This paper investigates tree species classification using the Sentinel-2 multispectral satellite image time series (SITS). Despite its importance for many applications and users, such mapping is often unavailable or outdated. The value of using SITS to classify tree species on a large scale has been demonstrated in numerous studies. However, many methods proposed in the literature still rely on a standard machine learning algorithm, usually the random forest (RF) algorithm. Our analysis shows that the use of deep learning (DL) models can lead to a significant improvement in classification results, especially in an imbalanced context where the RF algorithm tends to predict the majority class. In our case study in central France with 10 tree species, we obtained an overall accuracy (OA) of around 95% and an F1-macro score of around 80% using three different benchmark DL architectures (fully connected, convolutional, and attention-based networks). In contrast, using the RF algorithm, the OA and F1 scores obtained were 92% and 60%, indicating that the minority classes are poorly classified. Our results also show that DL models are robust to imbalanced data, although small improvements can be obtained by specifically addressing this issue. Validation on independent in situ data shows that all models struggle to predict in areas not well covered by training data, but even in this situation, the RF algorithm is largely outperformed by deep learning models for minority classes. The proposed framework can be easily implemented as a strong baseline, even with a limited amount of reference data.

This article studies the monitoring of oak dieback in forests of the Centre-Val de Loire region (France), where drought-induced dieback has become a major concern due to climate change. The main objective of the study is to evaluate the applicability of multispectral satellite time series for operational monitoring of forest dieback. Using in situ data collected from 2017 to 2022 on approximately 2700 oak plots, a multiyear mapping of the analyzed region was performed using the random forest algorithm and Sentinel-2 images. Our results show that it is possible to detect oak dieback accurately (average overall accuracy = 80% and average balanced accuracy = 79%). A spatial cross-validation analysis also evaluates the performance of the model on regions that were never encountered during training, across all years, resulting in a slight decrease in accuracy ( $\sim$ 5%). The study also highlights the importance of measuring the stability and performance of the classification model over time, in addition to standard cross-validation metrics. A feature analysis shows that the shortwave infrared part of the spectrum is the most important for mapping forest dieback, while the red-edge portion of the spectrum can increase the stability of the model over time. Overall, both in situ data and model predictions showed evidence of forest decline in many areas of the study region. Our results suggest that large areas of forest can decline over short periods of time, highlighting the interest of satellite data to provide timely and accurate information on forest status.

Forest decline and dieback are growing phenomena worldwide, resulting in severe, large‐scale degradation of the canopy. This can profoundly alter the provision of trophic resources and microhabitats for canopy‐dwelling arthropods. In 2019, we assessed the effect of oak decline on the community of canopy‐dwelling Hymenoptera. We selected 21 oak stands and 42 plots, located in three forests in France, presenting contrasting levels of decline. Insects were sampled at the canopy level with green multi‐funnel and flight‐interception traps. We collected a particularly diverse community of 19,289 insect individuals belonging to 918 taxa, 10 larval trophic guilds and 7 nesting guilds. Oak decline had no effect on the abundance or richness of the overall community but significantly reshaped the community assemblages. Decline had contrasting effects depending on the taxa and guilds considered. Specialist parasitoids were more abundant at intermediate levels of decline severity while generalists were negatively affected. Taxa depending on ground‐related resources and microhabitats were promoted. Saproxylic taxa were more abundant while xylophagous insects were negatively impacted. Reduced leaf area index promoted several guilds and the diversity of the overall community. While an increasing tree mortality rate enhanced the abundance and diversity of deadwood resources, it had negative impacts on several Hymenoptera guilds. Our results suggest that micro‐environmental changes at the ground level due to canopy decline have major cascading effects on the communities of canopy‐dwelling Hymenoptera. Our study highlights the relevance of studying Hymenoptera communities to investigate the outcomes of disturbances on forest biodiversity.

Ce travail a pour ambition de caractériser l’ampleur du phénomène de dépérissement des chênes dans la région Centre-Val de Loire à l’aide d’images satellites Sentinel-2 et d’un modèle de classification supervisée. Cette étude propose d’analyser en détail comment le modèle de classification utilise les séries temporelles d’indices Sentinel-2 pour détecter le dépérissement. Il apparaît que de faibles teneurs en eau (corrélées à l’infrarouge moyen) pendant l’été sont déterminantes pour détecter le dépérissement. Les cartes produites ont également été croisées avec d’autres données environnementales qui montrent que les dépérissements des chênes sont fortement associés à l’intensification de l’évapotranspiration de ces dernières années. Messages clés :- Il est possible de détecter avec précision les dépérissements forestiers du chêne à l’aide d’images satellites Sentinel-2.- Une cartographie de l’état sanitaire des chênaies a été produite sur plusieurs années pour l’ensemble de la région Centre-Val de Loire et ses environs.- Notre analyse met en avant l’importance du contenu en eau de la canopée en été comme facteur le plus déterminent pour détecter les dépérissements.