Here be dragons: important spatial uncertainty driven by climate data in forecasted distribution of an endangered insular reptile

Abstract

The effect of future climate change is poorly documented in the tropics, especially in mountainous areas. Yet, species living in these environments are predicted to be strongly affected. Newly available high-resolution environmental data and statistical methods enable the development of forecasting models. Nevertheless, the uncertainty related to climate models can be strong, which can lead to ineffective conservation actions. Predicted studies aimed at providing conservation guidelines often account for a range of future climate predictions (climate scenarios and global circulation models). However, very few studies considered potential differences related to baseline climate data and/or did not account for spatial information (overlap) in uncertainty assessments. We modelled the environmental suitability for Phelsuma borbonica, an endangered reptile native to Reunion Island. Using two metrics of species range change (difference in overall suitability and spatial overlap), we quantified the uncertainty related to the modelling technique (n = 10), sample bias correction, climate change scenario, global circulation models (GCM) and baseline climate (CHELSA versus Worldclim). Uncertainty was mainly driven by GCMs when considering overall suitability, while for spatial overlap the uncertainty related to baseline climate became more important than that of GCMs. The uncertainty driven by sample bias correction and variable selection was much higher when assessed based on spatial overlap. The modelling technique was a strong driver of uncertainty in both cases. We eventually provide a consensus ensemble prediction map of the environmental suitability of P. borbonica to identify the areas predicted to be the most suitable in the future with the highest certainty. Predictive studies aimed at identifying priority areas for conservation in the face of climate change need to account for a wide panel of modelling techniques, GCMs and baseline climate data. We recommend the use of multiple approaches, including spatial overlap, when assessing uncertainty in species distribution models.

Choice of climate data influences current and future global invasion risks for twoPhelsumageckos

“…We identified local differences between predicted invasion risks using different climate data sources. Differences may be driven by the selection of different variables (e.g., models calibrated with CHELSA data selected ‘Daily temperature range’ but not with Worldclim for P. grandis ); however, a recent study showed that differences can persist even when the same predictors are selected (Dubos et al 2022a; see also Jiménez-Valverde et al 2021). The mismatch may be better explained by the methods used to compute the climatologies.…”

“…Since we had no a priori on which climate data source is best for ecological modelling, we based the ranking on the mean value between predictions obtained between CHELSA and Worldclim. To account for uncertainty, we penalised the mean prediction by subtracting its standard deviation (mean – SD), following the approach developed by Kujala et al (2013) applied to single species (Dubos et al 2022a). This enabled us to prioritise areas where the invasion risk is most-consistently identified as high across climate data sources and model replicates.…”

“…We projected the predicted values of our models on future climate data. For each species, climate data, GCM, and scenario individually, we quantified the difference between current and future predicted distributions with the two aforementioned complementary approaches, i.e., difference in total (unpenalised) suitability scores and spatial overlap (Dubos et al 2022a). We computed SRC as the difference between the summed suitability scores of future projections and current distribution and show the proportion of increase/decrease relative to current suitability scores, following Buisson et al (2010) and Baker et al (2016).…”

“…Secondly, we used an approach that takes into account spatial information, i.e., spatial overlap (Muscatello et al 2021; Petford and Alexander 2021; Dubos et al, 2022a). We computed the Schoener’s D overlap between projections of current invasion risk between predictions based on the two climate datasets considered.…”

“…These factors can strongly affect SDMs and induce uncertainty in assessments of climate suitability (Buisson et al 2010). More recently, the choice of the source of climate data (e.g., CHELSA or Worldclim; Fick and Hijmans 2017; Karger et al 2017) used for model calibration has been identified as a major source of uncertainty in SDMs (Baker et al 2016; Dubos et al 2022a). In spite of this, most SDM studies aimed at projecting future distributions use climate data from a single source.…”