Table 1.
Details of the nine general circulation models (GCMs) used in this comparative study.
Fig 1.
Standard deviation of the climatic conditions predicted by the GCMs.
Standard deviation of the temperature variables (BIO1-BIO11) and the precipitation variables (BIO12-BIO19) predicted by the nine General Circulation Models for the last glacial maximum. In general, temperature predictions are more robust for the oceans than for the continents, while precipitation errors are distributed in both seas and continents. Temperature predictions for the last glacial maximum are highly heterogeneous for cold climates, including the mountains, while predictions for tropical, warm and desert environments are more similar between models. On the other hand, tropical areas have highly heterogeneous predictions about precipitation, while predictions for the temperate and cold environments show better agreement.
Fig 2.
Maps identifying the areas where the bioclimatic predictions for the last glacial maximum show between-model agreement (in dark red) separated by (A) temperature layers (BIO1 to BIO11) and (B) precipitation layers (BIO12-BIO19).
Pink areas are those with more climatic uncertainty (GCMs predict different values for temperature and precipitation). These maps are based on the quartile coefficient of dispersion (see methods), which takes into account the dispersion of the predictions related to the actual range of the predictions.
Fig 3.
Standard deviation of BIO1 (annual mean temperature) (A) and BIO12 (annual precipitation) (B) in relation to latitude.
Our analysis indicates that these variables show an opposite latitudinal distribution of their uncertainties. Temperature predictions diverge at high latitudes, while precipitation predictions have high standard deviations in the tropics.
Fig 4.
Boxplot showing the correlation values for the temperature variables (in red) and the precipitation variables (in blue), of each General Circulation Model (GCM) compared with the rest of the layers from the other GCMs.
Although there are discrepancies in certain variables, temperature variables are highly congruent between models. On the other hand, precipitation variables show more discrepancies between models. COSMOS is the most different model in relation to its predictions about precipitation. Points are outliers (located 1.5 times the interquartile range above the upper quartile and bellow the lower quartile, which is the default definition of outlier in the R function boxplot).
Fig 5.
Hierarchical cluster grouping the nine GCMs by the correlation of their predictions for all 19 bioclimatic variables.
Table 2.
Results from the hierarchical clustering analysis (k = 4) identifying the groups of general circulation models (GCMs) that have similar predictions for each variable at a global scale.
Fig 6.
Boxplot showing the correlation between GCM predictions for each bioclimatic variable.
Fig 7.
Areas with high agreement between models.
Red areas show high agreement across the 19 bioclimatic variables (quartile coefficient of dispersion lower than 0.5 all variables at each cell).