynthesis/regulation (e.g., shared transcriptional control such as the D locus R2R3‑MYB) may scale pigment abundance and spectral characteristics together so hue, saturation and brightness co-vary Jung et al. 2009 — D locus (anthocyanin regulator).
2. Sampling selection: narrow germplasm (closely-related breeding lines and elite checks) can compress multi-dimensional variation into one dominant axis; the authors themselves note that unselected F1s showed weaker correlations (data not shown), supporting sampling artifact as contributor Authors.

Reproducibility checklist (what to report/standardize when repeating or scaling this work)

1. Provide raw images and ImageJ macro(s)/scripts and exact threshold values per trait; include 2015 images and show cross-year calibration method (the paper provides supporting info but no explicit accession numbers) Data availability note.
2. Use standardized color calibration (MacBeth card) in all years and report white-balance procedure/matrices; include reference patches in every photo to permit cross-year normalization (2016 used MacBeth, 2015 did not).
   
3. Report per-image masks and per-plot summary tables (n tubers used, excluded images) and annotate excluded defect regions to quantify threshold-exclusion bias.
   
4. Add analytical chemistry (total anthocyanins, anthocyanidin profiles by HPLC-MS) for a subsample to link HCL changes to pigment abundance/chemistry (literature: methods for anthocyanin extraction/identification are standard and informative) HPLC anthocyanin methods.

Practical recommendations for breeders and phenotyping engineers

• Image-based HCL measures are usable as plot-level selection targets because of high h2 and BLUP reliability — implement at scale with standardized light box + color card and automated ImageJ macros; target storage retention (chroma) as a selectable trait since storage effect is consistent but has low G×E in this panel Storage effect and selection implications.
• For genetic mapping or QTL discovery, expand germplasm to segregating populations (F1 progeny, diversity panels) and increase environment diversity (multiple locations/years) to capture allelic variation and environment‑specific effects (the authors note mapping in larger F1s is ongoing).
  

What would change my interpretation (falsification tests)

• Release of raw images and ImageJ macros showing inconsistent thresholding across genotypes or years (indicating the high h2 was an artifact) would weaken conclusions about phenotype reliability.
• Chemical assays showing anthocyanin amounts and composition do not correlate with HCL changes (e.g., a genotype loses anthocyanin but HCL remains similar) would break the assumed link between pigment chemistry and image-derived color metrics.
• Repetition across diverse locations revealing strong G×E (low genetic correlations) would limit the use of single-location BLUPs for broader selection.

Concise action plan to extend the work (3 priority experiments)

1. Cross-site replication: Run identical imaging pipeline (same camera, lightbox, MacBeth card, ImageJ macros) at 3 contrasting production locations (different soils/climates) with a larger, partially overlapping genotype set (selected elites + segregating F1s). Outcomes: test genetic correlations across real-world environments and G×E for color/skin set. (Cost: moderate; timeline: 2 seasons.)
2. Chemistry–image linkage: For a strategically chosen subset (~6 genotypes spanning PC1), sample tuber skin for HPLC‑MS anthocyanin profiling (total anthocyanins, main aglycones, acylation patterns) at harvest and 6-week storage; correlate pigment metrics with HCL to quantify chemical drivers of chroma/lightness changes. (Cost: moderate; timeline: one season plus lab analyses.)
3. Automated segmentation & uncertainty quantification: Replace semiautomatic thresholding with a reproducible ML segmentation (U-Net or classical color clustering with standardized calibration) and propagate segmentation uncertainty into trait variance components (e.g., bootstrap masks) to test robustness of heritability estimates to image-mask variability. (Cost: low–moderate; timeline: months.)

Conclusions — balanced, evidence-weighted

The study convincingly demonstrates that straightforward image-based HCL color metrics and area-based skinning percentage, measured with ImageJ from standardized photographs, produce high plot-level heritability and consistent BLUPs for a small, selected set of red potato clones at a single research station; storage causes measurable, genetically consistent loss of chroma and lightness. The method is therefore ready for internal breeding use and as a phenotype for genetic mapping within similar germplasm. However, to claim broad generality (other cultivars, locations, and chemical mechanisms), the next step must link image phenotypes to pigment chemistry and expand the germplasm and environmental breadth; otherwise the observed collinearity and stability conclusions risk being specific to this breeding program's material.

Author reviews (quick links)

All claims above are grounded in the focal paper and methodological references cited inline; see the main DOI for full methods and supporting information:
Caraza‑Harter & Endelman 2020 (primary)