The products were Beyer, CHELSA-TraCE21k, EcoClimate, PALEO-PGEM-Series and WorldClim. Their spatial resolutions ranged from 0.008° for CHELSA-TraCE21k to 1° for PALEO-PGEM-Series, while their temporal coverage, underlying climate models and downscaling approaches also differed. The analysis did not determine whether the observed discrepancies originated in the climate simulations, downscaling procedures, proxy records or a combination of these factors.

To assess modelled cooling, the researchers extracted temperature-change estimates at 174 locations represented by two proxy compilations by Farrera et al. (1999) and Bartlein et al. (2011). The records consisted of fossil pollen and plant macrofossil data from radiocarbon-dated sediment records. Fifty-nine locations were inside mountain ranges delineated by the Global Mountain Biodiversity Assessment inventory, and 115 were outside them.

At each location, the authors compared the difference between present and LGM temperatures in each dataset with the corresponding proxy-derived difference. The five products generally estimated warmer LGM conditions than the proxies, indicating less cooling. Outside mountain ranges, median deviations from the proxy estimates for WorldClim, Beyer and EcoClimate were −0.08 °C, 0.28 °C and 0.14 °C, respectively. CHELSA-TraCE21k, PALEO-PGEM and the global grid cooling comparison had larger deviations of 1.11 °C, 1.35 °C and 1.33 °C.

Within mountain ranges, all five datasets had warmer median LGM temperatures than the proxy records. Median deviations were 2.1 °C for CHELSA-TraCE21k, 0.8 °C for WorldClim, 1.27 °C for Beyer, 0.9 °C for EcoClimate and 1.38 °C for PALEO-PGEM. The global grid cooling method, which subtracted a uniform 4.673 °C cooling estimate from the present-day CHELSA-TraCE21k raster, had the smallest median deviation in mountains at 0.33 °C. Its difference from the proxy estimates was significantly smaller within mountains than outside them at p < 0.01; the corresponding in-mountain versus out-of-mountain effect was not significant for the other methods.

A hierarchical generalised additive model using restricted maximum likelihood examined how dataset–proxy differences varied with latitude. The geographic pattern was not uniform. The datasets generally underestimated LGM cooling in the tropical mountains of Central and South America, including the northern and central Andes, and in South Africa. In parts of western and northwestern North America, continental Asia, Southeast Asia and Oceania, they instead overestimated cooling relative to the proxies. Differences were relatively stable across the Southern Hemisphere but more variable in the Northern Hemisphere. Some large negative differences occurred mainly at northern high-latitude locations, where proxy-derived LGM temperatures were 8 °C and 13 °C higher than corresponding dataset estimates.

To demonstrate the consequences of the temperature differences, the researchers reconstructed temperature-delimited treelines. For eligible mountain ranges, they extracted the temperature associated with the present-day treeline location from CHELSA v2.1, then located the LGM elevation at which each paleoclimate dataset and the global grid cooling method reached that temperature, allowing a margin of ±0.25 °C for resolution constraints. Elevation data were resampled to the spatial resolution of each climate product, and minimum, maximum, mean and standard deviation treeline elevations were calculated for each range.

Reconstructed treeline elevations differed by thousands of metres among datasets in some ranges, including the Tian Shan, Hindu Kush and central Andes, while other ranges showed closer agreement. Maximum differences exceeding 3,000 m occurred especially in the Americas, including the Rocky Mountains, Pacific Coast Ranges, Cordillera de la Costa and southern Andes. The global grid cooling method yielded a treeline estimate for 70 of 75 mountain ranges, compared with 67 for CHELSA-TraCE21k. Coarse-resolution products did not consistently capture smaller ranges and islands.

Spatial resolution also affected how products represented temperature gradients with elevation. In the Central European Highlands, Beyer, EcoClimate and PALEO-PGEM assigned single temperature values across broad elevation ranges, whereas CHELSA-TraCE21k and WorldClim showed a clearer decline in temperature with increasing elevation. The study was a global comparison and sensitivity analysis rather than an evaluation of a single reconstruction method. Its limitations included uneven proxy coverage, sparse records in many mountain ranges and uncertainties in proxy estimates. The results show that mountain and biome reconstructions based on paleoclimate products can depend materially on dataset resolution, regional bias and the method used to estimate LGM temperature change.

Subject of Research: Comparative assessment of Last Glacial Maximum temperature datasets and their effects on mountain treeline reconstructions

Article Title: Global variability in LGM cooling amongst paleoclimate datasets affects biome reconstructions in mountains

Article References: Rentier, E. S., Mottl, O., Pacheco-Riaño, L. C., Schultz, L., Seguinot, J., Wiersma, A. T., Grytnes, J.-A., & Flantua, S. G. A. (2025). Global variability in LGM cooling amongst paleoclimate datasets affects biome reconstructions in mountains. Frontiers of Biogeography, 18, Article e135871. https://doi.org/10.21425/fob.18.135871

Image Credits: AI Generated

DOI: 10.21425/fob.18.135871

Keywords: Last Glacial Maximum, paleoclimate modelling, treeline reconstruction, mountain biomes, mean annual temperature, proxy records, spatial resolution, biogeography, climate cooling, comparative analysis, Global, variability