Palaeoreconstruction of the Baltic Ice Lake in the Eastern Baltic - The Baltic Sea Basin - page 191

Fig. 9.1 ) . Pärna ( 1960 ) also suggested that water level lowered ca. 30 m when the
ice retreated from the Pandivere/Neva to the Palivere ice-marginal zone (12,700
corrected varve years BP; Kalm 2006 ; Fig. 9.1 ) . However, this low water level is
marked only by few glaciofluvial flat plains near Tallinn (Pärna 1960 ; Fig. 9.1 ) .
Similar low water level, the so-called g-delta level, also existed in Finland (Sauramo
1958 ) as a result of a connection between the Baltic Sea and White Sea. Later studies
(Kvasov and Raukas, 1970 ) , however, showed that such a connection was unlikely
and there was no physical reason for a low water level before the BIL stage BI
(Donner 1995 ) .
The aim of the current chapter is to correlate the Baltic Ice Lake coastal
formations in eastern Baltic using a shoreline database and GIS analyses. The
palaeogeographical reconstructions were used to study drainage routes of proglacial
lakes and BIL, especially how and when Lake Peipsi and Võrtsjärv isolated from
the BIL.

9.2 Methods

Reconstruction of the BIL shorelines and bathymetry in eastern Baltic area were
based on GIS analysis, by which interpolated surfaces of water levels were removed
from the modern digital terrain model (DTM; Rosentau et al. 2004 ) .
The interpolated surfaces of water levels were derived using the late-glacial
shoreline database (Vassiljev et al. 2005 , Saarse et al. 2007 ) . The late-glacial shore-
line database covers more than 1,200 sites from eastern Baltic, including shore
displacement data for Estonia (Vassiljev et al. 2005 , Saarse et al. 2007 ) , Latvia
(Grinbergs 1957 , Veinbergs 1979 ) , NW Russia (Markov 1931 , Shmaenok et al.
1962 ) and southern Finland (Donner 1978 ) . However, statistical analyses showed
that roughly half of this data does not match water level reconstruction require-
ments due to inaccurate elevations or erroneous correlation of different shore
marks. The reliability of shoreline data was verified by different methods. First,
sites with altitudes not matching with neighbouring sites were eliminated. Then,
point kriging interpolation with linear trend was used to interpolate water level
surfaces. Kriging is advantageous because it interpolates accurate surfaces from
irregularly spaced data and it is easy to identify outliers in the data set. Residuals,
the difference between the actual site altitude and the interpolated surface, were
calculated and used to check the shoreline data reliability so that sites with resid-
uals more than 1 m were discarded. Finally, interpolated water level surfaces

were calculated using for A 1 - 52, A 2 - 77, BI - 111, BII - 88 and BIII -
164 sites.
BIL stage A 1 correlates with Pandivere/Neva ice-marginal zone (Fig. 9.1 ) dated
13,300 corrected varve years BP (Saarnisto and Saarinen 2001 , Hang 2003 , Kalm
2006 ) . BIL stage A 2 correlates with Palivere ice-marginal zone (Fig. 9.1 ) dated
12,700 corrected varve years BP (Hang 2003 , Kalm 2006 ) . The age of stage A 2
in earlier studies was older than Palivere stade; however, recent studies (Rosentau
et al. 2007 ) indicate that it formed during the Palivere stade.

±