top of page

Assessing Deforestation in Arid Areas with Satellite Image Time Series


Context and Issues

Masae was requested by the AFD to identify deforested zones and assess the volume of available ressources in an area which is famous for being the main wood supply basin for Niamey.


Masae processed likewise:

  1. Extraction of pixel-level information from Sentinel-2 imagery for Niamey wood supply basin

  2. Use of Machine learning to classify each pixel of the different images obtained depending on 8 land use classes

  3. Pre/post analysis of the images allowing the team to identify areas where land use had changed between 2019 and 2020, thereby identifying deforested areas

The machine learning algorithm used to classify pixels by land use class achieved 99.5% accuracy. The team was able to detect deforested and degraded areas consistent with ground truth collected by a third party.

A web platform was developed to visualise the data. It is available here.


bottom of page