Search

International audience

Monitoring forest–agriculture mosaics is crucial for understanding landscape heterogeneity and managing biodiversity. Mapping these mosaics from remotely sensed imagery remains challenging, since ecological gradients from forested to agricultural areas make characterizing vegetation more difficult. The recent synthetic aperture radar (SAR) Sentinel-1 (S-1) and optical Sentinel-2 (S-2) time series provide a great opportunity to monitor forest–agriculture mosaics due to their high spatial and temporal resolutions. However, while a few studies have used the temporal resolution of S-2 time series alone to map land cover and land use in cropland and/or forested areas, S-1 time series have not yet been investigated alone for this purpose. The combined use of S-1 & S-2 time series has been assessed for only one or a few land cover classes. In this study, we assessed the potential of S-1 data alone, S-2 data alone, and their combined use for mapping forest–agriculture mosaics over two study areas: a temperate mountainous landscape in the Cantabrian Range (Spain) and a tropical forested landscape in Paragominas (Brazil). Satellite images were classified using an incremental procedure based on an importance rank of the input features. The classifications obtained with S-2 data alone (mean kappa index = 0.59–0.83) were more accurate than those obtained with S-1 data alone (mean kappa index = 0.28–0.72). Accuracy increased when combining S-1 and 2 data (mean kappa index = 0.55–0.85). The method enables defining the number and type of features that discriminate land cover classes in an optimal manner according to the type of landscape considered. The best configuration for the Spanish and Brazilian study areas included 5 and 10 features, respectively, for S-2 data alone and 10 and 20 features, respectively, for S-1 data alone. Short-wave infrared and VV and VH polarizations were key features of S-2 and S-1 data, respectively. In addition, the method enables defining key periods that discriminate land cover classes according to the type of images used. For example, in the Cantabrian Range, winter and summer were key for S-2 time series, while spring and winter were key for S-1 time series.

The conservation of tropical forest cover is a key to ensuring sustainable provision of multiple ecosystem services. In human-modified landscapes, forest conservation must also be reconciled with agricultural productivity. However, increasing demography, demand for agricultural products and changes in land uses are affecting forest sustainability through degradation processes. A first step to tailor effective forest management by local decision makers is to identify most vulnerable forests and to characterize what is driving this vulnerability. The objective of this thesis is to develop a multidimensional approach using remote sensing to assess forest degradation and the relations with the broader dynamics of land use/cover towards the evaluation of forest ecological vulnerability. The thesis was applied in old-deforestation fronts of Paragominas (Brazil) and Di Linh (Vietnam) where large-scale deforestation driven by commercial agriculture shaped the landscape into land use mosaics with increasing degradation pressures. In Paragominas, degradation is linked with long-term accumulation of selective logging and fire implying changes in forest structure. We estimated the potential of multisource remote sensing to map forest aboveground biomass (AGB) from large-scale field assessment of carbon stock. We improved the accuracy of AGB mapping compared to pantropical datasets and revealed that 87% of forest was degraded. At a lower scale, we investigated the consequences of 33 years of degradation history from Landsat on forest structures using Unmanned Aerial Vehicle. We found that canopy textures captured canopy grain, heterogeneity and openness gradients, correlated with forest structure variability and could be used as proxies to characterize degraded forests. We also assessed the potential of very high resolution satellite images and derived canopy textures to upscale texture-structure relations at the municipality scale. Based on environmental, geographical factors and landscape structure metrics derived from land use/cover classification, we demonstrated that 80% of forest degradation was mainly driven by accessibility, geomorphology, fire occurrence and fragmentation. The drivers of degradation acted together and in sequence and clustering analysis disentangled different cascades of effects. Changes in landscape structure allowed reconstructing trajectories informing on agricultural frontier dynamics. The combination of current forest state, landscape dynamics and distribution of degradation drivers would be at the basis of ecological vulnerability assessment. In Di Linh, degradation mostly concerns forest edges and is driven by encroachment of coffee-based agriculture. Field inventory of the different forest types and other landscape elements combined with Sentinel-2 images allowed to map with high precision the current land cover. We then mapped land cover changes over 45 years using Landsat time series. We constructed trajectories of landscape structure dynamics from which we characterized the expansion of the agricultural frontier and highlighted heterogeneous agricultural encroachment on forested areas. We also identified degradation and fragmentation trajectories that affect forest cover at different rates and intensity. Combined, these indicators pinpointed hotspots of forest ecological vulnerability. Most vulnerable forest areas were experiencing rapid and recent forest cover loss associated with landscape fragmentation, land use competition due to coffee production and degradation. Through the developed remote sensing approaches and indicators at forest and landscape scales, we provided a holistic diagnosis of forests in human-modified landscapes encompassing forest state and broader dynamics and drivers. This thesis aims to pave the way for tailored and prioritized management of degraded forests at the landscape scale.; La conservation du couvert forestier tropical est essentielle pour assurer la fourniture durable de services écosystémiques. Dans les paysages anthropisés, la conservation des forêts doit également être conciliée avec la productivité agricole. Toutefois, l'accroissement de la démographie, la demande de produits agricoles et les changements dans l'utilisation des terres affectent la durabilité des forêts. Une première étape pour adapter la gestion efficace des forêts par les décideurs locaux consiste à identifier les forêts les plus vulnérables et à caractériser ce qui la génère. L'objectif de cette thèse est de développer une approche multidimensionnelle utilisant la télédétection pour évaluer la dégradation des forêts et les relations avec la dynamique de l'utilisation des terres afin d’estimer la vulnérabilité écologique des forêts. La thèse a été appliquée à Paragominas (Brésil) et Di Linh (Vietnam), où la déforestation à grande échelle due à l'agriculture commerciale a façonné le paysage en mosaïques d'utilisation des terres. A Paragominas, la dégradation est liée à l'accumulation de l’exploitation sélective du bois et au feu impliquant des changements dans la structure forestière. Nous avons estimé le potentiel de la télédétection multisource pour cartographier la biomasse forestière aérienne (AGB) à partir de données de stock de carbone. Nous avons amélioré la précision de la cartographie de l'AGB par rapport aux données pantropicales et révélé que 87 % des forêts étaient dégradées. À une plus petite échelle, nous avons étudié les conséquences de 33 ans de dégradation sur les structures forestières à l'aide de drone. Nous avons constaté que les textures de la canopée capturaient le grain, l'hétérogénéité et les gradients d'ouverture de la canopée, corrélés à la variabilité de la structure forestière et pouvaient être utilisés comme indicateurs pour caractériser les forêts dégradées. Nous avons également évalué le potentiel des images satellites à très haute résolution pour cartographier les structures des forêts dégradées à l'échelle de la municipalité. En nous basant sur des facteurs environnementaux, géographiques et de structure du paysage dérivés de la classification de l'utilisation des terres, nous avons démontré que 80 % de la dégradation des forêts était principalement due à l'accessibilité, la géomorphologie, la fréquence des incendies et à la fragmentation. Les facteurs de dégradation sont interconnectés et agissent en séquence au sein de différentes cascades d'effets. L'évolution de la structure du paysage a permis de reconstituer des trajectoires informant sur la dynamique des frontières agricoles. La combinaison de l'état actuel des forêts, de la dynamique du paysage et de la distribution des facteurs de dégradation permettra d’évaluation la vulnérabilité. A Di Linh, la dégradation concerne principalement les lisières forestières et est due à l'empiètement de l'agriculture (café). L'inventaire sur le terrain des différents types de forêts et d'autres éléments, combiné aux images Sentinel-2, a permis de cartographier avec une grande précision la couverture terrestre actuelle. Nous avons cartographié l'évolution de la couverture terrestre sur 45 ans à l'aide de séries chronologiques Landsat. Nous avons construit des trajectoires de dynamique paysagère afin de caractériser l'expansion de la frontière agricole et mis en évidence l'empiétement agricole sur les zones forestières. Nous avons également identifié des trajectoires de dégradation et de fragmentation qui affectent le couvert forestier à différentes intensités. Ensemble, ces indicateurs ont mis en évidence des points chauds de vulnérabilité. Grâce aux approches et aux indicateurs élaborés à multiples échelles, nous avons fourni un diagnostic holistique des forêts dans les paysages anthropisés, englobant l'état des forêts et des dynamiques à plus larges échelles. Cette thèse vise à orienter une gestion adaptée des forêts dégradées à l'échelle du paysage.

Forest degradation accounts for 68.9% of overall forest carbon losses and changes forest structure. Due to the diversity (nature, intensity and frequency) of anthropogenic disturbances, multiple forest structures can be observed within the forest landscape mosaic. Remote sensing offers a unique opportunity to identify and characterize these structures. Very High Spatial Resolution (VHRS) optical imagery has the capacity to model forest canopy. SAR VHSR data which are independent of visibility conditions are capable to retrieve forest biomass, depending on the radar frequency and forest density. However, optical and SAR images have rarely been combined to study degraded forest structure. The aim of this study is i) to characterize the structure of a wide range of degraded forests types using optical UAV imagery and ii) to evaluate the potential of VHSR SAR imagery for scaling up.The investigated area is located in the human-dominated landscape of Paragominas municipality (Para state, Brazil). 52 UAV images of 25 ha (10 cm resolution) were acquired in September 2017 along the whole forest degradation gradient (from "undisturbed" to logged and/or burned forests) and to derive Digital Elevation Models. Four TerraSAR-X images were acquired in October 2017 (1.5m resolution) with different incident angles and polarizations. Three textural indexes based on the spatial variation in pixel radiance were derived from VHSR images (Fourier Transform Textural Ordination spectrum, the lacunarity and the skewness indexes) within different window sizes (from 50 to 200m). Generalized Linear Models were established between the UAV and SAR derived indexes and the forest canopy heights. Models and robustness were assessed by calculating the coefficient of determination (R-squared), the Akaike information criterion (AIC) and the root mean square error (RMSE). The combination of the three textural indexes from UAV images were able to capture the full gradient of canopy crown size distribution and canopy gaps and show a positive relation with the opening and lowering of the canopy. TerraSar X derived indexes showed most accurate results with the 26° incidence angle and HH-HV polarization. This quantification is highly informative for forested land use planning and policy makers to better understand and characterize degraded forests status from local to regional scale.

International audience; In the Brazilian Amazon, degraded forests dominate the landscapes on the agricultural frontiers. This region is now facing a major challenge: halting degradation and sustainably managing these forests. Today, degraded forests represent a class of forest in their own right. They can nevertheless play a key role in combating climate change, and can also help to improve the ecological functioning of the different territories. Implementing public policies with the twin objectives of reducing degradation and promoting these forests implies strong support from research. In this Perspective, we focus on four research priorities: developing methods to characterise and monitor degraded forests; drafting specific forest management plans; understanding the role played by all social actors; and supporting policies at the territorial level.; En Amazonie brésilienne, les forêts dégradées dominent les paysages des fronts pionniers. Un grand défi attend désormais cette région : stopper la dégradation et gérer durablement ces forêts. Les forêts dégradées représentent aujourd’hui une catégorie de forêt à part entière. Elles peuvent néanmoins jouer un rôle majeur pour lutter contre le changement climatique. Elles peuvent aussi contribuer à un meilleur fonctionnement écologique des territoires. Développer des politiques publiques visant le double objectif de réduire la dégradation et de valoriser ces forêts implique un appui fort de la recherche. Dans ce Perspective, nous mettons l’accent sur quatre priorités de recherche : développer des méthodes de caractérisation et de suivi des forêts dégradées, élaborer des plans d’aménagement spécifiques, comprendre le rôle joué par tous les acteurs sociaux et accompagner les politiques à l’échelle territoriale.; Na Amazônia brasileira, as florestas degradadas dominam as paisagens das frentes pioneiras. Atualmente, esta região enfrenta um grande desafio: parar a degradação e fazer a gestão sustentável dessas florestas. As florestas degradadas são atualmente parte integrante de uma categoria de florestas. No entanto, elas podem desempenhar um papel importante na luta contra as mudanças climáticas e podem contribuir para um melhor funcionamento ecológico dos territórios. Desenvolver políticas públicas com o duplo objetivo de reduzir a degradação e melhorar a condição dessas florestas implica um forte apoio da pesquisa. Neste Perspective, nós focalizamos em quatro prioridades de pesquisa: desenvolver os métodos de caracterização e monitoramento das florestas degradadas, elaborar planos específicos de ordenamento do território, compreender o papel desempenhado por todos os atores sociais e acompanhar as politicas em escala territorial.

The Amazonian pioneer front region is a mosaic of different forests types and agricultural landscapes resulting from the colonization of the region through forest conversion into pasture and agricultural lands. Fearnside and Guimaraes (1996) showed that 47% of the deforested area is rapidly abandoned. It also appears that logged forests surface is equivalent to deforested areas (Asner et al., 2005). Consequently a degradation gradient exists from low-impacted logged forests (depending of the logging intensity) to young secondary (regrowth) forests. To obtain more accurate estimation of carbon stocks, it is important today to take into account the degraded forest gradient including all degraded forest stages between mature intact forests and non-forest areas. The first main challenge is to identify and to characterize the various stages. The identification of forest degradation is still a complex and expansive problem even if it has been focused until now only on logged tropical rainforest (Asner, 2009; Gond and Guitet, 2009; Desclées et al., 2006; Asner et al., 2005; Souza et al., 2003). In parallel estimation of biomass loss in the degraded forest is little-studied. Within temperate and boreal forests some estimation are made by Solberg et al., (2013). The combination of optical remotely sensed data (Landsat-8), radar (Terra-Sar-X) and Lidar (IceSat) have to be studied to analyze the potential of the multi-sensors techniques to characterize the tropical rainforest degradation (Betbeder et al., 2014). The study presents the first results obtained during the field work at Paragominas (Pará, Brazil) on different forest degradation intensities (Bérenguer et al., 2014). This field database is then compared with multi-sensors remote sensing to better understand multiple interactions and to establish a forest degradation typology. (Texte intégral)

Recently, several remote sensing methods have been developed to quantify the degradation of tropical forests. However, it still lacks finest spatial and temporal analysis to define trajectories of forest degradation i.e. a temporal analysis of the impacts on forest integrity. This communication aims to explore this issue and proposes a set of operational indicators to monitor forest degradation, which can constitutes a decision tool to support forestry managers and policy makers. We studied the trajectories of forest degradation in the municipality of Paragominas – PA in the eastern Brazilian Amazon between 1995 and 2009, with a focus on the forestry company Cikel (400 000 ha certified by FSC since 2001). First, we developed a semi-automatic remote sensing methodology to detect forest degradation using multi-temporal Landsat images (spatial resolution of 30m) covering the 1995-2009 period. This method included two steps: 1) Identification of logging tracks and log landings using an algorithm of Bourbier et al. (2013). This algorithm uses spectral indices and morphological filters to strengthen the spectral contrasts between bare soil and forest cover. 2) Identification of logging gaps - which are characterised by senescent vegetation due to trees fall - using a Spectral Mixture Analysis carried out in CLASlite (Asner et al., 2009) and a fraction index (Souza et al., 2013). So, we obtained annual maps identifying these three major impacts. Secondly, we calculated annual landscape metrics of forest degradation using the R package "SpatialEco". Then, we calculated indicators which synthetize information about logging impacts and logging frequencies over the period from these annual degradation metrics. Finally, we selected a set of 6 indicators and statistically analysed the trajectories of degradation occurring in Paragominas using ACP and CAH. Our results emphasize four major degradation trajectories from well managed forests to highly-logged forests. They clearly show a difference between legal and illegal logging in terms of forest degradation. Moreover, they indicate that impacts of FSC certification on forest degradation was positive. Degradation was statistically lower in the certified logged plots compared to the uncertified plots. These set of indicators are adequate to monitor forest degradation through space and provide guidance to policy-makers for a better management of forest resources. (Texte intégral)

The Amazonian pioneer front region is a mosaic of different forests types and agricultural landscapes resulting from the colonization of the region through forest conversion into pasture and agricultural lands. Fearnside and Guimaraes (1996) showed that 47% of the deforested area is rapidly abandoned. It also appears that logged forests surface is equivalent to deforested areas (Asner et al., 2005). Consequently a degradation gradient exists from low impacted logged forests (depending of the logging intensity) to young secondary (regrowth) forests. To obtain more accurate estimation of carbon stocks, it is important today to take into account the degraded forest gradient including all degraded forest stages between mature intact forests and non-forest areas. The first main challenge is to identify and to characterize the various stages. The identification of forest degradation is still a complex and expansive problem even if it has been focused until now only on logged tropical rainforest (Asner, 2009; Gond and Guitet, 2009; Desclées et al., 2006; Asner et al., 2005; Souza et al., 2003). In parallel estimation of biomass loss in the degraded forest is little-studied. Within temperate and boreal forests some estimation are made by Solberg et al., (2013). The combination of optical remotely sensed data (Landsat-8), radar (Terra-Sar-X) and Lidar (IceSat) have to be studied to analyze the potential of the multisensors techniques to characterize the tropical rainforest degradation (Betbeder et al., 2014). The study presents the first results obtained during the field work at Paragominas (Pará, Brazil) on different forest degradation intensities (Bérenguer et al., 2014). This field database is then compared with multi-sensors remote sensing to better understand multiple interactions and to establish a forest degradation typology. (Texte intégral)

ln the Brazilian Amazon, deforestation and forest degradation have resulted in a complex mosaic of forest types. Nearly 20% of the Brazilian Amazonian forests have been cleared. ln this area, abandon ment of fields led to regrowth of secondary forests of varying ages. A fraction of the remaining forested land has also suffered from anthropic degradation (mainly over-logging and fire). Human-modified Amazonian landscapes are therefore an assemblage of these various forests associated with pastures and agricultural lands. These landscapes are now at the centre of political concerns. Coercive measures taken by the Brazilian government to curb deforestation, associated with private initiatives (soy and beef moratoria) drastically reduced deforestation rates. The colonization of the Amazonian territory through agricultural expansion over forest areas is now severely restricted. Consequently, conciliation between agricultural production and environmental conservation should be pursued in all human-impacted forests. However these secondary and degraded forests have not received the necessary attention. While identification and characterization of degraded forests became a critical task, there is an overall limitation in the remote sensing analyses developed so far.To define management plans for these areas and to understand their role in the maintenance of ecological services, the first challenge is to identify and characterize the forests that result from different disturbance trajectories. We carried out a study aiming at classifying the large spectrum of degraded forests into forest classes based on degradation levels using satellite data. The study area took place in the municipality of Paragominas (eastern Amazonia). A large range of captors (optical, radar and lidar) have been tested combining with ground-truth validation. This classification has important applications in ecological studies as well as in supporting decisions for land-use planning. (Texte intégral)

In the Brazilian Amazon, deforestation and forest degradation have resulted in a complex mosaic of forest types. Nearly 20% of the Brazilian Amazonian forests have been cleared. In this area, abandonment of fields led to regrowth of secondary forests of varying ages. A fraction of the remaining forested land has also suffered from anthropic degradation (mainly over-logging and fire). Human-modified Amazonian landscapes are therefore an assemblage of these various forests associated with pastures and agricultural lands. These landscapes are now at the centre of political concerns. Coercive measures taken by the Brazilian government to curb deforestation, associated with private initiatives (soy and beef moratoria) drastically reduced deforestation rates. The colonization of the Amazonian territory through agricultural expansion over forest areas is now severely restricted. Consequently, conciliation between agricultural production and environmental conservation should be pursued in all human-impacted forests. However these secondary and degraded forests have not received the necessary attention. While identification and characterization of degraded forests became a critical task, there is an overall limitation in the remote sensing analyses developed so far.To define management plans for these areas and to understand their role in the maintenance of ecological services, the first challenge is to identify and characterize the forests that result from different disturbance trajectories. We carried out a study aiming at classifying the large spectrum of degraded forests into forest classes based on degradation levels using satellite data. The study area took place in the municipality of Paragominas (eastern Amazonia). A large range of captors (optical, radar and lidar) have been tested combining with ground-truth validation. This classification has important applications in ecological studies as well as in supporting decisions for land-use planning.

Le territoire amazonien présente actuellement une mosaïque de paysages forestiers et agricoles issus de plusieurs décennies de déforestation. Concernant les forêts, il s'agit d'un mélange complexe de types forestiers, résultats d'usages variés et répétés dans le temps aboutissant à une large gamme de perturbations. Tout un gradient existe : depuis les forêts matures, peu ou pas impactées par des activités anthropiques, jusqu'à de jeunes forêts secondaires, partie intégrante du système d'agriculture sur brûlis et constituant un de ses états de transition. La mutation agro-environnementale en cours en Amazonie marquée par une très forte réduction de la déforestation met désormais cette mosaïque forestière au coeur de nombreux enjeux environnementaux. La gestion de ce territoire implique d'en identifier les principaux types à travers des techniques de télédétection multisource combinées aux analyses de terrain. Nous avons mené une étude visant à estimer un gradient de dégradation forestière au sein du municipe de Paragominas (Para, Brésil). Nous avons tout d'abord étudié la capacité d'estimer une typologie de la dégradation à partir d'une typologie évaluée sur le terrain. Par la suite, des relevées de biomasse et une large gamme de variables de télédétection (optique et radar) ont servis à établir un modèle de régression de la biomasse par l'algorithme Random Forest. Une forte corrélation (RMSE=38.8 Mg.ha-1) a été observée entre la biomasse aérienne théorique et estimée. La typologie de terrain a servi de donnée de validation pour la prédiction de la biomasse à l'échelle de la commune laissant apparaitre de nouvelles perspectives quand à l'utilisation d'une typologie afin d'étudier la dégradation forestière. Enfin, nous avons montré que l'analyse du grain de canopée constitue une piste intéressante dans la mise en relation de la texture et de paramètres forestiers pour étudier la dégradation.

In the Amazonian “arc of deforestation”, landscapes are a result of 50 years of systematic deforestations, driven by a "race to the land", where land appropriation was the main objective of the migrants. In this context, land accumulation was a priority over land production, and even more over sustainable management of natural resources. In consequence, current landscapes include huge areas of systematic deforestation, and disconnected forest remaining patches. Both are degraded. Agronomic productivities are decreasing, due to extractive management systems. Ecosystemic services also are low, due to forest fragmentation, soil degradation and deforestation of vulnerable areas. Since 2005 - 2008, federal policies against deforestation and forest act have fixed the usable agricultural areas, and economic growth must now take place in a limited land extension. Natural resources management becomes essential, which induces changes in farming practices. In this transition, land suitability becomes a key to organize a new farming design, zoning areas dedicated to economic production, and areas for ecosystemic services. In the first ones, improved and intensified agricultural practices are developed, valorizing best soils and or topographic units. The other areas, less productive but with a high connectivity and potentiality for water cycle and soil protection, are abounded and could go back to forest regeneration. A new forest matrix is being built, as well as a new map of agricultural production. Landscape efficiency is also defined by the optimized combination of these two maps. Land suitability defines the tradeoff between economic production and ecosystemic services. This presentation explains the Paragominas experience, in Brazilian eastern amazon, to develop a landscape restoration plan, in a jurisdictional approach. In a first step, land suitability is mapped at municipal level, using topographic data, and derived information's about soils, declivity and hydrographic network. All categories are related to farmer's practices, and discussed with them. In a second step, land uses are also mapped, especially forest cover and forest degradation, using middle resolution remote sensing data. In the third step, land uses and suitability's are spatially crossed in a GIS, in order to define strategic areas for ecosystemic services by reforestation or forest restauration, and other strategic areas for economic production, by improved practices (especially tree-crop-livestock integration). In a fourth step, this cartography become a municipal law for land use, discussed and voted in the municipal congress. In a fifth step, every farmer can develop a new farming design, according to this mapping, and possibly accessing a specific credit line and facilities. By this way, forest restoration is growing, especially in the strategic areas, and with a higher connectivity index. It is possible to improve ecosystemic services, above the Brazilian NDC in COP 21 for example, not only for carbon sequestration but also for water cycle, soil protection and biodiversity. This win-win strategy, based on land suitability, is attractive for farmers and all private sector, and also can achieve higher results than only command and control policies. In the other hand, it needs a strong monitoring capacity in the jurisdiction, and good governance mechanisms.