Predicting redistribution of tree species and communities under recent climatic change: Benchmarking temporal predictions from correlative and mechanistic models

Mr David Uribe1, A.Prof. Patricio Pliscoff2, Prof. Solomon Dobrowski3, A.Prof. Gurutzeta Guillera-Arroita1, Prof. Brendan Wintle1

1The University Of Melbourne, School of Biosciences, Parkville, Australia, 2Pontificia Universidad Católica de Chile, Departamento de Ecología & Instituto de Geografía, Santiago, Chile, 3University of Montana, Department of Forest Management, Missoula, United States of America


Reliable predictions of species range redistribution are required to ensure the success of conservation planning under future global change scenarios. Recently, big efforts have been made to improve the theory and statistical approaches of ecological models for predicting range dynamics. However, most of the exercises rely on forecasting distribution changes into the future, while benchmarking predictive skills between different modeling approaches using observed redistributions remains elusive. This study makes a call to better understand the predictive performance of correlative and mechanistic modeling approaches through retrospective forecasting exercises (i.e. predicting observed changes using historical datasets to train the models into the past and testing their predictive performance contrasting temporal predictions against current distributions). We used four different modeling approaches: correlative Species Distribution Models (SDM), dynamic hybrid SDM, community-level joint SDM and community-level Occupancy-Detection Models to predict observed range shifts of 28 Californian tree species during a 70 years period. We assessed the accuracy and discrimination power at the species and community level by contrasting predictions and observations using a temporal-independent dataset and analyzed using generalized linear regressions. We discuss the results in order to emphasize the improvements in predictive performance as colonization processes, detection probability, species interactions or community-level modeling are explicitly incorporated into the range-dynamics models. Finally, we present recommendations to optimize the reliability of temporal predictions of species redistribution and community reassembly.


Beca Doctorado Becas Chile N° 72170498-2016; Fondecyt N° 1181677.


I’m a Chilean ecologist currently based at the Quantitative and Applied Ecology Group, The University of Melbourne, Australia. My PhD (2017-2021) is being supervised by Prof. Brendan Wintle and A/Prof. Gurutzeta Guillera-Arroita. My research interests are as wide as global change biology, but at the moment focused on conservation biogeography. I attempt to design it to support conservation practice. So, it is mostly applied and usually incorporates explicit recommendations to improve environmental management. My current research is focused on the effects of anthropogenic climate change, land use change, biological invasions and emerging diseases on biodiversity redistribution and persistence.

Similar Posts