Mr. Avioz Dagan

Nitrogen (N) is often regarded as the most critical nutrient and the growth-limiting factor in soil for ‎‎plant growth and often needs to be supplemented by N-fertilization to minimize yield ‎‎loss. However, the over-application of N fertilizers may cause nutrient imbalance ‎and contribute ‎to groundwater ‎contamination by nitrate (NO3-) leaching and NOx air pollution. Today the ‎evaluation of plant ‎nutritional status (including ‎N content) is mainly based on chemical ‎analysis of leaf ‎samples in the lab’. These are costly and time-consuming, not represent the agricultural system’s spatial and temporal variability. This study suggests ‎investigating the ability to combine Sentinel-2 ‎and VENµS data with unmanned aerial vehicles (UAV) to derive ‎ canopy nitrogen content ‎‎(CNC) in citrus orchards. A new ‎framework to infer the N content in citrus-tree ‎canopy-level using spectral data ‎and vegetation indices with the ML algorithm wad ‎suggested. The framework includes six ‎steps (1) leaf sampling for N content data, (2) ‎image preprocessing of UAV and ‎satellite data, (3) segmentation of canopies and estimation of plant area index, (4) feature extraction, (5) model ‎calibration and validation based on ML models, ‎and (6) the development of site-specific N ‎management model. We integrated data from Sentinel-2 ‎and VENµS satellites ‎and UAV images using bi-monthly data collected in the past three years to estimate CNC. Several ‎machine learning algorithms were tested to assist tree CNC derived from UAV data. The suggested model was proven flexible and could include different or additional variables, enabling the delineation of site-specific nitrogen management (SSNM) zones in orchards. These could be used to ‎reduce the need for chemical analysis of the leaf tissue and optimizes the CNC monitoring ‎by considering the spatial and temporal variability in the citrus ‎orchard on different scales.