A summer temperature proxy from height increment of Scots pine since 1561 at the northern timberline in Fennoscandia

Height increments of 60 Scots pine trees were used to reconstruct mean June—August temperature variability at interannual to decadal scales from 1561 to 2004. Three standardization methods (67%, 33% flexible splines, and a fixed 22 years spline) were compared in building chronologies in order to optimize the frequency response in relation to major climatic forcing factors. The height-growth chronology built using the 33% spline standardization proved to have the most consistent and time-stable relationship with the summer temperatures. Among the monthly precipitation and temperature variables from previous June to current August, previous July shows the highest correlation with height growth. In addition, both previous June and previous August have significant positive correlations. Our final transfer model accounts for 32.5% of the dependent instrumental temperature variance between 1909 and 2004. The Fourier spectra of the height-growth chronology and mean summer temperature are very similar in appearance, both series having peaks at 2.7—3.2 years, 6.7 years and 15.7 years. Thus, the 444 years long summer temperature reconstruction is limited to high and medium frequencies. The coldest three summers in this record were experienced in years 1601, 1790 and 1903. Correspondingly, the summers of 1626, 1689 and 1598 were the warmest. The 1820s experienced the warmest 10-year mean, while the first decade of the twentieth century was the coldest. Among the 14 non-overlapping 30-year periods between 1561 and 1980, the period 1621—1650 was the warmest and the period 1591—1620 the coldest.