Site map and lengths of individual segments included in final chronology. Green triangles are tree-ring sites and red diamond is location of Duckhole Lake from which a sediment core has previously been obtained (Saunders et al. 2013). Duckhole Lake is, on average, ca. 17 km away from the tree-ring sites. Site names listed are A: Abrotonella Rise; K: King Billy Saddle; M: Moonlight Ridge; Ms: Mesa; O: Ooze Lake; R: Reservoir lakes; W: Mt Wylly. Photos of the sediment core and the lake it came from are also inset in the figure.

The 12 different age-aligned growth classes used to produce the initial mRCS chronology. The classes are defined by three criteria, the first is the length/age of the series; if the age is less or greater than the mean age of all trees. The second criterion is the juvenile growth rate of the trees, whether that rate is above or below the mean for the age class. The final criteria is the rate of growth during the middle third of the series; if that rate is less, greater or equal to the mean growth rate during the same biological age of all the trees of similar length. For example, the top left panel contains old trees with high initial and mid-life growth rates. The lower right panel contains younger trees, having an initial growth rate that falls within the mean of that for the age-class and a lower-than-average mid-life growth rate.

Process of piecewise adaptive smoothing of the mRCS chronology. a. Individual indexed series making up the chronology. 100-year smoothing spline applied to all series. Although there are many series still with low growth into the modern era, there is an increasing number of series with higher index values towards the modern end of the collection, thus leading to the steady increase in chronology index values (next panel). b. The mRCS chronology prior to piecewise adaptive smoothing. c. Fitting of smoothed segments to original mRCS chronology, tie points of piecewise segments have been smoothed with a 60-year smoothing spline. d. Final chronology (gray) after piecewise adaptive smoothing and variance stabilization. Thick black line is the series smoothed using a 20-year Gaussian smooth. e. Sample depth with RBar calculated for 100-year segments successively lagged by 20 years. Note that this differs in its calculation from the interseries correlation reported above from which all medium- to low-frequency variability has been removed by a 32-year spline used for crossdating purposes.

Wavelet spectrum for the new chronology. The ‘classic’ Morelet wavelet spectrum base has been used. Significant spectral power at the p < 0.1 level is outlined in black.

Distribution of tree growth initiation years (blue) and termination years (red) for all trees at all sites (locations on map) in southern Tasmania. 50-year bins have been used to account for time taken to grow to coring height and rings to centre, as well as wood that has rotted away from the outer surface of the tree. Sites of L. franklinii (green diamonds) and P.aspleniifolius (orange triangles) are all represented in the histogram. The two L. franklinii sites mentioned in the text, Spero and Pearces Basin are the more northerly sites of the species.

Relationship between mean monthly and seasonal temperatures over Tasmania and the new chronology (AGCD data for southern Tasmania: 43–44°S, 146.3–146.8°E. A ‘p’ in front of the month indicates a month that is prior to the nominal growing season, whereas a ‘c’ indicates the current growing season. There is a ‘p’ in front of both seasons beginning in July.

Stability of relationship between the new chronology and mean temperature over time for three different seasons: Jul–Oct, Dec–Feb and Jul–Feb. Three different climate data sets have been used. AGCD is the Australian Gridded Climate Data; BRK is Berkeley Earth data; and CRU is the Climate Research Unit 4.05 data. All data are for Tasmania only. The vertical dashed line marks 1920 when the number of meteorological stations available for inclusion in the gridded data sets rapidly increases, and hence the quality of the instrumental data likely improves considerably.