Natural mortality estimation using tree-based ensemble learning models

Empirical studies are popular in estimating fish natural mortality rate (⁠M⁠). However, these empirical methods derive M from other life-history parameters and are often perceived as being less reliable than direct methods. To improve the predictive performance and reliability of empirical methods, we develop ensemble learning models, including bagging trees, random forests, and boosting trees, to predict M based on a dataset of 256 records of both Chondrichthyes and Osteichthyes. Three common life-history parameters are used as predictors: the maximum age and two growth parameters (growth coefficient and asymptotic length). In addition, taxonomic variable class is included to distinguish Chondrichthyes and Osteichthyes. Results indicate that tree-based ensemble learning models significantly improve the accuracy of M estimate, compared to the traditional statistical regression models and the basic regression tree model. Among ensemble learning models, boosting trees and random forests perform best on the training dataset, but the former performs a slightly better on the test dataset. We develop four boosting trees models for estimating M based on varying life-history parameters, and an R package is provided for interested readers to estimate M of their new species.