Student information

MSc thesis topic: Combining Sentinel-1 Imagery and Environmental Variables for Deforestation Prediction in Tropical Forests Using Deep Learning

Deforestation in tropical forests represents one of the most critical environmental challenges of our era, resulting in devastating consequences, including the loss of habitat, reduction in biodiversity, and significant contributions to global warming due to the release of stored carbon. Addressing this issue is of high importance, not only for conserving biodiversity but also for mitigating climate change impacts. With advancements in the field of machine learning and remote sensing, the potential to predict and monitor deforestation has significantly improved [1]. Current solutions for predicting deforestation largely focus on the use of environmental and socio-economic variables [2][3]. However, the potential of combining these variables with raw satellite data remains underexplored. Integrating environmental variable with high-resolution Sentinel-1 data could provide a more comprehensive view of deforestation dynamics, allowing for the identification of areas at high risk of deforestation before the damage occurs.

This research employs deep learning to combine environmental variables and raw satellite data for deforestation prediction in tropical forests. In this thesis, Sentinel-1 images, environmental variables, and satellite alerts (GLAD [4] and RADD [5] alerts) will be used to predict deforestation in tropical forests using a fully convolutional neural network model.

Software: Pytorch (python) and Google Earth Engine

Objectives

  • Literature review on deforestation prediction in tropical forests.
  • Develop a deep learning model that integrates Sentinel-1 imagery with environmental variables for deforestation prediction.
  • Evaluate the predictive accuracy of the model in comparison to a model that relies solely on environmental data.
  • Generate a map of future deforestation for target regions: Laos and Colombia.

Literature

  1. Mayfield, Helen J., et al. "Considerations for selecting a machine learning technique for predicting deforestation." Environmental Modelling & Software 131 (2020): 104741.
  2. DE SOUZA, Rodrigo Antônio; JUNIOR, Paulo De Marco. Improved spatial model for Amazonian deforestation: An empirical assessment and spatial bias analysis. Ecological modelling, v. 387, p. 1-9, 2018.
  3. Sánchez, Alexander Cotrina, et al. "Peruvian Amazon disappearing: Transformation of protected areas during the last two decades (2001–2019) and potential future deforestation modelling using cloud computing and MaxEnt approach." Journal for Nature Conservation 64 (2021): 126081.
  4. Hansen, Matthew C., et al. "Humid tropical forest disturbance alerts using Landsat data." Environmental Research Letters 11.3 (2016): 034008.
  5. Reiche J, Mullissa A, Slagter B, Gou Y, Tsendbazar N, Odongo-Braun C, Vollrath A, Weisse M, Stolle F, Pickens A, Donchyts G, Clinton N, Gorelick N & Herold M, (2021), Forest disturbance alerts for the Congo Basin using Sentinel-1, Environmental Research Letters.

Requirements

  • Advanced Earth Observation course
  • Machine learning course
  • Geo-scripting course (Good knowledge in scripting is an asset; e.g. python, Google Earth Engine)

Theme(s): Sensing & measuring; Modelling & visualisation