High-elevation inter-site differences in Mount Smolikas tree-ring width data

We present the longest high-elevation tree-ring width dataset in the Mediterranean reaching back to the 6th century CE. The network includes 101 living and 92 relict Pinus heldreichii Christ trees from four differently exposed sites in the 2100-2200 m a.s.l. elevation range of Mt. Smolikas in the Pindus Mountains in Greece. Though the sites were all sampled within a distance of< 1 km, inter-site correlations are surprisingly low (r(1550-2014) = 0.65-0.87), indicating site exposure might affect tree-ring formation. We here explore the consequence of exposure differences on the climate signals in an eastern Mediterranean treeline ecotone. Temporally stable growth/climate relationships reveal similar seasonal patterns among the four sites, but differences in signal strength. P. heldreichii growth at Mt. Smolikas is significantly controlled by temperature in April (r(1951-2014) = 0.33-0.50) and precipitation in June-July (r(1951-2014) = 0.23-0.42), which emphasizes the overall importance of an early growth onset and subsequent moisture conditions. The association between stem growth and April climate is strongest in the South-facing stand, supporting the significance of higher insolation rates at this thermally privileged site. Strongest summer precipitation signals are found in the NE-facing stand, where trees seem to benefit least from an early growth onset and where reduced meltwater supply may enhance the dependency on early summer precipitation. The significance of spring temperature on tree growth in all four sites constrains the emergence of a distinct summer precipitation signal in the Mt. Smolikas high elevation ecotone. Exploration of the site-specific influences on a new millennium-long tree-ring width dataset is an important step towards an improved understanding of long-term climate variability in the Eastern Mediterranean. Site-related differences in climate sensitivity in the high-elevation tree-ring network at Mt. Smolikas indicate that both temperature and precipitation during different seasons could potentially be reconstructed if distinct site exposures (S versus NE) are considered.