To obtain robust quantitative reconstructions of climate changes in NW Amazonian lowlands, four transfer function methods will be adopted, which have been used extensively for Holocene climate change in Europe and Asia, proving the reliability of these approaches: Modern Analogue Technique (MAT) compares past assemblages to modern pollen assemblages (analogues) and assesses the degree of similarity between the two datasets; Weighted-Averaging Partial-Least-Squares (WA-PLS), relies on calibration between climatic variables and pollen assemblages, and uses a regression method to suppose relationships between fossil pollen and climate parameters; Random Forest (RF), based on the estimation and combination of a large number of regression trees from on a randomised group of modern pollen samples; Boosted Regression Tree (BRT) is a very new machine learning method only recently adopted for palaeoecology, and uses random binary splitting and cross-validation to predict the relationship between climatic variables and pollen assemblages. These models will be run using Rstudio, using scripts that I have already access to for European-based reconstructions, and applying data from existing pollen records available on the public repository Neotoma, as well as data available from supervisor and collaborators. New advanced techniques include GAMs, which will require a secondment at Aarhus University, and investigating the applicability of the latest developed transfer functions (i.e. BRT and RF) to tropical environments, during a short stay at the University of Montpellier.

Once the climatic variables have been reconstructed for each fossil pollen record (e.g. Laguna Pindo, Ayauch, Lake Sauce, Kumpak), we will compare the results at each site to understand whether there are any shared responses to past changes in temperature, precipitation and seasonality, and determine the degree of influence of the reconstructed palaeoclimatic parameters on vegetation dynamics. We will utilise any other available datasets from the records (geochemical, isotopic, sedimentological) to identify evidence of any fluvial response to climate that may have modulated changes in vegetation. The timing of vegetation change between records will also be compared, to understand if there are any coeval responses to change which may be related to regional variations in climate. Emphasis will be placed to study the vegetation responses to the Mid Holocene Dry event (9–5 cal kyr BP), as a potential past analogue of the projected scenarios for the ongoing global change in terms of hydrological differences.

The reconstructions will be compared to existing records of climate change in the region, such as speleothem records from Amazonian lowlands such Cuíca Cave, Paraíso Cave and Santiago Cave. Regional records from higher altitudes may be used for regional comparison, such as oxygen isotope (𝛿18O) records of lake Pumacocha and the Husacarán ice core from the Andes, the record of El Nino Southern Oscillation (ENSO) from Laguna Pallcacocha, and the calcium record of glacial advance from Queshquecocha. This will allow us to evaluate whether the reconstructions are consistent with existing records, but also enable the identification of any disparities between the climatic records and thus help understand how the NW Amazonian lowlands have reacted to past changes. This represents an important step in starting to fill our knowledge gaps regarding the heterogeneity in response to external drivers throughout the region as the effects of precipitation and seasonality have barely been explored so far.

These will be determined by examining the records of vegetation change and their reactions to the reconstructed climate parameters. We will attempt to determine rates of change, identify whether there are any time lags in the response of vegetation assemblages, understand whether certain species are more sensitive to external changes than others, and establish if there are specific combinations of favourable conditions to enable the resilience of forest communities before the crossing of tipping points. By working with local institutions and possibly getting into contact with local and international conservation agencies, a short documentary, photographic articles and reports will be produced outlining the findings of the project and advising practices to help ensure the protection of Amazonian lowlands.