Search

Timely and accurate knowledge of species-level biomass is essential for forest managers to sustain forest resources and respond to various forest disturbance regimes. In this study, maps of species-level biomass in Chinese boreal forests were generated by integrating Moderate Resolution Imaging Spectroradiometer (MODIS) images with forest inventory data using k nearest neighbor (kNN) methods and evaluated at different scales. The performance of 630 kNN models based on different distance metrics, k values, and temporal MODIS predictor variables were compared. Random Forest (RF) showed the best performance among the six distance metrics: RF, Euclidean distance, Mahalanobis distance, most similar neighbor in canonical correlation space, most similar neighbor computed using projection pursuit, and gradient nearest neighbor. No appreciable improvement was observed using multi-month MODIS data compared with using single-month MODIS data. At the pixel scale, species-level biomass for larch and white birch had relatively good accuracy (root mean square deviation < 62.1%), while the other species had poorer accuracy. The accuracy of most species except for willow and spruce was improved up to the ecoregion scale. The maps of species-level biomass captured the effects of disturbances including fire and harvest and can provide useful information for broad-scale forest monitoring over time. Key words: species-level biomass, MODIS, Chinese boreal forest, Random Forest (RF), kNN. Une connaissance precise et en temps opportun de la biomasse de chaque espece est essentielle pour permettre aux amenagistes forestiers d'effectuer un amenagement durable des ressources forestieres et pour s'ajuster aux divers regimes de perturbations forestieres. Dans cette etude, des cartes de biomasse par espece ont ete generees dans les forets boreales chinoises en integrant des images MODIS (spectroradiometre imageur a resolution moyenne) a des donnees d'inventaire forestier au moyen de l'approche des k plus proches voisins (kNN) et evaluees a differentes echelles. Les performances de 630 modeles kNN ont ete comparees en fonction de differentes metriques de distance, valeurs de k et variables predictives temporelles de MODIS. Les forets d'arbres decisionnels (RF) ont fait ressortir les meilleures performances parmi les six metriques de distance : la distance RF, la distance euclidienne, la distance de Mahalanobis, le plus proche voisin dans l'espace de correlation canonique, le plus proche voisin calcule par poursuite de projection, et le plus proche voisin par gradient. L'utilisation de donnees MODIS sur plusieurs mois n'a apporte aucune amelioration notable en comparaison de l'utilisation des donnees MODIS d'un seul mois. A l'echelle des pixels, la biomasse par espece avait une precision relativement bonne pour le meleze et le bouleau blanc (ecart quadratique moyen < 62,1%), tandis que la precision etait plus faible pour les autres especes. Pour la plupart des especes, a l'exception du saule et de l'epinette, la precision s'est amelioree jusqu'a l'echelle de l'ecoregion. Les cartes de la biomasse par espece ont capte les effets des perturbations, y compris les feux et la recolte, et peuvent fournir des informations utiles pour la surveillance des forets a une vaste echelle au fil du temps. [Traduit par la Redaction] Mots-cles: biomasse par espece, spectroradiometre imageur a resolution moyenne (MODIS), foret boreale chinoise, forets d'arbres decisionnels (RF), methode des k plus proches voisins (kNN)., 1. Introduction The boreal forest is the second largest terrestrial biome in the world, covering 33% of forest area and holding 23% of terrestrial carbon stocks (Carlson et al. 2009; [...]

Forest landscape models (FLMs) are an important tool for assessing the long-term cumulative effects of harvest over large spatial extents. However, they have not been commonly used to guide forest management planning and on-the-ground operations. This is largely because FLMs track relatively simplistic vegetation information such as age cohort presence/absence, forest type, and biomass that are incompatible with tree density and size on which most harvest prescriptions are based. We describe and demonstrate the newly developed harvest module of the LANDIS PRO FLM, which tracks density, size, basal area, and stocking by species age cohorts for each site (cell). With this quantitative information, the module can simulate basal area controlled harvest, stocking-level controlled harvest, and group selection harvest. Through user-specified harvest year (frequency), stand ranking, and species and age preference, the new module can simulate a wide variety of harvest prescriptions such as thinning from above and below, shelterwood, clear-cutting, and group selection. We applied the LANDIS PRO harvest module to a large (17 000 [km.sup.2]) central hardwood forest landscape in Missouri. The simulated harvest prescriptions produced realistic stand-scale results when plotted on Gingrich stocking charts. The harvest module improves on previous versions by allowing partial treatment of individual age-classes within a cell and reporting results in metrics commonly used in stand-scale silviculture. It provides a closer link between landscape-scale simulation methods and stand-scale management. Resume: Les modeles de paysage forestier (MPF) sont des outils importants pour evaluer les effets cumulatifs a long terme de la recolte de bois sur de vastes superficies. Cependant, ils n'ont pas ete communement utilises comme guide pour la planification de l'amenagement forestier et les operations sur le terrain. Cela est surtout du au fait que les MPE font le suivi de donnees relativement simplistes sur la vegetation, telles que l'absence ou la presence de groupes d'age, le type forestier et la biomasse qui sont incompatibles avec la densite et la taille des arbres, donnees sur lesquelles sont fondees la plupart des prescriptions de recolte. Nous decrivons et illustrons le module de recolte qui vient d'etre developpe dans le MPF LANDIS PRO et qui fait le suivi de la densite, de la taille, de la surface terriere et de la densite relative par groupes d'age de chaque espece dans chaque station (cellule). Avec cette information quantitative, le module peut simuler une coupe limitee par la surface terriere, une coupe limitee par la densite relative ou une coupe de jardinage par groupe. a partir de l'annee de la recolte (frequence), du classement des peuplements, ainsi que des preferences concernant l'espece et l'age determines par l'utilisateur, le nouveau module peut simuler une vaste gamme de prescriptions de recolte, telles que l'eclaircie par le haut ou le bas, la coupe progressive, la coupe a blanc et la coupe de jardinage par groupe. Nous avons applique le module de recolte de LANDIS PRO a un vaste (17 000 [km.sup.2]) paysage de foret feuillue centrale au Missouri. La simulation de prescriptions de recolte a produit des resultats realistes a l'echelle du peuplement lorsqu'ils etaient representes sur des diagrammes de densite de Gingrich. Le module de recolte constitue une amelioration par rapport aux versions precedentes parce qu'il permet le traitement partiel de classes d'age individuelles a l'interieur d'une cellule et la presentation des resultats a l'aide de metriques communement utilisees pour la sylviculture a l'echelle du peuplement. Il procure un lien plus etroit entre les methodes de simulation a l'echelle du paysage et l'amenagement a l'echelle du peuplement. [Traduit par la Redaction], Introduction Harvest influences forest composition and structure from fine to broad spatial scales (Papaik et al. 2010). Forest landscape models (FLMs) are an important tool for understanding the cumulative effects [...]

Abstract: We compared four fire spread simulation methods (completely random, dynamic percolation, size-based minimum travel time algorithm, and duration-based minimum travel time algorithm) and two fire occurrence simulation methods (Poisson [...]

Abstract: We used the LANDIS model to study the long-term cumulative effects of postfire 10-year management (harvest and reforestation) on species abundance, age structure, and spatial pattern in the Tuqiang [...]

The LANDIS model simulates ecological dynamics, including forest succession, disturbance, seed dispersal and establishment, fire and winddisturbance, and their interactions. We describe the addition to LANDIS of capabilities to simulate forest vegetation management, including harvest. Stands (groups of cells) are prioritized for harvest using one of four ranking algorithms that use criteria related to forest management objectives. Cells within a selected stand are harvested according to the species and age cohort removal rules specified in a prescription. These flexible removal rules allow simulation of a wide range of prescriptions such as prescribed burning, thinning, single-tree selection, and clear-cutting. We present a case study of the application of LANDIS to a managed watershed in the Missouri (U.S.A.) Ozark Mountains to illustrate the utility of this approach to simulate succession as a response to forest management and other disturbance. The different cutting practices produced differences in species and size-class composition, average patch sizes (for patches defined by forest type or by size class), and amount of forest edge across the landscape. The capabilities of LANDIS provide a modeling tool to investigate questions of how timber management changes forest composition and spatial pattern, providing insight into ecological response to forestmanagement. [ABSTRACT FROM AUTHOR]

Natural disturbance patterns can provide useful information for ecosystem management. Our objective was to provide a detailed spatial picture of the pre-European settlement vegetation cover for the northwestern Wisconsin Pine Barrens and to compare it with the present vegetation cover. We analyzed the presettlement conditions using an extensive data set comprised of U.S. General Land Office surveyor records from the mid-19th century and related it to the vegetation cover in 1987 as depicted in a Landsat satellite forest classification. Changes were quantified by calculating differences in abundance and relative importance of tree species at presettlement time and today. Our results revealed a strong decline of jack, red, and white pine (Pinus banksiana Lamb., Pinus resinosa Ait., and Pinus strobus L., respectively),accompanied by an increase of oak (Quercus spp.), trembling aspen (Populus tremuloides Michx.), and other hardwood species. Certain vegetation types, e.g., red pine and oak savannas, were removed from the landscape. The forest density gradient of the presettlement landscape with open savannas and woodlands in the South and denser forests in the North disappeared. These changes, especially the increase in forest cover, are ecologically significant because numerous species are adapted to open habitat, which was previously created by fire, and their populations are declining. [ABSTRACT FROM AUTHOR]