Prediction of near-bottom water salinity in the Baltic Sea using Ordinary Least Squares and Geographically Weighted Regression models

Abstract

A map of spatial salinity distribution in the bottom water layers of the Baltic Sea is presented in this paper. The map has been constructed based on the data obtained from the ICES Dataset on Ocean Hydrography. The typical salinity values and the depth of halocline location in the major basins of the Baltic Sea are also presented.While the spatial salinity distribution is commonly derived by interpolation from the available data set, the linear regression model has been applied in this work. The analyzed data cover the period between 1913 and 2011, with a spatial resolution of ca. 10 km. In order to prepare the salinity map for the bottom water layers in the Baltic, the relationships between the salinity, depth and the distance from the Danish Straits have been determined by using Geographically Weighted Regression (GWR). Next, the salinity map was created by using the maps of regression coefficients, the digital elevation model (DEM) of the Baltic Sea, and the map of Euclidean distance from the Danish Straits. Subsequently the salinity values in the water layer above and below the halocline that are typical for the specific Baltic basins as well as the depth of location of the halocline were calculated based on the data extracted from the map by random point sampling.The calculated salinity values for the upper layer were similar to the values reported in the current publications on the subject of the Baltic Sea. On the other hand, the obtained salinity values for the layer below the halocline were slightly lower than those found in the literature, which is attributable to different methodology used. The obtained results demonstrate that GWR is a reliable tool for estimating the natural variation of salinity in the Baltic Sea. At the same time, we conclude that the Ordinary Least Squares regression should not be used to analyze similar data.