Table 1.

Informative datasets (features, target variables) and system boundaries used to run machine learning and compositional data analysis models on fruit production systems at a plot scale.

Method	System Closure	Feature	Target Variable
Machine learning	Plot boundary	Tissue test (diagnostic tissue): N, P, K, Mg, S, Cu, Fe, Zn, Mn, B, Na, Al, … Soil test: pH, sand, silt and clay, organic matter, P, K, Ca, Mg, exchangeable acidity, … Others: cultivar (clone), rootstock, year, semester, plot number, well number, soil classification, management practices (tillage, cover crop, training method, pest management, …), meteorological data, harvest method, …	Fruit yield Fruit yield class Nutrient offtake Fruit quality
Compositional data analysis	Measurement unit	Tissue test (diagnostic tissue): N, P, K, Mg, S, Cu, Fe, Zn, Mn, B, Na, Al, … Soil test (0–20 cm, sometimes 20–40 cm): sand, silt and clay, organic matter, P, K, Ca, Mg, exchangeable acidity, …	Euclidean distance Clr differences Perturbation vector