An evaluation of transferability of ecological niche models

Abstract

Ecological niche modeling (ENM) is used widely to study species’ geographic distributions. ENM applications frequently involve transferring models calibrated with environmental data from one region to other regions or times that may include novel environmental conditions. When novel conditions are present, transferability implies extrapolation, whereas, in absence of such conditions, transferability is an interpolation step only. We evaluated transferability of models produced using 11 ENM algorithms from the perspective of interpolation and extrapolation in a virtual species framework. We defined fundamental niches and potential distributions of 16 virtual species distributed across Eurasia. To simulate real situations of incomplete understanding of species’ distribution or existing fundamental niche (environmental conditions suitable for the species contained in the study area; N* F ), we divided Eurasia into six regions and used 1–5 regions for model calibration and the rest for model evaluation. The models produced with the 11 ENM algorithms were evaluated in environmental space, to complement the traditional geographic evaluation of models. None of the algorithms accurately estimated the existing fundamental niche (N* F ) given one region in calibration, and model evaluation scores decreased as the novelty of the environments in the evaluation regions increased. Thus, we recommend quantifying environmental similarity between calibration and transfer regions prior to model transfer, providing an avenue for assessing uncertainty of model transferability. Different algorithms had different sensitivity to completeness of knowledge of N* F , with implications for algorithm selection. If the goal is to reconstruct fundamental niches, users should choose algorithms with limited extrapolation when N* F is well known, or choose algorithms with increased extrapolation when N* F is poorly known. Our assessment can inform applications of ecological niche modeling transference to anticipate species invasions into novel areas, disease emergence in new regions, and forecasts of species distributions under future climate conditions.

Collinearity in ecological niche modeling: Confusions and challenges

“…Our study clarifies whether, when, and how collinearity affects model performance in Maxent. First, we show decreased model performance in model transfer scenarios, a well‐known phenomenon observed in many studies (Fitzpatrick et al, ; Owens et al, ; Qiao et al, ). The potential underlying mechanisms are likely the degree of predictor collinearity, collinearity shift, and environmental novelty.…”

“…Previous studies have shown that model extrapolation in novel environmental conditions can lead to decreased performance (Fitzpatrick et al, ; Owens et al, ; Qiao et al, ). Therefore, we quantified environmental novelty, in essence environmental distance, between testing and training data.…”

Modeling the distribution of a wide‐ranging invasive species using the sampling efforts of expert and citizen scientists

“…The use of citizen data in SDMs is often criticized due to uncertainties associated with underlying sampling processes (Mair & Ruete, (Qiao et al, 2018). As the novelty of the environmental conditions in the region being evaluated increases, the model transferability performance scores decrease (Sequeira et al, 2016) while the probability of obtaining erroneous predictions increases ).…”

Physiology in ecological niche modeling: using zebra mussel's upper thermal tolerance to refine model predictions through Bayesian analysis

“…Another reason for the relative narrowness of native occurrence data is adaptation of introduced populations to previously limiting conditions in the invaded range (Prentis et al 2008, Moran andAlexander 2014). The poor performance of the native model in the global context is analogous to decreased accuracy of a model calibrated with present climate data and projected to future climate data (Moreno-Amat et al 2015, Fitzpatrick et al 2018, since model extrapolation is involved in both scenarios (Qiao et al 2019). However, the reason for the narrower range of native data does not affect the improvement in physiologically informed models, rather our conclusion is supported by incompleteness of native data and effectiveness of physiological information.…”

“…To estimate potential distributions, transferring an ecological niche model through space and time is becoming a norm in the discipline (Peterson et al 2007, Murray et al 2009, Wenger and Olden 2012, Moreno-Amat et al 2015, Duque-Lazo et al 2016), but the underlying uncertainties of model transfers warrant our attention (Murray et al 2011, Sequeira et al 2016. Without adequate data, model prediction in novel environmental conditions (beyond range of conditions used in model calibration) will experience extrapolation, which is problematic (Peterson et al 2011, Owens et al 2013, Qiao et al 2019. For example, occurrences from the species' native range may be inadequate in explaining the relationship between the species and abiotic conditions in the introduced range; same concern applies to the adequacy of use of species' present occurrences to predict species' future ranges when novel or no-analog climate conditions will emerge in the future (e.g.…”