DUGi

Compositional Meta-analysis of the Nutrient Profile of Potato Cultivars

http://dugi.udg.edu/item/http:@@@@hdl.handle.net@@2072@@273646

While several potato (Solanum tuberosum L.) cultivars of different maturity groups (e.g.early, mid-season, late) are being selected each year as a result of successful breeding fordisease resistance and market requirements, their nutrient management is based on pastexperience and few experiments. Nutrient profiles from leaf analysis can guide fertilizationand liming programs of potato cultivars. Since leaf analytical data are strictly positive andcompositional, nutrient profiling using raw data is spoiled by non normal distribution,resonance and spurious correlations. Compositional data analysis provides log ratiotransformations that avoid such problems. Our objective was to derive nutrient profiles fromtissue analysis using isometric log ratio (ilr) coordinates and meta-analysis for classificationof cultivars into uniform nutrient management groups. The dataset comprised 678 potatofields producing more than 28.5 Mg marketable tuber ha-1, i.e. above Quebec average, of theearly-, mid-, and late-season cultivars. The first mature leaf from top was sampled at thebeginning of flowering for N, P, K, Ca, and Mg analysis. Anionic (N, P) and cationic (K, Ca,Mg) nutrients were arranged into binary partitions representing positive and negative nutrientinteractions. Groups of cultivars were compared to ‘Superior’ using ilr mean and standarddeviation in the mixed model of meta-analysis. We minimized the within-group heterogeneity(I2value) by allocating cultivars iteratively between ilr groups. We derived group-specific ilrnorms to compute the Aitchison distance. The critical value for nutrient imbalance was 0.38.To guide correcting nutrient deficiencies with appropriate nutrient management techniques,nutrient composition can be altered numerically by a perturbation vector on nutrients that leadto the largest and most negative ilr differences from ilr norms until the Aitchison distancefalls below critical value

Universitat de Girona. Departament d’Informàtica i Matemàtica Aplicada