Pollen and climate dataset at southern Patagonia: evaluating weaknesses and strengths for quantitative palaeoclimatic reconstructions

TONELLO, M. S; MANCINI, M. V.; DE PORRAS, M. E.; BAMONTE, F.; SOTTILE, GONZALO

Valdivia

Simposio; II International Symposium PAGES “Reconstructing Climate Variations in South America and the Antartic Península over the last 2000 years”; 2010

One of the important advances in pollen-based climate reconstructions hasbeen the increase of efforts to improve the statistical precision of pollenbasedreconstruction models. In this sense, pollen and climate dataset evaluationsare presented for southern Patagonia. The original pollen datasetconsisted of 221 georeferenced surface soil samples uniformly distributedbetween 46°-52°S and 67°-73°W and included 81 pollen types. Manual proceduresand graphical tools were used to identify duplicate samples and/orto evaluate taxonomic consistency in the pollen identifications. This qualityassessment led to the bestPollen dataset (bPD) that includes 155 samplesand 57 pollen types. A further selection consisted of eliminating samplesfrom over-represented vegetation units, generating a minimalPollen dataset(mPD) of 87 samples and 43 pollen types. Climate data were comparedbetween instrumental data and different Global Climate databases in orderto accurately represent modern climatic features. Some data overestimatedprecipitation values (50-240 mm) at locations close the Andes, while underestimatedvalues (50-100 mm) are registered at coastal stations. Temperatureestimations did not reveal significant differences between databases, andgenerally showed good agreement with the instrumental data. Calibrationmodels for annual precipitation (Pann) as well as annual (Tann), July (Tjul) andJanuary (Tjan) temperature were elaborated based on bPD and mPD, applyinga transfer function (WA-PLS) and the Modern Analog Technique (MAT). Theresults of cross-validation (r2, RMSEP, MaxBias) demonstrated that MAT modelsperformed slightly better than WA-PLS models although there were nosignificant differences based on using bPD or mPD. The models presented agood inference power to estimate Pann and Tjan, but moderate and poor successat estimating Tann and Tjul, respectively. In order to validate the calibration modelspotential, Pann and Tjan were reconstructed from two fossil pollen recordsof the Late Holocene, providing therefore key information on past regionalclimate variability at southern Patagonia.