High-frequency reconstruction of mid-summer temperature from annual height increment of Scots pine at the northern timberline since 1846

Annual height increments of 35 Scots pine (Pinus sylvestris L.) trees from the northern timberline (68°30′N, 27°30′E, 220 m a.s.l., Laanila, North Finland) and monthly climate data from two meteorological stations, Sodankylä (from 1908 to present) and Ivalo (from 1958 to present) were used in climate and growth comparisons. The measured growth series were standardized using 67% splines. A height-increment chronology was built by averaging the indices. This chronology was further divided into high- and low-frequency components using reciprocal filters. Among the temperature variables, mean July temperature of the previous year correlated most significantly with height growth. We compared several simple linear reconstruction models based on the three height-growth chronologies (the unfiltered, high-pass and low-pass filtered chronologies) individually as predictors of the mean July temperature. The high-frequency reconstruction showed superior model performance in calibrations. However, only calibrations using climate data from the nearest Ivalo station were time stable and showed reasonable reconstruction skill. The coefficient of determination (R2) in the final model during calibration period (1958-1998) is 0.67.