Abstract:
A new reconstruction of the vegetation cover for New Zealand at the Last Glacial Maximum (LGM) is presented, based primarily on a database of 66 pollen site records and a more limited range of plant macrofossil and coleopteran records. Extensive forest is evident only from Auckland northwards. Conifer–broadleaf forest similar to that in the region today, but with Agathis australis scarce, persisted in the far north, whilst Nothofagus trees and a range of shrub taxa characterised the more open forests elsewhere in Northland. Survival of Nothofagus-dominated forest in coastal and exposed continental shelf locations to the southwest of Auckland and northwestern South Island is also indicated. Beyond these regions, vegetation cover comprised shrubland- and grassland-dominant communities, with the latter more prominent in eastern areas, to the south and presumably at higher altitudes. Nevertheless the survival of forest trees is indicated unambiguously in most regions apart from the eastern South Island. Thus the concept of ‘micro glacial forest refugia’ in New Zealand remains supported by this latest glacial vegetation reconstruction and we draw possible parallels with the developing but contentious concept of ‘northern cryptic refugia’ in Europe. Recent assertions that pollen and beetle reconstructions of the New Zealand LGM vegetation patterns diverge significantly are not supported by this analysis. Rather, the two proxies are readily reconciled if the term ‘woody’ as indicated by coleoptera is not restricted to tall forest trees but extended to the widespread woody shrub and small tree elements of the New Zealand flora. Regional distinctions in the LGM vegetation reconstruction concur broadly with the contemporary vegetation pattern, suggesting that, along with temperature depression and likely drier growing conditions, a zonal circulation regime with prominent southern westerly winds was important at 21 ka, as it is today. Pollen–climate modelling of the extent of temperature change is limited by the lack of LGM analogues and the relatively small number of cool climate indicator taxa in the modern (predeforestation) pollen training set. Nevertheless, a group of 17 sites which display a >90% match with the modern training set, and therefore should provide more reliable estimates, give an average reconstructed mean annual temperature depression of −6.01 ± 1.91 °C using the partial least squares method. This value, towards the upper end of the range of most independently derived estimates, accords with recent reconstructions of LGM temperature depression derived from glacial modelling.