SiZer - A Tool for Inferring Significant Features in Environmental Reconstructions
by
Panu Erästö
Rolf Nevanlinna Institute, University of Helsinki
Coauthors: Lasse Holmström (Rolf Nevanlinna Institute, University of Helsinki), Atte Korhola (Department of Ecology and Systematics, University of Helsinki), Jan Weckström (Department of Ecology and Systematics, University of Helsinki)

Often a reconstruction of an environmental variable exhibits considerable variation and establishing features such as trends, minima and maxima can be difficult. Moreover, it is not easy to say how much of the seeming variation is really statistically significant. To answer such questions we propose a principled inference approach based on the recently introduced SiZer method that uses the scale space idea from computer vision. A family of smooths of the reconstructed environmental variable is considered simultaneously making possible inferences about significant trends at different time scales. The approach is applicable to situations where either the modern or the historical data set or both are regarded as sources of randomness.

We apply the proposed approach to Holocene temperature reconstruction based on a diatom fossil data set collected in the Finnish Lapland. The inferred cooling and warming periods match well with climatic history established from other records.

Date received: April 27, 2001

Copyright © 2001 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cahi-27.