Yet another assessment of climate change in the Baltic Sea area: Breakpoints in climate time series

The aim of the present study is to assess changes in the Baltic Sea climate based on different available meteorological data sources (ERA40 and ERA-INTERIM) and various published Baltic Sea climate indices. This regional assessment will be presented in relation to global climate change and assessments available from the literature. The climate of the Baltic Sea which is located between 50N and 70N is mainly influenced by the competition of westerly humid air flow and easterly continental type air masses and is therefore highly variable. We are investigating air temperature, wind speed, cloud cover, solar radiation and precipitation. Comparisons to climate indices of general relevance as the Baltic ice cover will be conducted. Using regression analysis we could confirm the following basic trends, increase in air temperature, increase in precipitation, increase in cloudiness. The increase in air temperature in the Baltic Sea area (0.02K/year) is much more rapid then the warming trend for the global air temperature (0.005K/year). The increase in cloudiness has resulted in an effective reduction of incoming solar radiation therefore the accelerated warming is not a result of increased solar radiation, but likely due to an increased net long wave radiation input. Further it has to be mentioned that not all available data sets confirmed the trend in cloudiness, ERA40 data show a nonsignificant decrease instead. No clear trend in the wind velocities could be detected, but wind velocities from ERA40 reanalysis project show an insignificant increase in wind speeds. Results from model runs with the GETM model (General Estuarine Transport Model, http://getm.eu) show sea surface warming consistent with the increase in heat flux forcing and with satellite observations. The warmer sea surface without an adequate warming in the deeper parts results in a much stronger vertical density stratification and consequently to reduced vertical mixing. A more thorough inspection of the available regional and global data provides some reasonable doubt concerning the application of least square regression analysis to the available time series. Indeed it can be shown by a test based on the F statistics that most of the analyzed time series cannot be considered as stationary and therefore drawing simple regression lines trough these datasets is statistically incorrect. Testing for structural breakpoints in these time series reveals for many investigated parameters and also for many tested climate indices the existence of such breakpoints in the 70–80ties of the last century. Therefore it has to be concluded that the simple trend estimation for many climate parameters is statistically incorrect. Instead for statistical investigations it has to be assumed that there exist either 2 different climate states with either 2 different means or alternatively with 2 different trends which have to be estimated separately.