Humboldtian Diagnosis of Peach Tree (Prunus persica) Nutrition Using Machine-Learning and Compositional Methods

Abstract

Regional nutrient ranges are commonly used to diagnose plant nutrient status. In contrast, local diagnosis confronts unhealthy to healthy compositional entities in comparable surroundings. Robust local diagnosis requires well-documented data sets processed by machine learning and compositional methods. Our objective was to customize nutrient diagnosis of peach (Prunus persica) trees at local scale. We collected 472 observations from commercial orchards and fertilizer trials across eleven cultivars of Prunus persica and six rootstocks in the state of Rio Grande do Sul (RS), Brazil. The random forest classification model returned an area under curve exceeding 0.80 and classification accuracy of 80% about yield cutoff of 16 Mg ha−1. Centered log ratios (clr) of foliar defective compositions have appropriate geometry to compute Euclidean distances from closest successful compositions in “enchanting islands”. Successful specimens closest to defective specimens as shown by Euclidean distance allowed reaching trustful fruit yields using site-specific corrective measures. Comparing tissue composition of low-yielding orchards to that of the closest successful neighbors in two major Brazilian peach-producing regions, regional diagnosis differed from local diagnosis, indicating that regional standards may fail to fit local conditions. Local diagnosis requires well-documented Humboldtian data sets that can be acquired through ethical collaboration between researchers and stakeholders.

Conditioning Machine Learning Models to Adjust Lowbush Blueberry Crop Management to the Local Agroecosystem

“…As illustrated by Parent [ 46 ], the shape of such space is irregular and blob- or cloud-like. Arbitrarily delimiting the contours of a prosperous or successful region [ 10 , 91 , 92 ] should be avoided. According to Parent [ 46 ], interpreting a perturbation between a nutritionally imbalanced specimen and its optimum or successful target “should be done with a multivariate and compositional data perspective in mind.…”

“…Arbitrarily delimiting the contours of a prosperous or successful region [ 10 , 91 , 92 ] should be avoided.…”

Nutrient Diagnosis of Eucalyptus at the Factor-Specific Level Using Machine Learning and Compositional Methods

“…While the Law of the Maximum relies on tens of growth factors and countless factor interactions [20], «Jardins do Eden», where all factors are at their optimum levels, are rarely encountered. On the other hand, «ilhas encantadas» (enchanting islands) [18,26,27], where controllable factors are close to their optima under given combinations of uncontrollable factors, can be documented as successful Humboldtian loci where several yield-limiting factors have been handled adequately by local growers.…”

“…The suitability of downscaling regional standards for application at a local scale where myriads of factor interactions occur has been minimally addressed [18]. Errors on interactions [19] that involve environmental factors, genetics, nutrients and time may reduce diagnostic efficiency because factor effects are averaged across factors at a regional scale.…”

“…Humboldtian patterns can be extracted using methods of artificial intelligence to solve complex problems that are beyond human capabilities [24,25]. A heuristically simple factor-specific diagnostic approach is to compare defective and successful Humboldtian loci across a set of features using compositional and classification or regression machine learning (ML) methods [18,26,27]. In such a case, the assumption that factors other than the ones being addressed are equal or at optimum levels [28] is replaced by the assumption that documented factors other than the ones being addressed are comparable.…”