Using an artificial intelligence-based model to identify the nitrogen state in a plant by spectroscopy to examine the relationships between minerals and metabolites and develop a tool for optimal fertilization in orchards
Research Funding: Chief Scientist, Israel Ministry of Agriculture and Rural Development
Research team: Dr. Or Sperling (PI), Prof. Uri Yermiyahu (CI), and Dr. Tarin Paz-Kagan (Co-PI)
MSc: Guy Mordechai Mizrahi
AI-Powered Spectral Diagnostics for Precision Fertilization in Orchard Crops
Excess nitrogen fertilization remains difficult to detect in orchard systems, primarily due to low nitrogen application rates and the lack of accessible, empirical tools for monitoring plant nutrient status. In our previous research, we identified a strong relationship between leaf nitrogen and starch concentrations, allowing for the early detection of nitrogen surplus. This led to the development of a novel diagnostic index (patent pending) to support nitrogen optimization in orchard crops. However, a key limitation of existing tools lies in their low temporal resolution; even in experimental settings, measurements are typically limited to a single or a few sampling dates per season. To overcome this constraint, we demonstrated in a separate study that leaf nitrogen and starch content can be accurately estimated using light reflectance spectroscopy in the 800–2,500 nm range. This technique offers a fast, cost-effective, and non-destructive alternative to traditional laboratory analysis. Building on this foundation, we propose creating a comprehensive system for tracking spatial and temporal variations in mineral and starch concentrations in orchard crops using spectral data and artificial intelligence. This research will position Israeli orchard agriculture at the forefront of: Advanced plant diagnostics using spectroscopy and AI—precision nutrition based on frequent, scalable, and non-invasive measurements. We aim to eliminate the dependency on costly, time-consuming chemical analyses, transforming laboratory-grade spectroscopy tools into industrial field solutions. This approach mirrors existing commercial applications such as oil quality analysis, feed protein estimation, and carbohydrate detection in flours. The development of high-frequency, branch-level measurement systems will result in an expansive database characterizing diverse growing conditions across Israel. These data will support dynamic,
