Student information
MSc thesis topic: Decoding the optimal time for detecting commodity crops after deforestation.
Commodity crops play a pivotal role in the livelihoods of a significant population across pantropical regions. However, their cultivation often accompanies environmental challenges, particularly deforestation. Timely detection and monitoring of commodity crops are imperative to ensure their cultivation occurs sustainably, without compromising forest resources. While satellite imagery and deep learning methods have been utilized for detecting and classifying commodity crops like cocoa, oil palm, and rubber, a critical gap remains regarding the optimal timing for their detection.
Early stages of commodity crop growth, especially tree crops, bear a striking resemblance to each other or to annual crops, posing challenges for their accurate detection solely through satellite imagery. Thus, the precise timing for detecting commodity crops becomes a pivotal factor in mapping efforts. Understanding the optimal window for detection is essential for effective forest management and sustainable development initiatives.
The recent advance and availability of earth observation data and deep learning technologies provide an opportunity to monitor the earth in high detail and all weather. In this thesis you will address this gap by investigating the optimal time for detecting commodity crops following deforestation by leveraging time series of satellite imagery, forest loss product, and deep learning techniques. The study seeks to delineate the most effective timeframe for accurate identification of commodity crops post-deforestation. The findings of this research will not only enhance our understanding of land use dynamics but also inform policy interventions and conservation strategies aimed at promoting sustainable agricultural practices while safeguarding forest ecosystems.
Software: Tensorflow, Pytorch (python) and Google Earth Engine
Objectives
- Literature on commodity crops (oil palm, rubber, cocoa, coffee and soy) and existing mapping effort
- Design and implement the training and testing strategy of deep learning models for detecting commodity crops based on multiple time steps of crop growth
- Report the optimal time the deep learning model accurately detect commodity crops post deforestation.
Literature
- Masolele, R.N., et al. (2024). Mapping the diversity of land uses following deforestation across Africa. Sci Rep 14, 1681.
- Descals, A. et al. (2021). High-resolution global map of smallholder and industrial closed-canopy oil palm plantations. Earth Syst. Sci. Data 13, 1211–1231.
- Nguyen, T., et al. (2019). Mapping Rubber Plantations in Southeast Asia Using Landsat and Sentinel-2 Data: A Comparison of Machine Learning Methods. GIScience & Remote Sensing, 56(3), 372-391.
- Kalischek, N., et al. (2023). Cocoa plantations are associated with deforestation in Côte d’Ivoire and Ghana. Nat Food 4, 384–393.
- Silva, L., et al. (2018). Mapping Soybean Fields in South America Using Landsat Time Series and Random Forest Classification. Remote Sensing of Environment, 210, 35-47.
- Perez, R., et al. (2019). Mapping Coffee Plantations in Central America Using Sentinel-2 Data and Object-Based Image Analysis. International Journal of Applied Earth Observation and Geoinformation, 81, 15-25.
- Masolele, R. N. et al. (2022). Using high-resolution imagery and deep learning to classify land-use following deforestation: a case study in Ethiopia. GISci. Remote Sens. 59(1), 1446–1472.
Requirements
- Advanced Earth Observation course
- Machine learning course
- Geo-scripting course (Good knowledge in scripting is an asset; e.g. python, Google Earth Engine, and java script)
Theme(s): Sensing & measuring; Modelling & visualisation