A dendroecological reconstruction of storm events in the forests of central and southern North Island, New Zealand: 1300-2000 AD

Abstract

Understanding of climate change in the New Zealand region is largely based upon palaeoclimatic reconstructions, due to the region’s short settlement history. Storms are a feature of New Zealand’s climate, and various proxies for storm history have been reconstructed. A review of these proxies, disturbances in New Zealand forests, and the state of dendrochronology in New Zealand suggested that a fine resolution record of storm events could be derived from tree-ring studies. The research had three primary aims: (1) to develop and apply a methodology based on forest structure and dendroecology for dating storm events, (2) to test this method against known storm events, and (3) to reconstruct a regional history of storms for central North Island. Secondary aims were to compare the reconstructed storm history with other storm proxies, and to gain further insights into the role that storms play in the ecology of North Island forests. The study was intended to identify major regional storms. Eighteen sites were studied in central and southern North Island. Sites ranged from c. 400 to 2400 m2 in area and were selected to have light demanding species, such as Libocedrus bidwillii and Nothofagus spp., which have distinct annual rings. Dating of past disturbance events depended on the detection of releases and suppressions in tree-ring series and estimates for the timing of cohort initiations, based on the age structures of the populations. The results are of two types: (1) those relating to the methodology employed, and (2) those relating to the history of storminess in the region. Methodological results included defining the filtering mechanism used to detect abrupt growth changes, and field sampling procedures. The results proper indicated periodicity of storminess over the last 700 years. The effects of Cyclone Bernie and the storm of 1936 were used as modern analogues of forest response to storm events. Extreme wind events were demonstrated to cause abrupt changes in ring-width and to initiate cohorts. Clusters of cohort initiations, probably attributable to storms or periods of increased storminess, were dated to 1290-1320, 1580-1640, 1720-1750, 1860- 1890, 1930-1950, and 1980-1990 AD. All cohorts dated during this research fell within these periods, which constituted c. 23% of the time from 1200 to 2000 AD. The recruitment of Libocedrus peaked regionally in the early 1600s, suggesting that canopy disturbance events were frequent and widespread at this time. Peaks in abrupt growth changes that could be attributable to the effects of individual storms occurred in the 1600s, 1660s, 1830s, 1740s, 1880s, 1900s, and 1930s. Thus the early 1600s and 1880s were characterized by both abrupt growth changes and recruitment. There was strong evidence for a storm event in the Ruapehu district in the 1880s, with tree-falls, recruitment pulses, and releases at multiple sites around Mount Ruapehu. The storm history reconstructed was combined with other proxies to strongly suggest a period of increased storminess in the late 1500s and early 1600s. Other storms or periods of storminess probably occurred in the early 1300s, early to mid 1700s and late 1800s. The results suggest that the most damaging winds are from the east, often associated with the passage of ex-tropical cyclones. The El Nino-Southern Oscillation influences cyclone occurrence in the New Zealand region. Periodicities of this climatic phenomenon may also cause periodic changes in forest disturbance regime. Changes in storm frequency are likely to play an important role in regional, long term forest dynamics. The identification of storm events is strongly reliant on sample depth and the detection of major sustained releases. Confidently attributing ring-width releases to storms is dependant on supporting evidence from population age structures and aligned tree-fall directions. This limits the length of reconstructed storm histories. Attributing cohorts and abrupt growth changes to storm events may be facilitated by a multi-proxy study within a single catchment. Sediment pulses synchronous with the initiation of cohorts would provide evidence for older storm events and additionally enable reconstruction of both rainfall and wind characteristics of past storms.