Hyperspectral imaging coupled with microclimatic-based alert help targeted management of downy mildew (Hyaloperonospora parasitica (Pers.:Fr) Fr.) of wild rocket (Diplotaxis tenuifolia L. [D.C.])

Wild rocket, Diplotaxis tenuifolia L. (D.C.) (Brassicaceae), is a widely cultivated baby-leaf salad crop for the fresh high-convenience food chain. Downy mildew, caused by Hyaloperonospora parasitica (Pers.:Fr) Fr. here identified by Internal Transcribed Spacer region sequencing, is the key disease of this crop, favored by environmental and agronomic conditions occurring during cultivation. Digital monitoring by optoelectronic sensors coupled with the assessment of microclimatic parameters can help more targeted disease management. A plastic tunnel trial, carried out comparing plot treatments with Trichoderma atroviride TA35, laminarin and cell-wall extract of Saccharomyces cerevisiae LAS117, to fungicides and no treatment, was subjected to hyperspectral imaging and recording of microclimatic parameters, with the aim of defining symptom spectral features on the canopy and setting phyto-pathological alert, respectively. Correlation analysis between spectral signatures and disease incidence allowed for the selection of four hyperspectral vegetation indices, ZTM, VOG1, Red-Edge NDVI and HVI (working in the range 692-750 nm) able to distinguish early symptoms from intermediate and advanced ones. A Random Forest machine learning model including a few of predictive variables (400, 536, 696, 756, 948, 963 and 962 nm), allowed to classify healthy or diseased samples with 92% accuracy. The analysis of the micrometeorological data allowed the spatial-temporal characterization of the experimental area, with the definition of the points necessary for the correct monitoring of the parameters for the evaluation of the spatial trend in disease risk. In addition, a leaf wetness estimation model and a disease alert model were implemented in MATLAB.