Syndrome “Basses Richesses” (SBR) is a rapidly emerging sugar beet disease in central Europe having a severe economic impact on the sugar beet industry and thus commanding a demand for its control. The cultivation of tolerant varieties is a promising method to reduce SBR. Digital plant phenotyping can support the screening process for tolerant varieties by characterizing traits of interest and quantifying tolerance. This research provides foundational work for digitally phenotyping SBR. Morphological and spectral traits were analyzed with machine learning, supporting disease monitoring and screening for tolerant varieties under controlled conditions. A susceptible sugar beet variety was infected with the dominant causal agent of SBR Candidatus Arsenophonus phytopathogenicus (ARSEPH). Hyperspectral images of the canopy were recorded weekly between 20 and 62 days after inoculation (dai) and segmented by leaves and petioles. Sixty-seven dai each leaf was two-dimensionally (2D), and each taproot three-dimensionally (3D) imaged by angle-corrected 2D imaging and structured-light 3D scans, respectively. The results indicate substantial decreases in leaf area (19.7%), leaf length (6.9%), leaf blade length (13.1%), and leaf blade width (12.1%) resulting from ARSEPH-infection. The most important wavelengths for machine-learning-classification of ARSEPH-infected sugar beet were from the petioles (97% accuracy) in the range 623 to 659 nm and 421 to 432 nm. The 22 most relevant taproot 3D parameters were evaluated with Boruta-SHAP based on their importance to characterize SBR-induced taproot-deformation. Certain value- and spatial-regions were characteristic, indicating thresholds for 3D parameters and taproot-regions to analyse when comparing varieties.
Detring et al. (Tue,) studied this question.