Fig 1.
Study areas and species occurrence data.
(a) The three nested study areas and occurrence data of target species (Iris boissieri and Taxus baccata) in (b) the Iberian Peninsula at 5km2 cell size, (c) the Iberian Northwest at 5km2 (empty squares) and 1km2 (filled squares), and (d) the Peneda-Gerês National Park at 1km2 cell size.
Table 1.
Occurrence data (number of grid cells) available per species at each combination of spatial resolution and spatial extent.
Fig 2.
Multi-scale modelling framework.
General framework to test the scale-dependence of the performance of satellite-derived Ecosystem Functional Attributes (EFAs) as predictors in Species Distribution Models (SDMs).
Table 2.
Specific testable hypotheses for comparison of the performance and scale-dependence (in terms of spatial extent and resolution) of ecosystem functional attributes (EFAs) against traditional climate and land-cover datasets in Species Distribution Models (SDMs).
Table 3.
Final sets of predictors used to calibrate models.
The description and attributes of the original datasets are also provided.
Table 4.
Rationale for the four groups of models included in the modelling setup.
Fig 3.
Comparison of relative performance of the Area Under the Curve (AUC) between traditional (climate and land-cover)-based and satellite-derived Ecosystem Functional Attribute (EFA)-based models across all scale combinations and test species.
Performance of individual models (boxplots) showing the AUCmedian, two hinges (first and third quartiles), and two whiskers of each model filtered at AUC≥0.7 (empty-triangle signs represent the AUCmean). Filled-circle dots and crosses represent the AUCmedian and the TSSmedian, respectively, of the ensemble models. Different letters indicate significant differences among models (multiple comparisons of means were performed using Tukey's test at the 0.05 significance level).
Table 5.
Summary table of performance of the best ensemble models based on traditional (climate -CLI- and/or land-cover -LC-) and on satellite-derived Ecosystem Functional Attribute (EFAs), considering those individual models filtered at Area Under the Curve (AUC) ≥0.7.
The AUCmedian±IQR (Inter Quartile Range) and the top variables above a threshold of importance contribution (% >0.10) are showed per species, extent, and spatial resolution. See complete names of variables and extents in Table 3.
Fig 4.
Response curves of predicted habitat suitability for Iris boissieri to the most important predictors.
Response curves for predictors with the highest importance in traditional (climate and land-cover)-based (left) and Ecosystem Functional Attribute (EFA)-based (right) ensemble models for Iris boissieri at all combinations of spatial extents and resolutions.
Fig 5.
Response curves of predicted habitat suitability for Taxus baccata to the most important predictors.
Response curves for predictors with the highest importance in traditional (climate and land-cover)-based (left) and Ecosystem Functional Attribute (EFA)-based (right) ensemble models for Taxus baccata at all combinations of spatial extents and resolutions.
Fig 6.
Spatial projections of habitat suitability for Iris boissieri derived from Species Distribution Models (SDMs) based on traditional predictors (climate and land-cover) and on satellite-derived ecosystem functional attributes (EFAs).
Overlay maps of current potential presence-absence distributions predicted using an ensemble modelling approach per combination of spatial extent (IP, NW and NP) and resolution (1km and 5km) for Iris boissieri. IP: Iberian Peninsula; NW: Northwest IP; NP: Peneda-Gerês National Park.
Fig 7.
Spatial projections of habitat suitability for Taxus baccata derived from Species Distribution Models (SDMs) based on traditional predictors (climate and land-cover) and on satellite-derived ecosystem functional attributes (EFAs).
Overlay maps of current potential presence-absence distributions predicted using an ensemble modelling approach per combination of spatial extent (IP, NW and NP) and resolution (1km and 5km) for Taxus baccata. IP: Iberian Peninsula; NW: Northwest IP; NP: Peneda-Gerês National Park.
Table 6.
Comparison of spatial projections from traditional (CLI/LC-based) models and from Ecosystem Functional Attribute (EFA-based) models for Iris boissieri at all scale combinations.
The proportion of predicted suitable area is shown in brackets. IP: Iberian Peninsula; NW: Northwest IP; NP: Peneda-Gerês National Park.
Table 7.
Comparison of spatial projections from traditional (CLI and/or LC-based) models and from Ecosystem Functional Attribute (EFA-based) models for Taxus baccata at all scale combinations.
The proportion of predicted suitable area is shown in brackets. IP: Iberian Peninsula; NW: Northwest IP; NP: Peneda-Gerês National Park.