Challenges in automatic forest change reporting through land cover mapping. (17th December 2022)

• Record Type: Journal Article
• Authors: Alonso, Laura; Rodríguez, Andrés; Picos, Juan; Armesto, Julia
• Editors: Fassnacht, Fabian
• Is Part Of: Forestry. Volume 96:Number 2(2023)
• Page Start: 155
• Page End: 169
• Publication Date: 2022-12-17
• Subjects: Forests and forestry -- Periodicals; Forests and forestry -- Great Britain -- Periodicals; 634.9
• DOI: 10.1093/forestry/cpac053
• Languages: English
• ISSNs: 0015-752X

Abstract

Abstract: Up-to-date knowledge about changes in forest resources and their spatial distribution is essential for sustainable forest management. Therefore, monitoring of forest evolution is increasingly demanded by national and international agencies to design forestry policies and to track their progress. Annually updated land cover maps based on open access satellite imagery may serve as a primary tool for monitoring forest surface evolution over time. Spatially detailed information about forest change might be obtained by comparing land cover maps over time. This study aims to better understand the processes underlying pixels whose land cover changes from 1 year's map to the next and to understand why errors occur. In this study, two annual land cover maps were produced using Sentinel-2 images and afterwards they were compared. The comparison was performed in terms of total surface occupied in each map by each of the classes (net comparison) and in terms of spatial agreement, comparing the results pixel to pixel. The study was performed for the entire region of Galicia (in the Northwest of Spain) for the years 2019 and 2020. Land cover maps obtained had overall accuracies of 82 and 85 per cent. Differences in the total surface of change were encountered when performing the net comparison and spatial agreement comparison. The detailed analysis performed in this study helps to better understand the processes underlying the maps' discrepancies revealing the processes leading to wrongly identified forest changes. Future studies could aim to integrate this knowledge into the monitoring system to improve the ultimate usability of land cover maps to retrieve information about forest changes.