Search

Characterization of forest structure is important for management-related decision making, monitoring, and adaptive management. Increasingly, observations of forest structure are needed at both finer resolutions and across greater extents to support spatially explicit management planning. Unmanned aerial system (UAS) based photogrammetry provides an airborne method of forest structure data acquisition at a significantly lower cost and time commitment than existing methods such as airborne laser scanning (LiDAR). This study utilizes nearly 5000 stem-mapped trees in ponderosa pine (Pinus ponderosa Lawson & C. Lawson) dominated forests to evaluate several algorithms for detecting individual tree locations and characterizing crown area across tree sizes. Our results indicate that adaptive variable window detection methods with UAS-based canopy height models have greater tree detection rates compared with fixed window analysis across a range of tree sizes. Using the UAS approach, probability of detecting individual trees decreases from 97% for dominant overstory to 67% for suppressed understory trees. Additionally, crown radii were correctly determined within 0.5 m for approximately two-thirds of sampled trees. These findings highlight the potential for UAS photogrammetry to characterize forest structure through the detection of trees and tree groups in open-canopy ponderosa pine forests. Further work should investigate how these methods transfer to more diverse species compositions and forest structures. Key words: detection, canopy height model, crown delineation, drone, forest structure. Il est important de caracteriser la structure forestiere pour eclairer les decisions de gestion, le suivi et la gestion adaptative. De plus en plus, des observations de la structure forestiere sont necessaires, tant a des resolutions plus fines que sur de plus vastes etendues, pour soutenir une planification de gestion spatialement explicite. La photogrammetrie basee sur des systemes aeriens sans pilote (SASP) fournit une methode aeroportee d'acquisition de donnees sur la structure forestiere a un cout et un investissement en temps nettement inferieurs a ceux des methodes existantes telles que le balayage laser aeroporte. Cette etude utilise pres de 5 000 tiges d'arbre cartographiees, dans des forets dominees par le pin ponderosa (Pinus ponderosa Lawson & C. Lawson), pour evaluer plusieurs algorithmes de detection de l'emplacement des arbres individuels et de caracterisation de la surface de la couronne des arbres de differentes tailles. Nos resultats indiquent que les methodes adaptatives de detection a fenetre variable avec des modeles de hauteur du couvert forestier sur la base des donnees de SASP ont des taux de detection d'arbres plus eleves que l'analyse a fenetre fixe pour un assortiment de tailles d'arbres. Avec l'utilisation des SASP, la probabilite de detecter des arbres individuels passe de 97 % pour l'etage dominant a 67 % pour les arbres opprimes du sous-etage. De plus, les rayons de la couronne ont ete correctement determines, a moins de 0,5 m, pour environ les deux tiers des arbres echantillonnes. Ces resultats mettent en evidence le potentiel de la photogrammetrie basee sur les SASP pour caracteriser la structure forestiere grace a la detection d'arbres et de groupes d'arbres dans les forets de pins ponderosa dont le couvert n'est pas ferme. Des travaux supplementaires devraient etudier comment ces methodes peuvent s'appliquer a des compositions d'especes et des structures forestieres plus diverses. [Traduit par la Redaction] Mots-cles : detection, modele de hauteur du couvert forestier, delimitation des couronnes, drone, structure forestiere., Introduction Since the early 2000s, ecological management of dry, lower montane forest systems has increasingly focused on variability and spatial arrangement of horizontal and vertical forest structure (Lydersen et al. [...]

Increasing frequency of droughts and wildfire are sparking concerns that these compounded disturbance events are pushing forested ecosystems beyond recovery. An improved understanding of how compounded events affect tree physiology and mortality is needed given the reliance of fire management planning on accurate estimates of postfire tree mortality. In this study, we use a toxicological dose-response approach to quantify the impact of variable-intensity drought and fire on the physiology and mortality of Pinus monticola and Pseudotsuga menziesii saplings. We show that the dose-response relationship between fire intensity and mortality shifts toward increased vulnerability under drought, indicating higher mortality with increasing drought at any fire intensity. The trajectory we observed in postfire chlorophyll fluorescence, an indicator of photosynthetic efficiency and stress, was an effective early warning sign of impending tree death. Postfire mortality modeling shows that accurate mortality classification can be achieved using prefire physiology and morphology metrics combined with fire intensity. Variable importance measures indicate that physiological condition and fire intensity have greater influence on the classification accuracy than morphological metrics. The wide range in drought and fire responses observed between this study and others highlights the need for more research on compound disturbance effects. Study Implications : An improved understanding of how drought and fire affect tree physiology and mortality is needed by natural resource managers looking to predict postfire tree mortality. This study advances our compound disturbance understanding by subjecting conifer saplings to variable drought and fire intensities and quantifying and modeling moderate-term recovery and mortality. The results show reduced physiological recovery and amplified mortality in saplings exposed to greater drought and fire intensity. Overall, this study highlights the importance of physiological condition when modeling tree mortality and could potentially be used to inform current postfire tree mortality models. [ABSTRACT FROM AUTHOR]

: Light detection and ranging, or LiDAR, effectively produces products spatially characterizing both terrain and vegetation structure; however, development and use of those products has outpaced our understanding of the errors within them. LiDAR's ability to capture three-dimensional structure has led to interest in conducting or augmenting forest inventories with LiDAR data. Prior to applying LiDAR in operational management, it is necessary to understand the errors in Li-DAR-derived estimates of forest inventory metrics (i.e., tree height). Most LiDAR-based forest inventory metrics require creation of digital elevation models (DEM), and because metrics are calculated relative to the DEM surface, errors within the DEMs propagate into delivered metrics. This study combines LiDAR DEMs and 54 ground survey plots to investigate how surface morphology and vegetation structure influence DEM errors. The study further compared two LiDAR classification algorithms and found no significant difference in their performance. Vegetation structure was found to have no influence, whereas increased variability in the vertical error was observed on slopes exceeding 30°, illustrating that these algorithms are not limited by high-biomass western coniferous forests, but that slope and sensor accuracy both play important roles. The observed vertical DEM error translated into [+ or -]1%-3% error range in derived timber volumes, highlighting the potential of LiDAR-derived inventories in forest management. Resume: Le lidar (la detection et la telemetrie par la lumiere laser) peut servir a cartographier efficacement la morphologie du terrain et la structure de la vegetation. Cependant, le developpement et l'utilisation de ces cartes ont devance notre comprehension des erreurs qu'elles contiennent. La capacite du lidar a cartographier la structure tridimensionnelle de la vegetation a suscite un interet pour realiser ou etoffer les inventaires forestiers a l'aide de donnees lidar. Toutefois, avant d'appliquer le lidar dans la gestion operationnelle, il est necessaire de comprendre les erreurs d'estimation des metriques d'inventaire forestier derivees du lidar (comme la hauteur des arbres). La plupart des metriques d'inventaire forestier deri-vees du lidar exigent la creation de modeles numeriques de terrain (MNT). Comme ces metriques sont calculees a partir de la morphologie du terrain, les erreurs dans les MNT se propagent dans les metriques qui sont produites. Cette etude combine les MNT du lidar et 54 placettes au sol pour etudier comment la morphologie du terrain et la structure de la vegetation influencent les erreurs dans les MNT. De plus, nous avons compare deux algorithmes de classification des donnees lidar et nous n'avons trouve aucune difference significative entre leur performance. La structure de la vegetation n'exerce aucune influence sur l'erreur des MNT. Par contre, lorsque la pente depasse 30°, la variation de l'erreur verticale augmente. Ces resultats montrent que ces algorithmes ne sont pas limites par les forets de coniferes a forte biomasse de l'ouest des Etats-Unis, mais que la precision du capteur et la pente influencent l'erreur des MNT de facon importante. L'erreur verticale des MNT s'est traduite par une marge d'erreur de ±1-3 % dans les volumes de bois qui ont ete calcules. Cette faible marge d'erreur met en evidence le potentiel des inventaires bases sur le lidar pour la gestion forestiere. [Traduit par la Redaction], Introduction Light detection and ranging (LiDAR) is becoming one of the most effective and reliable means of characterizing surface topography and vegetation structure (Falkowski et al. 2009a). LiDAR is being [...]