Publicação
Validation of CCDC Use with Sentinel-2 Time Series for Deforestation Detection in Portugal
| Resumo: | Forest management planning and monitoring are complex tasks that require regular oversight. However, the widespread presence of smallholdings in Portugal and increasing rural abandonment make this task particularly challenging. The Continuous Change Detection and Classification algorithm is notable for enabling the analysis of trends in satellite image time series. While it has been widely used for continuous land cover change monitoring with Landsat data, it has seen limited application in Europe. The use of Sentinel-2 time series, offering higher spatial and temporal resolution and now covering a significant historical period, could be key to enabling CCDC for forest monitoring in Portugal and across Europe. This study aims to validate recently published parameter settings and processing optimizations for applying CCDC with Sentinel-2 data to detect deforestation events in Portugal. The methodology is compared against reference data on forestry activities provided by The Navigator Company. Results show strong agreement with the reference data, with eucalyptus stand harvest events detected with an F1-score of 0.86 and a detection lag of 19 days for 80%. There is no statistically significant difference in detection accuracy for smaller forest stands, suggesting the method is highly promising for monitoring in regions where smallholding forestry presents management challenges. |
|---|---|
| Autores principais: | Louro, Filipe |
| Outros Autores: | Costa, Hugo; Caetano, Mário |
| Assunto: | Continuous Change Detection and Classification Deforestation Land Cover Monitoring Sentinel-2 Sustainable Forest Management SDG 13 - Climate Action SDG 15 - Life on Land |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | comunicação em conferência |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Forest management planning and monitoring are complex tasks that require regular oversight. However, the widespread presence of smallholdings in Portugal and increasing rural abandonment make this task particularly challenging. The Continuous Change Detection and Classification algorithm is notable for enabling the analysis of trends in satellite image time series. While it has been widely used for continuous land cover change monitoring with Landsat data, it has seen limited application in Europe. The use of Sentinel-2 time series, offering higher spatial and temporal resolution and now covering a significant historical period, could be key to enabling CCDC for forest monitoring in Portugal and across Europe. This study aims to validate recently published parameter settings and processing optimizations for applying CCDC with Sentinel-2 data to detect deforestation events in Portugal. The methodology is compared against reference data on forestry activities provided by The Navigator Company. Results show strong agreement with the reference data, with eucalyptus stand harvest events detected with an F1-score of 0.86 and a detection lag of 19 days for 80%. There is no statistically significant difference in detection accuracy for smaller forest stands, suggesting the method is highly promising for monitoring in regions where smallholding forestry presents management challenges. |
|---|