Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/774604

embarcadero: Species distribution modelling with Bayesian additive regression trees in R

Carlson, C. J.
View original source ↗Export metadata ↓

Abstract

O_LIembarcadero is an R package of convenience tools for species distribution modelling with Bayesian additive regression trees (BART), a powerful machine learning approach that has been rarely applied to ecological problems. C_LIO_LILike other classification and regression tree methods, BART estimates the probability of a binary outcome based on a set of decision trees. Unlike other methods, BART iteratively generates sets of trees based on a set of priors about tree structure and nodes, and builds a posterior distribution of estimated classification probabilities. So far, BARTs have yet to be applied to species distribution modelling. C_LIO_LIembarcadero is a workflow wrapper for BART species distribution models, and includes functionality for easy spatial prediction, an automated variable selection procedure, several types of partial dependence visualization, and other tools for ecological application. The embarcadero package is available open source on Github and intended for eventual CRAN release. C_LIO_LITo show how embarcadero can be used by ecologists, I illustrate a BART workflow for a virtual species distribution model. The supplement includes a more advanced vignette showing how BART can be used for mapping disease transmission risk, using the example of Crimean-Congo haemorrhagic fever in Africa. C_LI

Source connections

Explore related subjects

ecologyPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Carlson, C. J.. 2019-09-19. embarcadero: Species distribution modelling with Bayesian additive regression trees in R. https://doi.org/10.1101/774604

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Validating static food web model diet matrices using stable isotope simulation

End-to-end ecosystem models, such as Ecopath with Ecosim (EwE), are valuable tools for resource management. Within these frameworks, trophic interactions, typically derived from stomach contents analyses, govern system-wide energy flow via consumption. However, stomach contents data may be unrepresentative of long-term average diets and assimilated consumption due to factors including differential digestion rates and identifiability of prey, introducing model uncertainty. Because trophic interactions define model structure, variation in modeled diets can alter predictions and management conclusions. Consequently, evaluating and reducing uncertainty in diet matrix compositions is critical for operational acceptance of ecosystem models. Stable isotope analysis provides complementary, time-integrated insight into average diets, but its application for EwE validation has largely been limited to evaluating trophic levels and isotopic niche space. Here, we present a quantitative framework for simulating consumer stable isotope signatures across food webs from parent ecosystem models using a matrix-based mixing model. Using a published Northern Gulf of Mexico EwE model as a case study, we demonstrate Monte Carlo simulation to propagate uncertainty from basal source signatures and trophic discrimination factors. Comparison of predicted signatures with empirical {delta}13C and {delta}34S measurements enables validation of primary carbon sources and benthic-pelagic pathway contributions, revealing agreement in isotope space alongside mismatch in energy pathways. Diagnostic residuals identify discrepancies in diet composition. Model scoring provides a reproducible framework for ranking alternative diet configurations. Redundancy analysis further identifies within-group species coverage differences between stomach contents and isotope datasets as a primary driver of mismatch, highlighting an important consideration for sampling design.

ecology↗
bioRxiv2026

Enhancing Germination and Assessing the Longevity of Native Plants' Seeds with Dormancy through Different Priming Techniques

Vachellia pachyceras and Convolvulus oxyphyllus are native species of arid and semi-arid desert regions, including Kuwait. They are inherently adapted to extreme conditions characterized by high temperature, prolonged drought and limited rainfall. They contribute significantly, to restore the ecosystem by providing food and shelter for wildlife. Despite their ecological importance, seed germination in both species is constrained by dormancy, and natural recruitment of seedlings is limited by narrow and unpredictable rainfall windows, severely restricting opportunities for establishment. In this context, seed priming is an effective pre-sowing technique that enables uniform and rapid germination within the short window of favorable conditions. This study aims to identify the optimal seed-priming techniques for these two keystone native species and to assess the effect of priming on seed longevity. The seeds were scarified to remove the physical dormancy prior to priming treatments. Relative seed longevity was measured for primed seeds through artificial ageing and analyzed using probit analysis and analysis of variance. Results showed that hormonal priming [(mechanical scarification + 100 ppm gibberellic acid (GA3)] for six hours and hydropriming (mechanical scarification + soaking in water) for 8 hours (h) were effective treatments among the tested priming treatments for V. pachyceras. The estimated p50 value (i.e., number of days to lose seed viability to 50%) was 22 days for hormonal priming and 21 days for hydropriming. Mechanical scarification followed by one and half hour of hydropriming was the most effective in C. oxyphyllus (p50 = 74 days). Based on these results, the longevity of V. pachyceras and C. oxyphyllus could be considered medium. Further studies are needed to determine the effectiveness of seed priming on additional ecologically and economically important native species to achieve success in revegetation programs.

ecology↗
bioRxiv2026

Restoration decisions in highly variable systems: Whether, when, and how to enhance kelp

In highly variable ecosystems, natural variability can obscure signals of decline that warrant intervention, as there is often a lack of long-term data needed to distinguish between declines that are transient and likely to recover naturally and those that would be permanent without intervention. Implementing restoration for a species involves deciding on a threshold condition at which to intervene, duration that this threshold needs to be met before intervening, life stages at introduction, and amounts to introduce. Current bull and giant kelp restoration efforts in California were spurred by persistent kelp loss that started in 2013 following a disease outbreak among predators of the primary herbivore and a marine heatwave. We asked how decisions about intervention cue, threshold, wait time, life stage at introduction, and amount affect population outcomes against a backdrop of environmental variability. We analyzed additional context-dependencies on grazing and climate intensity. To address these questions, we built and parameterized a stochastic stage-structured model for bull and giant kelp, in which we simulated kelp enhancement approaches under a range of grazing and climate scenarios. In a global sensitivity analysis of kelp enhancement decisions, we found that grazing intensity and life stages at introduction were the most influential determinants of kelp biomass and persistence outcomes. The kelp enhancement approach involving temporarily introduced juveniles restored the germ banks of bull and giant kelp, but the two species differed in their responses to that intervention, highlighting that optimal intervention approach is species-dependent. Waiting longer for natural recovery before intervening reduced total intervention effort but also reduced biomass recovery potential. These results support empirical findings from kelp enhancement pilot projects in California and highlight context-dependencies that inform restoration decision-making in kelp forests and other highly variable ecosystems.

ecology↗