Fitting forestry models using generalized additive models: a taper model example
Dates
Year
2011
Citation
Robinson, Andrew P., Lane, Stephen E., and Thérien, Guillaume, 2011, Fitting forestry models using generalized additive models: a taper model example: Canadian Journal of Forest Research, v. 41, no. 10, p. 1909-1916.
Summary
Nonparametric and semiparametric modelling methods are commonly applied in many fields. However, such methods have not been widely adopted in forestry, other than the most similar neighbour and nearest neighbor methods. Generalized additive modelling is a flexible semiparametric regression method that is useful when model-based prediction is the main goal and the parametric form of the model is unknown and possibly complex. Routines to fit generalized additive models (GAMs) are now readily available in much statistical software, making them an attractive option for forest modelling. Here, the use of GAMs is demonstrated by the construction of a taper model for six tree species in British Columbia, Canada. We compare the results with [...]
Summary
Nonparametric and semiparametric modelling methods are commonly applied in many fields. However, such methods have not been widely adopted in forestry, other than the most similar neighbour and nearest neighbor methods. Generalized additive modelling is a flexible semiparametric regression method that is useful when model-based prediction is the main goal and the parametric form of the model is unknown and possibly complex. Routines to fit generalized additive models (GAMs) are now readily available in much statistical software, making them an attractive option for forest modelling. Here, the use of GAMs is demonstrated by the construction of a taper model for six tree species in British Columbia, Canada. We compare the results with an existing flexible parametric taper model. We assess the performance of the models using the 0.632+ bootstrap method according to five key attributes: whole-stem volume, merchantable volume, number of logs, small-end diameter of the first log, and volume of the first log. The results show that the GAMs and the flexible taper function yielded similar accuracy for all attributes and all species. (English) ABSTRACT FROM AUTHOR]; Les méthodes de modélisation non paramétriques et semi-paramétriques sont couramment appliquées dans de nombreux domaines. Cependant, à part les méthodes des voisins les plus similaires et des voisins les plus proches, ces méthodes n'ont pas été largement adoptées en foresterie. La modélisation additive généralisée est une méthode de régression semi-paramétrique souple et utile lorsque la prédiction constitue l'objectif principal et que la forme paramétrique du modèle est inconnue et possiblement complexe. Les modèles additifs généralisés constituent une option intéressante pour la modèlisation forestière maintenant que les routines d'ajustement de ces modèles sont facilement disponibles dans beaucoup de logiciels statistiques. L'utilisation de modèles additifs généralisés est illustrée par la construction d'un modèle de défilement pour six essences forestières en Colombie-Britannique, au Canada. Les résultats sont comparés à ceux d'un modèle paramétrique flexible et existant de défilement. La performance des modèles est évaluée par la méthode d'auto-amorçage 0,632 + en fonction de cinq attributs clés : le volume total de la tige, le volume marchand, le nombre de billes, le diamètre au fin bout de la première bille et le volume de la première bille. Les résultats de la comparaison montrent que les modèles additifs généralisés et la fonction paramétrique flexible de défilement donnent une précision similaire pour tous les attributs et toutes les essences. (French) ABSTRACT FROM AUTHOR]; Copyright of Canadian Journal of Forest Research is the property of Canadian Science Publishing and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)