The HADES Yield Prediction System - A Case Study on the Turkish Hazelnut Sector
Simone Bregaglio1, Kim Fischer2, Fabrizio Ginaldi1
1CREA - Council for Agricultural Research and Economics, Research Centre for Agriculture and Environment, Bologna, Italy.
Frontiers in Plant Science
|June 24, 2021
Summary
A new hazelnut yield prediction system, HADES (HAzelnut yielD forEcaSt), combines simulation and machine learning. This hybrid approach provides accurate forecasts, enhancing hazelnut sector sustainability.
Area of Science:
- Agricultural Science
- Machine Learning Applications
- Crop Modeling
Background:
- Accurate crop yield forecasting is crucial for agricultural decision-making.
- Traditional methods rely on agro-climatic data and simulation models.
- Emerging trends integrate machine learning for enhanced yield prediction.
Purpose of the Study:
- To introduce HADES (HAzelnut yielD forEcaSt), a novel hybrid system for hazelnut yield prediction in Turkey.
- To combine process-based modeling with machine learning for improved forecasting accuracy.
- To provide timely and robust information for the global hazelnut sector.
Main Methods:
- Statistical analysis of biennial bearing in historical yield data.
- Calibration of a process-based hazelnut simulation model.
- Deployment of a Random Forest algorithm using model outputs and agro-meteorological data.
Main Results:
- The HADES system demonstrated balanced predictive ability in calibration and validation.
- Relative root mean square error was below 20%.
- R-squared and Nash-Sutcliffe efficiency values exceeded 0.7 across all municipalities.
Conclusions:
- The HADES system offers a next-generation approach to yield prediction.
- Hybridization of process-based and machine learning models enhances forecasting robustness.
- This system can significantly improve the sustainability of the hazelnut industry.
