A Finnish chironomid- and chaoborid-based inference model for reconstructing past lake levels

Abstract

In the present study I utilize subfossil chironomid and chaoborid distributions in surface sediments of 68 shallow lakes. The aim is to develop a calibration model for past water-level reconstructions by applying weighted averaging-partial least squares (WA-PLS) techniques and to evaluate its potential applications and limitations. This study considers water depth at sampling sites, rather than maximum lake depth. The best of the water depth models developed uses three components and has a cross-validated coefficient of determination ( r 2 jack) of 0.68 and root-mean-squared error of prediction (RMSEP) of 0.78 m. The model performance is tested on the sediment sequence of a previously studied lake from southern Finland that is known to have experienced past fluctuations in its water level. The water levels inferred are compared with results of chironomid-inferred air temperature reconstruction to ease separation of the effects of the variables. The reconstruction shows consistent results similar to those of previously published cladoceran planktonic:littoral ratios (P:Ls) from the same lake. However, the results indicate that factors other than depth and temperature, such as pollution, may possibly distort the inference results. The results suggest that in applying the midge-based water depth calibration model, it would be advantageous to use it together with an inference model for temperature and preferably in a multiproxy content where changes in water chemistry may be detected. The model can be useful in studies on past effective moisture variability that is closely related to climatic changes.