Monitoring road development in Congo Basin forests with multi-sensor satellite imagery and deep learning
Notice bibliographique
Résumé
Road development has affected many remote tropical forests around the world and has accelerated human-induced deforestation, forest degradation and biodiversity loss. The development of roads in tropical forests is largely driven by industrial selective logging, which can provide a sustainable source of revenue for developing countries while avoiding more detrimental forms of forest degradation or deforestation. Understanding the dynamics and impacts of road development is challenging, because road inventories in remote tropical forests have been largely incomplete or outdated. In this study, we present novel remote sensing-based methods for automated monitoring of road development and apply them across the Congo Basin forest region, an area characterized by increasing road development rates driven by logging activities. We trained a deep learning model with Sentinel-1 and -2 satellite imagery to map road development on a monthly basis at 10 m spatial scale, leveraging the complementary value of radar and optical imagery. Applying the model across the Congo Basin forest, we present a vectorized map of road development from January 2019 until December 2022, demonstrating an F1-score of 0.909, a false detection rate of 4.2% and a missed detection rate of 14.9%. In total, we mapped 35,944 km of road development in the Congo Basin forest during the four years, with at least 78% apparently related to logging activities, mainly located in the western part of the region. We estimate that 30% of the detected road openings were previously abandoned logging roads that were reopened. In addition, 23% of detected road development was located in areas considered to be intact forest landscapes. The road monitoring methods demonstrated in this study can facilitate several crucial forest management and conservation objectives in the tropics, such as assessing ecological and climate impacts related to selective logging, monitoring illegal or unsustainable activities, and providing a basis for improved understanding and evaluation of human impacts on forests at large scale. More information, including a full overview of the Congo Basin forest road map, can be found at: https://wur.eu/forest-roads . • Sentinel-1 and -2 data and deep learning were used to map monthly road development. • A false detection rate of 4.1% and a missed detection rate of 14.9% were achieved. • 35,944 km of road development was mapped in the Congo Basin forest for 2019-2022. • At least 78% of road development was related to industrial selective logging.
Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.
Comment cette classification a été obtenuedéplier
Prédiction machine sur la base complète
Imitation des enseignantsNi prévalence calibrée, ni vérité terrain. Validation humaine à venir. Le volet Gemma est une étiquette directe du modèle pour chaque travail de la base, lue sur la notice réduite au titre. Le volet Codex est un classifieur appris des 10 348 étiquettes directes de Codex et calibré sur les taux pondérés de l'échantillon; les champs sans appui suffisant ne portent aucun appel Codex. Le mode candidate est l'union des deux volets; le consensus est leur intersection. Ces sorties portent le statut machine_predicted_unvalidated et ne sont pas des étiquettes humaines.
Scores du classifieur distillé par catégorie (deux têtes)
| Catégorie | Codex | Gemma |
|---|---|---|
| Métarecherche | 0,000 | 0,001 |
| Méta-épidémiologie (sens strict) | 0,001 | 0,000 |
| Méta-épidémiologie (sens large) | 0,000 | 0,000 |
| Bibliométrie | 0,002 | 0,001 |
| Études des sciences et des technologies | 0,000 | 0,000 |
| Communication savante | 0,001 | 0,001 |
| Science ouverte | 0,000 | 0,001 |
| Intégrité de la recherche | 0,000 | 0,000 |
| Charge utile insuffisante (le modèle a refusé de juger) | 0,000 | 0,000 |
Scores machine (provisoires)
Les deux têtes enseignantes du modèle étudiant, lues sur ce travail. Un score ordonne la base pour la relecture; il n'affirme jamais une catégorie, et le statut de validation accompagne chaque rangée tel quel.
Scores de référence d'un modèle non mature (critères de maturité non atteints, 7 itérations). Un score ordonne; il n'affirme jamais une catégorie.
score_only:v0-immature-baseline · tel quel depuis la passe de notation : score_only signifie que le nombre peut ordonner les travaux, et qu'aucune étiquette de catégorie n'en découleClassification
machine, non validéePrédiction automatique; un appel candidat d’une seule source (Gemma direct ou Codex distillé), pas un consensus.
Le détail, modèle par modèle et score par score, se trouve en fin de page sous « Comment cette classification a été obtenue ».