Operational Mapping of Poplar Plantations at the National Scale using Sentinel-2 Time Series Satellite Imagery

Yousra Hamrouni a présenté les travaux de cartographie de Dynafor sur les peupliers ainsi que les cartes produites sur les six années (2017-2022) à la 27ième session de l’IPC 2024 (Commission internationale sur les peupliers et autres arbres à croissance rapide), organisée par la FAO qui se tenait cette année à Bordeaux du 22 au 25 octobre.

Hamrouni, Y. 1; Paillassa,E. 2; Chéret, V. 1; Monteil, C.1; Sheeren,D. 1 (2024) Operational Mapping of Poplar Plantations at the National Scale using Sentinel-2 Time Series Satellite Imagery. 27th Session of the International Commission on Poplars and Other Fast-Growing Trees Sustaining People and the Environment (IPC27), Bordeaux, 22-25 octobre.

Affiliations:

1-Université de Toulouse, INRAE, UMR DYNAFOR, Castanet-Tolosan, France

2-Centre National de la Propriété Forestière, Institut pour le Développement Forestier, 33075 Bordeaux, France

Abstract: In light of the challenges associated with accurately estimating the French poplar resource, this work leverages Sentinel-2 satellite capabilities for operational mapping of poplar plantations at a national scale. In collaboration with the research unit Dynafor, the National Poplar Council (CNP), and the National Forest Property Center (CNPF), we have developed an innovative approach integrating artificial intelligence and remote sensing to accurately map poplar plantations on a large scale.

This approach encompasses two complementary strategies: (i) the adoption of artificial intelligence techniques for efficient learning of image classification models from limited reference samples1, and (ii) the development of a novel spectral index, the Poplar Index (PI)2, harnessing specific spectral bands in the short-wave infrared (SWIR) and red edge domains available in Sentinel-2 imagery. The PI has demonstrated its relevance by achieving over 90% precision in discriminating poplar plantations from other deciduous tree species.

This methodological framework, focusing on the Poplar Index, has made it possible to produce six-year national maps from 2017 to 2022, demonstrating the ability to accurately identify poplar stands over large areas. It is noteworthy that this work represents a significant advance in the monitoring of poplar plantations through remote sensing. However, the precise assessment of surface areas remains pending as part of our ongoing endeavours. Further post-processing analyses are underway within the current initiatives to improve the accuracy of these estimates before their release to end users.

These efforts are part of an operational framework aimed at a systematic and regular production of poplar maps to be distributed through the Theia platform, thus ensuring thorough and recurrent monitoring for the poplar industry.

1. Hamrouni, Y., Paillassa, E., Chéret, V., Monteil, C., & Sheeren, D. (2021). From local to global: A transfer learning-based approach for mapping poplar plantations at national scale using Sentinel-2. ISPRS Journal of Photogrammetry and Remote Sensing, 171, 76-100.

2. Hamrouni, Y., Paillassa, E., Chéret, V., Monteil, C., & Sheeren, D. (2022). Sentinel-2 poplar index for operational mapping of poplar plantations over large areas. Remote Sensing, 14(16), 3975.