Ensemble learning in ecological niche modeling of the Siderastrea (Cnidaria:Scleractinia) Atlantic complex - an information theory approach

Sep 17, 2026

Ecological niche modeling (ENM) is a key method in biodiversity science, yet real-world studies still lack standards when combining multiple modeling strategies into one prediction set. This paper presents an information-theory ensemble approach using the Siderastrea Atlantic Complex (SAC) as a working example, testing nine ENM models across two species for cross-area transferability. Shannon’s entropy is used to assess overfitting, enabling a cell-by-cell weighting strategy and producing reliability maps that integrate occurrence probability with ensemble efficiency. Results identify suitable habitats for endemic and Caribbean species and highlight additional suitability along the Western African coast, consistent with reported observations.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages62

Reading Time2 h 36 min

AccessFree to download

Introduction

Ensemble learning and ecological niche modeling overview

Model classes and supervised learning framing

Ensemble forecasting and probabilistic interpretation

Motivation and framework for the proposed approach

DOI: xxx/xxxx O R I G I NA L A RT I C L E 1 Ensemble learning in ecological niche modeling of the 2 Siderastrea (Cnidaria:Scleractinia) Atlantic complex: 3 an information theory approach 4 Henrique J. Schipanski MSc1 Lucas B. Perroni2 Caio F. Sguario3 5 Marcos S. Barbeitos PhD1 6 1Departamento de Zoologia, Universidade Federal do Paraná, Curitiba, PR, Brazil 2Departamento de Física, Universidade Federal do Paraná, Curitiba, PR, Brazil 3Departamento de Informática, Universidade Federal do Paraná, Curitiba, PR, Brazil Correspondence Marcos S. Barbeitos PhD, Departamento de Zoologia, Universidade Federal do Paraná, Curitiba, PR, 81531980, Brazil. [email:msbarbeitos@gmail.com](email:msbarbeitos@gmail.com) Funding Information This research was supported by the Conselho Nacional de Pesquisa (CNPq), PhD Scholarship Number: 140363/2023-8; Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), MSc Scholarship Number: 1810146 Abstract Ecological niche modeling (ENM) is an invaluable tool in biodiversity science and many algorithms were developed for or incorporated into this area over the years. Back in 2007, Araújo and New 2007 outlined a framework to integrate different sources of variation into robust model projections. Almost 20 years later, surveys of real world applications show that standards are still lacking, particularly when it comes to combining different modeling strategies into a single set of predictions. In this paper we introduce a novel information theory approach to ensemble learning using the Siderastrea Atlantic Complex (SAC) as a working example. We employed nine models (BioClim, Gower, and Mahalanobis distances, GAM, GLM, MARS, ANN, MaxEnt, and SVM) to test for cross-area transferability of two species in SAC. We used Shannon’s entropy to evaluate model overfitting to the data. Our method reveals that overfitting is not spatially uniform, thus, we developed a weighting strategy that reduces the influence of overfitted models on ensemble predictions on a cell-by-cell basis. Overfitting is obscured by traditional measures, such as AUCand TSS, that provided inflated estimation of real model performances across projection areas. We also introduce reliability maps, analogous to the Maps of Biogeographical Ingnorance (7MoBIS), combining occurrence probability and ensemble eﬀiciency on each cell. Our approach uncovered the Yucatán peninsula as an adequate habitat for the endemic Brazilian species Siderastrea stellata and, conversely, the Northeastern Brazilian coast as a suitable area for the Caribbean species S. siderea, supporting previously published records. It also uncovered a stretch of Western African coast as highly suitable habitat for the genus, a result that was recently corroborated by citizenscience reports. K E Y W O R D S areas of accessibility, confusion matrix, coral, cross-validation, extrapolation, interpolation, machine learning,”No Free Lunch” Theorem, reef, sensitivity, Shannon’s entropy, species distribution modeling, specificity, testing, training, transferability 1 INTRODUCTION 8 Ecological niche modeling (ENM) has broad applications in agriculture, public health, biodiversity, and conservation (Soley- 9 Guardia et al 2024) . The goal of ENM is to construct a function that maps the input (environmental) data to its output pair, 10 Journal 2023;00:1–21 [wileyonlinelibrary.com/journal/](wileyonlinelibrary.com/journal/) © 2023 Copyright Holder Name 1 11 usually a continuous target variable expressing the adequacy of any given locality to the existence of a species (Peterson et al 12 2011). ENM algorithms are grouped in three main categories (Thuiller 2024): (i) environmental ”envelopes”, such as BioClim 13 (Booth et al 2014) and Mahalanobis distances (Farber and Kadmon 2003); (ii) statistical models, including generalized linear 14 models (GLM- Guisan et al 2002) and multivariate adaptive regression splines (MARS-Friedman 1991); and (iii

FAQs

What problem does the paper address in ecological niche modeling ensembles?

The paper targets the lack of standards in real-world applications when combining different ENM strategies into a single set of predictions, especially regarding how overfitting affects ensemble outputs.

How does the proposed method evaluate and mitigate model overfitting?

It uses Shannon’s entropy to measure overfitting and applies a cell-by-cell weighting strategy so that overfitted models contribute less to ensemble predictions on each grid cell.

What additional diagnostic outputs does the paper introduce beyond traditional performance metrics?

It introduces reliability maps that combine occurrence probability and ensemble efficiency per cell, showing cases where traditional measures like AUC and TSS can inflate perceived performance across projection areas.

Sep 17, 2026

Uploaded by Bill Black