Modelling of codling moth-canopy interactions as a new approach to variable-rate pesticide application in apple orchards

Variable-rate pesticide application (VRA) technologies have advanced considerably in recent years; however, current approaches continue to overlook plot-scale pest distribution, which may compromise control efficacy. This study assesses the relationship between Cydia pomonella spatial patterns and canopy variability in an apple orchard to evaluate the feasibility of prescribed VRA based on pest-canopy interactions. A total of 18 georeferenced traps were deployed across a 1.40 ha plot, and tree canopies were scanned over two consecutive years (2024–2025) using ground-based LiDAR. Treatment-threshold exceedance was modelled via standard and mixed-effects logistic regression, incorporating LiDAR-derived canopy features as predictors, with model selection guided by residual spatial autocorrelation (Moran's I) and predictive performance. Three VRA scenarios were evaluated: (i) canopy-driven, (ii) canopy- and pest-driven, and (iii) model-driven selective VRA. Scenarios (i) and (ii) used block kriging and k-means clustering to delineate management zones, while scenario (iii) generated prescription maps directly from model outputs. Canopy cross-sectional area yielded very satisfactory prediction results, indicating that pest incidence was predominantly associated with smaller trees. These results indicated that pest–canopy interaction must be properly quantified and spatially modelled to enhance decision-making when opting for VRA. More targeted approaches may further restrict treatments to only those areas with a high likelihood of exceeding treatment thresholds.