Thesis subject

MSc thesis topic: Sentinel data and deep learning for cocoa mapping in West Africa

West Africa is the world's second-largest tropical rainforest and presents favorable environments for cocoa plants. Due to highly valued nutritional and health benefits, cocoa is one of the key commodities globally and is the principal cash and food crop in West africa. The cocoa plantations in the west Africa is a mixture of free standing and shaded cocoa.

Remote sensing play an important role in mapping and monitoring cocoa plantations due to its repeated wide area coverage. With their high spatial and temporal resolution Sentinel-1 radar and Sentinel-2 optical satellite data offer unprecedented capabilities to map cocoa plantations. In the analysis domain, machine learning and deep learning methodologies now enable to analyse complex remote sensing datsets for improved landcover mapping  (Zhang et al. 2019).

For this research we will be using Sentinel-1/2 images from google earth engine and deep learning to map coconut plantations in the Congo basin. We will also use a high quality cocoa map to generate our training labels.

Software: Google Earth Engine and Python

Objectives

  • Generate a reference labels for the study area
  • Train a model that is able to accurately map both shaded and freestanding cocoa plantation in the Congo basin
  • Perform accuracy assessment.

Literature

  • A. G. Mullissa, C. Persello, and A. Stein, “Polsarnet: A deep fully convolutional network for polarimetric sar image classification,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 12, no. 12, pp. 5300–5309, Dec 2019.
  • Castillo, J.A.A.; Apan, A.A.; Maraseni, T.N.; Salmo, S.G., III. Estimation and mapping of above-ground biomass of mangrove forests and their replacement land uses in the philippines using sentinel imagery. ISPRS J. Photogramm. Remote Sens. 2017, 134, 70–85.
  • Zhang, C.; Sargent, I.; Pan, X.; Li, H.; Gardiner, A.; Hare, J.; Atkinson, P.M. Joint deep learning for land cover and land use classification. Remote Sens. Environ. 2019, 221, 173–187.

Requirements

  • Advanced Earth Observation course
  • Geo-scripting course (Good knowledge in scripting is an asset; e.g. R, python, java script)

Theme(s): Modelling & visualisation; Integrated Land Monitoring