Modeling annualized occurrence, frequency, and composition of ingrowth using mixed-effects zero-inflated models and permanent plots in the Acadian Forest Region of North America

Forest tree ingrowth is a highly variable and largely stochastic process. Consequently, predicting occurrence, frequency, and composition of ingrowth is a challenging task but of great importance in long-term forest growth and yield model projections. However, ingrowth data often require different statistical techniques other than traditional Gaussian regression, because these data are often bounded, skewed, and non-normal and commonly contain a large fraction of zeros. This study presents a set of regression models based on discrete Poisson and negative binomial probability distributions for ingrowth data collected from permanent sample plots in the Acadian Forest Region of North America. Models considered here include regular Poisson, zero-inflated Poisson (ZIP), zero-altered Poisson (ZAP; hurdle Poisson), regular negative binomial (NB), zero-inflated negative binomial (ZINB), and zero-altered negative binomial (ZANB; hurdle NB). Plot-level random effects were incorporated into each of these models. The ZINB model with random effects was found to provide the best fit statistics for modeling annualized occurrence and frequency of ingrowth. The key explanatory variables were stand basal area per hectare, percentage of hardwood basal area, number of trees per hectare, a measure of site quality, and the minimum measured diameter at breast height of each plot. A similar model was developed to predict species composition. All models showed logical behavior despite the high variability observed in the original data. Le recrutement d'arbres en foret est un processus tres variable et fortement stochastique. Par consequent, la prevision de l'occurrence, de la frequence et de la composition du recrutement d'arbres est une tache difficile, mais d'une grande importance pour les projections a long terme des modeles de croissance et de production forestiere. Cependant, les donnees de recrutement necessitent souvent des techniques statistiques autres que la traditionnelle regression gaussienne parce que ces donnees sont generalement delimitees, asymetriques et non normales en plus de contenir couramment une grande proportion de valeurs nulles. Cette etude presente une serie de modeles de regression bases sur des distributions de probabilite discretes de Poisson et binomiales negatives etalonnes a partir de donnees de recrutement provenant de placettes echantillons permanentes etablies dans la region forestiere acadienne de l'Amerique du Nord. Les modeles consideres dans cette etude incluent la distribution de Poisson reguliere, de Poisson a exces de zeros (PEZ), de Poisson tronquee a zero (PTZ; Poisson a obstacle), binomiale negative reguliere (BN), binomiale negative a exces de zeros (BNEZ) et binomiale negative tronquee a zero (BNTZ; BN a obstacle). Les effets aleatoires a l'echelle de la placette ont ete introduits dans chacun de ces modeles. Le modele BNEZ avec effets aleatoires a produit les meilleurs ajustements statistiques pour modeliser l'occurrence et la frequence annualisees du recrutement. Les variables explicatives les plus importantes etaient la surface terriere du peuplement a l'hectare, la proportion de la surface terriere en feuillus, le nombre d'arbres a l'hectare, une mesure de la qualite de la station et le DHP minimal mesure dans chaque placette. Un modele similaire a ete mis au point pour predire la composition en especes. Tous les modeles ont produit un comportement logique malgre la grande variabilite observee dans les donnees originales. [Traduit par la Redaction]
Introduction Ingrowth is defined as trees in a sample plot that have grown into a required threshold size (usually measured by tree height or diameter at breast height (dbh)) over [...]