Changes in summer-temperature correlation of tree-ring width from 1931–1960 to 1961–1990 at 525 ITRDB sites north of 55°N were not larger than stationary sampling noise

Question: Did the interannual sensitivity of tree-ring width to summer temperature decrease at northern sites (the 'divergence problem'), and are changes in climate-growth correlation ('non-stationarity') larger than sampling noise? Approach: We built 3,183 chronologies north of 30°N (525 north of 55°N) from raw ITRDB ring-width measurements with one method (30-year spline high-pass, biweight mean). We correlated them with CRU TS 4.09 June-August temperature and precipitation in adjacent 30-year windows. A stationary null splits the same 60 years at random into two sets of 5-year blocks, with one split for all sites, so it keeps spatial dependence. Results: North of 55°N, the mean correlation changed from 0.184 (1931-1960) to 0.144 (1961-1990); the mean change in Fisher z was -0.041 (95% CI -0.085 to +0.006; null p = 0.21). The cross-site variance of the change (0.067) was not larger than the null variance (median 0.074; p = 0.78). The SD of true site changes is below 0.095 in z, and at most 6.5% of sites can have a true decrease of 0.4 in z. At the 134 temperature-sensitive sites (full-period r >= 0.3), the mean r did not decrease (0.395 to 0.412). A 'significant in one window only' criterion flags 26.9% of sites, and the stationary null flags 26.3%. Sites selected by their early correlation lost 0.165 in r; the null with the same selection gives 0.150, so this loss is mostly regression to the mean. For all 3,182 sites from 30°N, the change in temperature and precipitation correlations was also inside the null range. Two secondary statistics (a cross-fitted selection test and an exploratory relation with local warming) suggest small real changes. For 1946-1975 vs 1976-2005, sensitive sites north of 55°N had a decrease of 0.074 in r (null p = 0.020), but the decrease was not significant with Berkeley Earth data and in 1951-1980 vs 1981-2010. Limitations: only total ring width, high-frequency variation and fixed JJA climate of grid cells; few northern chronologies continue after 2005; low-frequency trend divergence was not tested.

tree rings · dendrochronology · ITRDB · divergence problem · non-stationarity · climate-growth relationship · boreal forest · regression to the mean · null model · CRU TS · temperature sensitivity