International Geologiical Congress - Oslo 2008

Home

Search Abstracts

Author Index

Symposia Programmes

Sponsors

Help

 

 

CGC-09 Glacial-interglacial vegetation dynamics

 

Modern pollen-climate calibration models for northern and eastern Europe

 

Heikki Seppä, University of Helsinki (Finland)
H. John B. Birks, University of Bergen (Norway)
J. Sakari Salonen, University of Helsinki (Finland)
 

 

Modern ecological and climatological knowledge is the basis for using proxy data for palaeoecological and palaeoclimatological inferences and reconstructions. Our aim has been to collect a high-quality modern pollen surface sample collection from Europe, to use the modern samples to develop pollen-climate calibration models, and to test the performance and inference power of the models by numerical cross-validation with modern climate data. We have collected over 400 samples from small- to medium size lakes from Norway, Sweden, Finland, Estonia and Russia, ranging from temperate forest to arctic tundra. Parts of this surface sample set have been used for constructing transfer functions for key climate variables, mainly July mean temperature and annual mean temperature. The transfer functions have been developed with weighted averaging partial least squares regression, using all terrestrial pollen and spore types. A calibration model for July mean temperature has been developed and used in the North-European tree-line region. The performance of this model is evaluated by leave-one-out cross-validation. The cross-validated root mean square error of prediction (RMSEP) of the model is 1.0°C and the coefficient of determination (r2) between observed meteorological July mean temperature values and those predicted by the model in leave-one-out cross-validation is 0.72. A calibration model for annual mean temperature, mainly used in central and southern Scandinavia and the Baltic region, has RMSEP 0.89°C and r2 0.88. These figures indicate high performance statistics for our transfer functions compared with other inference models. This is probably because of standardization of our surface-sampling and pollen-analytical procedures, careful selection of the surface sample sites with consideration of the relevant pollen source area, the simple patterns of vegetation zones in the study area, and the mostly natural floristic composition of the forest in northern Europe.

 

CD-ROM Produced by X-CD Technologies