Search

The long history of gradient analysis is anchored in the observation that species turnover can be described along elevation gradients. This model is unsatisfying in that elevation is not directly relevant to plants and the ubiqui- tous "elevation gradient" is composed of multiple intertwined environmental factors. We offer an approach to landscape- scale vegetation analysis that disentangles the elevation gradient into its constituent parts through focused field sam- pling and statistical analysis. We illustrate the approach for an old-growth watershed in the Oregon Western Cascades. Our initial model of this system supports the common observation that forest community types are highly associated with specific elevation bands. By replacing elevation and other crude environmental proxy variables with estimates of more direct and resource gradients (radiation, temperature, and soil moisture), we create a vegetative model with stron- ger explanatory power than the proxy model in both cross-validation analysis and validation using an independent data set. The resulting model is also more biologically interpretable, which provides more meaningful insight into potential forest response to environmental change (e.g., global climate change scenarios). Acquiring a better mechanistic under- standing of the relationship between plant communities and environmental predictor variables presents the next great challenge to community ecologists conducting gradient studies at landscape scales. Resume : L'analyse de gradient s'appuie depuis toujours sur le fait que le renouvellement des especes peut etre decrit en suivant des gradients d'altitude. Ce modele n'est pas satisfaisant puisque l'altitude n'est pas directement pertinente pour les plantes et que le terme omnipresent « gradient d'altitude » est compose de multiples facteurs environnementaux entremeles. Les auteurs proposent une approche d'analyse de la vegetation a l'echelle du paysage qui decompose le gradient d'altitude en ses parties constituantes en ciblant l'echantillonnage de terrain et l'analyse statistique. Ils illustrent cette approche a partir d'un bassin versant compose de vieux peuplements de l'ouest des Cascades, en Oregon. Leur modele initial de ce systeme est en accord avec l'observation courante que les types forestiers sont etroitement associes aux bandes specifiques d'altitude. En remplacant l'altitude et d'autres variables environnementales indirectes par des es- timations de gradients de ressources plus directs (radiation, temperature et humidite du sol), ils creent un modele de vegetation ayant un pouvoir explicatif plus grand que celui du modele indirect, ce qui a ete verifie dans les cas d'une analyse de validation interne et d'une validation a l'aide d'un fichier de donnees independantes. Ce modele facilite aussi l'interpretation biologique des resultats, ce qui permet d'approfondir nos connaissances sur la reaction potentielle de la foret en fonction de changements environnementaux (p. ex., des scenarios de changement global du climat). L'acquisition d'une meilleure comprehension fonctionnelle des relations entre les communautes vegetales et les variables environne- mentales de prediction constitue le prochain grand defi pour les ecologistes des populations qui s'interessent aux etudes de gradient a l'echelle du paysage. (Traduit par la Redaction) Lookingbill and Urban 1753

Fire is a major agent of spatial pattern formation in forests, as it creates a mosaic of burned and unburned patches. While most research has focused on landscape-level patterns created by crown fires, millions of hectares of forests in North America are subject to surface fire regimes. A spatially explicit forest gap model developed for the Sierra Nevada was used to evaluate the influence of surface fire regimes on the heterogeneity of forest structure and composition within forest stands. Forest pattern was evaluated for a wide range of topographic positions in Sequoia National Park, California, to determine if repeated surface fires amplify existing spatial patterns. The spatial heterogeneity of some forest characteristics increased under a simulated fire regime relative to scenarios without fire. Although a distinct and regular fire-generated spatial pattern was not detected with an analysis of spatial autocorrelation, simulated surface fires did alter the spatial heterogeneity within a forest stand, primarily by degrading a regular structure that is imposed by competition for light in the absence of fire. The interaction between surface fires and forest pattern may be qualitatively different from that which occurs in forests subject to crown fires. As such, what has been learned about forests dominated by crown fires may not apply to forests subject to surface fire regimes.