About the project

Project titlePrediction of maize yield potential using machine learning models based on vegetation indices and phenological metrics from Sentinel-2 multispectral satellite imagery
AcronymAgroVeFe
Contract no.581-UNIOS-30
Host institutionJosip Juraj Strossmayer University of Osijek, Faculty of Agrobiotechnical Sciences Osijek
Principal investigatorProf. Ivan Plaščak, PhD
Duration1 October 2025 – 30 September 2029 (48 months)
BudgetEUR 130,000.00
FundingEuropean Union – NextGenerationEU, National Recovery and Resilience Plan 2021–2026

Project summary

Global challenges such as climate change, shrinking agricultural land and a growing world population call for innovative solutions that increase the efficiency and sustainability of agriculture. Precision agriculture can optimise agronomic inputs, but its adoption in Croatia is limited by the high cost of sensors and yield mapping systems. AgroVeFe addresses this barrier by predicting maize yield potential with machine learning models based on vegetation indices and phenological metrics derived from free Sentinel-2 imagery. Reference yield data come from precision harvesting, and the spatial relationship between yield potential and soil electrical conductivity (EC) will be analysed to optimise field management. The project will identify the optimal combination of indices for reliable prediction, reduce costs by removing the need for expensive sensors, and strengthen cooperation with the agricultural sector, supporting the green and digital transition of Croatian agriculture.

Objectives

  • robust machine learning models for predicting maize yield potential from vegetation indices and phenological metrics
  • optimised precision agriculture combining satellite data, soil electrical conductivity and ISOXML zone maps
  • research infrastructure for field measurements, data analysis and dissemination
  • education and knowledge transfer to scientists, students and farmers
  • open datasets and tools following open science principles
  • validation of the models across regions and agro-ecological conditions

Work packages

WP1Field and laboratory equipment – Prof. Ivan Plaščak
WP2Collection and harmonisation of yield mapping data – Asst. Prof. Ivan Vidaković
WP3Sentinel-2 satellite image time series – Prof. Ivan Plaščak
WP4Vegetation indices and phenological metrics – Asst. Prof. Dorijan Radočaj
WP5Machine learning modelling of yield potential – Asst. Prof. Dorijan Radočaj
WP6Yield potential prediction and zoning, ISOXML maps – Asst. Prof. Željko Barač
WP7Soil EC scanning and correlation analysis – Prof. Ivan Plaščak
WP8Dissemination of research results – Asst. Prof. Domagoj Zimmer

Expected results by 2029

  • at least 15 scientific papers in journals indexed in Web of Science and Scopus
  • a model with accuracy R² ≥ 0.7, validated on at least 100 ha
  • at least 50 spectral-phenological covariates in the prediction
  • 3 open datasets and 2 open-source digital tools
  • 5 formal collaborations with farms and 1 patent application
  • 2 international collaborations and 11 science outreach activities

AgroVeFe (581-UNIOS-30) is funded by the European Union – NextGenerationEU through the National Recovery and Resilience Plan 2021–2026. Views and opinions expressed are those of the authors only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.