Retrieving the stand age from a retrospective detection of multinannual forest changes using Landsat data. Application on the heavily managed maritime pine forest in Southwestern France from a 30-year Landsat time-series (1984-2014)
Résumé
The availability of Landsat data (Landsat 4, 5, 7 and 8) from ~30 years makes it possible to analyze the forest long term dynamics at high resolution (30m). The performances of the Landsat time-series have been already demonstrated for mapping and monitoring the annual clear-cuts and the storm damage in the Landes Forest, that covers ∼1 million ha in southwestern France and that is heavily managed with even-aged stands with rather short rotations after clear-cut harvesting. Our objectives aimed at improving, automating, and enriching these previous methods. This was to operationally produce over the whole Landes Forest not only (1) the annual maps of clear-cutting from 1984 up the current year but also (2) the map of the current age that was derived from the forest change detected every year since 1984. The developed methodology used the time-series of surface reflectance and cloud mask provided for Landsat by USGS and sought to cope the possible absence of cloud-free image during the interest season or the numerous missing data in Landsat 7 images after 2002. The retrospective processing of the Landsat time-series from 1984 to 2014 made it possible the prediction of actual current age with a satisfactory accuracy.