Geomatic Techniques to Support Phytosanitary Products Tests whithin the EPPO Standard Framework

Background and Research Gap: European agricultural research for Plant Protection Product (PPP) development relies on statistical hypothesis testing under European and Mediterranean Plant Protection Organization (EPPO) standards. Traditional statistical approaches recognized by EPPO attempt to address environmental variability through experimental design features such as randomized controls and blocking. These methods are fundamentally limited by their reliance on a priori identification of variance sources, where experimentalists must subjectively identify environmental variability patterns before data collection based solely on human experience and field observation. Geostatistical methods offer a mathematically rigorous alternative by enabling the estimation of spatial environmental variability after data collection through variograms or spline fitting techniques. While these methods are recognized by EPPO and have demonstrated superior performance in modeling environmental heterogeneity, they require spatially referenced observations (each data point must have precise spatial coordinates). This requirement creates a significant implementation barrier, as traditional manual assessment methods used in PPP trials typically do not capture spatial coordinates, making it practically impossible to apply geostatistical approaches despite their theoretical advantages. This research gap (the absence of practical methods to generate spatially referenced datasets that would enable geostatistical analysis within EPPO-compliant trials) has prevented the widespread adoption of more robust statistical approaches in agricultural field studies. Research Objectives: This research addresses the identified gap by investigating the applicability of geomatics technologies for recording spatially referenced observations in compliance with EPPO standards. The objective is to establish practical methods for generating georeferenced datasets that enable geostatistical analysis, demonstrating how these techniques can facilitate the adoption of more robust statistical approaches in agricultural research within the EPPO standard framework. Methodology: This work considered three aspects of geomatics technologies applicability in this context: 1. counting, using deep learning object detectors to count maize seedlings on orthomosaics; 2. scoring, using machine learning regressors to score phytotoxicity via photogrammetric multispectral imaging and custom feature extraction; 3. classify, using anomaly detection to classify healthy or deseased plant tissues via pre-trained models.