Answers are grounded in this paper and cite their sources.
Physiological fruit drop is a major determinant of crop load and yield in ‘Rojo Brillante’ persimmon and is closely linked to irrigation management, yet its severity has not been addressed as an ordinal prediction problem. This study evaluated whether fortnightly fruit-drop severity, recorded in 2015–2019 in trees grafted onto Diospyros lotus and Diospyros virginiana under different irrigation treatments, could be predicted one fortnight in advance with four ordinal classification algorithms (cumulative link model, LogitBoost, random forest and support vector machine). Nested and reduced predictor sets were evaluated with leave-one-year-out cross-validation against a persistence and a seasonal baseline, with Quadratic Weighted Kappa (QWK) as the pre-specified primary metric. LogitBoost with the complete predictor set achieved the highest agreement (QWK = 0.718), outperforming the persistence baseline (QWK = 0.510) in all five years and the seasonal baseline (QWK = 0.591) in four. Reduced models with cumulative growing degree-days and rootstock reached similar agreement, indicating that predictive ability arises mainly from seasonal phase. In contrast, the previous fruit-drop count and additional agroclimatic predictors improved the discrimination of low and moderate drop. Because irrigation enters the models through the cumulative hydric balance, the approach could support the evaluation of irrigation scenarios, once validated in other orchards.
No takes yet. Share an insight, caveat, or question.
Román et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: