Résumé
Quantitative climate reconstructions are fundamental to better understanding past environmental changes and evaluating climate simulations. The proliferation of pollen-based reconstructions during the past decades has been instrumental in improving our understanding of past climate dynamics across various spatial and temporal scales. However, this knowledge has been mainly concentrated in North America and parts of Eurasia, and very few quantifications exist in the tropics. This global data imbalance is partly due to the sparser network of supporting pollen records in these regions and, in equal proportions, to the limitations of the most commonly used reconstruction techniques (e.g. the analogue technique or WA-PLS) in dealing with the specificities of tropical vegetation. To address this problem and produce the much-needed climate quantification from tropical regions, we propose using the probabilistic method CREST (Climate REconstruction SofTware) that uses probability density functions ("pdfs") fitted on modern occurrence plant data to reconstruct environmental parameters. CREST, which has been successfully employed in Africa and South America, offers many advantages over the classical approaches, including 1) a higher flexibility of application, 2) a better capacity to estimate uncertainties, and thanks to the recent developments of a dedicated R package crestr that includes a global calibration dataset, 3) CREST is applicable in every environment where plants currently grow. Considering these advantages, the large-scale application of CREST to quantitatively reconstruct important climate parameters from the existing tropical fossil pollen records should 1) help better integrate and interpret regional proxy compilations, 2) shed light on the spatiotemporal climate variability of tropical regions, and 3) determine the main modes of tropical climate variability. With this contribution, we will showcase the use of the crestr package with a novel temperature reconstruction derived from the 270,000-year-long, high-resolution pollen record from Laguna Fuquene in Colombia and discuss how this type of analysis could be generalised to determine spatial patterns of climate change from multi-record reconstructions [Copyright Author(s) 2023. CC Attribution 4.0 License: https://creativecommons.org/licenses/by/4.0/legalcode]