Warm-season temperatures since 1600 BC reconstructed from Tasmanian tree rings and their relationship to large-scale sea surface temperature anomalies

We describe an improved tree-ring reconstruction of mean warm-season (Novem ber-April) temperatures for Tasmania from Huon pine. This record extends ba ck to 1600 BC and is based on a tree-ring chronology that was processed to retain as much low-frequency variance as possible. The resulting reconstruc tion explains 46.6% of the variance and verifies significantly when compare d to withheld instrumental data. Cross-spectral analysis of actual and esti mated temperatures over the 1886-1991 common period indicates that most of the unexplained variance is at periods <12 years in length. At periods >12 years, the squared coherency ranges between 0.6-0.8, and the cross-spectral gain indicates that the amplitude of the reconstruction is a nearly unbias ed estimate of the true temperature amplitude. Therefore, this reconstructi on should be especially useful for studying multi-decadal temperature varia bility in the Tasmanian sector of the Southern Hemisphere over the past 359 2 years. To this end, we examined the time evolution of low-frequency tempe rature amplitude fluctuations and found evidence for a 35% amplitude reduct ion after AD 100 that persisted until about AD 1900. Since that time, the l ow-frequency temperature amplitude has systematically increased. We also sh ow how this reconstruction is related to large-scale sea surface temperatur es (SST) in the Indian Ocean and eastward to the dateline. Pointwise correl ations between the Tasmanian record and SSTs reveal a relationship that ext ends across the southern Indian Ocean and towards the Arabian Sea. This pat tern is largely determined by inter-decadal temperature variability, with c orrelations in this >10-year bandwidth commonly exceeding 0.6 over most of the southern Indian and southwestern Pacific sectors. A rotated empirical o rthogonal function analysis reveals that the pattern of pointwise correlati ons found between the temperature reconstruction and SSTs is largely explai ned by the linear combination of three orthogonal modes of SST variability.