Résumé
Transfer functions are an efficient tool for the quantitative reconstruction of past climate from low to mid elevation pollen sites. However, the application of existing methods to high-altitude pollen assemblages frequently leads to unrealistic results. In the aim of understanding the causes of these biases, the standard "best modern analogue" method has been applied to two high altitude pollen sequences to provide quantitative climate estimates for the Lateglacial and Holocene periods. Both pollen sequences (Laghi dell'Orgials, 2130 m, S-W exposition and Lago delle Fate, 2240 m, E exposition) are located in the subalpine belt, on opposing sides of the St. Anna di Vinadio Valley (Italian Maritime Alps). Different results were obtained for the two sequences. The largest differences occurred in palaeotemperature reconstruction, with notable differences in both the values and trends at each site. These biases may be attributed to: 1) a lack of high elevation "best modern analogues" in the database of modern samples; 2) the problem of pollen taxa that have multiple climatic significance; 3) problems relied to the complexity of mountainous ecosystems, such as the phenomenon of uphill transport of tree pollen by wind. Possible improvements to the reconstruction process are discussed.