This project aims to investigate how deep learning can be used with data from the European Commission’s MARS Crop Yield Forecasting System and Eurostat to build a robust, explainable regional crop yield forecasting model. You will frame the task as classification or regression, and work with a real world dataset.
Field surveys, crop growth models, remote sensing, statistical models and their combinations have been commonly used to predict crop yield. Machine learning takes a data-driven or empirical modeling approach to learn useful patterns and relationships from input data and provides a set of techniques to improve crop yield predictions. In this project, you will combine deep learning with time series data (including crop model simulation outputs, weather observations and remote sensing indicators) used in an operational crop yield forecasting system.
In this project, you will design, implement and evaluate a deep learning forecasting system and compare its performance against the baseline implementation. In your research you will evaluate alternative deep learning architectures implemented in Python. This thesis builds upon the recent work of Paudel et al. (2020), and aims to learn explainable features for regional crop yield forecasting problem.
Application of deep learning to a real-world problem using geodata from the MARS Crop Yield Forecasting System and Eurostat.
- Become familiar with the data used by MARS Crop Yield Forecasting System to forecast crop yield.
- Learn to define the problem setup by considering operational requirements and as well as the working of deep learning architectures.
- Understand the usefulness of deep learning methods for regional crop yield forecasting.
- Paudel, D., et al. (2021). Machine learning for large-scale crop yield forecasting. Agricultural Systems, 187, 103016.
- MARS Crop Yield Forecasting System
- Machine Learning (FTE-35306) / Deep Learning (GRS-34806)
- Programming in Python / Geoscripting
Theme(s): Modelling & visualisation