Fig 1.
The South American study domain.
A: Population distribution in 2010 [inhabitants per grid cell based on a 2.5 arc-minute resolution] [41]. B: Elevation [meters] [42]. C: Multi-year mean monthly temperature based on ERA5Land [∘Celsius] [43]. D: Temperature seasonality [-] based on bioclimatic variable 4 (BCV4). Panels C and D refer to for the base period (1991-2020) and additionally show the locations of all selected weather stations (totaling 216 in Brazil and 20 in Colombia), collected form the Brazilian National Institute of Meteorology (INMET, [44]) and the Colombian Institute of Hydrology, Meteorology, and Environmental Studies (IDEAM, contacto@ideam.gov.co), and ValAr-P (4 in Brazil and 2 in Colombia), respectively. The boundaries of all areas are based on the Database of Global Administrative Areas, GADM (version 4.1) [45]. All maps display values across the chosen study domain in South America (north of 40∘ South).
Table 1.
Summary of the selected global gridded temperature data sets (GGTDs). All data sets were downloaded at their native resolution and with a (sub-)eakdaily time resolution, except for Climatic Research Unit Time-Series data set (CRUTS), for which only monthly data were available. All data sets were accessed and extracted in 2023.
Table 2.
Definitions and calculations of temperature extremes, based on the Expert Team on Climate Change Detection and Indices (ETCCDI), and of bioclimatic variables (BCVs). ETCCDI calculations utilized daily maximum (TX) and minimum (TN) temperature timeseries; therefore, only data sets with daily resolution were selected (excluding Climatic Research Unit Time-Series data set, CRUTS). BCVs were calculated from monthly climatological averages and hence include all data sets. Indices and variables were computed using the global gridded temperature data sets at their native spatial resolutions as well as on a common 0.5∘ grid to account for uncertainties across different data sets.
Fig 2.
Bioclimatic variable 1 (BCV1).
A: Mean annual temperature [∘C] for ERA5Land. B-D: For all other global gridded temperature data sets (GGTDs), the difference [∘C] compared to ERA5Land is shown. All values represent averages over the base period (1991-2020). The maps are presented on a common 0.∘ grid.
Fig 3.
Bioclimatic variable 6 (BCV6).
A: Minimum temperature of the coldest month [∘C] for ERA5Land. B-D: For all other global gridded temperature data sets (GGTDs), the difference [∘C] compared to ERA5Land is shown. All values represent averages over the base period (1991-2020). The maps are presented on a common 0.∘ grid.
Fig 4.
Tropical nights index per time period (TR).
A: Number of tropical nights [-] for ERA5Land (TR). B-C: For all other global gridded temperature data sets (GGTDs), the difference in number [-] compared to ERA5Land is shown. All values represent averages over the base period (1991-2020). The maps are presented on a common 0.5∘ grid. Note that the Climatic Research Unit Time-Series data set (CRUTS) was excluded from the analysis, as TR was calculated using daily values.
Table 3.
Statistical evaluation of monthly temperature timeseries from grid cells against ground observations. The evaluation presents the Pearson correlation (PCC), the mean absolute error (MAE), and the root mean square error (RMSE) during the period 2011-2020 for each global gridded temperature data set (GGTD), averaged across each primary validation area (ValAr-P). Mean values represent averages across all validation areas (ValAr). Note that the number of weather stations evaluated varies across the different ValAr, as detailed in Tables S1 Table in the Supplement. Only stations with correlations statistically significant at the 95% confidence level were included in the analysis. Results are provided for all GGTDs at their native spatial resolution and on a common 0.∘ grid (in brackets).
Fig 5.
Spatial maps of evaluation metrics.
Comparison of grid cell and station values based on A: Pearson correlation coefficient (PCC). B: Mean absolute error (MAE). C: Root mean square error (RMSE) between ERA5Land and ground-based observations at the corresponding station on a monthly time scale in Brazil and Colombia.
Fig 6.
Boxplots comparing the temperature distributions of monthly timeseries based on different global gridded temperature data sets (GGTDs), averaged across each primary validation area (ValAr-P) in Brazil and Colombia, for the base period (1991-2020).
The figures are organized as follows: A: Amazonas (BRA4) B: Rio de Janeiro (BRA19) C: Rio Grande do Sul (BRA21) D: Sergipe (BRA26) E: Boyacá (COL7) F: Magdalena (COL20) in Brazil (BRA) and Colombia (COL), respectively. These boxplots are based on area-level temperature timeseries derived from GGTDs at their native spatial resolution.
Fig 7.
Climatological annual cycles of monthly temperature (∘C) for the six primary validation areas (ValAr-P) selected across Brazil and Colombia, calculated over the base period (1991-2020).
The figures are organized as follows: A: Amazonas (BRA4) B: Rio de Janeiro (BRA19) C: Rio Grande do Sul (BRA21) D: Sergipe (BRA26) E: Boyacá (COL7) F: Magdalena (COL20) in Brazil (BRA) and Colombia (COL), respectively. These climatological annual cycles are based on area-level temperature timeseries derived from global gridded temperature data sets (GGTDs) at their native spatial resolution.
Fig 8.
Calculated values of reproduction number (B) for the six primary validation areas (ValAr-P) selected across Brazil and Colombia, calculated over the base period (1991-2020).
Displayed values are monthly averages calculated from daily values calculated from environmental covariates listed in Yellow fever data and model, with temperature suitability calculated from daily temperature values taken from the four GGTDs. For CRUTS, only monthly temperature values were available, so reproduction number values were calculated directly on a monthly basis.