Quantitative Calibration of Remote Mountain-Lake Sediments as Climatic Recorders of Air Temperature and Ice-Cover Duration

Abstract

A combination of empirical modeling and a diatom-based transfer function was developed to reconstruct air temperature and ice-cover duration through the study of lake sediments. By using a thermal degree-day modeling approach, ice-cover duration on European mountain and sub-Arctic lakes is found to be very sensitive to temperature change. For example, our model, which incorporates a weather generator, predicts a 100-day shortening in ice-cover duration for a 3°C temperature rise for catchments at elevations of 1500 m in the Southern Alps and the Pyrenees. For the more maritime lakes of Scotland, 30% higher sensitivities (130 d per 3°C) are found, whereas lakes in northwest Finland, in a more continental setting, have only half the sensitivity (50 d per 3°C). A pan-European data set of the species abundance of 252 diatom taxa in 459 mountain and sub-Arctic lakes has been compiled and taxonomically harmonized. Transfer functions were created that relate both seasonal air temperature and ice-cover duration to diatom species composition on the basis of a weighted averaging–partial least squares (WA-PLS) approach. Cross validation was used to test the transfer functions. For ice-cover duration the pan-European data set yields an R-squared value of 0.73, a jack-knifed R-squared value of 0.58, and a residual-mean-square error of prediction (RMSEP) of 23 days. A regional, northern Fennoscandian transect (151 lakes, 122 taxa) yields a jack-knifed R-squared value of 0.50 and an RMSEP of 9 days. For air temperature the pan-European database displayed greatest skill when reconstructing winter or spring temperatures. This result contrasts with the summer temperatures normally studied when using local elevation gradients. The northern Fennoscandian transect has a remarkably low winter RMSEP of 0.73°C.