Search

National audience; This presentation presents the results of a study to automatically detect American Maple from Lidar and multispectral data. The model selects the standard deviation of the heights and the vegetation index Red / (Green + Blue). Acer negundo is detected at over 90%.; Cette présentation retrace les résultats d'une étude visant à détecter automatiquement l'Erable américain à partir de données Lidar et multispectrales. Le modèle sélectionne l'écart-type des hauteurs et de l'indice Rouge/(Vert+Bleu). Acer negundo est détecté à plus de 90 %.

International audience; Dans le cadre du projet BIOMAREAU II, nous avons survolé les îles de la réserve naturelle de Mareau-aux-Prés (Loiret) avec un drone équipé d'un Lidar. Les résultats montrent que nous détectons à plus de 90 % l'érable américain qui est une espèce introduite et qui colonise rapidement les grèves sableuses.

International audience; Aim Presence records from surveys with spatially heterogeneous sampling intensity are a key challenge for species distribution models (SDMs). When sex groups differ in their habitat association, the correction of the spatial bias becomes important for preventing model predictions that are biased toward one sex. The objectives of this study were to investigate the effectiveness of existing correction methods for spatial sampling bias for SDMs when male and female have different habitat preferences. Location: Jura massif, France. Methods We used a spatially sex‐segregated virtual species to understand the effect of three sampling designs (spatially biased, uniform random, and systematic), and two correction methods (targeted background points, and distance to trajectories) on estimated habitat preferences, sex ratios, and prediction accuracy. We then evaluated these effects for two empirical Capercaillie (Tetrao urogallus) presence‐only datasets from a systematic and a spatially biased sampling design. Results Sampling design strongly affected parameter estimation accuracy for the virtual species: noncorrected spatially biased sampling resulted in biased estimates of habitat association and sex ratios. Both established methods of bias correction were successful in the case of virtual species, with the targeted correction methods showing stronger correction, as it more closely followed the simulated decay of detectability with distance from sampling locations. On the Capercaillie dataset, only the targeted background points method resulted in the same sex ratio estimate for the spatially biased sampling design as for the spatially unbiased sampling. Main conclusions We suggest that information on subgroups with distinct habitat associations should be included in SDMs analyses when possible. We conclude that current methods for correcting spatially biased sampling can improve estimates of both habitat association and subgroup ratios (e.g., sex and age), but that their efficiency depends on their ability to well represent the spatial observation bias.

International audience; Advances in numerical simulation and analysis of real-size field experiments have supported the development of process-based rockfall simulation models. Availability of high resolution remote sensing data and high- performance computing now make it possible to implement them for operational applications, e.g. risk zoning and protection structure design. One key parameter regarding rock propagation is the soil response to block impact. In the RockyFor3D model, it is taken into account as the normal coefficient of restitution during rebounds of rocks on the ground, which depends on the soil type. Soil type is usually determined by field experts, but its mapping in large areas is time-consuming, especially in mountain environments. Land cover maps can be derived from remote sensing data, with increasing spatial and semantic details thanks to the improvements in spatial and spectral resolution and to the advances in analysis techniques. Relying on the assumption that vegetation cover can be used as a proxy for soil type, it is possible to estimate the normal coefficient of restitution based on remote sensing-derived land cover maps. The objective of this work is to evaluate the relevance of using land-cover as a proxy for the coefficient of normal restitution in rockfall simulations. Land cover map OSO, derived from Sentinel-2 data, is used (Inglada et al., doi:10.3390/rs9010095). Two complementary aspects are investigated. First, on a test area of 600 ha, rockfalls are simulated with RockyFor3D using either expert input for soil types, or coefficient of restitution values depending on land cover type. Rockfall frequency, energy and runnout areas are compared. Second, the practicability of using OSO land cover map for large area rockfall simulation is tested in the valley of Chamonix (France) in order to identify problems that might arise due to data limitations and compatibility between input types. Guidelines drawn from the results are expected to help experts figure out when remote sensing products can be safely used for cost-efficient estimation of input data in rockfall simulations.

International audience; The island mosaic of Mareau-aux-Prés (Loiret, France) on the Loire river is host to the European beaver (Castor fiber) which is a protected species. The four islands, with a total area of 10 ha, are mainly dominated by black poplar (Populus nigra). Due to its impact on poplar, the presence of beaver indirectly favors ashleaf maple (Acer negundo), which is an introduced specie. Due to the accessibility constraints of protected river islands, remote sensing appears as a promising tool to monitor the dynamics of ashleaf maple in a conservation perspective. Unmanned Aerial Vehicules (UAV) make it possible to survey small areas more frequently and with very high spatial resolution. The objective of this work is to evaluate the potential of lidar and multispectral UAV remote sensing for the detection of ashleaf maple. Lidar data was acquired by a YellowScan surveyor and orthophotos with 4 bands were generated from aerial images. A reference dataset was created by manually classifying 300 discs of 1 m diameter as poplar or maple in the orthophotos. Two workflows were compared for maple detection. In the pixel-based workflow, lidar and optical metrics arecomputed for each pixel at 1 m resolution, and for the reference discs. A species classification model is built from the reference discs with all metrics as predictors. The model is then applied to all pixels for species mapping. The object-based workflow is similar, except that a preliminary single tree segmentation is performed on the canopy height model derived from lidar data. Lidar and optical metrics are computed for each segment instead of pixels. The two approaches yield similar classification accuracy (approx. 90%). However, pixel-based maps appear noisier which may hinder the use by practitioners. Moreover, this approach seems more prone to errors in shadow areas. This work showed the applicability of UAV remote sensing for species classification in riparian environments. Research perspectives are to further take advantage of the very high spatial resolution of UAV data to detect signs of decay in poplar trees affected by beavers and monitor its temporal evolution.

National audience; Sustainable forest management requires a precise knowledge of the forest resource and existing infrastructure, to consider harvesting scenarios. LIDAR, a remote sensing technology, is used by many countries to model and map the forest resource . In France, forests are more heterogeneous (in species and structures). An area-based method has been adapted to calibrate empirical models relating dendrometric parameters measured in the field with metrics derived from LIDAR (thanks to a research program ANR-Foresee). The results show that it is possible to characterize forest resources in different forests: from plain with beech or pine regular forest to mountain covered by heterogeneous fir-spruce forests. The accuracy of LIDAR inventories is equivalent to that of traditional field methods. Moreover, LIDAR allows to achieve a continuous mapping of modeled variables, and calculations at different scales (plots, forest). If thousands of square kilometers have been mapped using LIDAR data, demonstrators are underway at ONF to evaluate the operational use of these innovative methods, before a possible implementation to large scales.

International audience; The protective effect of forest against rockfall can be quantified thanks to the integration of tree impacts in rockfall simulation models. However, this requires to input spatially-explicit forest data, which is very labor-intensive to collect in operational conditions. Here we compare rockfall simulation results obtained with the RockyFor3D model in three different forest stands, and with different input data sources based on combinations of field and airborne laser scanning (ALS) data. For high rock energies (>200 kJ), the contribution of forest to hazard mitigation is minor compared to topography effects. The rockfall intensity is well estimated (within 16% of reference value) whatever the input data, whereas the frequency is generally over-estimated. Remote-sensing based methods yield a satisfactory assessment of forest protection, but the method based on tree detection has a tendency to under-estimate. ALS turns out to be an interesting option for cost-efficient inventory of high resolution input data for stand-level assessment of forest protection in rockfall simulations. Remote sensing also has potential for a better integration of the spatial heterogeneity of species and stem density, but results have to be evaluated at the hillside level.

International audience; In mountainous areas, forest resources are spatially heterogeneous, difficult to quantify and harvest because of topographical constraints. In the framework of the PlanEter project, we implemented a case study with the objective to map the quantity and mobilisation cost of fuelwood in six communes (600 km 2 ), thanks to LiDAR remote sensing data, a GIS-based accessibility model and an economical model. The main challenge was to design an easily reproducible methodology for larger application, thanks to the use of available public data as inputs. Wood growing stock was mapped at 20 m resolution with models calibrated with national LiDAR and forest inventory data. The technical accessibility of forests to common machinery was mapped with the GIS-based model Sylvaccess using national digital terrain and landscape models. Local forest statistics and cover maps were used to derive the annual timber and fuelwood potential. The results are used at two different levels. For strategic planning, a simplified model of forest management is used to map the annual timber and fuelwood harvest and the corresponding costs, depending on the technical accessibility and forest structure. For operational planning, resource and accessibility information are used as inputs along with expert information, into the detailed model CalCouFor for cost calculation of harvesting, yarding, transport, storage and chipping operations. This case study shows that it is possible to bridge the gap between regional forest statistics and field level analysis through the use of existing datasets. Additional remote sensing information could provide information about annual forest increment and species which would improve the model of forest management, taking into account other socio-economic services provided by forests. It would then be possible to map the effects of various factors (energy price, labour costs, public subsidies) on forest biomass mobilisation.

International audience; In mountainous areas, forest resources are spatially heterogeneous, difficult to quantify and harvest because of topographical constraints. In the framework of the PlanEter project, we implemented a case study with the objective to map the quantity and mobilisation cost of fuelwood in six communes (600 km 2 ), thanks to LiDAR remote sensing data, a GIS-based accessibility model and an economical model. The main challenge was to design an easily reproducible methodology for larger application, thanks to the use of available public data as inputs. Wood growing stock was mapped at 20 m resolution with models calibrated with national LiDAR and forest inventory data. The technical accessibility of forests to common machinery was mapped with the GIS-based model Sylvaccess using national digital terrain and landscape models. Local forest statistics and cover maps were used to derive the annual timber and fuelwood potential. The results are used at two different levels. For strategic planning, a simplified model of forest management is used to map the annual timber and fuelwood harvest and the corresponding costs, depending on the technical accessibility and forest structure. For operational planning, resource and accessibility information are used as inputs along with expert information, into the detailed model CalCouFor for cost calculation of harvesting, yarding, transport, storage and chipping operations. This case study shows that it is possible to bridge the gap between regional forest statistics and field level analysis through the use of existing datasets. Additional remote sensing information could provide information about annual forest increment and species which would improve the model of forest management, taking into account other socio-economic services provided by forests. It would then be possible to map the effects of various factors (energy price, labour costs, public subsidies) on forest biomass mobilisation.

L'objectif de ce mandat est : - d'une part de mettre en valeur les données LiDAR fédérales pour effectuer une cartographie de la ressource forestière sur l'ensemble du Valais et d'en tirer des estimations statistiques par unité administrative (triage forestier, Canton, arrondissement). - d'autre part de cartographier les modes de débardage avec le modèle automatique Sylvaccess.

National audience; Regarding the accuracy of forest estimates from airborne laser scanning data, several important parameters have already been investigated separately, and in various forest contexts. Comparing the results in order to determine the optimal setting for the accuracy and cost of LiDAR-based inventories thus remains difficult. This article presents an assessment of the sensitivity of prediction accuracy with regards to ground truth data and to the use of spectral information from LiDAR intensity or RapidEye satellite imagery. With data fusion the error of basal area estimates is 16.5%, 23.7% for stem density and 11.7% for mean diameter. The size and diameter threshold of the forest plots greatly impact prediction accuracy.The position accuracy and sample number of forest plots are crucial points as they influence both prediction error and its estimation.; Les différents critères qui influencent la précision de la méthode surfacique d’estimation de paramètres forestiers à partir de données LiDAR ont été étudiés le plus souvent séparément et dans des contextes forestiers variés. Cela limite les possibilités de comparaison en vue de la détermination des critères prépondérants du point de vue de la précision et du coût de ce type d’inventaire. Cet article présente une évaluation de l’influence des caractéristiques des relevés de terrain sur la précision des modèles d’estimation, ainsi que de l’intérêt de l’utilisation de l’information spectrale LiDAR ou issue de fusion avec des données satellitaires RapidEye. L’erreur obtenue en combinant les données LiDAR et RapidEye est de 16,5% pour la surface terrière, 23,7% pour la densité de tiges et 11,7% pour le diamètre moyen. Le choix de la taille de la placette de terrain et du diamètre de recensement ont un impact important sur l’erreur d’estimation. La précision du géoréférencement et le nombre de placettes apparaissent comme cruciaux dans la mesure où ils influencent non seulement l’erreur d’estimation mais également la précision de son évaluation.

National audience; Les données laser aéroporté à haute résolution permettent désormais de modéliser finement la propagation des blocs rocheux en intégrant les effets de la topographie et de la végétation.

Silvilaser 2015, La Grande Motte, FRA, 28-/09/2015 - 30/09/2015; International audience; This study presents a new open-access dataset of 18 plots from alpine forests of the Alpine Space. Eight detection algorithms were tested and evaluated against forest inventory data using a novel, automated matching procedure. Forest structure remains a key issue limiting tree detection, and algorithms would probably benefit from an adaptive tuning in order to achieve a better trade-off between omission and commission errors.

National audience; While the use of field plots and airborne LiDAR data for the estimation of forest parameters has been intensively investigated in the past ten years, the issue of the evaluation of their accuracy at the compartment level (surface of a few hectares)remains poorly documented. Based on a full-calliper inventory of 35 compartments representing 380 ha, the present study compares different strategies for the mapping of LiDAR predictions and their aggregation by compartment. Results show that the prediction error decreases between estimations at the plot and those at the compartment levels : from 15 to 6.4%for basal area, 26 to 7.7% for stem density and 6.5 to 3.4% for mean diameter. At the compartment level, a LiDAR-based inventory thus displays an accuracy similar to a full-calliper inventory, for basal area. For the mapping step, it is crucial to respect the size of field plots used for calibration, whereas for the aggregation step the handling of compartment borders remains tricky for all forest parameters.; Alors que les méthodes de modélisation des paramètres forestiers à partir de placettes de terrain et de données LiDAR aéroporté ont fait l’objet de nombreuses publications dans la dernière décennie, la question de leur utilisation et de leur évaluation à l’échelle de la parcelle forestière (surface de l’ordre de quelques hectares) reste peu documentée. La présente étude s’appuie sur un jeu de donnée d’inventaire en plein de 35 parcelles forestières sur 380 ha pour comparer différentes stratégies de spatialisation lors de l’application des modèles et d’agrégation des estimations de paramètres forestiers par parcelle. Les résultats montrent une diminution des erreurs entre les estimations à l’échelle de la placette et celles à l’échelle de la parcelle de 15 à 6.4% pour la surface terrière, de 26 à 7.7% pour la densité de tiges et de 6.5 à 3.4% pour le diamètre. À l’échelle de la parcelle, la précision de l’inventaire basé sur les données LiDAR se révèle similaire à celle d’un inventaire en plein, pour la surface terrière. Pour la spatialisation lors de l’application des modèles, le plus important est de respecter la taille des placettes de calibration terrain, alors que pour l’agrégation par parcelle le traitement des bordures reste délicat quel que soit le paramètre forestier.

NEWFOR (NEW technologies for a better mountain FORest timber mobilization) is a research project funded by the alpine space programme. It brings together 14 institutes of the 6 countries of the alpine space with the aim of improving forest timber evaluation and mobilisation. The project considers the whole wood supply chain, from forests to wood yards, with a particular emphasis on new remote sensing technologies and geographical information systems. The report presents the major findings of the project and outlines some key issues for a better mobilisation of alpine timber ressoures.

International audience; Identifying and characterizing the conditions of access to the forest resources are strategic points for their sustainable and multifunctional management. The model Sylvaccess presented in this article is designed to automatically map the accessibility of forests with the tree main logging techniques currently used in France: skidder, forwarder and cable yarder. The model is based on spatial information and specific parameters of each logging technique. It can also integrate physical or environmental obstacles in the analysis. The outputs of the model can be used for many applications ranging from forest management and planning of logging operations to the comparison and selection of new forest roads projects.; Identifier et caractériser les conditions d'accès aux ressources forestières sont des points stratégiques pour assurer leur gestion durable et multifonctionnelle. Le modèle Sylvaccess présenté dans cet article permet de cartographier automatiquement l'accessibilité des forêts en fonction des principaux systèmes de débardage actuellement utilisés en France : le tracteur forestier (ou « débusqueur »), le porteur forestier et le débardage par câble. Le modèle s'appuie sur des sources d'information spatiale et des paramètres techniques propres à chaque système de débardage. Il offre aussi la possibilité d'intégrer des obstacles physiques ou environnementaux dans l'analyse. Les résultats du modèle sont utilisables pour de nombreuses applications forestières allant de l'aménagement et de la planification des opérations d'exploitation jusqu'à la comparaison et la sélection de projets de desserte.

Ce rapport présente la méthodologie et les résultats de la tâche visant à produire une cartographie des ressources bois-énergie accessibles. La zone d'étude était située dans le Bas-Valais (Suisse). La méthodologie s'appuie notamment sur les développements récents de méthodes de cartographie des forêts basées sur les données de télédétection par scanner laser aéroporté, ainsi que sur les méthodes automatiques de traitement de données SIG pour caractériser les conditions techniques d’accessibilité des forêts.

Cette étude d’une durée totale de 18 mois présente deux principaux volets. Le premier consiste à valoriser les données Lidar acquises par le CRAIG sur les départements du Cantal et de la Haute-Loire pour l’estimation de paramètres forestiers comme par exemple le volume de bois sur pied et la hauteur dominante des peuplements. Cette estimation peut être complétée à l’aide des données du CRPF et plus précisément des Plans de Développement de Massif (PDM) afin d’obtenir une estimation plus précise des volumes prélevables sur un massif forestier. Le second objectif est de cartographier à grande échelle (100 000 ha) l’accessibilité et l’exploitabilité des forêts à l’aide du modèle Sylvaccess développé au centre Irstea de Grenoble. Ce modèle, initialement conçu pour le tracteur forestier et le débardage par câble aérien, a été complété par un module pour le porteur forestier. Sylvaccess fournit de nombreuses données utiles à la réflexion globale sur la planification de la desserte et de l’exploitation forestière.

NEWFOR (NEW technologies for a better mountain FORest timber mobilization) is a research project funded by the alpine space programme. It brings together 14 institutes of the 6 countries of the alpine space with the aim of improving forest timber evaluation and mobilisation. The project considers the whole wood supply chain, from forests to wood yards, with a particular emphasis on new remote sensing technologies and geographical information systems. The report presents the major findings of the project and outlines some key issues for a better mobilisation of alpine timber ressoures.

Ce rapport présente la méthodologie et les résultats de la tâche visant à produire une cartographie des ressources bois-énergie accessibles. La zone d'étude était située dans le Bas-Valais (Suisse). La méthodologie s'appuie notamment sur les développements récents de méthodes de cartographie des forêts basées sur les données de télédétection par scanner laser aéroporté, ainsi que sur les méthodes automatiques de traitement de données SIG pour caractériser les conditions techniques d’accessibilité des forêts.

INTERPRAEVENT International Symposium on Natural Disaster Mitigation to Establish Society With Resilience 2014, Nara, JPN, 25-/11/2014 - 28/11/2014; International audience; Rockfall simulation models are now able to take into account impacts on trees. Forest maps at operational scale can be derived from statistical inventories or remote sensing such as airborne laser scanning (ALS). The present study compares rockfall simulations results obtained with different forest inputs for three different stands. Results obtained with field and ALS-derived inputs showed little difference regarding energy values but sometimes large ones concerning the proportion of stopped blocks. It also turns out that the spatial heterogeneity and coniferous proportion of trees are critical issues which might be difficult to address at operational scale.

National audience; La majorité des travaux de recherche sur l'estimation de paramètres forestiers par l'approche lidar surfacique sont focalisés sur l'aspect "modélisation" et validé à l'échelle de la placette. Dans cette étude, nous nous intéressons à l'effet sur la précision des estimations de deux étapes cruciales pour la production de cartes à l'échelle de la parcelle : la spatialisation des estimations et l’agrégation des résultats par unité de gestion. Les résultats obtenus ont été comparés avec un inventaire pied à pied sur 35 parcelles représentant 380 ha et montrent que les estimations à la parcelle obtenues par lidar sont d'une précision équivalente à celles obtenues par des inventaires en plein, notamment pour la surface terrière.

International audience; Cette étude compare différentes méthodes pour l'estimation de paramètres forestiers à partir de données de télédétection, incluant le lidar, sur une vaste zone situées dans les Vosges haut-rhinoises.

International audience; Continuous maps of forest parameters can be derived from airborne laser scanning (ALS) remote sensing data. A prediction model is calibrated between local point cloud statistics and forest parameters measured on field plots. Unfortunately, inaccurate positioning of field measures lead to a bad matching of forest measures with remote sensing data. The potential of using tree diameter and position measures in cross-correlation with ALS data to improve co-registration is evaluated. The influence of the correction on ALS models is assessed by comparing the accuracy of basal area prediction models calibrated or validated with or without the corrected positions. In a coniferous, uneven-aged forest with high density ALS data and low positioning precision, the algorithm co-registers 91% of plots within two meters from the operator location when at least the five largest trees are used in the analysis. The new coordinates slightly improve the prediction models and allow a better estimation of their accuracy. In a forest with various stand structures and species, lower ALS density and differential Global Navigation Satellite System measurements, position correction turns out to have only a limited impact on prediction models.

[Departement_IRSTEA]Territoires [TR1_IRSTEA]SEDYVIN; International audience; While significant progress has been achieved regarding the modeling of rockfall propagation, including the interaction with forest cover, the estimation of the spatial distribution of rockfall frequency in release areas remains a challenging task. Indeed, the very low number of observed events makes any statistical modeling difficult, so that estimations are generally based on expert knowledge regarding the rock structure of start zone and a few recorded events. Meanwhile, recent dendrochronological studies have shown that trees, as silent witnesses of numerous events, bring information about the spatial distribution of rockfall trajectories. This data can be compared to simulated trajectories in order to reconstruct the most probable start zones. This study presents a methodological framework for the resolution of this problem and a real-case example. Consider a forest patch f located below a cliff with Ns potential rockfall start points. The expected value of the number of impacts in this forest patch during lapse T can be calculated as (1) : Impacts(f)=T*SUM(Pstart(i)*Ppropagation(i,f)) with Pstart(i) the probability of rockfall for the start point I, Ppropagation(i,f) the probability that a rockfall from start point i impacts at least one tree in the forest patch f . The matrix formulation for the whole forest divided in Nf forest patches is (2): Impact=T*Pstart*Propagation. Ppropagation can be estimated with numerical simulations, by taking into account the current state of the forest stand if trees are measured and georeferenced, and if the volume of rockfall is known. Impact can be also be observed on the field. Under the assumption that the forest stand and the start frequencies do not change during lapse T, the equation system (2) can be solved for the coefficients of Pstart. This framework was tested on a rockfall area located in Valdrôme, France, where more than 1000 trees were measured and geolocated in a one hectare stand located below a cliff with rockfall activity. The number of impacts on each tree was observed and starting points were determined based on a slope criterion applied to a digital terrain model. The rockfall trajectories were simulated with the RockyFor3D software. The system (2) was solved with an optimization routine. Results show that the rockfall frequencies display high spatial variability and that 21% of starting points account for more than 50% of rockfall releases. The coefficients of Pstart can then be used to weight the simulated trajectories in order to map the spatial variability of rockfall trajectories and corresponding kinetic energies, which is an important step for a better identification of hazard areas. The proposed framework can thus lead to a numerical solution, but it relies on numerous assumptions. The spatial variations of rockfall volume are not taken into account and the forest stand is supposed time-invariant. This last point could be tackled by simulating trajectories during past stages of the stand evolution with growth simulation models.

In the presented study established single tree detection methods are benchmarked and investigated. In total eight airborne laser scanning (ALS) based detection methods were applied to a unique dataset originating from different regions of the Alpine Space covering different study areas, forest types and structures. The evaluation of the detection results was carried out in a clear and reproducible way by automatically matching the detection results to precise in-situ forest inventory data. Quantitative statistical parameters such as the percentages of correctly matched trees and omission and commission errors are presented. The benchmarking results are prepared in complementary levels of information, starting with the analysis based on study area as well as detection method. Additionally investigations per forest type and an overall performance of the benchmark are presented. The best matching rate was obtained for single layered coniferous forests. Trees in lower height layers were challenging for all tested methods. The overall performance shows a matching rate of 47% which is comparable to results of other benchmarks performed in the past for other forest types. The study brings new hindsight regarding the potential and limits of tree detection with ALS and underlines some key aspects regarding the choice of method when performing single tree detection for the various forest types encountered in alpine regions.

International audience; Améliorer la connaissance et la description fidèle des écosystèmes forestiers est un prérequis indispensable pour parvenir à relever le défi de la gestion durable des forêts. L’organisation spatiale de la végétation joue un rôle fondamental dans le fonctionnement de ces écosystèmes. L’utilisation d’un système Lidar en milieu forestier permet de caractériser la structure en 3D de la végétation, ainsi que la topographie du terrain sous le couvert. L’objectif de cette intervention est de présenter des méthodes développées à l’Irstea et à l’Université de Sherbrooke pour analyser les données Lidar et en déduire des indicateurs de structure ou des paramètres biophysiques au niveau des arbres ou à l’échelle des peuplements. Ces approches nous permettent d’évaluer la ressource forestière en termes de volume de bois, de biomasse et de densité (e.g. surface terrière, profil vertical de végétation ou taux de couvert) sur des peuplements de complexité variable. Les capacités du Lidar sont aussi évaluées pour l’étude du lien entre la structure de la végétation forestière et la biodiversité, afin d’identifier des pratiques sylvicoles favorables à la biodiversité. Des systèmes Lidar terrestre et aérien sont utilisés et leur complémentarité est évaluée pour des études limitées aux massifs forestiers ou à l’échelle d’une région. Des projets de Lidars spatiaux sont aussi à l’étude en France dans le but de caractériser la ressource forestière à l’échelle mondiale et de permettre d’améliorer la quantification des stocks et flux de carbone des écosystèmes forestiers. Ces informations sont nécessaires aux modèles utilisés pour prédire l’évolution du climat à moyen et long terme et renforcer le rôle des forêts dans la régulation de ce changement global.

National audience; Le scanner laser (ou LiDAR) aéroporté est une technique de télédétection qui a connu un fort développement ces dix dernières années. Elle permet de caractériser en trois dimensions le sol et les éléments de la surface terrestre (végétation, bâtiments...). Ses applications en gestion forestière sont nombreuses, depuis la caractérisation de l'accessibilité des parcelles à la cartographie de la structure des peuplements. Certains produits, comme le modèle numérique de terrain et le modèle de la hauteur de la canopée (résolution métrique), sont directement utilisables en gestion opérationnelle. La cartographie de paramètres comme le volume ou la surface terrière nécessite la calibration de modèles d'estimation à partir de placettes de calage inventoriées sur le terrain. La modélisation géométrique des forêts obtenue par télédétection LiDAR permet en effet de cartographier des paramètres quantitatifs des peuplements, ce qui n'était possible qu'indirectement par photogrammétrie. Cette spatialisation de l'estimation de la ressource, qui peut ensuite être croisée avec une modélisation de l'accessibilité et des conditions économiques de récolte, répond à l'enjeu crucial de la cartographie opérationnelle des ressources mobilisables. Cette technologie est couramment utilisée pour la gestion forestière dans les pays Nordiques, mais n'en est qu'à ses débuts en France où la complexité des forêts pose encore des questions à la fois méthodologiques et organisationnelles.

International audience; Numerous methods have been proposed for the detection of single trees in airborne laser scanning (ALS) data. Most of them are highly dependent on the initial settings of the algorithm, as parameters (e.g. smoothing of the digital canopy height model) will affect the overall detection performance, and more particularly the trade-off between omission and commission errors. To tackle this issue, the use of prior information about the forest stand is possible when ground truth for the area is available. Alternatively, adaptive parametrization in the course of the detection procedure requires more complex algorithms which might have trouble when processing large areas. In this article a procedure for automated parametrization is presented. It is based on the unsupervised training of the detection algorithm with reference forest plots including coregistered field and ALS data. The local maxima filtering algorithm is adopted as it is simple and fast. The training step consists in evaluating the detection performance of the algorithm on the reference plots for several parameter combinations. Detection quality is evaluated as a trade-off between the number of correctly detected trees and the number of false detections. When trees are to be detected in a newly surveyed area, two possibilities for algorithm parametrization are compared. The first option is to use the parameter combination that is the more robust when used on the training set (“average” setting). The second option is to use the combination that yields the best detection on the ALS point cloud from the training set that best resembles the new data. The matching criterion is based on the Fourier spectrum of the canopy height model computed from the point cloud. 26 forest plots located in seven different ALS surveys of mountainous areas are used to test the workflow. Plots have a minimum area of 0.25 ha and represent various stand structures and tree species. To compare the two parametrization options and evaluate their sensitivity to the training set size (number of reference plots), a cross validation procedure based on repetitive sampling of the training set among the available 26 plots is performed. Results show that for training sets with less than 15 plots, the “average” setting performs better, whereas with training sets larger than 20, the matching procedure yields better detection performance. This method for unsupervised training is quite flexible as it can be used with any detection algorithm that requires initial parametrization. Moreover, the detection performance criterion can be modified in order to reflect the end-user preference regarding detection results. This study is an example of how single tree methods can benefit from an area-based analysis. Further work should investigate whether metrics usually computed for area-based methods (e.g. height quantiles) could also improve the point cloud matching.

Ce rapport scientifique dresse un état des lieux des connaissances sur l'utilisation du scanner laser aéroporté pour les applications forestières.

International audience; Knowledge of spatial distribution of wood quantity and quality in a region along with information on forest accessibility and road network is a fundamental component for sustainable and efficient utilization of forest resources. In the Alpine region the role of accessibility analysis of forest areas is extremely important. Classic remote sensing techniques alone provide only part of the information which is necessary for a correct management of the resource. Laser scanning techniques provide added value by providing data which can be used to estimate volume, structure and accessibility. The objective of project NEWFOR (NEW technologies for a better mountain FORest timber mobilization) is to use laser scanning technology as support to forest management and harvesting for a more efficient and sustainable use of forest resources.

Ce rapport scientifique dresse un état des lieux des connaissances sur l'utilisation du scanner laser aéroporté pour les applications forestières.

International audience; La gestion des forêts à fonction de protection est facilitée par la télédétection par scanner laser aéroporté. Les modèles de terrain et informations forestières qu'elle permet de cartographies sont en effet des variables cruciales pour la cartographie des forêts de protection et pour leur caractérisation par la simulation de chute de blocs.

Numerous studies have shown the potential of airborne laser scanning for the mapping of forest resources. However, the application of this remote sensing technique to complex forests encountered in mountainous areas requires further investigation. In this thesis, the two main methods used to derive forest information are tested with airborne laser scanning data acquired in the French Alps, and adapted to the constraints of mountainous environments. In particular, a framework for unsupervised training of treetop detection is proposed, and the performance of support vector regression combined with dimension reduction for forest stand parameters estimation is evaluated.; De nombreux travaux ont montré le potentiel de la télédétection par scanner laser aéroporté pour caractériser les massifs forestiers. Cependant, l'application aux forêts complexes de montagne reste encore peu documentée. On se propose donc de tester les deux principales méthodes permettant d'extraire des paramètres forestiers sur des données acquises en zone montagneuse et de les adapter aux contraintes spécifiques à cet environnement. En particulier on évaluera d'une part l'apport conjoint de la régression à vecteurs de support et de la réduction de dimension pour l'estimation de paramètres de peuplement, et d'autre part l'intérêt d'un apprentissage non supervisé pour la détection d'arbres.

Numerous studies have shown the potential of airborne laser scanning for the mapping of forest resources. However, the application of this remote sensing technique to complex forests encountered in mountainous areas requires further investigation. In this thesis, the two main methods used to derive forest information are tested with airborne laser scanning data acquired in the French Alps, and adapted to the constraints of mountainous environments. In particular, a framework for unsupervised training of treetop detection is proposed, and the performance of support vector regression combined with dimension reduction for forest stand parameters estimation is evaluated.; De nombreux travaux ont montré le potentiel de la télédétection par scanner laser aéroporté pour caractériser les massifs forestiers. Cependant, l'application aux forêts complexes de montagne reste encore peu documentée. On se propose donc de tester les deux principales méthodes permettant d'extraire des paramètres forestiers sur des données acquises en zone montagneuse et de les adapter aux contraintes spécifiques à cet environnement. En particulier on évaluera d'une part l'apport conjoint de la régression à vecteurs de support et de la réduction de dimension pour l'estimation de paramètres de peuplement, et d'autre part l'intérêt d'un apprentissage non supervisé pour la détection d'arbres.

Dans ce livre, le Cemagref, institut de recherche en sciences et technologies pour l'environnement, fait le point de ses travaux scientifiques menés sur la forêt. Pourquoi une recherche sur la forêt ? D'abord parce qu'elle occupe plus d'un quart de la surface de notre territoire métropolitain. Ensuite, parce qu'elle fournit tout un ensemble de services variés et, pour certains, essentiels à notre société : la production de bois, la protection contre les aléas naturels, la chasse et la promenade en sont de très bonnes illustrations. Enfin, parce qu'elle constitue un milieu naturel riche en biodiversité et occupe une place notable dans les débats sur les enjeux environnementaux. Forts de ce constat, les chercheurs et experts du Cemagref apportent leur contribution en abordant des problématiques majeures pour la gestion durable des forêts : état et dynamique de la forêt, de sa biodiversité, vulnérabilité des forêts aux changements globaux et stratégies d'adaptation, gestion multifonctionnelle des forêts, rôle de protection contre les aléas naturels, place de la forêt dans les territoires et enjeux sociaux. L'interaction entre forêt et gestion se situe ainsi au c½ur de nos travaux.

National audience; En montagne, les aléas naturels comme les chutes de blocs sont une des principales menaces pesant sur la population et les infrastructures. Un document cartographique des zones concernées par un aléa et une qualification précise du risque sont nécessaires pour que les autorités publiques puissent prendre les décisions en matière d'aménagement du territoire. En France, les Plans de Prévention des Risques naturels prévisibles (PPR) sont des documents qui cartographient les zones de risques et qui s'imposent aux documents d'urbanisme pour une durée moyenne de 10 ans. Lors de la révision du PPR de Veyrier-du-Lac (Haute-Savoie, France), une nouvelle méthodologie utilisant la géomatique pour l'estimation du risque de chute de blocs a été développée. L'analyse conjointe de données à haute résolution LiDAR, de relevés terrain et de simulations trajectographiques 2D et 3D sous SIG a permis de réaliser des cartes d'aléa précises en combinant les fréquences et intensités des phénomènes. Cette méthode prend en compte l'éventuel effet protecteur du couvert forestier dans un environnement urbanisé. Le SIG a permis aux autorités de tester différents seuils de risque avant de valider le zonage PPR. Cette approche SIG permet une coopération plus facile entre experts scientifiques et autorités publiques.

National audience; La mobilisation des bois en montagne est handicapée par la méconnaissance des volumes existants et les difficultés de prospection. Le scanner laser est un technique de télédétection permettant de caractériser avec précision les éléments de la surface terrestre. Dans cet article, nous avons utilisé des données acquises par scanner laser pour reconstituer une image du couvert forestier puis identifier le sommet et la hauteur des principaux arbres. On obtient ainsi une cartographie du volume sur pied qui peut-être utilisée pour optimiser les opérations d'exploitation forestière.