DYNAMICS OF UPLAND PODOCARP BROADLEAVED FOREST ON MAMAKU PLATEAU, CENTRAL NORTH-ISLAND, NEW-ZEALAND/

Forest composition was examined on one hectare of upland Dacrydium cup ressinum-Prumnopitys taxifolia/Weinmannia racemose-Beilschmiedia tawa forest on south Mamaku Plateau, central North Island, New Zealand, a s ite of catastrophic volcanic disturbance c. 1900 years ago. The phasic model (gap, building, mature) of the forest growth cycle was used. Me an age of sampled stems differed significantly between gap (13 yr), bu ilding-phase (68 yr), and mature forest (252 yr), which comprised 10%, 50%, and 40% respectively of the area and persisted on average for 20 yr (gaps) and 60 yr (building phase). 'Expanded' gaps averaged 0.008 ha in size. Tree ferns - mostly Dicksonia squarrosa and Cyathea smithi i - were numerically dominant throughout, and dominated basal area in younger (< 80 yr) forest; D. cupressinum dominated basal area in older (> 80 yr) forest. Current diameter growth rates among broadleaved spe cies followed the ranking: Griselinia littoralis, Elaeocarpus dentatus >B. tawa>Ixerba brexioides, W. racemosa, and were significantly faster during building (3.9 mm yr(-1)) than mature (2.5 mm yr(-1)) phases. E mergent D. cupressinum and P. taxifolia sampled were aged between simi lar to 400 and similar to 1000 yr; established seedlings of both speci es were rare. Much younger Prumnopitys ferruginea occurred commonly bu t only up to small tree size; Podocarpus totara was present only as lo ng-dead fallen trees. Broadleaved species were all much younger (<simi lar to 250 yr). Shade-tolerant P. ferruginea, B. tawa, and I. brexioid es regenerate continuously in the understorey of high forest, although P. ferruginea requires some canopy opening to develop beyond sapling size. Less tolerant G. littoralis, D. squarrosa, C. smithii, and to so me extent W. racemosa are gap invaders; D. squarrosa, C. smithii, and W. racemosa also continue to establish during the building phase. Elae ocarpus dentatus maintains a pool of 'advance growth' which develops r apidly if light levels increase, as in gaps. Despite similar populatio n structures, marked differences in the microsites preferred for estab lishment are evident between some species. Larger-scale studies over l onger time periods and over larger areas are needed to determine defin itively the regeneration strategies of the conifers. The unexpectedly high proportion of building-phase forest suggests a period of substant ial canopy collapse within the past century. Widespread fallen podocar ps on the forest floor indicate the presence of dense conifer forest o n the site in the discernible past, lending support to the extension o f the 'lozenge' regeneration model suggested initially for Agathis aus tralis to tall podocarps as well.