A review of evidence about use and performance of species distribution modelling ensembles like BIOMOD

Abstract

Aim:The idea of combining predictions from different models into an ensemble has gained considerable popularity in species distribution modelling, partly due to free and comprehensive software such as the R package BIOMOD. However, despite proliferation of ensemble models, we lack oversight of how and where they are used for modelling distributions, and how well they perform. Here, we present such an overview. Location: Global.Methods: Since BIOMOD is freely available and widely used by ensemble species distribution modellers, we focused on articles that apply BIOMOD, filtering the initial 852 papers identified in our structured literature search to a relevant final subset of 224 eligible peer-reviewed journal articles. Results: BIOMOD-based ensembles are used across many taxa and locations, with terrestrial plants being the most represented group of species (n = 72) and Europe being the most represented continent (n = 106). These studies often focus on forecasting distributions in the future (n = 109), and commonly use presence-only species data (n = 139) and climatic environmental predictors (n = 219). An average of six models are used in ensembles, and approximately half of ensembles weight contributions of models by their cross-validation performance. However, discussion about choices made in the modelling process and unambiguous information on the performance of ensemble models versus individual models are limited. The use of independent data to validate model performance is particularly uncommon. Main conclusions:We document the breadth of ensemble applications, but could not draw strong quantitative conclusions about the predictive performance of ensemble models, due to lack of unambiguous information reported. Understanding how and where ensembles are best used when modelling species distributions is important for enabling best choices for different applications. To enable this objective to be achieved, we provide recommendations for thorough reporting practices in a BIOMOD-based ensemble workflow. K E Y W O R D SBIOMOD, consensus forecast, ecological niche models, ensemble, habitat suitability models, species distribution model

Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models

“…This is consistent with previous investigations into ensemble performance showing that they perform well at interpolating tasks (Marmion et al ), but are not always the best at transfer tasks (Crimmins et al , Zhu and Peterson ). This is worth further investigation, because ensembles are popular with such transferring tasks (Hao et al ). In our latitudinal tests, the observed decrease in ensemble performance relative to others may be explained by a combination of: 1) spatial dependencies between train and test sets are decreased in latitudinal blocking, and 2) complex models can overfit spatial patterning, erroneously attributing geographic patterns of SAC to environmental covariates (Merow et al , Roberts et al ).…”

“…stepwise selection, or use of a full model). In practice, the evidence points towards common use of default tunings with biomod2 (Hao et al ). To be consistent with that, here we also used ‘biomod2’ in‐built functions to build individual models as well as ‘biomod2’ default tuning choices (see our archived data for our R code, including tuning parameters used).…”

“…Ensemble SDM workflows can be carried out using other toolkits such as BioEnsembles (Diniz-Filho et al 2009) or without specialised tool sets (Hardy et al 2011, Crimmins et al 2013). However, we choose 'biomod2' because it is most popular among ensemble SDM users and is therefore representative of the typical ensemble SDM workflow (Hao et al 2019a).…”

“…Ensemble modelling is widely applied by SDM users (Araújo and New , Hao et al ) and many modellers believe ensemble models are superior for prediction tasks compared to individual models (Hao et al ). However, there is limited empirical investigation about how well ensemble models predict compared to individual SDMs.…”

Model complexity affects species distribution projections under climate change

“…Varying parameterization complexity has not been employed routinely in the recent literature. Instead, algorithms were mostly run with default parameterizations or else with simplifications of the default flexibility (see also Hao, Elith, Guillera-Arroita, & Lahoz-Monfort, 2019). Instead, algorithms were mostly run with default parameterizations or else with simplifications of the default flexibility (see also Hao, Elith, Guillera-Arroita, & Lahoz-Monfort, 2019).…”