Development of a forest structural complexity index based on multispectral airborne remote sensing and topographic data

This paper presents development of a multivariate forest structural complexity index based on relationships between field-based structural variables and geospatial data. Remote sensing has been widely used to model individual forest structural attributes at many scales. As opposed to, or in addition to, individual structural parameters such as leaf area index or tree height, overall structural complexity information can enhance forest inventories and provide a variety of information to forest managers, including identifying damage and disturbance as well as indicators of habitat or biodiversity. In this study, a multivariate modelling technique, redundancy analysis, was implemented to derive a model incorporating both horizontal and vertical structural attributes as predicted by an ensemble of high-resolution multispectral airborne imagery and topographic variables. The first redundancy analysis axis of the final model explained 35% of the total variance of the field variables and was used as the complexity index. With a root mean squared error of 19.9%, the model was capable of differentiating four to five relative levels of complexity. This paper presents the forest ecological and modelling aspects of the research. A related paper presents the remote sensing aspects, including application of the model to map predicted structural complexity, map validation, and testing of the method at multiple scales. Resume: Cet article presente le developpement d'un indice multivarie de complexite de la structure de la foret base sur les relations entre des variables structurales mesurees sur le terrain et des donnees geospatiales. La teledetection a ete largement utilisee pour modeliser les attributs structuraux de la foret a plusieurs echelles. Contrairement a, ou en plus, des parametres structuraux individuels comme l'indice de surface foliaire ou la hauteur des arbres, une information globale au sujet de la complexite structurale peut ameliorer les inventaires forestiers et fournir une variete de renseignements aux amenagistes forestiers, incluant l'identification des dommages et des perturbations aussi bien que des indicateurs d'habitat ou de biodiversite. Dans cette etude, une technique de modelisation multivariee, l'analyse de redondance, a ete appliquee pour deriver un modele qui incorpore des attributs, tant horizontaux que verticaux de la structure, predits par un ensemble d'images aeriennes multispectrales a haute resolution et des variables topographiques. Le premier axe de l'analyse de redondance du modele final expliquait 35% de la variance totale des variables terrain et a ete utilise comme indice de complexite . Avec une erreur quadratique moyenne de 19,9 %, le modele etait capable de differencier quatre a cinq niveaux de complexite. Cet article presente les aspects de la recherche relies a la modelisation et a l'ecologie forestiere. Un autre article presente les aspects relies a la teledetection, incluant l'application du modele pour cartographier la complexite structurale predite, la validation de la cartographie et l'essai de la methode a plusieurs echelles. [Traduit par la Redaction]
Introduction Understanding patch-level structural variations across a forest can enhance inventories and provide important information to forest managers. Specifically, information with respect to damage and disturbance (e.g., Cosmopoulos and King [...]