Characterizing height-diameter relationships for Caribbean trees using mixed-effects random forest algorithm

doi:10.1016/j.foreco.2022.120507

Return

Characterizing height-diameter relationships for Caribbean trees using mixed-effects random forest algorithm

2022-11-01

Tamara Heartsill Scalley

Humfredo Marcano‐Vega

DOI：10.1016/j.foreco.2022.120507

Original

Save

View PDF

Abstract

En 中文

Accurately characterizing the relationships between tree height and diameter at breast height (DBH) is vital to quantify productivity and predict tree growth for forest management. This study makes use of data from forest inventories in the Caribbean islands of Puerto Rico, Trinidad and the U.S. Virgin Islands to compare the efficacy of parametric mixed models, random forest and mixed-effects random forest algorithms (MERF) for modeling tree species-specific H-D relationships. MERF as a semi-parametric approach is advantageous for handling high dimensional and complex datasets using the random forest algorithm, as well as to provide species-specific predictions with random effects in the mixed model framework. A total of 37,124 observations used in ana-lyses were collected from 645 permanent sample plots across three Caribbean islands, including 309 species in Puerto Rico, 121 species in the U.S. Virgin Islands and 34 species in Trinidad.Results showed that MERF can be used to reliably predict total tree height for diverse species, which produced more accurate predictions than the other two models. In addition to DBH, competition index was found to be the most important variable among all other stand and environmental variables in all islands. Adding climate-related variables can improve the prediction accuracy of total tree height in parametric mixed models, especially when sample size is small. Precipitation variables were selected most often, confirming the overall importance of precipitation to forest height. The results of this work will not only provide species-specific H-D relationship models for Caribbean trees, but also advance knowledge in modeling H-D relationships with broad application for mixed-species forests.

Keywords:

Puerto Rico

Virgin Islands

Trinidad and Tobago

Tree allometry

Environmental variables

Machine learning

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.