Improving tree survival prediction with forecast combination and disaggregation

Abstract

The tree mortality model plays an important role in simulating stand dynamic processes. Past work has shown that the disaggregation method was successful in improving tree survival prediction. This method was used in this study to forecast tree survival probability of Chinese pine (Pinus tabulaeformis Carrière) in Beijing. Outputs from the tree survival model were adjusted from either the stand-level model prediction or the combined estimator from the forecast combination method. Our results show that the disaggregation approach improved the performance of tree survival models. We also showed that stand-level prediction played a crucial role in refining outputs from a tree survival model, especially when it is a very simple model. Because the forecast combination method produced better stand-level prediction, we prefer the use of this method in conjunction with the disaggregation approach, even though the performance gain in using the forecast combination method shown for this data set was modest.Résumé : La modélisation de la mortalité des arbres joue un rôle important dans la simulation des processus dynamiques de la croissance forestière. Les travaux antérieurs ont montré que la méthode de désagrégation pouvait améliorer la prédic-tion de la survie des arbres. Cette méthode a donc été utilisée ici pour prédire la probabilité de survie du pin de Chine (Pinus tabulaeformis Carrière) à Pékin. Les extrants du modèle de survie des arbres ont été ajustés à partir soit de la prédiction du modèle à l'échelle du peuplement, soit de l'estimateur combiné de la méthode de combinaison des prédictions. Nos ré-sultats montrent que l'approche de désagrégation a amélioré la performance du modèle de survie des arbres. Nous avons également montré que la prédiction à l'échelle du peuplement a joué un rôle crucial dans le raffinement des extrants du modèle de survie des arbres, surtout lorsque le modèle est très simple. Comme la méthode de combinaison des prédictions pré-dit le mieux les attributs du peuplement, nous préférons l'utiliser conjointement avec la méthode de désagrégation, même si le gain de performance était modeste pour l'ensemble de données considéré.[Traduit par la Rédaction]

Development and evaluation of an individual tree growth and yield model for the mixed species forest of the Adirondacks Region of New York, USA

“…Regardless, the findings highlights the sensitivity of models projections to mortality and likely a more sophisticated approach like a linked standand tree-level approach (e.g. Zhang et al 2011) is warranted. However, a linked approach would require the development of a stand-level mortality equation, which can be problematic for multi-cohort, mixed-species stands (Weiskittel et al 2011a).…”

“…Li et al 2011), and assessing alternative approaches to forecasting mortality (e.g. Zhang et al 2011).…”

Modeling tree mortality in relation to climate, initial planting density, and competition in Chinese fir plantations using a Bayesian logistic multilevel method

“…The larger the value of the AUC, the better the model performs (Fielding and Bell 1997). A rough guideline for the AUC follows the traditional academic point system as follows: 0.9-1 = excellent (A), 0.8-0.9 = good (B), 0.7-0.8 = fair (C), 0.6-0.7 = poor (D), and 0.5-0.6 = fail (F) (Zhang et al 2011).…”

“…Tree mortality, one of the main components of forest succession, is important for the maintenance of biological and structural diversity in forest ecosystems (Bigler and Bugmann 2003;Zhang et al 2011). A complex of endogenous and exogenous factors acting as induction agents lead to tree death during forest succession (van Mantgem et al 2009).…”

Predicting total and component biomass of Chinese fir using a forecast combination method

“…Again, SUR was used to estimate parameters of this system of equations. Zhang et al (2010Zhang et al ( , 2011a applied the forecast combination method to combine different types of models for predicting stand basal area and stand survival. A similar approach was applied in this study to predict total tree biomass by combining: (a) direct prediction from the regression model (eqn.…”