Detecting tropical cyclones from climate- and oscillation-free tree-ring width chronology of longleaf pine in south-central Georgia

摘要

The knowledge of past tropical cyclone (TC) activity is vital to understanding patterns of current and future TCs and how they will impact society, infrastructure, and the natural system. Various historical, biological, and geological proxies are commonly used to reconstruct TC behavior; however, these records vary significantly in their temporal resolution. Tree rings are known to provide high-resolution data and have been proven as a valuable proxy record of past TC activity. Here we aim to produce a TC reconstruction based on tree-ring width data of longleaf pine (Pinus palustris Mill.) located in Lake Louise near Valdosta, Georgia. Results of growthclimate analysis showed that all climate and oscillation variables influence the tree growth, and removing this climate influence may be necessary to unmask the TC signal. For this purpose, we used stepwise linear regression to iteratively model tree growth with monthly climate factors (precipitation, temperature, drought) and monthly oscillation indices (ENSO, AMO, NAO, PDO) to obtain residual chronologies. Tree-ring chronologies were compared to TC data from the National Hurricane Center's North Atlantic Hurricane Database (HURDAT2) for a 150-km buffer zone around Lake Louise between 1894 and 1999. Of the storms that entered the buffer, 13 of the 17 -0.3 growth rings and 8 of the 10 -0.4 growth rings occurred in the year directly after a TC event. Our results revealed the strong climate and oscillation influence in the raw chronology and no significant TC signal. However, the removal of temperature and precipitation influence from raw chronology unmasked the TC signal (p 0.05). The TC signal was even more pronounced if influence of oscillation indices on growth had been also removed. The most pronounced TC signal was identified between most intense TCs ( 33 ms-1) and the climate and oscillation signal-free chronology. We emphasize that future research is needed for validation and elimination of the influence of other external growth drivers, as well as replication studies at multiple sites using similar statistical applications.