Ring widths at 455 ITRDB sites north of 55° N fell as far below a 1921–1960 summer temperature fit in 1901–1920 as in 1961–2000

Question: did tree growth at high northern latitudes lose its link to summer temperature after 1960 (the divergence problem), or is the post-1960 offset an effect of the calibration period, detrending end effects or site screening? Approach: we used raw ring widths from 4,417 Northern Hemisphere sites of the International Tree-Ring Data Bank, CRU TS 4.09 and Berkeley Earth temperature, and ModNegExp and spline detrending. We fitted the index on June-August temperature in 1921-1960 and compared the mean residual after (1961-2000) with the mean residual before (1901-1920). We also cut series at the window end to test end effects, and selected sites in both time directions to test regression to the mean. All intervals come from a spatial block bootstrap. Results for sites at or north of 55° N: the 1961-2000 index was 0.37 SD below the 1921-1960 fit (95% CI -0.51 to -0.16; 463 sites). But at 455 sites with full 1901-2000 coverage, the residual was -0.44 SD in 1901-1920 and -0.38 SD in 1961-2000 (difference +0.06, CI -0.08 to +0.22). In 16 variants of season, temperature product and detrending, the difference was -0.05 to +0.13 SD and no interval excluded zero. The anomaly is a growth high in 1921-1950 (+0.41 SD above the flanking periods), at the same calendar years in sites sampled from 1990 to after 2010. Cutting series at the window end made the offset 0.10-0.13 SD less negative. Screened sites lost about 0.1 in r in both time directions; the estimated true change in r was -0.01 (CI -0.05 to +0.04). At 59 sites sampled to 2015 or later, the residual moved to -0.16 SD in 2001-2015. Limitations: with a 1931-1960 calibration and a 1901-1930 pre-period, the post-1960 residual was 0.28 SD lower than the pre-period residual, so the 1920s carry much of the contrast. Regions differ (a small post-1960 excess in North America with spline detrending, the opposite in Europe). We tested ring width only, with a linear single-season model and individual-series detrending.

tree rings · divergence problem · ITRDB · dendroclimatology · boreal forest · temperature sensitivity · detrending end effect · regression to the mean · CRU TS · out-of-sample test